Here - Stuff

Volume 82, Number 6 November/December 2011

Volume 82, Number 6, November/December 2011Editor in ChiefJonathan M. LeesEastern Section EditorMartin C. ChapmanAssociate EditorsJennifer S. HaaseSusan E. HoughErol KalkanEduQuakes EditorAlan KafkaElectronic Seismologist EditorJohn N. LouieHistorical Seismologist EditorJohn EbelElectronic Supplements EditorKim OlsenManaging EditorMary GeorgeTypesetterRodney SauerCopy EditorLaura CarusoEditorial AssistantMelissa HouleSubscriptions to Seismological ResearchLetters (SRL): The subscription rate fornew institutions and other nonmembers inNorth America is $150 ($160 outside NorthAmerica). Members of the SeismologicalSociety of America receive SRL as a perquisiteof membership. Individuals may applyfor membership using the form printed nearthe back of this issue.Single copies: Many back issues of SRL areavailable from SSA Headquarters.Seismological Research Letters (ISSN 0895-0695) is published bimonthly in January,March, May, July, September, and Novemberby the Seismological Society of America,201 Plaza Professional Building, El Cerrito,California 94530. Periodicals postage paid atEl Cerrito, California and at additional mailingoffices.Seismological Research Letters was formerlypublished as Earthquake Notes (ISSN 0012-8287) from 1929–1986.Postmaster: Send address changes toSeismological Research Letters (SRL), 201Plaza Professional Building, El Cerrito,California 94530-4003.Communications regarding publications,apart from submission of manuscripts,should be addressed to the SeismologicalSociety of America, 201 Plaza ProfessionalBuilding, El Cerrito, California 94530.© 2011 by the Seismological Society ofAmerica. Printed in the U.S.A. by TheSheridan Press, Hanover, Pennsylvania.The Seismological Society of America201 Plaza Professional BuildingEl Cerrito, California 94530+1-510-525-5474; Fax +1-510-525-7204http://www.seismosoc.orgSeismological Research Letters—SubmissionsSeismological Research Letters (SRL) is a journal containing articles and itemsof broad appeal on topics in seismology and earthquake engineering. Articlessubmit-ted to SRL should be informational in nature and should be of currentinterest to a cross-section of SSA membership. Articles expressing some particularview about seismology or seismological research also will be accepted. Articlesthat contain original research results should be submitted to the Bulletin of theSeismological Society of America (BSSA). News and notes, special reports on particularearthquakes, seismic network summaries, information on computer hardwareor software pertinent to seismology, seismological equipment information, bookreviews, and letters to the editor also are solicited for publication in SRL.Consult the SRL Information for Authors at http://www.seismosoc.org/publications/srl/srl-authorsinfo.phpfor details about making submissions. In general,articles should not exceed 20 pages of double-spaced text (excluding figures) unlessapproved by the editor. Electronic supplements can be considered for SRL; theelectronic supplement policy is posted at http://www.seismosoc.org/publications/esupps.php. The SRL Editor in Chief is Jonathan M. Lees, srled@seismosoc.org.Upload submissions via SRL’s electronic submission system at http://srl.edmgr.com. Direct questions about the system to the managing editor at srl@seismosoc.org.Submissions to the Eastern Section of the SSA (ES-SSA) Section of SRLThe ES-SSA Section of SRL is devoted to the seismology of continental interiors.Articles pertaining to eastern North American earthquakes, intraplate seismotectonics,and earthquake engineering are particularly encouraged. Upload submissionsto the Eastern Section via SRL’s electronic submission system at http://srl.edmgr.com. The ES-SSA editor is Martin C. Chapman, Dept. of GeologicalSciences, Virginia Polytechnic & State Univ., 4044 Derring Hall, Blacksburg, VA24061; telephone +1-540-231-5036; fax +1-540-231-3386; e‐mail mcc@vt.edu.Appropriate review articles and tutorials are encouraged, as well as news andnotes pertaining to the Eastern Section of SSA. Referees are sought on all ES-SSApapers exceeding four published pages in length. Page charges for articles in theES-SSA Section are $25 for each printed page. The editor may allow exceptions tothe page charges under certain circumstances.On the CoverOn 22 February 2011 a magnitude 6.2 earthquake struck southeast ofChristchurch, New Zealand; this issue of SRL focuses on that event, the deadliestand most disastrous in New Zealand since 1931. Front: More than 100,000 buildingswere damaged in the Christchurch earthquake, from high-rise office buildingsto timber-frame homes like this one, located in the city’s heavily impacted centralbusiness district. The structure has been tagged with a red card, indicating that itis set for demolition (photo by İ.E. Bal; see more at Smyrou et al., 882–892).Back: The Christchurch earthquake provided an unprecedented dataset for testingthe effectiveness of 3D numerical modeling tools. Shown here are the spatialdistributions of peak ground velocity values for two simulations described inGuidotti et al., 767–782.Authorization to photocopy items for internal or personal use, or for the internal or personaluse of specific clients, is granted by the Seismological Society of America providedthat the appropriate fee of $3 per copy is paid directly to Copyright Clearance Center,ISSN 0895-0695, 222 Rosewood Drive, Danvers, MA 01923, USA; telephone 978-750-8400. Prior to photocopying items for educational classroom use, please contact CopyrightClearance Center at the above address. Consent for reproduction as described above doesnot extend to other types of copying, such as copying for general distribution, for advertisingor promotional purposes, for creating new collective works, or for resale. For permission toreprint material, please read “Permission to Reproduce Material from SRL” on page 994.

News and Notes (continued)Nominations Open Now for SSA AwardsSSA members are invited to submit nominations forthe following four SSA awards by 15 February 2011.Electronic submission is encouraged, though nominationsmay be submitted in hard copy.Please note that the principal nominator shouldintegrate the nomination letters and send one nominationpackage to ensure that all letters of endorsementreach the decision makers on time. Previousrecipients of all awards are listed on the SSA Website, www.seismosoc.org. No current member of theSSA Board of Directors shall be eligible for awardnomination.Electronic submissions should be e‐mailed in.TXT, .PDF or .DOC files to awards@seismosoc.org.Electronic submissions are encouraged, but hardcopies may be mailed or FAX’d to:Secretary, Seismological Society of Americac/o Susan Newman201 Plaza Professional BuildingEl Cerrito, California 94530 U.S.A.Fax: +1-510-525-7204You will receive a confirmation of receipt of yournomination. Names of the award winners will beannounced at the SSA Annual Meeting Luncheonin San Diego, California, 17 April 2012. The awardswill be presented at the 2013 SSA Annual MeetingLuncheon in Salt Lake City, Utah.Harry Fielding Reid MedalThe Harry Fielding Reid Medal of the SeismologicalSociety of America, formerly known simply as “TheSSA Medal,” is the Society’s highest honor. It isawarded for outstanding contributions in seismologyor earthquake engineering. A Reid Medal nominationpackage should include letters of nominationfrom at least two but no more than five Society members.Each nominating letter may have more than onesignatory, but each signatory should sign only oneletter. A single curriculum vitae and bibliography ofthe nominee may be included. To simplify communicationswith the Secretary, nominators of a particularnominee should select among themselves a chiefnominator for correspondence purposes. Submittingone integrated package insures that all endorsementletters will be included in the packet to the Board ofDirectors.For more information about the Reid Medalnomination process, contact Bob Engdahl, chairmanof the Reid Medal Subcommittee, at engdahl@colorado.edu.Charles F. Richter Early Career AwardThe Charles F. Richter Early Career Award honorsoutstanding contributions to the goals of the Societyby a member early in her or his career. A nomineemust satisfy the following criteria: 1) Regular orHonorary Member of the Society in good standing,2) the most recent academic degree must have beenawarded no more than six years prior to 18 April ofthe year that she or he is selected for the award, and3) not more than 40 years old on 18 April of the yearthat she or he is selected for the award.Any member of the SSA who is not on the RichterAward Subcommittee may nominate a candidate forthe Richter Award. A single nomination packagemust be submitted to the Secretary of the Societyat the above address no later than 15 February ofeach year. The package should contain 1) a letter ofnomination no more than 2 pages long summarizingthe nominee’s significant accomplishments, 2) a curriculumvitae including bibliography, 3) two to foursupporting letters no more than two pages long, atleast two letters of which must come from individualsnot currently employed at the nominee’s currentinstitution or the institution from which the nomineereceived her or his most recent degree, and 4)an eligible birth date and date of degree. Questionsmay be directed to Charlotte Rowe, chairman of theRichter Award Subcommittee, char@lanl.gov.Frank Press Public Service AwardThe Frank Press Public Service Award honors anyindividual, combination of individuals, or any organizationthat has served the profession of seismologyor the advancement of public safety or publicinformation relating to seismology. The Press AwardSeismological Research Letters Volume 82, Number 6 November/December 2011 761

TransitionsEdwin Victor Apel III recently moved to Risk ManagementSolutions from AMEC Geomatrix.Jonathan Bray, a professor of geotechnical engineering atUniversity of California Berkeley, has been selected as the 2012Joyner Lecturer.Ethan D. Brown recently moved to RDRTec from ScienceApplications International Corp.Andres Chavarria recently moved from Vialogy to SR2020Borehole Seismic Services.Vedran Lekic recently moved to the University of Marylandfrom Brown University.Wayne D. Pennington was inducted as the new presidentAmerican Geosciences Institute (AGI) at the recent GeologicalSociety of America Annual Meeting in Minneapolis,Minnesota. He is chairman of the Department of Geologicaland Mining Engineering and Sciences at MichiganTechnological University.Stephane Rondenay moved recently from the MassachusettsInstitute of Technology to University of Bergen.Takahiko Uchide recently moved from Scripps Institutionof Oceanography at University of California San Diego to theDisaster Prevention Research Institute at Kyoto University inJapan.Peter Yanev recently founded Yanev Associates, a consultingfirm that specializes in earthquake risk management and engineering.Your contributions of items for “Transitions” are most welcome.Please tell us about your awards, promotions, job changes,reorganizations, and retirements. Send announcements to SRLEditor Jonathan M. Lees in care of the SRL managing editor atsrl@seismosoc.org.764 Seismological Research Letters Volume 82, Number 6 November/December 2011 doi: 10.1785/gssrl.82.6.764

Preface to the Focused Issue on the 22 February2011 Magnitude 6.2 Christchurch EarthquakeErol KalkanErol KalkanU.S. Geological Survey, Menlo ParkThe 22 February 2011 magnitude 6.2 Christchurch earthquake,centered southeast of Christchurch, was part of the aftershocksequence that has been occurring since the September 2010magnitude 7.1 quake near Darfield, 40 km west of the city. TheChristchurch earthquake killed more than 180 people, damagedor destroyed more than 100,000 buildings, and is NewZealand’s most deadly disaster since the earthquake that struckthe Napier and Hastings area on 3 February 1931.This special focused issue of Seismological Research Letters,which I had the fortune to edit, contains a selected set of 19original technical papers. These papers cover different aspectsof the 2011 Christchurch earthquake from seismological, geodetic,geological, and engineering perspectives.The first eight papers focus on earthquake source modeling,fault stress variation, and aftershock sequence. The paper byGuidotti et al. presents three-dimensional numerical simulationsof the Christchurch earthquake by comparing different faultand interface models. Using data from a dense network of strongmotion instruments, Holden et al. presents the inversion schemefor constraining the source kinematics of the Christchurchevent. The constrained geodetic source model is presented nextby Beavan et al. using a large amount of ground-displacementdata. The following paper by Zhan et al. concentrates on howapplicable the static Coulomb stress triggering mechanism is tothe 2011 Christchurch aftershock, and it examines the sensitivityof the stress changes to mainshock slip distribution and aftershockfault orientation. Along the same line, Barnhart et al. performsinversions of optical imagery data for spatial distributionof fault slip that occurred during the Darfield and Christchurchearthquakes, and assesses the potential contribution of the staticCoulomb stress change during the Darfield event to the eventualrupture of the Christchurch event. The next paper, by Sibsonet al., evaluates how the complex earthquake sequence of theregion likely has arisen through reactivation under the contemporarytectonic stress field of a mixture of comparatively newlyformed and older inherited fault structures. The paper by Fryand Gerstenberger presents apparent stresses of the three largestregional earthquakes, and compares them to global and regionaldata to improve future seismic hazard estimates due to similarhigh-stress events. In order to better understand the regionalcomplex fault system, Bannister et al. provides relocation analysisof aftershocks that have occurred since the February earthquakethrough May 2011.The next three papers concentrate on recorded strongground motions and their engineering implications. Fry et al.investigates characteristics of recorded horizontal and verticalwaveforms and their correlation with the observed nonlinearsite response. The following paper, by Bradley and Cubrinovski,provides a preliminary assessment of the near-source groundmotions recorded in the Christchurch region by examiningtheir spatial distribution including source, path, and siteeffects. The next paper of this series is by Segou and Kalkan,which evaluates the performance of global ground-motion predictionmodels using the strong motion data obtained from theDarfield and Christchurch earthquakes in order to improvefuture seismic hazard assessment and building code provisionsfor the Canterbury region.The next set of eight papers focus on observed structuraland geotechnical damages associated with strong ground shakingduring both the Darfield and Christchurch earthquakes.The paper by Iizuka et al. investigates the damage around theseismic stations to determine the relationship between structuraldamage and strong motions during the Christchurchearthquake. Similarly, Smyrou et al. evaluates the strongground motions of this event in an effort to broadly explain andquantify the observed structural and geotechnical damages.The next paper, by Zupan et al., summarizes the key field observationsmade following the Christchurch earthquake regardingthe effects of soil liquefaction on building performance inthe central business district. Along the same line, Orense et al.compares the Darfield and Christchurch earthquakes accordingto the results of the reconnaissance works with emphasis onthe geotechnical implications of liquefaction-observed damagein the affected areas. Using the ambient noise measurementsfollowing the Christchurch earthquake, Mucciarelli investigatesthe relationships with previous microzonation studies,liquefaction, and soil nonlinear response. Green, Wood et al.compare the observed versus predicted liquefaction occurrenceduring the Darfield and Christchurch earthquakes usingDCP and SASW tests; and Green, Allen et al., summarizes theperformance of the levees along the Waimakariri and Kaiapoirivers during these two events. The final paper of this specialfocused issue is by Wotherspoon et al. and presents a summaryof field observations, and subsequent analyses on the damage tosome of the bridges in the Canterbury region as a result of theChristchurch earthquake.doi: 10.1785/gssrl.82.6.765Seismological Research Letters Volume 82, Number 6 November/December 2011 765

Numerical Study on the Role of Basin Geometryand Kinematic Seismic Source in 3D GroundMotion Simulation of the 22 February 2011M W 6.2 Christchurch EarthquakeRoberto Guidotti, Marco Stupazzini, Chiara Smerzini, Roberto Paolucci, and Paolo RamieriRoberto Guidotti, 1 Marco Stupazzini, 2 Chiara Smerzini, 1Roberto Paolucci, 1 and Paolo Ramieri 3INTRODUCTIONAlmost six months after the M W 7.1 Darfield (Canterbury)earthquake, on 22 February 2011 at 12:51 p.m. (localtime), an M W 6.2 earthquake struck the city and suburbs ofChristchurch—the largest city on the South Island of NewZealand, with about 400,000 inhabitants. The Christchurchearthquake can be considered one of the greatest natural disastersrecorded in New Zealand. The death toll was more than180, with around 2,000 people injured, and structures alreadyweakened by the Darfield event and its aftershocks were badlyaffected (Cubrinovski and Green 2010; Tonkin and Taylor Ltd.2010; Kam et al. 2011). The earthquake was generated by anoblique thrust fault located between the Australian and Pacificplates, within about 6 km of the city center. It is worth recallingthat prior to the Darfield event there was no surface evidenceof the fault that generated the Christchurch earthquakeon February 2011, nor of the Greendale fault, recognized asresponsible for the September 2010 earthquake (Quigley et al.2010). During the last decade a set of seismic surveys across theCanterbury Plains had been carried out (Green et al. 2010), butthey did not reveal any convincing evidence of the Greendalefault and there was no clear indication that a major earthquakewas imminent in this particular region. Beyond the effects andthe consequences of the seismic event, the attention of the scientificcommunity was drawn to two aspects that had a primaryrole in the Christchurch earthquake: 1) the extremelysevere, strong ground shaking observed, especially on the verticalcomponent; and 2) the widespread liquefaction phenomenaacross the city (Cubrinovski and Green 2010; Green et al.2011, page 927 of this issue).1. Department of Structural Engineering, Politecnico di Milano,Milan, Italy2. Munich RE, Munich, Germany3. Consorzio Interuniversitario Lombardo per l’ElaborazioneAutomatica (CILEA), Segrate, Milan, ItalyBetween September 2010 and June 2011 the Canterburyarea experienced three major earthquakes with M W ≥ 6.0 anda large number of aftershocks (Gledhill et al. 2011; Bannisteret al. 2011, page 839 of this issue). The Christchurch earthquakewas recorded by several digital stations of the permanentnetwork operated by the Institute of Geological and NuclearSciences (GNS; data available at the GeoNET Data Centre:http://www.geonet.org.nz/). Peak ground motion accelerationsin the epicentral region of the earthquake range up to 1.261 gon the horizontal component and up to 1.629 g on the verticalcomponent. Table 1 shows a list of the accelerometric stationslocated within a 40-km-radius from the epicenter withthe corresponding values of peak ground acceleration (PGA)and peak ground velocity (PGV) on both the horizontal andvertical components (data from the Center for EngineeringStrong Motion Data, CESMD: http://www.strongmotioncenter.org/;bandpass filter transition bands are 0.1–0.25 Hz and24.50–25.50 Hz). The ground accelerations recorded withinthe city of Christchurch are among the largest ever recordedfor a New Zealand earthquake, with exceptionally high verticalground acceleration (Bradley and Cubrinovski 2011, page853 of this issue). The unusual severity of the ground shakingcan be explained as a combination of four major effects: 1) theproximity of the causative fault to the city, 2) the directivity ofground motion toward the urban area, 3) the strong amplificationeffects of the soft alluvial sediments beneath the city, and4) the hanging wedge effect, causing a significant increase ofground shaking on the hanging wall.The availability of this unprecedented dataset of nearfaultstrong ground motion, combined with the peculiargeological configuration of the Christchurch area, makes theChristchurch earthquake a relevant benchmark to test theeffectiveness of 3D numerical tools for the prediction of thevariability of strong ground motion in near-fault conditions.To this end, numerical simulation of seismic wave propagationwithin the Canterbury Plains, extending from the northwesternportion of the Southern Alps mountain range to thedoi: 10.1785/gssrl.82.6.767Seismological Research Letters Volume 82, Number 6 November/December 2011 767

TABLE 1Peak ground acceleration (PGA) and peak ground velocity (PGV) values recorded for the stations within a 40-km radius fromthe epicenter. Stations located on rock are typed in italic (data from CESMD, Center for Engineering Strong Motion Data:http://www.strongmotioncenter.org/; bandpass filter transition bands are 0.1–0.25 Hz and 24.50–25.50 Hz). R e denotes theepicentral distance.NameStationsIdR e[km]PGAEW[g]PGANS[g]PGAUP[g]PGVEW[cm/s]PGVNS[cm/s]Heathcote Valley Primary School HVSC 1 1.230 1.261 1.466 79.96 89.12 38.51Lyttelton Port Company LPCC 4 0.919 0.771 0.413 37.27 40.97 16.43Pages Road Pumping Station PRPC 6 0.665 0.589 1.629 82.26 74.83 49.48Christchurch Cathedral College CCCC 6 0.415 0.375 0.692 64.75 43.58 21.53Christchurch Cashmere High School CMHS 6 0.369 0.403 0.796 41.69 46.79 15.17Christchurch Resthaven REHS 8 0.719 0.372 0.529 87.30 45.54 21.29Christchurch Hospital CHHC 8 0.357 0.337 0.511 65.17 56.58 20.87Christchurch Botanic Gardens CBGS 9 0.536 0.432 0.271 63.95 43.12 13.42Shirley Library SHLC 9 0.324 0.319 0.500 52.73 51.74 21.73Hulverstone Drive Pumping Station HPSC 9 0.232 0.155 0.858 36.21 26.75 33.93Riccarton High School RHSC 12 0.297 0.254 0.188 29.95 23.63 12.11Christchurch Papanui High School PPHS 12 0.187 0.243 0.195 36.73 28.24 16.82Styx Mill Transfer Station SMTC 14 0.182 0.144 0.176 28.55 20.49 12.31McQueens Valley MQZ 15 0.147 0.098 0.072 7.10 5.44 4.05Christchurch Canterbury Aero Club CACS 18 0.182 0.180 0.185 15.85 13.88 11.84Lincoln Crop and Food Research LINC 19 0.081 0.164 0.084 7.59 16.17 5.41Templeton School TPLC 19 0.091 0.099 0.136 10.36 10.09 8.16Kaiapoi North School KPOC 23 0.211 0.201 0.057 14.19 19.39 5.71Rolleston School ROLC 26 0.164 0.163 0.072 6.30 7.77 4.28Swannanoa School SWNC 29 0.251 0.143 0.056 14.69 13.94 4.73Selwyn Lake Road SLRC 33 0.086 0.088 0.049 7.25 8.32 2.92Ashley School ASHS 35 0.088 0.076 0.037 7.28 5.44 2.30Cust School CSTC 39 0.075 0.078 0.042 6.46 6.71 3.09PGVUP[cm/s]Lyttelton-Akaroa volcanic region, was performed throughthe software package GeoELSE (http://geoelse.stru.polimi.it). Based on the Spectral Element formulation proposed byFaccioli et al. (1997), GeoELSE is designed to perform linearand nonlinear dynamic wave propagation analyses in heterogeneousmedia, exploiting in 3D its implementation in parallelcomputer architectures. Examples of application of GeoELSEto seismic wave propagation studies in complex geologicalconfigurations can be found in Stupazzini et al. (2009) andSmerzini et al. (2011). Different 3D numerical models wereconstructed for the Christchurch earthquake, to check thedependence of the results on: 1) the kinematic source model,based on the information retrieved from recent seismic sourceinversion studies, and 2) the shape of the alluvial-bedrockinterface within the Canterbury Plains. To check the accuracyof the numerical models, the synthetic results are comparedagainst the strong ground motion records. The misfit betweensimulated and recorded waveforms is evaluated in a quantitativeway in a format suitable for engineering applications,making use of the criteria proposed by Anderson (2004). Togive insights into the variability of surface earthquake groundmotion, due to the interaction between near-fault conditionsand strong geological variations, ground-shaking maps andsnapshots of the velocity wavefield are shown.GEOLOGY OF THE CHRISTCHURCH REGIONThe area under study extends from the Southern Alps tothe Lyttelton-Akaroa volcanic region and includes the cityof Christchurch and parts of the Canterbury Plains and theBanks Peninsula. Information about the geology of this areacomes from the 1:250,000 geological map of Christchurch,produced by GNS (Forsyth et al. 2008). Based on the geologicalmap and on the available cross-sections, it was possible toinfer a preliminary stratigraphy for the alluvial cover fillingthe Canterbury Plains in the area of interest (Figure 1). Thebasement rock of the whole region is the Torlesse compositeterrane, a deformed package of Carboniferous to Cretaceoussedimentary rocks. Late Miocene volcanism forms the twomajor overlapping volcanoes Lyttelton and Akaroa on BanksPeninsula, today largely eroded. Erosion of the shallow landmassand glacial-interglacial climatic fluctuations led to the768 Seismological Research Letters Volume 82, Number 6 November/December 2011

▲ ▲ Figure 1. A) Area of the model under study, including part of the Canterbury Plains and the Lyttelton-Akaroa volcanic region, alongwith the strong ground motion stations. B) Geological A-A’ cross-section.widespread decomposition of unconsolidated Quaternarysediments that constitute the Canterbury Plains. The alluvialsequence is formed by coal, clay, limestone, and sand (Forsythet al. 2008).It is possible to consider the deep crustal model as basinsediments overlying a layer of Torlesse graywacke down toaround 5 km, overlying, in turn, a layer of Haast Schist downto around 20 km depth. The lower crust is interpreted as mafic(diorite, diabase, and gabbro) Mesozoin ocean crust (Reynersand Cowan 1993; Godfrey et al. 2001, 2002; Melhuish et al.2005). This deep crustal layer subdivision is confirmed by thehigh-resolution seismic wide-angle data, collected within theframework of the South Island Geophysical Transect (SIGHT)experiment (Kleffmann et al 1998; Mortimer et al. 2002; Longet al. 2003; Scherwarth et al. 2003).3D NUMERICAL SIMULATIONS OF THE SEISMICRESPONSE OF THE CANTERBURY PLAINS3D numerical simulations consist of the following features:1) kinematic description of the close-by seismic source, 2)horizontally layered deep geological model, 3) a simplifiedbut realistic description for the Cretaceous-Cenozoic alluvialCanterbury Plains, and 4) a linear visco-elastic soil behavior.Note that in these preliminary analyses we considereda relatively rough model for the soil behavior by assuming alinear-visco elastic constitutive law, with a quality factor Q proportionalto frequency (further details about the implementationof the visco-elastic soil behavior model can be found inStupazzini et al. 2009). Different 3D numerical models werebuilt for the Christchurch earthquake to achieve the best fitwith the ground motion observations, combining: 1) two differentkinematic seismic fault solutions, based on recent seismicsource inversion studies, and 2) two simplified models for theshape that defines the interface between the alluvial soft soilsediments and the rigid volcano materials. The 3D model ofthe region of the South Island of New Zealand covers an area ofapproximately 60 x 60 x 20 km around the city of Christchurch,including the information available in the geological map andthe 2D cross-sections, shown in Figure 1 and described in theprevious section.Two different models were constructed to approximatethe complex geological configuration of the Canterbury Plains.The models, referred to hereafter as “step-like” and “smooth,”basically differ in the transition between the alluvial soft sedimentsand the rigid volcano materials, as sketched in Figure 2.In the “step-like” model (Figure 2A), a rough approximationof the alluvial-bedrock shape is adopted. More specifically, thethickness of the alluvial basin is assumed constant across thewhole area under study and equal to 1.5 km. In the “smooth”model, the shape of the interface between the soft soil and thevolcanic materials is improved, with constraints inferred fromthe topography of the volcano (Figure 2B). For both models,the alluvial basin consists of three different layers with V Sranging from 300 m/s in the top 300 m to 1,500 m/s at theinterface with the volcanic materials (top three layers of TableSeismological Research Letters Volume 82, Number 6 November/December 2011 769

▲▲Figure 2. A) Sketch of the “step-like” and B) “smooth” model, referred to as an approximation of the transition between the softsediments of the Canterbury Plains and the rigid volcano materials. Note that the alluvial basin has a maximum thickness of 1.5 km andconsists of three horizontal layers with V s ranging from 300 m/s to 1,500 m/s.LayerTABLE 2Soil profile adopted in the numerical simulations. The top three layers constitute the Canterbury Plains.Depth[m]Thickness[m]V p[m/s]V s[m/s]ρ[kg/m 3 ]1 0–300 300 600 300 1700 702 300–750 450 1870 1000 2000 1003 750–1500 750 2800 1500 2300 1004 0–5000 5000 5500 3175 2600 2005 1500–5000 3500 5000 2890 2700 2006 5000–20000 15000 6000 3465 2700 250Q2). In the absence of direct measurements for the consideredarea, V S values adopted in Table 2 are in reasonable agreementwith available geotechnical data and engineering geologicalmodels, although in a different region of New Zealand (Boonet al. 2011; Semmens et al. 2011). The volcano region (layer 4in Table 2), with V S = 3,175 m/s, extends down to a maximumdepth of 5 km. As regards the background geology, a horizontallylayered crustal model was assumed, as summarized inTable 2 (layers 5 and 6 in Table 2). The frequency proportionalquality factor Q values given in Table 2 correspond to a frequencyf equal to 0.67 Hz, according to the following equation:fQ ( f ) = Q 0, (1)f 0where Q 0 = πf 0 /γ, γ is an attenuation parameter, and f 0 is a referencevalue representative of the frequency range to be propagated,herein equal to 2 Hz (Stupazzini et al. 2009).We considered two different preliminary static faultsolutions, proposed by the Istituto Nazionale di Geofisicae Vulcanologia (INGV; http://www.sigris.it/) and by GNS(http://www.gns.cri.nz/Home/News-and-Events/Media-Releases/Most-damaging-quake-since-1931/Canterbury-quake/Hidden-fault). The source parameters adopted for the two modelsare summarized in Table 3, while the slip distribution acrossthe fault plane, along with the hypocenter location, is presentedin Figure 3 for both the INGV and GNS seismic source inversions.The source model proposed by INGV was identified bythe ASI-SIGRIS system, exploiting COSMO-SkyMed images(Atzori and Salvi 2011). The “GNS model,” published as a pressrelease, is a static model derived from pre-earthquake and postearthquakegeodetic data using both InSAR and GPS data. Anupdated version of the model is given by Beavan et al. (2011,page 789 of this issue) and a true kinematic source modelbased on inversions of strong-motion data has been developedby Holden (2011, page 783 of this issue). While the two finitefault solutions have similar sizes and slip patterns, they differquite significantly in strike angle and location of the hypocenter.It is important to emphasize that the “kinematic sourcemodels” used in this work have been obtained by turning thesestatic models into kinematic models by assuming a rupture770 Seismological Research Letters Volume 82, Number 6 November/December 2011

TABLE 3Kinematic source parameters adopted for the simulation of the 22 February 2011 Christchurch earthquake. V R denotes thevelocity rupture and τ the rise timeHypocenter°N °E ZL x W[km]Strike[°]Dip[°]Rake[°]Depth ofupper points[km]V R[km/s]INGV –43.58°N 172.68°E –10.3 km 14 x 10 45 67 145 1.5 2.4 0.9GNS –43.56°N 172.70°E –6.47 km 18 x 9 58 68 145 1 2.4 0.9τ[s]▲ ▲ Figure 3. Slip distribution according to the A) INGV and the B) GNS fault solutions. The superimposed star denotes the hypocenterlocation.velocity, rise time, and slip origin, and simplifying the “GNSmodel” shown on the GNS Web site by making the rake constant,equal to 145°. A value V R equal to 2,400 m/s is assumedas rupture velocity. The slip source time function is given by anapproximate Heaviside function, as follows:M 0( )( t ) = 1 ⎡ − 2τ1+ erf 2.0t2 ⎣⎢ τ / 2⎤⎦⎥ , (2)where erf( ) is the error function and τ = 0.9 s is the rise time,assumed to be constant across the fault plane. Figure 4 shows aplot of moment rate function and its spectrum.The 3D spatial discretization by spectral elements of thearea requires the design of a large-scale unstructured mesh ofhexahedral elements. The computational domain is subdividedinto small chunks; each of them is meshed starting from thealluvial basin down to the bedrock. The mesh was constructedmaking use of the software CUBIT (available at http://cubit.sandia.gov/), according to the technique already described byCasarotti et al. (2007). Note that four different numerical modelswere constructed to attain the different hypotheses regarding,on one side, the seismic source (INGV vs. GNS) and, onthe other side, the alluvial-volcano interface (see above). Bothseismic source models, either INGV or GNS, have been testedwith the: 1) “step-like” and 2) “smooth” approximation for thealluvial-volcano interface. The 3D hexahedral spectral elementmesh adopted for the numerical simulations by GeoELSE consistsof about 476,000 and 496,000 elements for the INGV andGNS model, respectively. The size of the elements ranges, inboth cases, from a minimum of about 150 m (at the top of thealluvial basin) up to 1,500 m at bedrock. The mesh is, hence,designed to propagate up to about 2 Hz, for spectral degreeequal to 4 (see Figure 5). The numerical simulations were performedon the Lagrange cluster located at CILEA. The maincharacteristics and the performances of the analyses are summarizedin Table 4.NUMERICAL RESULTS IN THE NEAR-SOURCEREGIONThis section aims to show the main results obtained throughthe numerical simulations by GeoELSE. We compare the3D synthetic seismograms with the observed waveforms at aset of stations located in the near-source region of the earthquake.In order to point out the role of the 3D geometry ofthe Canterbury Plains and of the kinematic seismic source,we proceed in the following way. At first, we address the issueregarding the effect of the model assumed for the alluvialbedrocktransition, “step-like” vs. “smooth,” considering thefault solution proposed by INGV; afterward, we evaluate thedependence of the results on the fault model, INGV vs. GNS,considering only the “smooth” alluvial-bedrock transition.Figures 6 and 7 show the comparison between 3D numericalsimulations with strong ground-motion observations, inSeismological Research Letters Volume 82, Number 6 November/December 2011 771

▲▲Figure 4. A) Adopted moment rate function and B) its spectrum.▲▲Figure 5. A) 3D geometry of the area under study, with depth contours of the contact between the alluvial soft sediments and therigid volcano materials (depth in meters). B) Zoom of the corresponding hexahedral spectral elements mesh.ModelTABLE 43D numerical models size and computational time. Data of CPU time refer to the Lagrange cluster located at CILEA.SDNumber ofSpectral ElementsNumber ofLGL NodesNumber ofcoresSimulation time(h)INGV – “Step-like” 4 475,992 ~31.6 × 10 6 64 ~75INGV – “Smooth” 4 475,992 ~32 × 10 6 64 ~80GNS – “Smooth” 4 495,385 ~33.3 × 10 6 128 ~107772 Seismological Research Letters Volume 82, Number 6 November/December 2011

▲ ▲ Figure 6. Comparison of 3D numerical simulations by GeoELSE with strong ground motion observations, in terms of three-componentvelocity time histories, obtained at stations: A) HVSC, at outcrop rock; B) REHS, on soft sediments in the CBD; C) SHLC, on softsediments at R e = 9 km; D) SLRC, on soft sediments at R e = 33 km, in the southwestern portion of the model. Observed and simulateddata are bandpass filtered between 0.1 and 2.0 HzSeismological Research Letters Volume 82, Number 6 November/December 2011 773

▲▲Figure 7. As in Figure 6 but in terms of velocity amplitude Fourier spectra.774 Seismological Research Letters Volume 82, Number 6 November/December 2011

terms of velocity time histories and corresponding Fourieramplitude spectra, respectively. The comparison is presentedfor four representative stations: HVSC, which lies on outcroppingrock; REHS, located on alluvial soil in the CentralBusiness District (CBD); SHLC, situated on alluvial soil at epicentraldistance R e = 9 km, close to the CBD; and SLRC, lyingon alluvial soil at R e = 33 km, southwest of the epicenter. Bothobserved and simulated waveforms have been processed with abandpass acausal Butterworth filter between 0.1 and 2 Hz.A quantitative estimation of the overall quality of thenumerical analyses can be inferred evaluating the misfitparameters proposed by Anderson (2004). Stating that a singleparameter is incomplete to assess the correspondence betweensimulated and observed time-histories, Anderson (2004) introduceda set of 10 parameters, each one evaluated for a specificfrequency band of interest: Arias duration (AD), energy duration(ED), Arias intensity (AI), energy integral (EI), peakacceleration (PA), peak velocity (PV), peak displacement (PD),response spectra (RS), Fourier spectra (FS), and cross-correlation(CC). A score between 0 and 10, with 0 indicating noagreement and 10 perfect agreement, is calculated for each ofthese parameters, yielding an overall goodness of fit.Figure 8 depicts the goodness of fit parameters computedin the frequency band 0.25–0.50 Hz for the three modelsunder study, “step-like” INGV, “smooth” INGV, and “smooth”GNS, and for three components of motion (EW, NS, and UD).The scores of the aforementioned parameters are shown for the23 stations summarized in Table 1.Dependence of Results on the Geometry of theCanterbury PlainsIn this section we address the issue regarding the role of the3D geometry of the Canterbury Plains, i.e., the “step-like” vs.“smooth” model, on the simulated waveforms. Referring toFigures 6 and 7, we note that, in spite of the rough approximationsbehind the “step-like” model, the overall agreementis fairly satisfactory. For most of the considered stations, the“step-like” model reproduces the first arrivals and the PGVswith reasonable accuracy. The agreement between syntheticsand observations is satisfactory for the stations located in thecentral-western portion of the Canterbury Plains, while it deterioratesfor the station located on the southern volcanic region.Nonetheless, such model tends to overestimate PGV valuesmeasured in the eastern area of Christchurch, where majorliquefaction effects were observed. This effect may be due tothe rough representation of the interface between the volcanicmaterial and the soft soil of the basin, leading to an excessivelylarge concentration of energy toward the city of Christchurchas a consequence of the high impedance contrast between thevolcano region and the surrounding soft sediments.The introduction of the “smooth” model yields significantimprovements of the simulated waveforms, in particular forthe reproduction of the coda-waves inside the alluvial plain.The smooth interface between volcanic rock and alluvial soilleads to a better agreement in terms of PGV; nonetheless, themodel still tends to overestimate the peak values measured inthe CBD and in the eastern-coastal area of Christchurch andto underestimate the peak values recorded far from the epicenter,especially in the southwestern region of the model.Figure 8 allows us to have a quantitative criterion to assessthe performance of the different numerical simulations. Whilethe “step-like” model shows at least six stations with goodscores for the whole set of parameters (Figure 8A), the INGV“smooth” model, with at least nine stations with good averagescores (Figure 8B), yields actual improvements of the numericalanalyses. As a general remark, for both simulations integralmeasures of Arias and energy duration (AD and ED) and peakground acceleration, velocity, and displacement values (PA, PV,and PD) present a good fit for all the considered stations, whilea poor score is achieved for intensity measures (AI and EI),spectral amplitudes, and cross-correlation (RS, FS, and CC).Effect of the Kinematic Seismic SourceAfter having illustrated the results obtained for the “step-like”and “smooth” model, we now turn to evaluating the effect ofdifferent kinematic seismic sources. To this end, we will showa comparison of the numerical results obtained for the INGVand GNS fault solutions (see Figure 3), relying on the “smooth”model, which turns out to produce satisfactory results as discussedin the previous section.The comparison in Figures 6 and 7 shows that the GNSfault model leads to a better agreement between recorded andsimulated ground motion velocities at the four stations underconsideration. For this kinematic source model, a good agreementis found both at stations located on alluvial soil a fewkilometers from the epicenter and at those stations locatedseveral kilometers farther away in the southwestern portion ofthe model. In spite of the rough assumptions behind the GNS“smooth” model, numerical simulations are able to reproducewith reasonable accuracy the PGVs within the CanterburyPlains. Nonetheless, the agreement between synthetics andobserved values is still quite poor for the station located in thevolcanic region. This is most likely due to the simplified modelassumed for the topography of the Banks Peninsula, which isapproximated as a smooth surface and does not capture thecomplex geometry of bays and coves that may play an importantrole in seismic wave propagation phenomena. Furthermore, ahomogeneous soil profile is assumed for the volcano region, sothat erosion and weathering phenomena of the surface rock layersare not taken into account.The analysis of the Anderson misfit criteria (Figure 8C)confirms the quality of the numerical simulations, showinggood average scores for almost all the stations under consideration.As mentioned previously, the figure highlights a goodagreement in terms of PGVs (parameter PV) for many stationsinside the computation domain. Figure 9 shows a comprehensivecomparison between recorded and simulated (GNS “smooth”model) velocity time histories at the entire set of accelerometricstations inside the computational model, ordered by epicentraldistance. A good agreement is found in terms of arrivaltimes, peak ground values, and attenuation with distance, inspite of the rough assumptions concerning the characteriza-Seismological Research Letters Volume 82, Number 6 November/December 2011 775

▲ ▲ Figure 8. Misfit parameters (Anderson 2004) of simulated ground motion for the three components of motion, evaluated at the 23stations under consideration (see Table 1 and Figure 1) in the frequency band between 0.25 and 0.50 Hz. The results are shown for theA) INGV “step-like” , B) INGV “smooth” (center panel), and C) GNS “smooth” model (bottom panel).776 Seismological Research Letters Volume 82, Number 6 November/December 2011

▲ ▲ Figure 9. Comparison between recorded and simulated (GNS “smooth”) velocity time histories (in cm/s), on the EW, NS, and UDcomponent for the whole set of stations at R e < 40 km. The label on the left vertical axis reports the peak value of the correspondingtime history.tion of soil mechanical properties. To give a broad picture ofseismic wave propagation effects, Figure 10 depicts the spatialdistribution of PGV values obtained through GeoELSE alongwith the observed values (superimposed filled dots) for boththe INGV and GNS “smooth” models. Furthermore, Figure11 shows some representative snapshots of the simulated faultnormal velocity wavefield, in which the seismic wave propagationfield with the high contrast between rock and alluvial soiland the directivity toward the city of Christchurch is clearlydistinguishable. In particular, looking at Figure 10 it is possibleto notice that the GNS fault model better reproduces the variabilityof PGV values on the whole modeled area. From Figures10 and 11 it is apparent that the INGV source model producesstrong “up-dip” directivity effects in the central-eastern part ofChristchurch, in reasonable agreement with the spatial distributionof observed damage and liquefaction phenomena. Thewave propagation pattern obtained with the GNS fault modelproduces noticeable directivity off the sides of the fault due tothe relatively shallow hypocenter (around 6.5 km) and becausethe rake is oblique (145°). This leads to larger ground motionamplitudes in the southwestern portion of the city.Comparison with Observed Standard Spectral RatiosAs a concluding check of the quality of 3D numerical simulations,in this section we compare simulations with observationsin terms of standard spectral ratios (SSRs). Most of the 23 stationsincluded in the numerical model are located on alluvialsoil, while three of them lie on outcropping volcanic rock,namely HVSC, LPCC, and MQZ. For our purposes, we consideredthe four stations located in the Christchurch CBD, onsoft alluvial sediments, namely CCCC, REHS, CHHC, andCBGS. Station LPCC, located on rock around 15 km southeastfrom the CBD, is considered as reference rock station.The SSRs computed as the ratio of the Fourier spectrumof the recordings at CCCC, REHS, CHHC, and CBGS (geometricmean of the horizontal components), over that at LPCCreference station, are shown in Figure 12, for 4 September 2010Darfield earthquake, with the epicenter located around 40 kmwest of the considered set of stations. We refer to the Darfieldearthquake because we believe that this event is more reliablethan the Christchurch one in evaluating the SSR because ofthe strong nonlinear soil behavior verified in the latter and theshort distance (less than 10 km) between the epicenter and theconsidered stations. It is worth noting that stable resonancepeaks are found at around 0.3, 0.6, 1.3, and, more consistently,at around 1.7 Hz. The recorded SSRs are compared, on the onehand, with the 1D analytical transfer function obtained for asystem of four layers (layers 1, 2, 3, and 5 in Table 2) over a halfspace(layer 6 in Table 2), and, on the other hand, with the SSRsobtained through 3D numerical simulation (GNS “smooth”model). Compared with the 1D amplification function, the 3DSSRs show a better agreement with the records, pointing outresonance frequencies at about 0.4, 0.7, 1.3, and 1.7 Hz.CONCLUDING REMARKSThe main aim of this paper was to perform 3D numerical simulationsof the M W 6.2 Christchurch earthquake on 22 FebruarySeismological Research Letters Volume 82, Number 6 November/December 2011 777

▲ ▲ Figure 10. Spatial variability of peak ground velocity (geometric mean of horizontal components) as estimated by 3D numericalsimulations with the A) INGV “smooth” model and B) GNS “smooth” model. The recorded PGV values (filled dots) are superimposedfor comparison purposes.778 Seismological Research Letters Volume 82, Number 6 November/December 2011

▲ ▲ Figure 11. Snapshots (t = 7, 9, 11, and 13 s, from top to bottom) of the simulated fault normal velocity wavefield with the A) INGV“smooth” model and B) GNS “smooth” model.Seismological Research Letters Volume 82, Number 6 November/December 2011 779

▲ ▲ Figure 12. Standard spectral ratios (SSRs) of the four stations of CBD: A) CCCC, B) REHS, C), and D) CBGS located on soft alluvialsediments, over the rock reference station LPCC. The SSRs obtained from strong ground motion recordings during the Darfield earthquakeare compared with the 1D analytical transfer function and with the SSRs computed with the 3D numerical simulations (GNS“smooth” model).2011, the most devastating and deadliest event of the seismicsequence that struck the Canterbury Plains, and particularlythe city of Christchurch, between September 2010 and June2011, and to compare the numerical results with strong groundmotion observations.The numerical simulations of seismic wave propagationwithin the Canterbury Plains, where widespread damage wasrecognized during the post-earthquake reconnaissance surveys,were performed by means of the Spectral Element codeGeoELSE (http://geoelse.stru.polimi.it). Based on the availablegeological and seismological data, 3D numerical simulationsof the Christchurch earthquake were carried out, combiningthe following features: 1) two different kinematic finite faultmodels, provided by INGV and GNS seismic source inversionstudies, and 2) two simplified models for the descriptionof the interface between the stiff volcanic rock of the BanksPeninsula and the soft materials within the Canterbury Plains,referred to as the “step-like” and the “smooth” model. As apreliminary assumption a linear visco-elastic soil behavior wasassumed. The comparison of the results obtained through 3Dnumerical simulations with the strong ground motion recordsin the epicentral area of the earthquake (R e < 40 km) shows agood agreement both in time and frequency domain, especiallyfor the “smooth” model with the GNS kinematic extendedfault model. It is worth remarking that the simplified assumptionof linear visco-elastic soil behavior cannot adequatelydescribe the amplification phenomena and the shift of fundamentalfrequency, clearly recorded in many stations located onthe alluvial soil of the Canterbury Plains. Although the GNS“smooth” model is found to produce the best agreement with780 Seismological Research Letters Volume 82, Number 6 November/December 2011

the observed waveforms, it should be noted that accounting fora more complex constitutive model could improve significantlythe results of the INGV smooth model.3D numerical simulations allow us to reproduce the mostsignificant features of surface earthquake ground motion inthe near-fault region. Ground motion shaking maps, in termsof PGV, and snapshots of simulated velocity wavefield are discussed,giving insights into seismic wave propagation effects inrealistic geological structures and under near-fault conditions.In spite of the simplified assumptions behind the numericalmodel, 3D numerical simulations represent a relevant tool topredict realistic earthquake ground motion in complex tectonicand geological environments, and for different seismicsource scenarios that may play a major role in seismic hazardassessment studies.ACKNOWLEDGMENTSThe authors acknowledge Simone Atzori of INGV for kindlyproviding the data about the seismic source inversion studies.John Beavan and Caroline Holden of GNS are greatlyacknowledged for their useful suggestions and remarks aboutthe GNS source inversion adopted in this work. Also gratefullyacknowledged are Pilar Villamor, Andrew King, RafaelBenites of GNS, Misko Cubrinovski, Brendon Bradley, JohnBerril of the Canterbury University, and Hugh Cowan of theEarthquake Commission of New Zealand (EQC). We are alsograteful to Anselm Smolka, Martin Käser, and AlexanderAllmann (Munich RE) for their fruitful comments. We deeplythank the research center CRS4 (http://www.crs4.it/) and inparticular Fabio Maggio and Luca Massidda, for the essentialcooperation in the development of GeoELSE. Finally, a particularthanks to Robert Graves, U.S. Geological Survey, forthe detailed, constructive criticism he devoted to our paper.REFERENCESAnderson, J. G. (2004). Quantitative measure of the goodness-of-fit ofsynthetic seismograms. In Proceedings of the 13 th World Conferenceon Earthquake Engineering, Vancouver, B.C., Canada. Paper no.243. Oakland, CA: Earthquake Engineering Research Institute.Atzori, S., and S. Salvi (2011). Preliminary source of the destructiveChristchurch earthquake identified by the SIGRIS system. SIGRISActivities for the Christchurch (New Zealand) Earthquake; http://www.sigris.it/.Bannister, S., B. Fry, M. Reyners, J. Ristau, and H. Zhang (2011).Fine-scale relocation of aftershocks of the 22 February M w 6.2Christchurch earthquake using double-difference tomography.Seismological Research Letters 82, 839–845.Beavan, J., E. Fielding, M. Motagh, S. Samsonov, and N. Donnelly(2011). Fault location and slip distribution of the 22 February 2011M W 6.2 Christchurch, New Zealand, earthquake from geodeticdata. Seismological Research Letters 82, 789–799.Boon, D., N. D. Perrin, G. D. Dellow, R. Van Dissen, and B. Lukovic(2011). NZS 1170.5:2004 Site Subsoil Classification of LowerHutt. Proceedings of the Ninth Pacific Conference on EarthquakeEngineering, Building an Earthquake-Resilient Society. 14–16April 2011, Auckland, New Zealand, paper no. 13. Auckland, NewZealand: 9PCEE.Bradley, B. A., and M. Cubrinovski (2011). Near-source strong groundmotions observed in the 22 February 2011 Christchurch earthquake.Seismological Research Letters 82, 853–865.Casarotti, E., M. Stupazzini, S. Lee, D. Komatitsch, A. Piersanti, andJ. Tromp (2007). CUBIT and seismic wave propagation basedupon the spectral-element method: An advanced unstructuredmesher for complex 3D geological media. In Proceedings of the16th International Meshing Roundtable, ed. M. L. Brewer and D.Marcum, 579–597. New York: Springer.Cubrinovski, M., and R. A. Green, eds. (2010). Geotechnical reconnaissanceof the 2010 Darfield (Canterbury) earthquake. Bulletin of theNew Zealand Society for Earthquake Engineering 43, 243–320.Faccioli, E., F. Maggio, R. Paolucci, and A. Quarteroni (1997). 2D and3D elastic wave propagation by a pseudospectral domain decompositionmethod. Journal of Seismology 1 (3), 237–251.Forsyth, P. J., D. J. A. Barrell, and R. Jongens (2008). Geology of theChristchurch Area. Institute of Geological and Nuclear Sciences.1:250 000 geological map 16, 1 sheet + 67 pp. Lower Hutt, NewZealand: GNS Science.Gledhill, K., J. Ristau, M. Reyners, B. Fry, and C. Holden (2011). TheDarfield (Canterbury, New Zealand) M W 7.1 earthquake ofSeptember 2010: A preliminary seismological report. SeismologicalResearch Letters 82, 379–386.Godfrey, N. J., F. Davey, T. A. Stern, and D. Okaya (2001). Crustal structureand thermal anomalies of the Dunedin region, South Island,New Zealand. Journal of Geophysical Research 106 (B12), 30,835–30,848.Godfrey, N. J., N. I. Christensen, and D. Okaya (2002). The effect ofcrustal anisotropy on reflector depth and velocity determinationfrom wide-angle seismic data: A synthetic example based on SouthIsland, New Zealand. Tectonophysics 355, 145–161.Green, A. G., F. M. Campbell, A. E. Kaiser, C. Dorn, S. Carpentier, J. A.Doetsch, H. Horstmeyer, D. Nobes, J. Campbell, M. Finnemore,R. Jongens, F. Ghisetti, A. R. Gorman, R. M. Langridge, and A.F. McClymont (2010). Seismic reflection images of active faults onNew Zealand’s South Island. In Fourth International Conference onEnvironmental and Engineering Geophysics, Chengdu, China, June2010.Green, R. A., C. Wood, B. Cox, M. Cubrinovski, L. Wotherspoon,B. Bradley, T. Algie, J. Allen, A. Bradshaw, and G. Rix (2011).Use of DCP and SASW tests to evaluate liquefaction potential:Predictions vs. observations during the recent New Zealand earthquakes.Seismological Research Letters 82, 927–938.Holden, C. (2011). Kinematic source model of the 22 February 2011M w 6.2 Christchurch earthquake using strong motion data.Seismological Research Letters 82, 783–788.Kam, W. Y., U. Akguzel, and S. Pampanin (2011). 4 Weeks On:Preliminary Reconnaissance Report from the Christchurch 22 Feb2011 6.3 M W Earthquake; http://db.nzsee.org.nz:8080/en/web/chch 2011/structural/.Kleffmann, S., F. Davey, A. Melhuish, D. Okaya, T. Stern, and theSIGHT Team (1998). Crustal structure in the central South Island,New Zealand, from the Lake Pukaki seismic experiment. NewZealand Journal of Geology and Geophysics 41, 39–49.Long, D. T., S. C. Cox, S. Bannister, M. C. Gerstenberger, and D. Okaya(2003). Upper crustal structure beneath the eastern Southern Alpsand the Mackenzie Basin, New Zealand, derived from seismicreflection data. New Zealand Journal of Geology & Geophysics 46,21–39.Melhuish, A., S. Holbrook, F. Davey, D. Okaya, and T. Stern (2005).Crustal and upper mantle seismic structure of the Australian plate,South Island, New Zealand. Tectonophysics 395, 113–135.Mortimer, N., F. J. Davey, A. Melhuish, J. Yu, and N. J. Godfrey (2002).Geological interpretation of a deep seismic reflection profile acrossthe Eastern Province and Median Batholith, New Zealand: Crustalarchitecture of an extended Phanerozoic convergent orogen. NewZealand Journal of Geology & Geophysics 45, 349–363.Seismological Research Letters Volume 82, Number 6 November/December 2011 781

Quigley, M., R. Van Dissen, P. Villamor, N. Litchfield, D. Barrell, K.Furlong, T. Stahl, B. Duffy, E. Bilderback, D. Noble, D. Townsend,J. Begg, R. Jongens, W. Ries, J. Claridge, A. Klahn, H. Mackenzie,A. Smith, S. Hornblow, R. Nicol, S. Cox, R. Langridge, and K.Pedley (2010). Surface rupture of the Greendale fault during theM W 7.1 Darfield (Canterbury) earthquake, New Zealand: Initialfindings. Bulletin of the New Zealand Society for EarthquakeEngineering 43, 236–242.Reyners, M., and H. Cowan (1993). The transition from subduction tocontinental collision: Crustal structure in the North Canterburyregion, New Zealand. Geophysical Journal International 115,1,124–1,136.Scherwath, M., T. Stern, F. Davey, D. Okaya, W. S. Holbrook, R.Davies, and S. Kleffmann (2003). Lithospheric structure acrossoblique continental collision in New Zealand from wide-angle Pwave modeling. Journal of Geophysical Research 108 (B12), 2,566;doi:10.1029/2002JB002286.Semmens, S., N. D. Perrin, G. Dellow, and R. Van Dissen (2011).NZS 1170.5:2004 Site subsoil classification of WellingtonCity. Proceedings of the Ninth Pacific Conference on EarthquakeEngineering, Building an Earthquake-Resilient Society, 14–16April 2011, Auckland, New Zealand, paper no. 7. Auckland, NewZealand, 9PCEE.Smerzini, C., R. Paolucci, and M. Stupazzini (2011). Comparison of3D, 2D and 1D numerical approaches to predict long period earthquakeground motion in the Gubbio plain, central Italy. Bulletinof Earthquake Engineering (June) 1–23; doi:10.1007/s10518-011-9289-8.Stupazzini, M., R. Paolucci, and H. Igel (2009). Near-fault earthquakeground motion simulation in the Grenoble Valley by a high performancespectral element code. Bulletin of the Seismological Society ofAmerica 99, 286–301.Tonkin and Taylor Ltd. (2010). Darfield Earthquake 4 September 2010,Geotechnical Land Damage Assessment & Reinstatement Report.Earthquake Commission. Stage 1 Report for the New ZealandEarthquake Commission. Christchurch, New Zealand: Tonkin &Taylor Ltd.Department of Structural EngineeringPolitecnico di MilanoPiazza Leonardo da Vinci, 32Milano 20133 Italyguidotti@stru.polimi.it(R. G.)782 Seismological Research Letters Volume 82, Number 6 November/December 2011

Kinematic Source Model of the 22 February2011 M w 6.2 Christchurch Earthquake UsingStrong Motion DataCaroline HoldenCaroline HoldenGNS ScienceINTRODUCTIONThe Canterbury earthquake sequence began in September2010 with the Mw 7.1 (source: GeoNet catalog, http://geonet.org.nz/canterbury-quakes/) Darfield earthquake that rupturedthe previously unknown 40-km-long Greendale fault 30 kmwest of Christchurch (Gledhill et al. 2011). Extreme groundaccelerations as high as 1.8 g near the epicenter were recorded.The event caused intense liquefaction in the eastern suburbs ofChristchurch as well as closer to downtown, near the courseof the Avon River. The Darfield earthquake was followed bya major aftershock on 22 February local time (21 FebruaryUTC) of magnitude Mw 6.2 (source: GeoNet), but Me 6.7(source: USGS, http://earthquake.usgs.gov/earthquakes/eqinthenews/2011/usb0001igm/neic_b0001igm_e.php). Thisearthquake was centred only a few kilometers south of theChristchurch city center. Extremely high accelerations (as highas 2.2 g) were also recorded near the epicenter (Kaiser, Beniteset al. 2011). In addition to the extreme liquefaction seen afterthe Darfield earthquake, this event also caused landslides,large rockfalls, widespread damage to earthquake-risk buildingsin Christchurch, and, most tragically, about 180 casualties.Another large aftershock of Mw 6.0 (source: GeoNet), butwith Me 6.7 (source: USGS), subsequently occurred on 13 Junelocal time (12 June UTC) just a few kilometers south of theFebruary event, causing further damage, landslides, rockfalls,and liquefaction.Following the Darfield earthquake, the GeoNet network(New Zealand National Hazard Monitoring Network) and itsregional component the CanNet network (Berrill et al. 2011)was supplemented by the deployment of 13 additional strongmotion instruments regionally (and another nine following theFebruary earthquake). We used this dense network of strongmotion instruments to constrain the source kinematics of theFebruary event. We present the inversion scheme and discussits limitations. These results are preliminary, since more thoroughdata processing is needed; however, they already provide akey model that will help in understanding the sequence of largeaftershocks that has developed near Christchurch. This work isstrongly dependent on other studies by Beavan et al. 2011, page789 of this issue; Fry et al. 2011; Bannister et al. 2011, page839 of this issue; Sibson et al. 2011, page 824 of this issue;and Kaiser, Benites et al. (2011).THE STRONG MOTION DATASETAt the time of the February earthquake there were 14 strongmotion GeoNet sites, from both the national and the regionalCanterbury network CanNet, within 20 km of the epicenter(Figure 1). However, there were strong site effects at stationsPRPC, SHLC, and HPSC, each of which sits on very softground and suffered intense liquefaction from the earthquake;therefore those three were excluded, leaving 11 stations to beincluded in the inversion scheme. The source-station distanceranges from 2 to 20 km.All of the recordings used in this study suffered from siteeffects to some degree. Stations on rock sites are found only onthe hills of Banks Peninsula (south of Christchurch) wherestrong topographic effects are the likely cause of an intensedamage pattern over the hills of Banks Peninsula as describedby Hancox et al. (2011). Stations on the plains suffered fromvery soft shallow layers inducing non-linear amplificationsand extreme phenomena such as liquefaction and trampolineeffects (Fry et al. 2011). Unfortunately, ground conditionswithin Christchurch are highly variable and will require furtherstudies for stations in this region to be included in themodeling (Kaiser, Holden et al. 2011).For our inversion study, the acceleration data has beenintegrated into velocity and filtered using a Butterworth bandpassfilter from 0.1 to 1.0 Hz. Since we are interested in thepolarity and amplitude of the first onset we used a causal filter.We applied the same filter to observed and synthetic data. Thedata from the CanNet stations were rotated from their originalorientation to north-south and east-west components.INVERSION SCHEMEWe inverted three-component data for 11 well-distributedstrong motion stations within 20 km of the epicenter. We useda fixed fault plane geometry of strike 59 and dip 67 as definedby Beavan et al. 2011 (page 789 of this issue) since processedInSAR data clearly shows deformation fringes resulting fromdoi: 10.1785/gssrl.82.6.783Seismological Research Letters Volume 82, Number 6 November/December 2011 783

slip on a 59-degree striking fault plane. Their solution is alsovery similar to the regional moment tensor solution of strike/dip/rake: 55/66/129 (Sibson et al. 2011, page 824 of thisissue). We constrained the fault plane location using the relocatedhypocenter of Bannister et al. 2011 (page 839 of thisissue) of latitude –43.571, longitude 172.703, and 6 km depth.The geometry of the fault plane is fixed, but the rupture isallowed to start anywhere on the fault plane. We didn’t constrainthe rupture to start at the hypocenter since we are invertingfor the time history of the low-frequency source of energyrelease, which can have a different origin time location thanthe hypocenter. In order to account for delays in the onset ofthe main slip patch, the slip distribution is independent of thestarting time. The modeled fault plane dimensions are 20 × 20km discretized into 400 1 × 1 km 2 subfaults. We invert for anelliptical distribution of slip on the fault plane (Di Carli et al.2010) described by nine parameters: the size (two semi-axislengths) and location (along strike and along dip) of the ellipse,location of rupture starting time (along strike and along dip),rake, maximum slip, and rupture velocity.Synthetic seismograms were computed using the discretewavenumber approach of Bouchon (1981), using the 1D velocitymodel from Reyners and Cowan (1993).To search for a minimum waveform misfit, we used thenon-linear Neighborhood Algorithm (Sambridge 1999). Themisfit function is a least-squares scheme applied to the first 10seconds of the recordings. The 10-second window is justifiedby the short rupture time of this moderate-size earthquakeand the proximity of the stations (less than 20 km). The followingsearch parameters are empirically based on many yearsof experience using the neighborhood algorithm for kinematicinversion purposes. In this study the parameters are tuned tobe more exploitative than explorative (Sambridge 1999). Theinversion scheme first computes 400 models then runs 600iterations. For each iteration, the algorithm computes 12 modelsand selects the four models with the lowest misfit to definethe parameter search for the next iteration.RESULTS: SOURCE MODELSA total of 7,600 models were computed during the inversion.Our final slip distribution is shown in Figure 2. The inversionprocedure has converged clearly (Figure 3) to a final misfitvalue of 45.1458. Each individual parameter has also convergedsharply, and the differences between the final values and valuesof the last 1,000 computed models are negligible as seenon Figure 4. The misfit value for model 6530 (iteration 511) is45.1476.The waveform fit is overall very good (Figure 5). All threeonset, amplitude, and polarities criteria are well-matched. Thematch is even better for vertical components, showing fewercomplexities than the horizontal components, where we alsofit the first wave envelopes of the synthetic and observed seismograms.The final slip distribution parameters are shown on Figure1 and Figure 2 and described in Table 1. The rupture is characterizedby a high rupture velocity of 2.8 km/s and a maximumslip value of 4.2 m. The maximum slip is located at about 4 kmdepth; there is still up to 1.5 m slip at 500 m depth. The rupturearea is 12 by 18 km 2 . The total rupture duration is less than4 seconds. The rake angle is 135 degrees. The direction of therupture is nearly vertical, oriented toward Christchurch, hencecontributing to the extreme ground shaking experienced inChristchurch city.DISCUSSIONThis proposed kinematic model shows consistent locationof a slip patch right below the Avon River estuary, depthrange (maximum slip around 4 km depth), and moment of3.46 × 10 18 Nm to the model obtained by Beavan et al. 2011(Mo 3.13 × 10 18 Nm) (page 789 of this issue). However, intheir one-fault model, Beavan et al. obtained a smaller maximumslip amplitude (2.5 m), although this value can increase bychanging the smoothing parameter in their inversion (Beavan,personal communication 2011). Our rake angle of 135 degreesshows a larger reverse component than their value of 154degrees; however, the rake angle of their main fault decreasesif they introduce a secondary strike-slip source. Our solution isvery close to the regional moment tensor solution of strike/dip/rake 55/66/129 (Mo 2.49 × 10 18 Nm) (Sibson et al. 2011, page824 of this issue).The proposed kinematic model is characterized by highrupture parameter values for an earthquake of its size such ashigh rupture velocity (2.8 km/s) and very large slip (maximum4.2 m), suggesting that this was a high stress drop event. Thehigh rupture velocity is also noted in Fry et al. 2011, page 833of this issue; based on data filtered up to 5 Hz, their modelrequires a rupture velocity of 3.2 km/s in order to reproducethe very high accelerations observed near source. Finally, thelarge difference between the energy magnitude of 6.7 (USGS)for this earthquake versus the M w magnitude of 6.2 supportsthe possibility of this event being a high stress drop event (Fryand Gerstenberger 2011, page 833 of this issue).Similar to the Darfield Mw 7.1 event, geodetic studies andseismic data observations of the February earthquake suggestthat this is a segmented event. First, geodetic studies (Beavanet al. 2011, page 789 of this issue) suggest that a secondstrike-slip source (strike/dip/rake 80/90/180) of magnitudeM w 5.9 was involved in the rupture process. However they arenot able to assign a chronology to the two events. Then, thefocal mechanism solution using first motions analysis of theraw data (B. Fry, personal communication 2011) indicates analmost pure right-lateral strike-slip mechanism (strike/dip/rake 95/75/180), strongly suggesting that the February earthquakestarted on a small strike-slip patch and then ruptureda larger oblique-reverse slip patch. Finally, the waveform dataused in the present kinematic inversion (Figure 5) show a dominantpeak, even larger on the horizontal components, cominga few seconds after the main rupture and prior to the 3-secondresonant site effect for the soft site stations (the 3-second siteeffect was observed also in the Darfield earthquake by Cousins784 Seismological Research Letters Volume 82, Number 6 November/December 2011

▲▲Figure 1. Final slip distribution on a plane oriented 59 o strike and 67 o dip. Only the non-zero slip patches are shown here. The reddots are relocated aftershocks from Bannister et al. (2011, this issue) up until 29 May 2011. The light gray region represents the areaof Christchurch city. The stations named xxxC belong to the regional CanNet strong motion network of GeoNet; others are GeoNetnational strong motion stations. Stations used for the inversions are in bold straight letters. The slip distribution is characterizedhere by a patch of maximum 4.2 m slip occurring north and up-dip of the relocated hypocenter (yellow star). Banks Peninsula volcanoextends from just south of HVSC and beyond MQZ.▲▲Figure 2. Slip and rake history for the fault plane orientedstrike 59 o , dip 67 o . Slip is shown in color, and the rake is representedby black vectors for each grid cell. Distances are inkilometers and rupture time iso-contours are in seconds. Theyellow star is the relocated hypocenter from Bannister et al.(2011, this issue).▲ ▲ Figure 3. Convergence of the inversion scheme after 600 iterations.Each iteration computes a waveform misfit value for 12models; a total of 7,600 models have been computed.Seismological Research Letters Volume 82, Number 6 November/December 2011 785

▲▲Figure 4. Convergence of inverted parameters for the fault plane described above after 7,600 models: A) rupture initiation locationalong strike (km), B) rupture initiation location along dip (km), C) ellipse starting point along strike (km), D) ellipse starting point alongdip (km), E) rake angle (degree), F) rupture velocity (km/s), G) and H) semi-axis lengths of the ellipse (km), I) maximum slip (m). They-axis is the parameter space range sampled during the inversion. The x-axis is the number of models run. The figure shows excellentconvergence of the parameters to their final values detailed in Table 1.and McVerry 2010). For central Christchurch stations this signalis even larger than the signal modeled from the main patch.This strongly suggests the presence of a large strike-slip sourceshortly following the main slip patch. Therefore preliminaryseismic observations indicate that at least three subevents wereinvolved in the overall rupture process. This is a subject ofongoing studies.SUMMARY AND FURTHER STUDIESThis model is based on a comprehensive kinematic inversionscheme: high-frequency velocity seismograms and well distributedvery-near-source stations. The results are consistent withother source models of the February earthquake and observedcharacteristics of a very energetic event. Simple waveformTABLE 1Source Parameters for the Final Source ModelMax Slip (m) 4.2Rake (degrees) 135Half-length of main axis (km) 8.86Half-length of secondary axis (km) 6.0Depth (km) @ max slip 4.2Min. depth (km) 0.5Max Depth (km) 9.7V r (km/s) 2.80M 0 (×10 18 Nm) 3.46M w 6.3786 Seismological Research Letters Volume 82, Number 6 November/December 2011

▲ ▲ Figure 5. Observed (black line) and synthetic (dashed line) velocity seismograms computed for the slip model described in the paper;the duration is 30 seconds. Values above the traces are the maximum observed absolute velocities (m/s). The model fits well the onsetpolarities and amplitudes, especially for the vertical components. The top four recordings are from stations located on rock sites orvery shallow soft sites (CMHS); the other ones are located on very soft sites in the plains. A resonant period of about 3 seconds isnoticeable on “soft site” horizontal recordings. A sharp single peak, not modeled by our solution, is also noticeable just prior to the3-second period signal (at 10 seconds on CCCC) on all horizontal components. The amplitude of the peak is actually twice as large forstations further away from the modeled fault plane.Seismological Research Letters Volume 82, Number 6 November/December 2011 787

observations and other source studies that suggest the occurrenceof at least one secondary event make this earthquakevery segmented for its magnitude. Such rupture segmentationhas already been observed regionally for the Darfield Mw 7.1earthquake, as well as in preliminary studies of another largeaftershock, the Mw 6.0 June earthquake (Beavan, personalcommunication 2011). Earthquakes in Canterbury are alsoparticularly energetic (larger ones all present a high Me/Mwratio (Fry and Gerstenberger 2011, page 833 of this issue). Theregion is characterized by the presence of a very dehydrated andbrittle structure, the Hikurangi plateau (Reyners and Cowan1993). This structure pushes the regional brittle-ductile transitiondeeper (35 km), fostering strain release following a largeevent through the generation of aftershocks rather than aseismicslip. The Christchurch area is also marked by the presenceof the intraplate volcanism formation of the now extinct BanksPeninsula volcano, about 11 My ago (Timm et al. 2009). Theintrusion of the volcano has not only highly segmented faultsin the region near Christchurch, but may have also broughtcloser to the surface the very brittle and dehydrated Hikurangiplateau. This segmented and energetic fault system may explainan event like the February earthquake.Finally, we hope to bypass issues arising from variable siteconditions by first obtaining a better defined local velocitymodel from aftershock studies (Reyners et al. 2011; Bannisteret al. 2011, page 839 of this issue), and second by using thelarge database of well-recorded aftershocks as empirical Green’sfunctions. This will allow us to increase the frequency bandwidthof the waveforms and hence define the slip history inmore details.ACKNOWLEDGMENTSThe author would like to acknowledge the anonymous reviewerfor significantly improving this manuscript. This study madeuse of SAC and GMT software.REFERENCESBannister, S., B. Fry, M. Reyners, J. Ristau, and H. Zhang (2011).Fine-scale relocation of aftershocks of the 22 February M w 6.2Christchurch earthquake using double-difference tomography.Seismological Research Letters 82, 839–845.Beavan, J., E. Fielding, M. Motagh, S. Samsonov, and N. Donnelly(2011). Fault location and slip distribution of the 22 February 2011M W 6.2 Christchurch, New Zealand, earthquake from geodeticdata. Seismological Research Letters 82, 789–799.Berrill, J., H. Avery, M. B. Dewe, A. Chanerley, N. Alexander, C. Dyer,C. Holden, and B. Fry (2011). The Canterbury AccelerographNetwork (CanNet) and some results from the September 2010, M7.1 Darfield earthquake. In Proceedings, Ninth Pacific Conference onEarthquake Engineering, NZSEE, Auckland, New Zealand paperno. 181.Bouchon, M. (1981). A simple method to calculate Green’s functions forelastic layered media. Bulletin of the Seismological Society of America71, 959–971.Cousins, J., and G. McVerry (2010). Overview of strong motion datafrom the Darfield earthquake. Bulletin of the New Zealand Societyfor Earthquake Engineering 43 (4), 222–227.Di Carli, S., C. François-Holden, S. Peyrat, and R. Madariaga (2010).Dynamic inversion of the 2000 Tottori earthquake based on ellipticalsubfault approximations. Journal of Geophysical Research 115,B12328; doi:10.1029/2009JB006358.Fry, B., R. Benites, M. Reyners, C. Holden, A. Kaiser, S. Bannister, M.Gerstenberger, C. Williams, J. Ristau, and J. Beavan (2011). Verystrong shaking in the New Zealand earthquakes. Submitted to Eos.Fry, B., and M. Gerstenberger (2011). Large apparent stresses from theCanterbury earthquakes of 2010 and 2011. Seismological ResearchLetters 82, 833–838.Gledhill, K., J. Ristau, M. Reyners, B. Fry, and C. Holden (2011).The Darfield (Canterbury, New Zealand) Mw 7.1 earthquake ofSeptember 2010: A preliminary seismological report. SeismologicalResearch Letters 82 (3), 378–386; doi:10.1785/gssrl.82.6.378.Hancox, G., C. Massey, and N. Perrin (2011). Landslides and relatedground damage caused by the Mw 6.3 Christchurch earthquakeof 22 February 2011. Geomechanics News (New Zealand) 81 (June2011), 53–67.Kaiser, A. E., R. A. Benites, A. I. Chung, A. J. Haines, E. Cochran, andB. Fry (2011). Estimating seismic site response in Christchurch city(New Zealand) from dense low-cost aftershock arrays. ExtendedAbstract of the Fourth IASPEI/IAEE International Symposium onthe Effects of Surface Geology on Seismic Motion, August 23–26,Santa Barbara, California.Kaiser, A., C. Holden, J. Beavan, D. Beetham, R. Benites, A. Celentano,D. Collett et al. (2011). The February 2011 Christchurch earthquake:A preliminary report. Submitted to New Zealand Journal ofGeology and Geophysics.Reyners, M. E., and H. Cowan (1993). The transition from subductionto continental collision: Crustal structure in the north Canterburyregion, New Zealand. Geophysical Journal International 115 (3),1,124–1,136.Reyners, M., D. Eberhart-Phillips, and S. Bannister (2011). Trackingrepeated subduction of the Hikurangi Plateau beneath NewZealand. Earth and Planetary Science Letters; doi:10.1016/j.epsl.Sambridge, M. (1999). Geophysical inversion with a neighbourhoodalgorithm—I. Searching a parameter space. Geophysical JournalInternational 138, 479–494.Sibson, R., F. Ghisetti, and J. Ristau (2011). Stress control of an evolvingstrike-slip fault system during the 2010–2011 Canterbury, NewZealand, earthquake sequence. Seismological Research Letters 82,824–832.Timm, C., K. Hoernle, P. Bogaard, I. Bindeman, and S. Weaver (2009).Geochemical evolution of intraplate volcanism at Banks Peninsula,New Zealand: Interaction between asthenospheric and lithosphericmelts. Journal of Petrology 50 (6), 989–1,023.GNS Science1 Fairway DriveLower HuttAvalon 5010 New Zealandc.holden@gns.cri.nz788 Seismological Research Letters Volume 82, Number 6 November/December 2011

Fault Location and Slip Distribution of the22 February 2011 M W 6.2 Christchurch, NewZealand, Earthquake from Geodetic DataJohn Beavan, Eric Fielding, Mahdi Motagh, Sergey Samsonov, and Nic DonnellyJohn Beavan, 1 Eric Fielding, 2 Mahdi Motagh, 3 Sergey Samsonov, 4 andNic Donnelly 5EOnline material: Additional figures showing interferogramsand fault models; data tablesINTRODUCTIONThe 22 February (local time) M W ~6.2 Christchurch earthquakeoccurred within the aftershock region of the 4 September2010 M W 7.1 Darfield (Canterbury) earthquake (Gledhill etal. 2011). Both the Darfield and Christchurch earthquakesoccurred on previously unknown faults in a region of historicallylow seismicity, but within the zone of plate boundarydeformation between the Pacific and Australian plates. TheDarfield earthquake caused surface rupture up to 5 m (Quigleyet al. 2010, forthcoming), but none has been observed associatedwith the Christchurch earthquake. Geodetic dataindicate that strain has been slowly accumulating within theregion (Wallace et al. 2007; Beavan et al. 2002), and the presenceof active subsurface faults was known or suspected (e.g.,Pettinga et al. 2001). Earthquakes of magnitude up to 7.2 inthis region had been allowed for in the national seismic hazardmodel (Stirling et al. 2002), but the observed high apparentstresses (Fry and Gerstenberger 2011, page 833 of this issue)and high ground accelerations (Fry et al. 2011, page 846 ofthis issue) had not been anticipated, particularly those experiencedin the Christchurch event. These and other factors(Fry and Gerstenberger 2011, page 833 of this issue; Fry et al.2011, page 846 of this issue; Holden 2011, page 783 of thisissue), plus the close proximity of the February earthquake to1. GNS Science, Lower Hutt, New Zealand2. Jet Propulsion Laboratory/Caltech, Pasadena, California, U.S.A.3. Helmholtz Centre Potsdam, GFZ German Research Centre forGeosciences, Potsdam, Germany; also at Department of Geomaticsand Surveying Engineering, University of Tehran, Tehran, Iran4. European Center for Geodynamics and Seismology, Walferdange,Luxembourg; now at Canada Centre for Remote Sensing, Ottawa,Canada5. Land Information New Zealand, Wellington, New ZealandChristchurch city center, were responsible for the major damagecaused by the earthquake (e.g., Kaiser et al. 2011).A large amount of geodetic ground-displacement datais available to constrain the source of the earthquake, in partbecause we reoccupied nearly 200 GPS sites that had beenobserved following the Darfield earthquake, and in partbecause a number of space agencies collected synthetic apertureradar (SAR) data over the source area that we were able to usein differential interferometric SAR (DInSAR) processing. Thegeodetic data were collected one day to seven weeks followingthe February earthquake, so they include ground deformationdue to aftershocks, in particular the M W 5.8 and M W 5.9 eventsthat occurred within two hours of the mainshock.To first order, the earthquake source can be modeled as aplanar fault striking ~59° and dipping ~69° to the southeast.The peak slip of 2.5–3 m is a mixture of reverse and right-lateralslip and is located ~7 km east-southeast of Christchurchcity center at a depth of ~4 km. Slip of ~1 m reaches within~1 km of the ground surface. The slip near the southwest endof the plane is approximately right-lateral with magnitude ~1m. The geodetic data are significantly better fit by two faultplanes, a compact region of oblique slip on the fault describedabove, plus right-lateral strike slip on a near-vertical fault to itssouthwest that coincides with the locations of the two majoraftershocks and with a trend of smaller aftershocks. A lobe ofground uplift seen in some of the SAR data (e.g., Figure 4) justwest of the main slip patch is not well modeled, and suggestssome slip may also have occurred elsewhere, perhaps on a splayoff the main fault plane.GEODETIC DATAWe use campaign GPS data collected between 28 Februaryand 14 April from 57 sites (Figure 1) that were also occupiedfollowing the September 2010 Darfield earthquake(Beavan, Samsonov, Motagh, et al. 2010). We also use continuousGPS (cGPS) data from five regional sites operated byGeoNet (http://www.geonet.org.nz) for Land Informationdoi: 10.1785/gssrl.82.6.789Seismological Research Letters Volume 82, Number 6 November/December 2011 789

-42. 5-43. 0-43. 5-44. 0-44. 5-43.2-43.4-43.6-43.8-44.0-43.45-43.50-43.55-43.60(A)172.50171.0(B)171. 5(C)METHWAIMMETH172.0172. 0172.60C5 km100 km173.025 km172. 5172.70KAI KChristchurc hMQZG37 mm/yrMQZGAHE173. 0172.80▲ ▲ Figure 1. GPS sites used in analysis. A) Regional GPS sites(red triangles) and active faults (purple lines). Station CHAT is800 km due east of Christchurch. Arrow shows Pacific platemotion relative to Australia. B) Red triangles show continuoussites (with LINZ sites labeled), blue triangles show 24-houroccupation sites, and green triangles show four-hour occupationsites (see text). Orange line shows surface rupturefrom 4 September 2010 Darfield earthquake. C) Crosses showshort-occupation sites in Christchurch city and also indicatethe approximate extent of the city. Displacements at sites withblack crosses are large outliers to the dislocation model predictionsand are downweighted in the inversion. AHE and C denotethe Avon-Heathcote estuary and Cashmere.New Zealand (LINZ) and eight sites in Christchurch operatedby private companies. In addition, we use lower-accuracycampaign GPS data collected between 14 April and 27 Aprilfrom 123 sites within Christchurch City and surroundingsuburbs, eight of which were also observed in the higher-accuracycampaign dataset. The GPS displacement data are listedin Table S1 in the Supplementary Information. We use differentialinterferometric synthetic aperture radar (DInSAR;Table 1) from the Italian Cosmo-SkyMed (CSK) X-band (3.1cm wavelength) radar satellite, both from an ascending trackwith images acquired on 19 February and 23 February, and adescending track with images acquired on 20 February and 16March. We also have available two ascending interferogramsfrom the Japanese ALOS/PALSAR L-band (23.6 cm wavelength)instrument, with time spans 10 January–25 February2011 and 27 October 2010–14 March 2011; we use only thesecond of these in our modeling as it has better spatial coverage.GPS ProcessingThe “high accuracy” campaign GPS dataset is composed of amix of 34 stations that have at least one 24-hour session bothbefore and after the earthquake and 23 stations with at least onefour-hour session both before and after (Figure 1). We processedthese data together with the cGPS data by standard methods(e.g., Beavan, Samsonov, Denys, et al. 2010) using Bernese version5.0 software (Dach et al. 2007) to give station coordinatesand their estimated covariances for each day of observation. Weplaced the coordinates in the IGS05 reference frame by a translationof each daily solution to best fit the IGS05 coordinates ofa set of regional Australian and Pacific sites at the epoch of survey.We combined the resulting pre-earthquake daily coordinateand covariance data using least-squares inversion software“adjcoord” (Bibby 1982; Crook 1992) to give minimally constrainedpre-earthquake coordinates and covariances by holdingstation CHAT (Chatham Island, Figure 1) fixed. Duringthis procedure, we multiplied the formal covariances from theBernese software by 25 (i.e., uncertainties are multiplied by 5)to account for unmodeled temporal correlation between the180-sec samples used in the final stages of the GPS processing(e.g., Darby and Beavan 2001). We used the same procedure onthe post-earthquake data. We are able to make two pre-earthquakeposition estimates for a number of the cGPS stations inChristchurch. In one case we use pre-earthquake data at thesame epoch as the pre-earthquake campaign data (September–October 2010); in the other we take pre-earthquake continuousdata from shortly before the earthquake (January–February2011). There are no significant systematic differences betweenthe two estimates after taking interseismic motion intoaccount (see Table S1). We also found that post-seismic deformationat GPS sites in the vicinity of the Darfield earthquakewas small (

TABLE 1InSAR DetailsData typeTrack*Dates,yyyymmddCSK asc — 2011021920110222CSK desc — 2011022020110316ALOS asc 336 2010102720110314ALOS asc335 20110110(not used)20110225* CSK track numbers are not available.Multi-looks, pixelsApprox groundBaseline, m Across-track Along-track resolution, m20 10 10 22123 16 16 302100 3 6 40359 3 6 40continuous GPS data (unpublished) from sites in Christchurchshow that post-seismic deformation from a few days to threemonths after the February earthquake is small, so we are justifiedin combining all the post-earthquake geodetic data intoa single post-earthquake solution. We combine the pre- andpost-earthquake coordinates and covariances using in-houseleast-squares inversion software “disp” to generate coseismicdisplacements and their estimated uncertainties (Table S1).Plate tectonic motion and reference frame rotation in the ~sixmonths between the pre- and post-earthquake coordinate setsis accounted for by solving in the inversion for the rigid-bodytranslation and rotation of three far-field sites nominally on thePacific plate: CHAT (Chatham Island), KAIK (Kaikoura),and WAIM (Waimate) (Figure 1A).The “low-accuracy” campaign GPS data consist of two30-minute observations in October 2010 and similar observationsin April 2011 (Figure 1C). The data were collected forLINZ by Opus International Consultants Limited and processedto baseline solutions using industry software. The baselineswere provided to LINZ, and baseline errors were estimatedfrom experience with similar surveys. The data were combinedusing LINZ’s geodetic adjustment software “SNAP” into a preearthquakeand a post-earthquake solution. In each solution,three stations common to the high- and low-accuracy surveys(MQZG, 5508, B87X; see Table S1) were tightly constrained totheir coordinates derived in the processing of the high-accuracysurveys. Based on adjustment results and experience, the resultingcoordinates were assigned horizontal 1s uncertainties of 20mm and a vertical 1s uncertainty of 30 mm.SAR ProcessingThe single-look complex CSK ascending (satellite moving northand looking eastward) data were processed using the SARscapesoftware (http://www.sarmap.ch/), while the descending (satellitemoving south and looking westward) CSK data wereprocessed from raw data with the Jet Propulsion Laboratory(JPL)/Caltech ROI_pac software. The raw ALOS PALSARdata were processed using the GAMMA software (http://www.gamma-rs.ch). The topographic contribution to the interferometricphase was removed using a 3 arc sec digital elevationmodel (DEM) from the Shuttle Radar Topography Mission(SRTM). The DInSAR interferogram phases were then filteredusing a weighted power spectrum technique (Goldsteinand Werner 1998). The CSK data were unwrapped using aminimum cost flow algorithm (Chen and Zebker 2002), andthe ALOS PALSAR data were unwrapped using a branch-cutregion growing algorithm (Rosen et al. 1994). Finally, all interferogramswere projected to a geographic grid using the SRTMDEM. Further details of the DInSAR processing are given inTable 1.Before modeling, each of the interferograms is sampledwith a quadtree algorithm (e.g., Jónsson et al. 2002). In thisprocedure, the scene is divided into four quadrants. If the rootmean square (rms) scatter about the mean in any quadrantexceeds a given threshold, the quadrant is divided into fournew quadrants, and the rms scatter about the mean is againtested. The process continues iteratively until convergence.Data reduced in this manner represent the statistically significantportion of the signal with far fewer sampling points thanthe original, 300–400 points per interferogram in this case(see Figures S1–S3).In the processed interferograms, the shorter wavelengthCSK interference patterns (Figures 2, S1, S2) are incoherentthroughout substantial parts of Christchurch city, presumablybecause the scattering properties of the ground changed due toground failure and/or building damage. The longer-wavelengthALOS data are also incoherent over some parts of the city, butdo show regions of coherence (Figure 3) where it is lacking in theCSK data (Figure 2). In particular, a coherent region of towardsatellitefringes is visible just west of the Avon-Heathcote estuaryin both the track 335 and track 336 ALOS interferograms(e.g., Figure 3). However, we were not able to reliably phase connectthis area with the rest of the interferogram, so it has beenomitted from the unwrapped image we use in our modeling(Figure S3) by setting a high coherence threshold of 0.7.We also performed pixel tracking, or sub-pixel correlationanalysis, on the CSK ascending and descending SARimage pairs using the ROI_pac software (Pathier et al. 2006).Cross-correlation of the SAR amplitude images was done with64 × 64 pixel windows, giving measurements of line-of-sightSeismological Research Letters Volume 82, Number 6 November/December 2011 791

“Product CSK © ASI, (ItalianSpace Agency), year ofacquisition, 2011, distributedby e-GEOS (an ASI/TelespazioCompany).”10 kmGround displacementaway from satelliteChristchurchcentral businessdistrictFlight directionRadar look directionIncidence angle = 36°Ground displacementtowards satellite▲ ▲ Figure 2. The colored image shows an interferogram derived from CSK X-band radar images acquired on 19 and 23 February 2011,overlaid on shaded topography. Each color cycle (“fringe”) represents 1.55 cm of ground displacement in the line-of-sight (LOS) directionfrom the ground to the satellite. The total LOS displacement between the western edge of the image and central Christchurch ismore than 20 cm. The image is incoherent over most of Christchurch city due to extensive ground and building damage. The order ofthe colors in the fringes indicates whether ground displacement is toward or away from the satellite, as shown by text in two regions.See Figure 3 and Figures S1–S4 for other interferograms.and along-track displacements with a spatial resolution of about100 m. This technique measures the horizontal displacementsin the along-track direction and the line-of-sight displacements(same component as InSAR) with a precision of about 5–20 cm(see Figure 4).EARTHQUAKE SOURCE MODELING FROMDISPLACEMENT DATAGPS data provide a 3D coseismic displacement vector at a setof points, while DInSAR provides coseismic ground displacementin the line-of-sight from the ground to the satellite atpoints throughout the interference pattern, provided the patternremains coherent. For the modeling we define five differentdatasets: 1) 4-hr and 24-hr GPS; 2) 1-hr GPS; 3) CSK ascendingimage; 4) ALOS ascending image; and 5) CSK descendingimage. We apply an overall weighting to each dataset so thateach one has approximately the same misfit ( χ 2 per degree offreedom) in the best-fitting model.Single-fault ModelsThe displacement data are modeled in two steps (Arnadottirand Segall 1994) to solve for the location and geometry of thefault plane and the amount and direction of slip. First, we usednon-linear least-squares inversion software “disloc99” (Darbyand Beavan 2001) to solve for the best-fitting uniform-slip,792 Seismological Research Letters Volume 82, Number 6 November/December 2011

ALOS/PALSAR data fromJapanese Space AgencyAHERegion of fringes omittedfrom modeling (see text)5 km▲▲Figure 3. Original (wrapped) version of the ALOS track 335 ascending interferogram of 10 January–25 February 2011. Each fringerepresents 11.8 cm of apparent ground motion in the line-of-sight direction to the satellite. A region of narrowly spaced fringes is visibledue west of the Avon-Heathcote estuary (AHE). We have been unable to reliably phase connect these fringes with the remainder ofthe interferogram, so have left this region out of our current modeling. Compare with Figure 4, where a toward-satellite displacementis observed in this region using sub-pixel correlation techniques.10 kmRegion of towardssatellitedisplacementalso seen in ALOSinterferogramsmetersradar look▲ ▲ Figure 4. Ground displacement observed along line of sight to CSK satellite on ascending track, using sub-pixel correlation. Thistechnique provides measurements in regions where the DInSAR technique fails due to loss of coherence, though at lower resolution.Compare with other ascending track images in Figures 2, 3, S1, and S3. In this image, red denotes displacement away from the satelliteand blue denotes toward (opposite convention to other images in the paper).Seismological Research Letters Volume 82, Number 6 November/December 2011 793

TABLE 2Solutions for fault location, geometry and slipUniform-slip, GPS only One fault Two faults, GPS+SARParameter 1σ uncert. GPS+SAR Plane 1 Plane 2lat –43.541° 0.1 km –43.545° –43.535° –43.575°lon 172.691° 0.2 km 172.690° 172.711° 176.666°strike 56° 1° 59° 58° 79.5°dip 61° 1° 66.5° 72° 87°rake 147° 2° see Fig. 5 see Fig. 6 see Fig. 6length 10.6 km 0.2 km 16 km 12 km 8 kmtop depth 1.4 km 0.1 km 1 km 1 km 2 kmwidth 5.3 km fixed 7 km 7 km 6 kmslip 2.1 m 0.1 m see Fig. 5 see Fig. 6 see Fig. 6M W 6.35 6.3 6.25 5.95Uniform-slip GPS-only solution has bottom depth of fault fixed at 6 km in this model.Lat and lon are the center of the fault trace if the fault plane were extended to the ground surface.M W calculated assuming 3 × 10 10 Nm –2 rigidity.Formal 1s uncertainties for the uniform-slip solution were calculated using the method recommended by the authors of thenon-linear least squares algorithm used by Darby and Beavan (2001).rectangular fault plane in an elastic half-space, using the GPSdata only (because disloc99 has not been modified to includeDInSAR data). We found that it was necessary to constrainonly the lower depth of faulting to obtain a solution for theother eight parameters describing the fault. Several sites,all in the eastern suburbs of Christchurch and other placeswhere ground damage was prevalent, had very large residuals.We increased the estimated uncertainties on these dataand repeated the inversion to give the solution in Table 2. Ifwe fix the bottom depth to different values we find a strongtrade-off between fault width and slip magnitude, but thecentroid depth is well constrained by the modeling at about3.5 km (i.e., if the bottom depth is fixed shallower, the solutionsfor top depth and slip magnitude are deeper and larger,respectively). The inferred M W is about 0.1 higher than seismologicalestimates (J. Ristau, personal communication 2011;see also Sibson et al. 2011, page 824 of this issue), but thisis not surprising given that deformation due to aftershocks isincluded in the geodetic solution.We extended the uniform-slip fault plane a few km inall directions and did a linear inversion for slip on this faultplane using both the GPS and DInSAR data, again assumingan elastic half-space model. We used inversion software basedon Jónsson et al. (2002) as described by Beavan, Samsonov,Denys, et al. (2010), solving for a linear ramp on each of theDInSAR datasets. We then did a grid search in the vicinity ofthis solution, varying the location, strike, and dip of the faultand repeating the inversion at each step to find the lowest χ 2solution. Some of the GPS data previously marked as outlierswere now reasonably good fits to the resulting model, sowe restored their original uncertainty values and repeated theinversion and grid search. The resulting fault location and slipdistribution are shown in Figures 5 and S5, the fault parametersare given in Table 2, and the slip solution is tabulated inTable S2. The fault runs from near Cashmere (Figure 1C) at itssouthwest end toward the Avon-Heathcote estuary and a fewkilometers offshore, with a strike of 59° and a dip of 66.5° to thesoutheast, a little steeper than in the uniform-slip, GPS-onlysolution. The main patch of slip, with maximum magnitude~2.5 m, is centered at a depth of ~4 km beneath the estuaryand is a mix of reverse faulting and right-lateral strike slip. Theslip on the southwestern part of the fault, with a maximummagnitude of ~1 m, is predominantly right-lateral strike slip.A preliminary, but very similar, version of this model using asubset of the data was published by Kaiser et al. (2011).Two-fault ModelThree observations suggest that a model consisting of morethan one fault might provide a significantly improved fit to theobservations: 1) the presence of a distinct lineation of smallaftershocks (Bannister et al. 2011, page 839 of this issue) tothe south of the inferred fault plane; 2) the location of twomajor aftershocks near the west end of this lineation; and 3)the pattern of misfits to some of the GPS data. We tried atwo-fault model with one plane coincident with the one-faultmodel and another coincident with the aftershock lineation,then did a grid search varying the location and geometry ofthe faults and solving for the slip distribution at each step. Theminimum misfit solution is shown in Figures 6 and S6, parametersfor both planes are listed in Table 2, and the full solutionis tabulated in Table S3. Again, some sites previously markedas misfits were now in good agreement with the model so wereturned their uncertainties to their original values before thefinal inversion run; it is this final set of uncertainties that isgiven in Table S1. In this solution the southern fault plane isclosely coincident with the aftershock lineation and with the794 Seismological Research Letters Volume 82, Number 6 November/December 2011

-43.45(A)degrees N-43.50-43.552.52.01.51.00.50.0Fault slip, mupper edge-43.60200 mm observed200 mm modelled172.55172.60172.65 172.70degrees E172.75172.80distance down dip, km02468(B)2.5 m2. 52. 01. 51. 00. 50. 0Slip, m0510 15 20SW distance along strike, km NE▲ ▲ Figure 5. A) Locations of model fault and its slip magnitude (colored rectangles) assuming a single planar fault, GPS displacementsobserved (blue arrows) and modeled (red arrows), and aftershocks since September 2010 (crosses). Central Christchurch shown bysolid black square. B) Slip distribution of hanging wall relative to footwall on model fault plane. Red-and-white four-pointed stars showlocations of mainshock on 22 February and (in A) the two major aftershocks to its southwest a few hours later.estimated hypocentral locations and focal planes of the twomajor aftershocks (S. Bannister, J. Ristau, personal communication2011; see also Sibson et al. 2011, page 824 this issue).The estimated M W is also close to that estimated seismologicallyfor the aftershocks. The mainshock plane is somewhatsteeper than in the single-fault solution, and the slip patch ismore concentrated. By changing the degree of smoothing inthe inversion the maximum slip can vary from less than 2.5 mto over 3 m, but the moment is stable.We perform an F-test to determine whether the two-faultsolution is significantly better than the one-fault solution. Theweighted residual sum of squares, number of data, and numberof parameters are 8,210, 1,657, 231 (one fault), and 5,400,1,657, 278 (two faults). These values give a tiny probability thatthe two models fit the data equally well. However, the truenumber of parameters is overestimated due to smoothing ofthe solution so this result is not definitive.Consistency of the DatasetsWe investigate the consistency of the datasets by re-running thesolution with different weightings for the GPS and DInSARdata sets. If the DInSAR data are strongly downweighted (i.e.,effectively a GPS-only solution), the location of maximum slipshallows by about 0.5 km but there is little other change. IfSeismological Research Letters Volume 82, Number 6 November/December 2011 795

-43.45(A)degrees N-43.50-43.552.52.01.51.00.50.0Fault slip, mupper edge-43.60200 mm observed200 mm modelled172.55172.60172.65 172.70degrees E172.75172.80distance down dip, km02468(B) Main shock2.5 m2. 52. 01. 51. 00. 50. 0Slip, m051015SW distance along strike, km NEdistance down dip, km02468(C) Aftershocks2.5 m2. 52. 01. 51. 00. 50. 0Slip, m051015W distance along strike, km E▲ ▲ Figure 6. A) Locations of model faults and their slip magnitudes (colored rectangles), GPS displacements observed (blue arrows)and modeled (red arrows), and aftershocks since September 2010 (crosses). Slip distribution of hanging wall relative to footwall onmodel fault planes of B) 22 February mainshock and C) 22 February aftershocks. Red-and-white four-pointed stars show locations ofmainshock and the two major aftershocks a few hours later.796 Seismological Research Letters Volume 82, Number 6 November/December 2011

-43.45km0 10 200-43.50-50degrees N-43.5503000200-43.60100200 mm observed200 mm modelled172.55172.60172.65 172.70degrees E172.75172.80▲ ▲ Figure 7. Observed (blue arrows) and modeled (red arrows) vertical displacements for the model of Figure 6. Predicted model displacementsare also shown as contours with 50 mm spacing. Central Christchurch shown by solid black square. An extensive regioneast of central Christchurch shows subsidence exceeding the model predictions, probably as a result of ground failure due to liquefaction,lateral spreading, and compactionthe GPS data are strongly downweighted, the maximum slipdecreases by 10–15% and its depth increases by about 0.5 km.In either case the goodness of fit of the non-downweighted datadoes not change significantly from the original solution. Wetake this as evidence that the solution is not strongly dependenton a particular dataset.DISCUSSIONThe Christchurch earthquake occurred within the wider aftershockregion of the September 2010 Darfield earthquake, andvery close to a strongly felt M W 5.1 aftershock (http://www.geonet.org.nz/earthquake/quakes/3368445g.html) that occurredwithin a few days of the Darfield mainshock. This indicatesthat stress changes due to Darfield almost immediately causedsignificant earthquake activity in the vicinity of the futureFebruary earthquake. Calculations by ourselves and others(e.g., Zhan et al. 2011, page 800 of this issue) show positive butvery small Coulomb stress changes from Darfield in the regionof the February quake; these results do not highlight easternChristchurch as a region of large Coulomb stress increase. TheChristchurch event seems less complex than Darfield, withmost of the surface deformation (away from the liquefactionregions) explicable by slip on two sub-parallel fault planes; theDarfield event involved several reverse fault segments in additionto the main strike-slip fault.An inversion for fault slip kinematics using strong-motiondata is reported by Holden (2011, page 783 of this issue),using a fault geometry based on the geodetic solution. As wellas revealing details of the rupture process, she finds a similarfault slip distribution, depth, and magnitude, though a slightlyhigher ratio of reverse faulting to strike-slip (rake 135° comparedto 145°–150° on the main slip patch of the geodetic model) anda larger maximum slip (more than 4 m compared to 2.5–3 mfor the geodetic model). This provides a degree of confidencein both the geodetic and strong-motion models, but indicatesthere are still differences to be resolved with future work.Both the CSK ascending and descending datasets fit wellwith the majority of the GPS data, leading to generally lowresiduals between model and observations (Figures S1, S2). Theascending ALOS data have a slightly worse fit (Figure S3), butthis mostly occurs in regions where the ALOS data are coherentand the CSK data are not. Some of the GPS stations in thelow-lying areas between central Christchurch and the coastalso have large residuals to the model (Figures 6–7). The GPSdata also show a significant region of ground subsidence in centralChristchurch (Figure 7) amounting to tens of centimetersin excess of what is modeled, even in regions where the modelSeismological Research Letters Volume 82, Number 6 November/December 2011 797

is still a good fit to the horizontal observations (Figure 6). Wesuspect this is due to liquefaction and compaction of underlyingsediments even in regions where major observable groundsurface damage did not occur.The longer-wavelength ALOS data are coherent in partsof central and eastern Christchurch where CSK lost coherencebecause of severe ground damage, including major liquefaction.In particular, there is a clear region of ALOS fringes(Figure 3) west of the Avon-Heathcote estuary and just west ofthe main slip patch in the models of Figures 5–6. We obtainedtwo ALOS pairs, one using a post-earthquake scene only twodays after the earthquake and the other from nearly a monthlater. The two images are quite similar to each other, suggestingthat ALOS is detecting real ground deformation (or at least realphase changes) in this area. In support of this, pixel-tracking(i.e., non-interferometric) analysis of the CSK ascending data(Figure 4) also shows a region of toward-satellite ground motionin this area. We have attempted to model these signals as a smallshallow fault splaying off the main fault plane. Although thisdoes improve the fit to the ALOS data without degrading fitsto the other data, we have not found a solution that provides aclearly significant improvement. More complicated fault geometrymay be necessary to fit all of the details of the surface deformationsouth and east of Christchurch. Also, the deformationdetected by satellite radar and GPS in these regions would nothave been purely fault related, but would include deformationdue to ground damage and phase changes due to variationsof water content in the near surface; the mixture of shallowground damage and deeper fault slip may be difficult to unravel.We have used a uniform elastic half-space for all our modeling.There is in fact significant topography in the region, andthere are both depth and lateral variations in structure. Northof the fault are flat lying gravels and muds over greywacke basement(Forsyth et al. 2008), beneath which is dehydrated oceanicplateau material (Reyners et al. 2011). To the south of thefault are hills that are the remnant of a late Miocene volcanothat formed through the oceanic plateau crust. Future geodeticmodeling should take into account this elastic structure andtopography. However, because there are so many near-field dataconstraining the fault location, we doubt that our conclusionson the fault geometry and slip will be greatly changed by moresophisticated modeling.The geodetic source model presented here is just one partof the still-unfolding story of the earthquake sequence thatbegan in September 2010 and is continuing at the time of thiswriting with a damaging M W 6 aftershock on 13 June 2011.Multiple different fault surfaces have been active so far, andeach of the larger earthquakes has produced radiated energywell above the average expected for the size of the fault. Howand why this large amount of energy has been released shouldbecome clearer with future research.ACKNOWLEDGMENTSWe thank GeoNet, Trimble Navigation NZ Ltd, GeosystemsNZ Ltd, and Global Survey Ltd for providing continuousGPS data, and Josh Thomas, Dave Collett, Paul Denys, KirbyMacLeod, and Linda Alblas for their assistance with the postearthquakeGPS surveys. We thank Stephen Bannister andCaroline Holden for providing comments on the manuscript,and an anonymous reviewer for a number of suggestions thathelped us improve the paper. CSK original data is copyright2011 Italian Space Agency; part was provided by e-GEOS, anASI/Telespazio company, and part was provided under CSKAO PI project 2271. ALOS original data is copyright 2010and 2011 METI and JAXA, distributed by GeoGRID andPASCO. The inversions used Igor Pro (http://www.wavemetrics.com/);figures were prepared using Igor Pro and GMT(http://gmt.soest.hawaii.edu/). Much of this research wasfunded by the New Zealand government. Part of this researchwas performed at the Jet Propulsion Laboratory, CaliforniaInstitute of Technology, under contract with the NationalAeronautics and Space Administration.REFERENCESArnadottir, T., and P. Segall (1994). The 1989 Loma Prieta earthquakeimaged from inversion of geodetic data. Journal of GeophysicalResearch 99, 21,835–21,855.Bannister, S., B. Fry, M. Reyners, J. Ristau, and H. Zhang (2011).Fine-scale relocation of aftershocks of the 22 February M w 6.2Christchurch earthquake using double-difference tomography.Seismological Research Letters 82 (6), 839–845.Beavan, J., S. Samsonov, P. Denys, R. Sutherland, N. Palmer, and M.Denham (2010). Oblique slip on the Puysegur subduction interfacein the 2009 July M W 7.8 Dusky Sound earthquake from GPSand InSAR observations: Implications for the tectonics of southwesternNew Zealand. Geophysical Journal International 183 (3),1,265–1,286; doi: 10.1111/j.1365-246X.2010.04798.x.Beavan, J., S. Samsonov, M. Motagh, L. Wallace, S. Ellis, and N. Palmer(2010). The Darfield (Canterbury) earthquake: Geodetic observationsand preliminary source model. Bulletin of the New ZealandSociety for Earthquake Engineering 43 (4), 228–235.Beavan, J., P. Tregoning, M. Bevis, T. Kato, and C. Meertens (2002).Motion and rigidity of the Pacific plate and implications for plateboundary deformation. Journal of Geophysical Research 107 (B10);doi:10.1029/2001JB000282.Bibby, H. M. (1982). Unbiased estimate of strain from triangulationdata using the method of simultaneous reduction. Tectonophysics82 (1–2), 161–174.Chen, C. W., and H. A. Zebker (2002). Phase unwrapping for large SARinterferograms: Statistical segmentation and generalized networkmodels. IEEE Transactions on Geoscience and Remote Sensing 40,1,709–1,719.Crook, C. N. (1992). ADJCOORD: A Fortran Program for SurveyAdjustment and Deformation Modelling. New Zealand GeologicalSurvey EDS Report 138, Department of Scientific and IndustrialResearch, Geology and Geophysics, Lower Hutt, New Zealand.Dach, R., U. Hugentobler, P. Fridez, and M. Meindl (2007). Bernese GPSSoftware Version 5.0. Bern, Switzerland: Astron. Inst., University ofBern, 612 pp.Darby, D. J., and R. J. Beavan (2001). Evidence from GPS measurementsfor contemporary interplate coupling on the southern Hikurangisubduction thrust and for partitioning of strain in the upper plate.Journal of Geophysical Research 106 (B12), 30,881–30,891.Forsyth, P. J., D. J. A. Barrell, and R. Jongens (2008). Geology of theChristchurch Area. Institute of Geological and Nuclear Sciences1:250,000 Geological Map 16, 1 sheet + 67 pp. Lower Hutt, NewZealand: GNS Science.798 Seismological Research Letters Volume 82, Number 6 November/December 2011

Fry, B., and M. Gerstenberger (2011). Large apparent stresses from theCanterbury earthquakes of 2010 and 2011. Seismological ResearchLetters 82, 833–838.Fry, B., R. Benites, and A. Kaiser (2011). The character of accelerationsin the M w 6.2 Christchurch earthquake. Seismological ResearchLetters 82, 846–852.Gledhill, K., J. Ristau, M. Reyners, B. Fry, and C. Holden (2011). TheDarfield (Canterbury, New Zealand) M W 7.1 earthquake ofSeptember 2010: A preliminary seismological report. SeismologicalResearch Letters 82 (3), 378–386; doi:10.1785/gssrl.82.6.378.Goldstein, R. M., and C. L. Werner (1998). Radar interferogram filteringfor geophysical applications. Geophysical Research Letters 25,4,035–4,038.Holden, C. (2011). Kinematic source model of the 22 February 2011M w 6.2 Christchurch earthquake using strong motion data.Seismological Research Letters 82, 783–788.Jónsson, S., H. Zebker, P. Segall, and F. Amelung (2002). Fault slip distributionof the Hector Mine earthquake estimated from satelliteradar and GPS measurements. Bulletin of the Seismological Societyof America 92, 1,377–1,389.Kaiser, A., C. Holden, J. Beavan, D. Beetham, R. Benites, A. Celentano,D. Collett et al. (2011). The February 2011 Christchurch earthquake:A preliminary report. Submitted to New Zealand Journal ofGeology and Geophysics.Pathier, E., E. J. Fielding, T. J. Wright, R. Walker, B. E. Parsons, andS. Hensley (2006). Displacement field and slip distribution of the2005 Kashmir earthquake from SAR imagery. Geophysical ResearchLetters 33, L20310; doi:20310.21029/22006GL027193.Pettinga, J. R., M. D. Yetton, R. J. Van Dissen, and G. L. Downes (2001).Earthquake source identification and characterisation for theCanterbury region, South Island, New Zealand. Bulletin of the NewZealand Society for Earthquake Engineering 34 (4), 282–317.Quigley, M., R. Van Dissen, N. Litchfield, P. Villamor, B. Duffy, D.Barrell, K. Furlong, T. Stahl, E. Bilderback, and D. Noble (forthcoming).Surface rupture during the 2010 M W 7.1 Darfield(Canterbury) earthquake: Implications for fault rupture dynamicsand seismic-hazard analysis. Geology 40 (1).Quigley, M., P. Villamor, K. Furlong, J. Beavan, R. Van Dissen, N.Litchfield, T. Stahl, B. Duffy, E. Bilderback, D. Noble, D. Barrell, R.Jongens, and S. Cox (2010). Previously unknown fault shakes NewZealand’s South Island. Eos, Transactions, American GeophysicalUnion 91, 469–472.Reyners, M., D. Eberhart-Phillips, and S. Bannister (2011). Trackingrepeated subduction of the Hikurangi plateau beneath NewZealand. Earth and Planetary Science Letters; doi:10.1016/j.epsl.2011.09.011.Rosen, P. A., C. W. Werner, and A. Hiramatsu (1994). Two-dimensionalphase unwrapping of SAR interferograms by charge connectionthrough neutral trees. Proceedings of the IGARSS’94, Pasadena, CA(8–12 August 1994).Sibson, R., F. Ghisetti, and J. Ristau (2011). Stress control of an evolvingstrike-slip fault system during the 2010–2011 Canterbury, NewZealand, earthquake sequence. Seismological Research Letters 82(6), 824–832.Stirling, M. W., G. H. McVerry, and K. R. Berryman (2002). A new seismichazard model for New Zealand. Bulletin of the SeismologicalSociety of America 92, 1,878–1,903; doi:10.1785/0120010156.Wallace, L., J. Beavan, R. McCaffrey, K. R. Berryman, and P. Denys(2007). Balancing the plate motion budget in the South Island,New Zealand, using GPS, geological and seismological data.Geophysical Journal International 168 (1); doi:10.1111/j.1365-246X.2006.03183.x.Zhan, Z., B. Jin, S. Wei, and R. W. Graves (2011). Coulomb stresschange sensitivity due to variability in mainshock source modelsand receiving fault parameters: A case study of the 2010–2011Christchurch, New Zealand, earthquakes. Seismological ResearchLetters 82, 800–814.GNS ScienceLower Hutt, New ZealandP. O. Box 30368Lower Hutt 5040 New Zealandj.beavan@gns.cri.nz(J. B.)Seismological Research Letters Volume 82, Number 6 November/December 2011 799

Coulomb Stress Change Sensitivity due toVariability in Mainshock Source Models andReceiving Fault Parameters: A Case Study ofthe 2010–2011 Christchurch, New Zealand,EarthquakesZhongwen Zhan, Bikai Jin, Shengji Wei, and Robert W. GravesZhongwen Zhan, 1,2 Bikai Jin, 3 Shengji Wei, 1 and Robert W. Graves 4INTRODUCTIONStrong aftershocks following major earthquakes present significantchallenges for infrastructure recovery as well as for emergencyrescue efforts. A tragic instance of this is the 22 February2011 M w 6.3 Christchurch aftershock in New Zealand, whichcaused more than 100 deaths while the 2010 M w 7.1 Canterburymainshock did not cause a single fatality (Figure 1). Therefore,substantial efforts have been directed toward understandingthe generation mechanisms of aftershocks as well as mitigatinghazards due to aftershocks. Among these efforts are the predictionof strong aftershocks, earthquake early warning, and aftershockprobability assessment. Zhang et al. (1999) reported asuccessful case of strong aftershock prediction with precursorydata such as changes in seismicity pattern, variation of b-value,and geomagnetic anomalies. However, official reports of suchsuccessful predictions in geophysical journals are extremelyrare, implying that deterministic prediction of potentiallydamaging aftershocks is not necessarily more scientifically feasiblethan prediction of mainshocks.A potentially more effective approach for aftershock hazardmitigation is described by Bakun et al. (1994) for the caseof the Loma Prieta earthquake. This approach relies on therapid detection of an aftershock using a dense observationnetwork in the rupture area of the mainshock and subsequentbroadcast of an alert to more distant sites. Recent progress inrapid determination of epicenter and magnitude involving a1. Seismological Laboratory, California Institute of Technology,Pasadena, California 91125 U.S.A.2. Mengcheng National Geophysical Observatory, School of Earthand Space Sciences, University of Science and Technology of China,Hefei 230026 China3. Key Laboratory of Dynamic Geodesy, Institute of Geodesy andGeophysics, Chinese Academy of Sciences, Wuhan 430077 China4. U.S. Geological Survey, 525 South Wilson Avenue, Pasadena,California, 91106 U.S.A.small number of stations and short time window of P waveforms(Allen and Kanamori 2003; Wan et al. 2009; Wang etal. 2009) make the approach of earthquake early warning moreeffective for regions not very close to the rupture area of themainshock. Such an approach might have been useful for aftershocksof the 2008 Wenchuan earthquake, where megacitiessuch as Chengdu are about 90 km away and 20 seconds wereavailable for rapid mitigation response. But in the case of the2011 Christchurch earthquake, the populated region is onlyabout 10 km from the epicenter, thus leaving little time forearly warning.For a situation such as Christchurch, aftershock probabilityassessment may provide a viable approach to address thehazard level. Several aftershock-triggering mechanisms, i.e., thestatic Coulomb stress theory (King et al. 1994; Stein 1999),the dynamic triggering theory (Felzer and Brodsky 2006),and viscoelastic relaxation theory (Freed et al. 2001), can beapplied to assess aftershock probabilities. In this paper we willconcentrate on how applicable the static Coulomb stress triggeringmechanism is to the 2011 Christchurch aftershock andexamine the sensitivity of the stress changes to mainshock slipdistribution and aftershock fault orientation. The Coulombstress theory has been broadly applied in aftershock studies(e.g., King et al. 1994; Parsons et al. 1999; Toda et al. 1998; Maet al. 2005), earthquake sequencing (Stein et al. 1997; Xionget al. 2010; Nalbant et al. 1998) and the triggering of large tomoderate earthquakes (Parsons et al. 2000). Previous studies(e.g., Harris 1998, 2000; Freed 2004; King et al. 1994; Stein1999) proposed a Coulomb stress change of 0.01 MPa to be thethreshold for potential earthquake triggering.The Coulomb stress triggering theory involves computingthe change in normal traction and shear traction on a fault(receiving fault) caused by changes of the stress field due to themainshock. Therefore, accurate information on the receivingfault geometry (strike, dip, rake, and focal depth) and sourcemodel of the mainshock are necessary for effective assessment800 Seismological Research Letters Volume 82, Number 6 November/December 2011 doi: 10.1785/gssrl.82.6.800

43.25˚S0 5 10 15 200 5 10 15 20M3 M5 M7depth(km)depth(km)GCMT/Mw7.0CAP/Mw6.343.5˚S43.75˚Skm0 10 20171.75˚E 172˚E 172.25˚E 172.5˚E 172.75˚E▲ ▲ Figure 1. Seismicity in the first three days after the 2010 Canterbury mainshock (color-scaled dots) and the 2011 Christchurchaftershock (dots in gray scale). The red lines are the free surface rupture trace from field observation. The big red star indicates theepicenter of the main event and the smaller red star is the location of the M w 6.3 aftershock. The GCMT solution of mainshock and thecut-and-paste (CAP) mechanism of the aftershock are shown as beach balls. The black rectangles are the free surface projection ofthe two-segment model, which are used in the teleseismic finite fault inversion. The yellow rectangles are for the four-segment modelused in the stochastic slip model.of aftershock probability. Due to inadequate coverage of seismicand geodetic observation systems and inaccurate 3D Earthstructure models, there are always errors in the source modelsof the mainshock. Moreover, fault geometries of future aftershocksare not precisely known, and aftershocks occurring onblind faults are particularly difficult to study due to lack ofgeological information about the faults. For example, the 1994Northridge earthquake occurred on a blind (buried) fault;the study of its potential triggering by the 1971 San Fernandoearthquake was only made possible after its rupture plane andhypocenter depth were resolved (Stein et al. 1994). The 2011Christchurch earthquake was another case of such a blindearthquake, which has not yet been associated with any knowngeological faults. Thus, this event is a valuable case study of howeffective the Coulomb stress mechanism is in triggering aftershocks,and its variability due to the uncertainties in receivingfault parameters and mainshock source models.In this paper we examine the sensitivity of computed staticCoulomb stress change levels to source parameterization byconsidering various combinations of mainshock rupture modelsand aftershock fault orientations for the 2010 Canterburyand 2011 Christchurch earthquakes. General constraints onthe mainshock source models and aftershock fault geometryare provided by teleseismic and geodetic data. We also investigatethe sensitivity of the results to aftershock focal depth andapparent coefficient of friction. We conclude with a discussionof how these results can be used in combination with focalmechanism studies to help constrain aftershock rupture assessmentusing Coulomb stress change calculations.GENERAL INFORMATION ON THE M w 7.1CANTERBURY EARTHQUAKE AND THE M w 6.3CHRISTCHURCH EARTHQUAKEThe Australian and Pacific plates converge obliquely at about40 mm yr −1 at New Zealand. Partly due to along-strike variationsin the orientations of both the plate boundary and thedirection of relative motion between the plates, the defor-Seismological Research Letters Volume 82, Number 6 November/December 2011 801

mation takes on a larger strike-slip component southward(Wallace and Beavan 2006). Accordingly, the style of deformationchanges southward, from subduction of the Pacificplate and back arc rifting in the North Island to nearly purestrike-slip in the Marlborough region to oblique convergencein the central South Island (causing formation of the centralSouthern Alps) and back to subduction of the Australian plateat the Fiordland subduction zone in the southwestern SouthIsland (Wallace and Beavan 2006). The earthquake sequencewe study in this paper occurred in the central South Island.The 2010 Canterbury earthquake occurred at 4:35 a.m. localtime on 4 September (16:35 UTC, 3 September), on a previouslyunrecognized fault system, the Greendale fault (Figure1) (Quigley et al. 2010). This M w 7.1 earthquake caused widespreaddamage throughout the area, but no deaths and onlytwo injuries were reported despite the epicenter’s locationabout 40 km west of Christchurch (population ~386,000),New Zealand’s second-most populated city (Quigley et al.2010). The 2011 Christchurch earthquake occurred at 12:51p.m. on 22 February 2011 local time (21 February UTC),causing widespread damage and more than 100 fatalities. Theearthquake was centered 2 km west of the town of Lytteltonand 10 km southeast of the center of Christchurch.SOURCE MODELS OF THE 2010 CANTERBURYEARTHQUAKEThe 2010 M w 7.1 Canterbury earthquake ruptured the previouslyunrecognized Greendale fault in an east-west direction for~30 km (Figure 1). The average displacement of this predominantlyright-lateral strike-slip event is ~2.5 m, with maxima of~5 m (Van Dissen et al. 2011). The first finite fault slip modelof the main event was published by the U.S. Geological Survey(http://earthquake.usgs.gov/earthquakes/eqinthenews/2010/us2010atbj/finite_fault.php), in which teleseismic body waveswere used for the inversion assuming a single fault plane. TheGPS and InSAR data were collected later on and a static slipmodel was derived by Beavan et al. (2010). Compared with thesingle fault plane model, this static slip model is composed ofsix segments, consisting of the strike-slip Greendale fault andseveral thrust faults.To address the variability of static triggering due to mainshockrupture models, we analyze: 1) a single fault plane modelwith uniform slip; 2) a two-segment slip model from teleseismicbody wave inversion; and 3) two stochastic slip models.Despite its simplicity, a uniform slip model can provide astraightforward physical picture and can explain the main featuresof some earthquakes (e.g., Talebian et al. 2006). Also, itis a good reference for comparison with results generated fromother slip models. First we use the Global Centroid MomentTensor (GCMT) solution to define the fault geometry andthe rake angle for the uniform slip model. We choose the faultplane with strike of 87° and dip of 85°, since the strike is consistentwith the rupture trace on the free surface (Figure 1).Slip with an amplitude of 3 m and rake of 172° is uniformlydistributed on the rectangle fault plane, which is 42 km alongstrike and 12 km along dip. However, because of the complexityof this earthquake, it is hard to fit the waveforms with thissimple slip model (Figure 2). Poor waveform fits are also shownin the USGS results (http://earthquake.usgs.gov/earthquakes/eqinthenews/2010/us2010atbj/finite_fault.php).To investigate the potential for additional complexity inthe rupture geometry, we derive a two-segment finite faultslip model by inverting teleseismic body waves. We collected27 teleseismic P waves and 15 SH waves from the earthquake.Stations are selected based on data quality and azimuthal coverage(Figure 3). To derive the finite fault model, we use theapproach developed by Ji et al. (2002a, 2002b), which allowsfitting of seismic waveforms in the wavelet domain. Nowadays,similar procedures are run routinely by several agencies, suchas the USGS (http://earthquake.usgs.gov/earthquakes/eqinthenews/)and the Caltech Tectonics Observatory (http://www.tectonics.caltech.edu/slip_history/). By examining the seismicityin the first three days after the 2010 Canterbury earthquake,we can observe a linear distribution of aftershocks in thenorth-south direction crossing roughly perpendicular to themapped surface rupture trace. The epicenter of the mainshockis also located within this linear band of seismicity (Figure1). This suggests the possibility that more than one fault wasinvolved in the rupture. Thus, we added one more fault planewith strike along this seismicity trend (strike of 345° and dipof 75°) into the finite fault inversion. The epicenter is specifiedto be on this fault plane with depth of ~7 km; thus we assumethe earthquake initiated on the north-south trending fault andpropagated to or triggered the rupture on the other fault lateron. The waveform fitting of the two-fault plane model is muchbetter than that of the single fault plane model, especially forthe beginning portion of some P-wave records (Figure 3). Forexample, station PSI’s P waveform, which is not fitted in thesingle fault plane inversion, is now fitted well. The slip distributionon the first segment shows mainly thrust motion, whichis required to fit positive first motions of some P waves. Thelargest slip patch is on the second fault plane and is dominatedby strike-slip motion. Some thrust motion is also shown in thewestern part of the second fault plane. The rupture length andthe location of thrust motion in our model are consistent withfield observations and static inversion results (Beavan et al.2010; Van Dissen et al. 2011).The stochastic slip model is another approach for characterizingthe slip distribution of an earthquake, and it has beenwidely applied in ground motion simulations (Mai and Beroza2002; Liu et al. 2006; Graves and Pitarka 2010). Lavalléeand Archuleta (2003) found that the slip distribution of the1979 Imperial Valley earthquake could be well modeled witha stochastic model assuming power law of k -n , where k is thewavenumber. A stochastic approach has to be taken in thefollowing two cases. The first case is when studying historicalearthquakes, which lack seismic waveform or geodetic datafor finite fault inversion, as with the 1811/1812 New Madridearthquakes. The other case is when characterizing the rupturefor scenario earthquakes. For the Canterbury earthquake, wefollow the procedure by Graves and Pitarka (2010) to gener-802 Seismological Research Letters Volume 82, Number 6 November/December 2011

P 9.4267BLDU45P 20.1213CASY39P 2.6205MAW57P 13.6181SBA34P 3.0181SNAA64P 13.8156PMSA63P 9.3128LCO87P 40.563PPTF41P 50.955RAR32P 52.139XMAS53P 31.330AFI32P 22.029KIP70P 17.220JOHN62P 28.71TARA44−10 0 10 20 30 40 50sSH 100.339XMAS53P 22.1354WAKE62P 35.0339HNR35P 9.9339ERM89P 11.7333MAJO85P 18.7330GUMO61P 34.1320PMG40P 29.0311COEN38P 41.1309CTAO31P 27.3298MTN46P 11.3291KSM70P 8.2283PSI79P 18.5283WRKA39P 13.4280MBWA48−10 0 10 20 30 40 50sSH 144.3354WAKE62SH 40.2311COEN38SH 64.8309CTAO31SH 81.8298MTN46SH 42.2291KSM70SH 57.2283PSI79SH 113.7283WRKA39SH 132.8267BLDU45SH 42.3213CASY39SH 184.6181SBA34SH 54.6181SNAA64SH 95.0156PMSA63SH 57.6128LCO87SH 106.863PPTF41−10 0 10 20 30 40 50s▲ ▲ Figure 2. Waveform fits for the single fault plane inversion. Black lines are data in displacement and red are synthetic. Station namesare displayed to the left of the traces along with the azimuths (above) and epicentral distances in degrees (below). Peak amplitude (inmicrons) for the data is indicated above the end of each trace.ate stochastic rupture models. We generated two models, onemodel with only a single fault segment (Figure 4) and the otherwith four fault segments, which were derived by simplifyingthe model of Beavan et al. (2010) (Figure 5).FAULT PARAMETERS OF THE 2011CHRISTCHURCH EARTHQUAKETo study the effect of receiving fault geometry on Coulombstress change, we need to determine the focal mechanismand focal depth of the 2011 Christchurch earthquake. Thereare many different approaches for studying earthquakesource parameters using regional or teleseismic waveforms.Two kinds of regional waveform data are generally used: surfacewaves and body waves. Since surface waves are generallymuch stronger than body waves, full waveform inversions aremainly controlled by surface waves. Dreger and Helmberger(1993) used the long-period body waves recorded by a regionalsparse network to invert for focal mechanism. Later, Zhao andHelmberger (1994) and Zhu and Helmberger (1996) developedthe “cut and paste” (CAP) technique, which breaks broadbandwaveforms into Pnl and surface wave segments and invertsthem independently, allowing for different bandpass filtering,time shifts, and weights. The CAP technique has been successfullyapplied to determine the depth and focal mechanism inmany regions (e.g., Tan et al. 2006). However, regional data arenot always accessible immediately after earthquakes, so inversiontechniques using teleseismic waveforms become importantand are routinely used to estimate source parameters forearthquakes of M 6 and above (e.g., the Global CMT solution,the USGS body wave moment tensor solution, and the USGSWphase solution). Most of these automatic approaches involveSeismological Research Letters Volume 82, Number 6 November/December 2011 803

P 9.4267BLDU45P 20.1213CASY39P 2.6205MAW57P 13.6181SBA34P 3.0181SNAA64P 13.8156PMSA63P 9.3128LCO87P 40.563PPTF41P 50.955RAR32P 52.139XMAS53P 31.330AFI32P 22.029KIP70P 17.220JOHN62P 28.71TARA44−10 0 10 20 30 40 50t(s)SH 100.339XMAS53P 22.1354WAKE62P 35.0339HNR35P 9.9339ERM89P 11.7333MAJO85P 18.7330GUMO61P 34.1320PMG40P 29.0311COEN38P 41.1309CTAO31P 27.3298MTN46P 11.3291KSM70P 8.2283PSI79P 18.5283WRKA39P 13.4280MBWA48−10 0 10 20 30 40 50t(s)SH 144.3354WAKE62SH 40.2311COEN38SH 64.8309CTAO31SH 81.8298MTN46SH 42.2291KSM70SH 57.2283PSI79SH 113.7283WRKA39SH 132.8267BLDU45SH 42.3213CASY39SH 184.6181SBA34SH 54.6181SNAA64SH 95.0156PMSA63SH 57.6128LCO87SH 106.863PPTF41−10 0 10 20 30 40 50t(s)F1: Strike= 345 degF2: Strike= 87 degkmkmDepth km01020-10 0 104101010-20 0 20101010101001020Depth km0 100 200 300▲ ▲ Figure 3. Results of the two-segment finite fault inversion. The upper panel shows the displacement waveform fits in black for thedata and red for the synthetic. Station names are displayed to the left of the traces along with the azimuths and epicentral distancesin degrees. Peak amplitude (in microns) for data is indicated above the end of each trace. The lower panel shows the cumulative slipdistribution (slip vectors with amplitude of slip also represented by color shading) and time contours of the rupture propagation asdetermined by the inversion. The rupture times are given relative to the origin time, and the red star indicates the epicenter. The strikeof each segment is shown on the top.cm804 Seismological Research Letters Volume 82, Number 6 November/December 2011

0Slip (cm) 0 / 160 / 55846005842464808W (km)1015108664220 5 10 15 20 25 30 35 40 45 50 55 602468101012360240120085 o Azi. (km)▲▲Figure 4. A stochastic single-segment slip model of the 2010 Canterbury earthquake. The background color shows the distributionof cumulative slip in centimeters and the green arrows show the slip vectors. Time contours of the rupture propagation are shown asblack contours. The rupture times (numbers on the contour lines) are given relative to the origin time. The strike of the fault plane isshown at the bottom.0Slip (cm) 0 / 165 / 58160051018480W (km)10866881416W (km)2360240150 5 1088100 5 10 15 200 5 10180 5 101200121 o Azi. (km)87 o Azi. (km)87 o Azi. (km)40 o Azi. (km)▲ ▲ Figure 5. Stochastic multiple-segment slip model of the 2010 Canterbury earthquake. The background color shows the distributionof cumulative slip in centimeters and the green arrows show the slip vectors. Time contours of the rupture propagation are shown asblack contours. The rupture times (numbers on the contour lines) are given relative to the origin time. The strikes of the fault segmentsare shown at the bottom of each panel.long-period waves so the solutions have poor resolution ofearthquake depth. Also when the earthquake is shallow, strongtrade-off among depth, focal mechanism, and magnitude cancause large uncertainties in source parameters (Dahlen andTromp 1998). To overcome these problems, we extend the ideaof the regional CAP technique to teleseismic cases (teleCAP).In teleCAP, we cut 10–50 s period band P-wave segments inthe vertical components and SH-wave segments in the transversecomponents and fit them independently, allowing differenttime shifts and weights. We choose the relative weightsbetween P and SH waves so that they contribute almost equallyto the final misfit function (e.g., Tan et al. 2006).Synthetic seismograms are calculated with a 1D sourcesidecrustal model obtained from CRUST 2.0 (Bassin etal. 2000). Figure 6 shows the seismic stations used in theinversion. These stations are chosen based on their signal-tonoiseratio (SNR) and azimuthal coverage. We find the bestwaveform-fitting source parameters by grid-searching earthquakemagnitude, focal mechanism (strike, dip, and rake),depth, and source duration. Figure 7 shows the best waveformfitting, the corresponding focal mechanism (strike/dip/rake = 174°/46°/42° or 52°/61°/128°) and magnitude (M w6.3). Compared with the Global CMT solution (strike/dip/rake/magnitude = 167°/57°/32° or 59°/64°/143°, M w 6.1),there is ~10 degree difference for strike/dip/rake and 0.2 differencein magnitude. These differences will be discussedlater. Both P and SH wave amplitudes and waveforms at allazimuths are fit very well. Figure 8 shows the waveform misfitas functions of centroid depth and source duration; the bestfitting depth is 5 km, which is the same as reported by NewZealand local seismologists using local stations. The best fittingsource duration is 6 s, the same as in the Global CMT solu-Seismological Research Letters Volume 82, Number 6 November/December 2011 805

ERMJNUMAJOINUJHJ2CBIJMIDWJOWYHNBTPUB SSLB YULBNACB TATO YOJHKPSGUMOPATSWAKEKWAJJOHNKIPQIZDAVMANURABLXMASKKM LDMSBMKSMKUM IPMBTDF KOMPSIUGMKAPIPMGCOENMTNCCTTAO AKNRA QISWRABHNRAFIRARPPTFTAOEXMISMBWAWRKACOCOGIRLMORW KMBLBLDUMUNBBOOCASYVNDA SBAPAFQSPAMAWLCOPLCACRZFPMSAEFITRQAHOPE▲ ▲ Figure 6. Seismic stations used in the inversion of fault parameters of the 2011 Christchurch earthquake. These stations are chosenbased on their signal-to-noise ratio (SNR) and azimuthal coverage.tion. For each candidate depth and source duration, we alsoplot its best fitting focal mechanisms and magnitudes in Figure8, to show the trade-off between magnitude and other sourceparameters. Obviously, the earthquake magnitude decreasesas depth increases, as expected from the free surface effects asdiscussed by Dahlen and Tromp (1998). If depth = 12 km (asdetermined in the Global CMT), the magnitude will be aboutM w 6.15, which is close to the M w 6.1 in the GCMT catalog.Due to the trade-off between depth and focal mechanism forshallow events, the GCMT’s focal mechanism may also besomewhat biased. We conclude that teleCAP provides higherresolutionsource parameters for the 2011 Christchurch earthquake.However, it should be noted that teleCAP assumes adouble-couple point source, as do most other approaches, so itcannot distinguish between the fault plane and auxiliary faultplane. The first three days’ aftershock distribution of the 2011Christchurch earthquake shows a clear linear trend from EENto WWS (Figure 1), which prefers the fault plane 52°/61°/128°.In the following discussion we will use only this fault plane.Epicenter location is another important source parameter thatwill greatly affect the computation of Coulomb stress change.In this paper, we use the epicenter location from New ZealandGeoNet, which is based on data from a dense local seismic networkand is presumably accurate.COMPUTATION OF COULOMB STRESS FORVARIOUS MAINSHOCK SOURCE MODELS ANDRECEIVING FAULT GEOMETRIESBased on the Coulomb failure criterion (Jaeger et al. 2007,475) and the theory of elastic dislocation (Okada 1992), wecalculate the coseismic Coulomb failure stress change (Δσ f )caused by the mainshock for different mainshock slip modelsand for different receiving fault geometries. Following King etal. (1994), Δσ f is given by Δσ f = Δτ s – μ′Δσ n , where Δτ s andΔσ n are the changes in shear and normal stress, respectively,due to the mainshock, and μ′ is the apparent coefficient of friction.Here we use the rock mechanics sign convention in whichcompressive is positive.In this study, we use the lithosphere model of dislocationsources embedded in an elastic multilayered half space (Wanget al. 2003, 2006) and adopt the program PSGRN/PSCMP(Wang et al. 2006) to compute the static Coulomb stresschange produced by the mainshock. Since the influence fromthe curvature of Earth’s free surface is small for this local study(Xiong et al. 2010), the Earth surface is treated as flat in ourmodel. The parameters of our multilayered model in Table 1are based on Crust 2.0. A moderate value of apparent coefficientof friction μ′ = 0.4 is used in our calculation (King et al.806 Seismological Research Letters Volume 82, Number 6 November/December 2011

P VSHP VSHP VSHP VSHMIDWTRQAPAFMBWA72.1/9.284.7/139.765.3/224.649.1/279.6−6.00553.03−3.00−1.00524.35−5.00928989798592JOHNEFIMUNWRKA62.2/19.375.3/150.145.3/265.140.5/282.4−5.00−3.00−4.00399.08−5.00361.52837593959687AFIPMSABLDUPSI32.5/29.063.1/156.345.7/267.080.2/283.2−3.00−1.00514.09−5.00400.43−5.00601.1680659491969672KIPHOPEMORWUGM70.1/29.079.2/163.147.0/268.464.6/284.0−5.00−2.00−4.00409.05−5.00516.71895992969385XMASQSPAKMBLBTDF52.8/38.746.5/180.041.5/269.675.9/285.8−3.00450.4015.00−4.00372.16−4.00579.4989736694968990RARSBACOCOIPM32.0/54.534.4/182.271.5/270.780.0/286.0−2.0015.00334.66−5.00−4.00601.04914394939379PPTFVNDAGIRLKOM41.0/62.834.3/184.252.1/273.976.1/286.2−1.00374.7015.00332.78−4.00442.52−5.00582.838791689091949194TAOEMAWBBOOKUM53.6/64.257.2/205.430.5/278.280.8/286.3454.762.00492.56−3.00298.02−5.00603.3185416692719475LCOCASYXMISKSM87.3/128.440.0/213.866.3/278.371.4/290.8−2.00640.332.00362.18−5.00526.55−4.00557.529283458296819287PLCACRZFXMIKNRA78.5/136.276.1/217.766.3/278.446.5/292.8−2.00595.84−1.00585.42−5.00526.51402.6187827879968087P VSHP VSHP VSHP VSHSBMCTAPMGHNR70.9/293.032.2/308.340.7/319.235.8/338.1−4.00555.40−3.00309.63−4.00−5.00886388718785KAPICTAOJOWERM60.3/293.732.2/308.381.1/320.789.3/338.4−4.00309.61−4.00−4.00648.518971949194WRABCOENMANUPATS39.7/294.138.9/310.347.1/324.351.9/341.7−5.00−4.00354.53−5.00407.62−5.00438.9195948787648872MTNTPUBJNUWAKE47.0/297.882.0/312.985.4/325.862.9/353.6−5.00407.95−5.00610.08−5.00625.06−6.00502.619089958294768966KKMSSLBGUMOKWAJ71.0/298.782.2/313.462.4/329.252.4/353.6−4.00554.84−5.00611.19−6.00−6.00437.9985749879908581QISYULBRABL35.9/299.281.7/313.543.3/329.3LDM−4.0096NACB−5.0089609.6781INU−5.0085FM 174 46 42 Mw 6.3068.8/299.982.2/314.185.1/331.4−3.00−5.00612.09−5.00625.357795909167QIZYHNBCBIJ84.8/302.382.7/314.375.8/332.1−3.00628.19−5.00613.69−8.009291929091DAVTATOJHJ266.0/307.282.9/314.582.1/332.6−7.00−5.00−6.00879393HKPSYOJMAJO84.8/307.581.7/315.485.8/332.8−4.00627.87−5.00608.63−5.009087888796▲ ▲ Figure 7. Best teleseismic waveform fitting and the corresponding focal mechanism (strike/dip/rake = 174°/46°/42°) and magnitude(M w 6.3). Black is the data and red is the synthetic. The red crosses on the focal mechanism beach ball show the locations of stations.1994), and parameter sensitivity will be discussed. To show theinfluence of the uncertainty of focal depth, we calculate the Δσ fcaused by the mainshock on several horizontal planes at 2, 5,10, and 15 km depths, respectively, each of which consists of101 × 101 grid points. In the next several subsections, we willshow the Δσ f results for these different cases and discuss theirvariability. It should be noted that the effect of viscoelasticrelaxation is not taken into account here. Because of the shorttime interval between the two events, its influence is believedto be relatively small, compared with the uncertainties of otherparameters above. More accurate results can be achieved bytaking this effect into account in the future.Coulomb Stress Change Caused by Different MainshockSlip ModelsThe selection of an appropriate mainshock slip model is importantfor the Δσ f distribution. Figure 9 displays the Δσ f distributionfor the four slip models discussed previously. In all cases,the slip models have a strong influence on the Δσ f distributionalong the Greendale fault in the near field. For example, theGreendale fault lies completely within a stress shadow in Figure9A, but there are some parts significantly loaded (>1MPa) usingthe other three models in Figure 9B–D. However, these differentslip models cause no significant difference in the far field. Atthe eastern and western ends of the Greendale fault, the stresschanges more than 0.01 MPa caused by each model, and the Δσ fdistributions, are very similar. In summary, it can be inferredthat a uniform slip model can explain the main features of theΔσ f distribution in the far field; the more complicated slip modelsmake the Δσ f distribution heterogeneous in the near-fieldalong the fault. At the hypocenter of the 2011 Christchurchearthquake, the Δσ f are 0.013, 0.044, 0.033, and 0.053 MPafor the four different slip models, respectively—all above 0.01MPa, the presumed threshold value for earthquake triggering.Seismological Research Letters Volume 82, Number 6 November/December 2011 807

(A)6.146.13(B)6.246.36Misfit6.156.186.656.196.466.206.236.396.266.256.276.296.34Misfit6.34 6.306.286.310 2 4 6 8 10 12 14Depth (km)1 2 3 4 5 6 7 8 9Duration (s)▲▲Figure 8. Misfit of teleseismic waveforms as a function of earthquake focal depth and duration. The best fitting focal mechanismsand magnitudes are also plotted to show the trade-offs between them.LayerTABLE 1Multilayered lithosphere model from CRUST 2.0.Thickness(km)V p(km·s –1 )Vs(km·s –1 )Density(kg·m –3 )Upper crust 14.3 6.0 3.5 2700Middle crust 9 6.6 3.7 2900Lower crust 11 7.2 4.0 3050Mantle 8.0 4.6 3300Effect of Focal Depth on Coulomb Stress ChangeThe sensitivity of our results to focal depth of the 2011Christchurch earthquake is explored by comparing the Δσ f distributionsresolved at depths of 2, 5, 10, and 15 km. Here thefocal depth is meant to be hypocentral depth, where ruptureof the Christchurch earthquake initiated. The depth inferredfrom waveform inversion is essentially centroid depth, and thehypocentral depth is typically difficult to resolve unless witha dense local seismic network. For these calculations, we usethe two-segment slip model of the mainshock along with thefocal mechanism of the receiving fault (52°/61°/128°) derivedfrom the teleCAP inversion. The results are shown in Figure10. Some local changes of the Δσ f distribution are observedin the far field, and complex changes happen in the near field.For example, the area with positive Δσ f at the east end of theGreendale fault is getting smaller when focal depth increases;and in the area of the mainshock thrust fault segment, Δσ fchanges polarity as the depth increases. The Δσ f at the 2011Christchurch earthquake epicenter are 0.035, 0.044, 0.065, and0.094 MPa at 2, 5, 10, and 15 km depth, respectively, increasingwith the depth and all above 0.01 MPa. Similar results areobtained for the other slip models, as shown in Table 2.Effect of Receiving Fault Geometry on Coulomb StressChangeTo analyze the impact of receiving fault geometry on the Δσ f ,we make some significant changes in the strike and dip of thereceiver fault (from focal mechanism) and compare their influenceon the resulting Δσ f distributions. The two-segment slipmodel of the mainshock and a focal depth of 5 km are adoptedin this calculation, and the results are shown in Figures 11(strike sensitivity) and 12 (dip sensitivity). In Figure 11, significantchanges in the Δσ f distribution can be observed as thestrike varies. When the strike of the receiving fault is rotatedcounterclockwise by 30 degrees, the area with positive Δσ fincreases in the near- and far-fields, compared with Figure 10.The opposite situation occurs when the strike of the receivingfault is rotated clockwise from the teleCAP solution. Similarchanges are also obtained when the dip angle of the receivingfault is changed by ± 20 degrees, as shown in Figure 12.The area with positive Δσ f outside the northwest corner ofthe Greendale fault grows with increasing dip angle. In thenear-field, the situation is a little more complex, but the wholeregion with positive Δσ f gets larger. From these two figures, weconclude that the Δσ f distribution can be quite sensitive to theassumed geometry of the receiving fault.Sensitivity of Coulomb Stress Change to Coefficient ofFrictionThe selection of an appropriate value for the apparent coefficientof friction μ′ is important because it controls the con-808 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A)(B)(C)(D)▲▲Figure 9. The coseismic Δσ f caused by the M w 7.0 earthquake with different slip models: A) a single fault plane model with uniformslip; B) a two-segment slip model; C) a stochastic model with single plane based on the National Earthquake Information Center (NEIC)solution; D) a stochastic model with four segments based on Beavan et al. (2010). The strike angle, dip angle, and rake angle of thereceiver fault are 52°, 61°, and 128°, respectively. The calculated depth is 5 km, and the apparent coefficient of friction μ′ is 0.4. The twofocal mechanisms show the location and mechanism of the mainshock and M w 6.3 aftershock.tribution of the normal stress change to the Δσ f (Xiong et al.2000; King et al. 1994). In general, μ′ is set to be differentvalues for different types of faults. Xiong et al. (2010) set μ′to a high value (0.8) for thrust faults, a moderate value (~0.6)for normal faults, and a lower value (0.2~0.4) for strike-slipfaults. Since the 2010 Canterbury earthquake mainshock isprimarily a strike-slip event (with some thrust component), wehave set μ′ at 0.4 in the previous calculations. Here we considervalues of μ′ of 0.0 and 0.8 to analyze their sensitivity on thecomputed Δσ f distribution. In these calculations, we again usethe two-segment slip model for the mainshock along with theteleCAP mechanism (52°/61°/128°) and focal depth (5 km)for the aftershock. The resulting Δσ f at the epicenter of the2011 Christchurch earthquake is 0.095, 0.044, and –0.008MPa for μ′ = 0.0, 0.4, and 0.8, respectively, as shown in Figure13. The decrease of Δσ f with increasing μ′ indicates that thechange of normal stress is positive (clamping the fault plane)Christchurch fault plane. Other areas, such as northwest of theGreendale fault, exhibit an increase in Δσ f with increasing μ′.Obviously, the polarity change of Δσ f can lead to significantuncertainty for evaluating seismic hazard, underscoring theneed for accurate constraints on mainshock faulting mechanismand estimation of μ′.DISCUSSION AND CONCLUSIONCoulomb stress triggering is physically straightforward and hasbeen widely applied in studying the distribution and probabilityof aftershocks. However, there can be substantial variabilitydue to uncertainty in mainshock slip models and fault orientationof the subsequent aftershocks. In this paper, we calculatethe coseismic static Coulomb stress change caused by the 2010Canterbury earthquake for several different mainshock slipmodels, and various permutations of receiving fault geometrySeismological Research Letters Volume 82, Number 6 November/December 2011 809

(A)(B)(C)(D)▲▲Figure 10. The coseismic Δσ f caused by the mainshock with the two-segment slip model at different depths, with μ′ = 0.4. A), B),C), and D) show results for depths of 2 km, 5 km, 10 km, and 15 km, respectively. The strike angle, dip angle, and rake angle of receiverfault are 52°, 61°, and 128°, respectively. The beach balls show the location and mechanism of the mainshock and M w 6.3 aftershock.TABLE 2Δσ f at the 2011 Christchurch earthquake epicenter, caused by different mainshock models with three µ ′ values at four focaldepths. The focal mechanism of the receiving fault is 52°/61°/128°.µ′2 km 5 km 10 km 15 kmUniform fault plane model 0.0 0.062 0.062 0.066 0.0810.4 0.010 0.013 0.021 0.0360.8 –0.042 –0.036 –0.023 –0.008Two-segment slip model 0.0 0.087 0.095 0.113 0.1430.4 0.035 0.044 0.065 0.0940.8 –0.018 –0.008 0.017 0.044Stochastic model with single 0.0 0.110 0.143 0.183 0.214plane based on NEIC solution 0.4 0.004 0.033 0.086 0.1310.8 –0.103 –0.077 –0.010 0.048Stochastic model with foursegmentsDepth0.0 0.120 0.119 0.114 0.1210.4 0.048 0.053 0.058 0.0680.8 –0.023 –0.013 0.002 0.015810 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A)(B)(C)(D)▲▲Figure 11. The coseismic Δσ f caused by the two-segment slip model at a focal depth of 5 km with μ′=0.4 for variations in receiverfault strike. A), B), C), and D) show results for strikes of 352°, 22°, 82°, and 112°, respectively. The dip and rake of the receiver fault areheld constant at 61° and 128°, respectively, in these calculations. The beach balls show the location and mechanism of the mainshockand M w 6.3 aftershock.at different focal depths with three values of apparent coefficientof friction. We find that different slip models can result insignificant differences in the amplitude and distribution of Δσ fin the near field of the fault, but no substantial difference inthe far field. On the other hand, focal depth and receiving faultgeometry play a much stronger role on Δσ f outside the immediatemainshock rupture zone. In our calculations for the 2011Christchurch earthquake, Δσ f can increase significantly (by afactor of 3) when the aftershock focal depth increases from 2 kmto 15 km. Additionally, our results show a change of 30 degreesin receiving fault geometry can even cause polarity changes inΔσ f . We also find the resulting Δσ f at the epicenter of the 2011Christchurch earthquake decreases significantly as the value ofapparent coefficient of friction (μ′) increases. This emphasizesthe need for careful consideration of the appropriate value ofμ′ for different faulting environments. It should be noted thatin this study we assume that the coseismic slip distribution ofthe Canterbury mainshock is responsible for the triggering ofthe Christchurch earthquake. Because the GPS measurementsafter the Canterbury earthquake show very little postseismicmotion (less than ~2% of coseismic) (Reyners 2011, this issue),postseismic deformation probably can be neglected. Howeverwe still cannot rule out the possibility that smaller aftershockstriggered the Christchurch earthquake as a secondary aftershockwith larger magnitude (e.g., Felzer et al. 2002).In general, we find the occurrence of the Canterburyearthquake with a reasonable set of parameter choices raisesthe Δσ f on the Christchurch fault plane beyond the 0.01MPa threshold, promoting the aftershock plane to break. Toimprove the accuracy of Δσ f analysis, and hence the probabilityassessment of aftershocks, it is helpful to carefully studysource parameters of historical earthquakes for each region tounderstand the potential receiving fault geometry. For M > 5.5earthquakes, the teleCAP technique used in this paper showspromise for obtaining accurate focal mechanism and depth.For M ~ 5 earthquakes not recorded with local broadbandseismic stations, teleseismic P waves are typically above noiselevel in the short-period band (~1Hz), and teleCAP can beSeismological Research Letters Volume 82, Number 6 November/December 2011 811

(A)(B)(C)(D)▲▲Figure 12. The coseismic Δσ f caused by the two-segment slip model at a focal depth of 5 km with μ′ = 0.4 for variations in receiverfault dip. A), B), C), and D) show results for dips of 21°, 41°, 61°, and 81°, respectively. The strike and rake of the receiver fault are heldconstant at 52° and 128°, respectively, in these calculations. The beach balls show the location and mechanism of the mainshock andM w 6.3 aftershock.also applied, although the variability of P-wave amplitudehas to be taken into account (Ni et al. 2010; Chu et al. 2011).For M ~ 5 earthquakes well recorded with local stations, thetraditional CAP technique can be applied to estimate sourceparameters. For even smaller earthquakes (M 2–4), Tan andHelmberger (2007) have proposed a new amplitude correctiontechnique to invert short-period (0.5–2 Hz) P waveforms forsource parameters, and achieved success in the 2003 Big Bearsequence. Since focal mechanism itself cannot distinguishbetween the fault plane and auxiliary fault plane, additionalinformation is needed, for example from aftershock distributionor earthquake rupture directivity (e.g., Luo et al. 2010).Paleoseismology and geology can also provide important informationon potential fault orientations, particularly for regionswithout active seismicity.ACKNOWLEDGMENTSConstructive reviews provided by Jeanne Hardebeck, MorganPage, and an anonymous reviewer were very helpful in revisingthe paper and making it acceptable for publication. Thiswork is supported by China Earthquake Administrationfund 200808078, and NSFC fund 40821160549, 41074032.REFERENCESAllen, R. M., and H. Kanamori (2003). The potential of earthquake earlywarning in southern California. Science 300 (5,620), 786–789.Bakun, W. H., F. G. Fischer, E.g. Jensen, and J. VanSchaack (1994). Earlywarning system for aftershocks. Bulletin of the Seismological Societyof America 84 (2), 359–365.Bassin, C., G. Laske, and G. Masters (2000). The current limits ofresolution for surface wave tomography in North America. Eos,Transactions, American Geophysical Union 81, F897.Beavan, J., S. Samsonov, M. Motagh, L. Wallace, S. Ellis, and N. Palmer(2010). The Darfield (Canterbury) earthquake: Geodetic observationsand preliminary source model. Bulletin of the New ZealandSociety for Earthquake Engineering 43 (4), 228–235.Chu, R., S. Ni, A. Pitarka, and D. V. Helmberger (2011). Inversion ofsource parameters for moderate earthquakes using short-period teleseismicP waves. Submitted to Geophysical Journal International.812 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A)(B)(C)▲▲Figure 13. The coseismic Δσ f caused by the two-segment slip model for different values of apparent coefficient of friction μ′ at adepth of 5 km with a receiver fault mechanism of 52°/61°/128°. A), B), and C) show results for μ′ of 0.0, 0.4, and 0.8, respectively.Dahlen, F. A., and J. Tromp (1998). Theoretical Global Seismology.Princeton, NJ: Princeton University Press.Dreger, D. S., and D. V. Helmberger (1993), Determination of SourceParameters at Regional Distances with Single Station or SparseNetwork Data, Journal of Geophysical Research 98, 8,107–8,125.Felzer, K. R., T. W. Becker, R. E. Abercrombie, G. Ekström, and J. R. Rice(2002), Triggering of the 1999 MW 7.1 Hector Mine earthquakeby aftershocks of the 1992 MW 7.3 Landers earthquake, Journalof Geophysical Research 107 B92190; doi:10.1029/2001JB000911.Felzer, K. R., and E. E. Brodsky (2006). Decay of aftershock densitywith distance indicates triggering by dynamic stress. Nature 441,735–738.Freed, A. M. (2004). Earthquake triggering by static, dynamic, andpostseismic stress transfer. Annual Review of Earth and PlanetarySciences 33, 335–367.Freed, A. M., and J. Lin (2001). Delayed triggering of the 1999 HectorMine earthquake by viscoelastic stress transfer. Nature 411, 180–183.Graves, R., and A. Pitarka (2010). Broadband ground motion simulationusing a hybrid approach. Bulletin of the Seismological Society ofAmerica 100, 2,095–2,123; doi: 10.1785/0120100057.Harris, R. A. (1998). Introduction to special section: Stress triggers,stress shadows, and implications for seismic hazard. Journal ofGeophysical Research 103, 24,347–24,358.Harris, R. A. (2000). Earthquake stress triggers, stress shadows, and seismichazard. Current Science 79 (9), 10.Jaeger, J. C., N. G. W. Cook, and R. W. Zimmerman (2007).Fundamentals of Rock Mechanics, fourth edition. Wiley-Blackwell.Ji, C., D. J. Wald, and D. V. Helmberger (2002a). Source description ofthe 1999 Hector Mine, California, earthquake, part I: Waveletdomain inversion theory and resolution analysis. Bulletin of theSeismological Society of America 92 (4), 1,192–1,207.Ji, C., D. J. Wald, and D. V. Helmberger (2002b). Source description ofthe 1999 Hector Mine, California, earthquake, part II: Complexityof slip history. Bulletin of the Seismological Society of America 92 (4),1,208–1,226.King, G. C. P., R. S. Stein, and J. Lin (1994). Static stress changes andthe triggering of earthquakes. Bulletin of the Seismological Society ofAmerica 84, 935–953.Lavallée, D., and R. J. Archuleta (2003). Stochastic modeling of slipspatial complexities for the 1979 Imperial Valley, California,earthquake. Geophysical Research Letters 30 (5), 1,245;doi:10.1029/2002GL015839.Liu, P., R. J. Archuleta, and S. H. Hartzell (2006). Prediction of broadbandground-motion time histories: Hybrid low/high frequencymethod with correlated random source parameters. Bulletin of theSeismological Society of America 96, 2,118–2,130.Seismological Research Letters Volume 82, Number 6 November/December 2011 813

Luo, Y., Y. Tan, S. Wei, D. Helmberger, Z. Zhan, S. Ni, E. Hauksson, andY. Chen (2010). Source mechanisms and rupture directivity of theMay 18, 2009 Mw 4.6 Inglewood, California earthquake. Bulletinof the Seismological Society of America 100 (6), 3,269–3,277.Ma, K.-F., C.-H. Chan, and R. S. Stein (2005). Response of seismicityto Coulomb stress triggers and shadows of the 1999 Mw = 7.6Chi-Chi, Taiwan, earthquake. Journal of Geophysical Research 110,B05S19.Mai, P. M., and G. C. Beroza (2002). A spatial random field model tocharacterize complexity in earthquake slip. Journal of GeophysicalResearch 107 (B11); doi:10.1029/2001JB000588.Nalbant, S. S., A. Hubert, and G. C. P. King (1998). Stress couplingbetween earthquakes in northwest Turkey and the north AegeanSea. Journal of Geophysical Research 103, 24,469–24,486.Ni, S., A. Pitarka, and D. V. Helmberger (2010). Rapid source estimationfrom global calibrated paths. Seismological Research Letters 81 (3),498–504; doi: 10.1785/gssrl.82.6.498.Okada, Y. (1992). Internal deformation due to shear and tensile faultsin a half-space. Bulletin of the Seismological Society of America 82,1,018–1,040.Parsons, T., R. S. Stein, R. W. Simpson, and P. A. Reasenberg (1999).Stress sensitivity of fault seismicity: A comparison between limitedoffsetoblique and major strike slip faults. Journal of GeophysicalResearch 104, 20,183–20,202.Parsons, T., S. Toda, R. S. Stein, A. Barka, and J. H. Dieterich (2000).Heightened odds of large earthquakes near Istanbul: An interaction-basedprobability calculation. Science 288, 661.Quigley, M., P. Villamor, K. Furlong, J. Beavan, R. Van Dissen, N.Litchfield, T. Stahl, B. Duffy, E. Bilderback, D. Noble, D. Barrell, R.Jongens, and S. Cox (2010). Previously unknown fault shakes NewZealand’s South Island. Eos, Transactions, American GeophysicalUnion 91, 469–472; doi:10.1029/2010EO490001.Reilinger, R. E., S. Ergintav, R. Bürgmann, S. McClusky, O. Lenk, A.Barka, O. Gurkan et al. (2000). Coseismic and postseismic faultslip for the 17 August 1999, M = 7.5, Izmit, Turkey earthquake.Science 289, 1,519–1,524.Reyners, M. (2011). Lessons from the destructive Mw 6.3 Christchurch,New Zealand, earthquake. Seismological Research Letters 82(3),371–372.Scholz, C. H. (1990). The Mechanics of Earthquakes and Faulting. NewYork: Cambridge University Press, 439 pp.Shan, B., X. Xiong, Y. Zheng, and F. Diao (2009). Stress changes onmajor faults caused by Mw7.9 Wenchuan earthquake, May 12,2008. Science in China, Series D—Earth Sciences 52 (5), 593–601;doi:10.1007/s11430-009-0060-9.Stein, R. S. (1999). The role of stress transfer in earthquake occurrence.Nature 402 (6,762), 605–609.Stein, R. S., A. A. Barka, and J. H. Dieterich (1997). Progressive failureon the North Anatolian fault since 1939 by earthquake stress triggering.Geophysical Journal International 128, 594–604.Stein, R. S., G. C. P. King, and J. Lin (1994). Stress triggering of the 1994M = 6.7 Northridge, California, earthquake by its predecessors,Science 265, 1,432–1,435.Talebian, M., J. Biggs, M. Bolourchi, A. Copley, A. Ghassemi, M.Ghorashi, J. Hollingsworth, J. Jackson, E. Nissen, B. Oveisi, B.Parsons, K. Priestley, and A. Saiidi (2006). The Dahuiyeh (Zarand)earthquake of 2005 February 22 in central Iran: Reactivation of anintramountain reverse fault. Geophysical Journal International 164(1), 137–148.Tan, Y., L. Zhu, D. V. Helmberger, and C. K. Saikia (2006). Locatingand modeling regional earthquakes with two stations. Journal ofGeophysical Research 111, B01306; doi:10.1029/2005JB003775.Tan, Ying, and Don Helmberger (2007). A New Method for DeterminingSmall Earthquake Source Parameters Using Short-Period P Waves.Bulletin of the Seismological Society of America 97 (4): 1,176–1,195;doi:10.1785/0120060251.Toda, S., J. Lin, M. Meghraoui, and R. S. Stein (2008). 12 May 2008M = 7.9 Wenchuan, China, earthquake calculated to increasefailure stress and seismicity rate on three major fault systems.Geophysical Research Letters 35, L17305.Toda, S., R. Stein, P. Reasenberg, J. Dieterich, and A. Yoshida (1998).Stress transferred by the Mw = 6.5 Kobe, Japan, shock: Effecton aftershocks and future earthquake probabilities. Journal ofGeophysical Research 103, 24,543–24,565.Van Dissen, R., D. Barrell, N. Litchfield, P. Villamor, M. Quigley, A.King, K. Furlong, et al. (2011). Surface rupture displacement onthe Greendale fault during the Mw 7.1 Darfield (Canterbury)earthquake, New Zealand, and its impact on man-made structures.Proceedings of the Ninth Pacific Conference on EarthquakeEngineering, 14–16 April 2011, 186–193.Wallace, L. M., and J. Beavan (2006). A large slow slip event on thecentral Hikurangi subduction interface beneath the Manawaturegion, North Island, New Zealand. Geophysical Research Letters33, L11301; doi:10.1029/2006GL026009.Wan, K., S. Ni, X. Zeng, and S. Paul (2009). Real-time seismology for the05/12/2008 Wenchuan earthquake of China: A retrospective view.Science in China, Series D—Earth Sciences 52 (2), 155–165.Wang, R., F. Lorenzo Martín, and F. Roth (2003). Computation ofdeformation induced by earthquakes in a multi-layered elasticcrust—FORTRAN programs EDGRN/EDCMP. Computers &Geoscience 29 (2), 195–207.Wang, R., F. Lorenzo-Martín, and F. Roth (2006). PSGRN/PSCMP—anew code for calculating co- and post-seismic deformation, geoidand gravity changes based on the viscoelastic-gravitational dislocationtheory. Computers & Geoscience 32, 527–541.Wang, W., S. Ni, Y. Chen, and H. Kanamori (2009). Magnitude estimationfor early warning applications using the initial part of Pwaves: A case study on the 2008 Wenchuan sequence. GeophysicalResearch Letters 36, L16305; doi:10.1029/ 2009GL038678.Xiong, X, B. Shan, Y. Zheng, and R. Wang (2010). Stress transfer andits implication for earthquake hazard on the Kunlun fault, Tibet.Tectonophysics 482, 216–225.Zhao, L. S., and D. V. Helmberger (1994). Source estimation from broadbandregional seismograms. Bulletin of the Seismological Society ofAmerica 84, 91–104.Zhang, G., L. Zhu, X. Song, Z. Li, M. Yang, N. Su, and X. Chen (1999).Predictions of the 1997 strong earthquakes in Jiashi, Xinjiang,China. Bulletin of the Seismological Society of America 89 (5),1,171–1,183.Zhu, L., and D. V. Helmberger (1996). Advancements in source estimationtechniques using broadband regional seismograms. Bulletin ofthe Seismological Society of America 86, 1,634–1,641.Seismological LaboratoryCalifornia Institute of TechnologyPasadena, California 91125, U.S.A.zwzhan@gmail.com(Z. Z.)814 Seismological Research Letters Volume 82, Number 6 November/December 2011

Insar and Optical Constraints on Fault Slipduring the 2010–2011 New Zealand EarthquakeSequenceWilliam D. Barnhart, Michael J. Willis, Rowena B. Lohman, and Andrew K. MelkonianWilliam D. Barnhart, Michael J. Willis, Rowena B. Lohman, andAndrew K. MelkonianCornell UniversityEOnline material: Images of InSAR data, model residuals; tablesincluding offsets and resampled InSAR dataINTRODUCTIONOur study used space-based interferometric synthetic apertureradar (InSAR) and feature tracking on sub-meter-resolutionoptical imagery pairs to characterize surface deformationresulting from the 4 September 2010 Mw 7.1 Darfield,22 February 2011 Mw 6.3 Christchurch, and 13 June 2011Christchurch earthquakes (dates in local time), each of whichoccurred in the Canterbury region of the South Island ofNew Zealand. A rapid, coordinated international emergencyresponse is often required when strong-motion earthquakes hiturban areas. Unfortunately in these cases relief workers oftenhave little information about the location or the extent of damage.Remote sensing can rapidly provide maps of certain keyvariables (i.e., building damage, potential loading of nearbyfaults, etc.) to relief workers on the ground. These maps cancover broad areas on time scales that are only limited by therevisit time of the satellite or aircraft. Critically, imagery typessuch as satellite-based synthetic aperture radar (SAR) have longrepeat times of up to 46 days at present, although the existenceof overlapping tracks and multiple satellite platforms effectivelyreduces the repeat time somewhat. Here we demonstratethe impact of commercial optical imagery that can be acquiredwithin hours to days after an earthquake, with the goal of supportingrelief efforts in future earthquakes on a more rapidtimescale than can be achieved with SAR imagery alone. Wedemonstrate that these sub-meter-resolution scenes are feasibletools for deriving near-fault surface displacements for use infault slip inversions, even in areas of heavy agricultural activity.The Darfield and Christchurch earthquakes present anopportunity to observe postseismic deformation related to multiplemoderate (

−44˚00' −43˚00'4-Sep-2010Mw 7.1Track 323Path 336NTrack 5113-June-201122-Feb-2011 Mw 6.0Mw 6.340 kmPath 335171˚00' 172˚00' 173˚00'Track 19536 mm/yr▲ ▲ Figure 1. Study region and spatial coverage of data. Focal mechanismsare Global Centroid Moment Tensor (GCMT; Dziewonskiet al. 1981) solutions for the 4 September 2010, 22 February 2011,and 13 June 2011 Canterbury earthquakes. Faults (thin lines) andseismicity (black dots) are from GNS Geonet (http://geonet.org.nz) with earthquake spanning the period 3 September 2010 to1 August 2011. Boxes indicate extent of InSAR data (black) andoptical imagery (white). Image overlaid on SRTM digital elevationmodel (Farr and Kobrick 2000). Inset shows map location withmajor tectonic features and Nuvel-1A plate motion of Australiarelative to fixed Pacific (DeMets et al. 1994).2). Another significant event (Mw 6.0) occurred 13 June 2011(hereafter 13-June) near the Christchurch earthquake epicenter,causing further damage in the city of Christchurch. TheDarfield earthquake exhibited a large stress drop of ~160 barswhile the Christchurch events exhibited more moderate stressdrops of 50–60 bars (Fry et al. 2011, page 846 of this issue).The city of Christchurch (Figure 1) is located on theeastern Canterbury Plains, an alluvial plain of Cretaceousthrough present sediments overlying the Late Paleozoic to MidCretaceous Torlesse terrane (e.g., Mackinnon 1983). The BanksPeninsula, an extinct Miocene volcanic structure, punctuatesthe eastern edge of the Canterbury Plains near Christchurch(Timm et al. 2009). Several other Cenozoic volcanic structuresexist throughout the South Island, including near the city ofDunedin. Paleoseismic and GPS studies suggest that up to 80%of the 38 mm/yr relative Australian-Pacific plate motion occurringwithin the central South Island of New Zealand (DeMetset al. 1994; Beavan et al. 1999; Wallace et al. 2007) is accommodatedby the Alpine fault (Berryman et al. 1992; Norris andCooper 2001) while the Porter’s Pass/Amberly fault system,north of our study area, accommodates ~10–15% (3–8 mm/yr, Beavan et al. 1999; Wallace et al. 2007). The rates and rateuncertainties in the central South Island allow for up to 10 mm/yr of unaccounted strike-slip motion, which has been attributedto model errors or uplift in the foothills of the Southern Alpsand strike-slip motion in the Canterbury Plains (e.g., Beavan etal. 1999; Sutherland et al. 2006; Wallace et al. 2007). Severallarge (>M 7.1) earthquakes are associated with the Porter’sPass fault zone in the Southern Alps foothills (Howard et al.2005) while other large Quaternary events are documentedto the north in the Marlborough fault system (Cowan 1991).Documentation of active faults in the Marlborough fault systemand Canterbury Plains reveals dominantly right-lateral andreverse-slip motion on shallow to steeply (>50 degrees) dippingplanes (see Pettinga et al. 2001 and associated references).Prior to the 2010–2011 earthquake sequence, the strongesthistorical ground motion in Christchurch was attributed to anM 7–8 event (Stirling et al. 1999), and the Canterbury Plains inthis focus area were characterized by low to moderate rates ofseismicity (e.g., Pettinga et al. 2001). Reflection-seismic surveysin the vicinity of the Darfield event revealed offsets and foldingof Quaternary sediments older than 24 ka by thrust faults (Dornet al. 2010), leading those authors to suggest that infrequentevents >M 7 with long recurrence intervals could be possible.DATA: AVAILABILITY AND PROCESSING RESULTSCharacteristics of the radar and optical data that we used inthis work are summarized in Tables 1 and 2, respectively.Multiple pairs of SAR imagery with at least two different lookangles span each earthquake (spatial coverage shown in Figure1), as well as the period in between them, allowing some redundancyin the data and the assessment of whether individualfeatures in the data are associated with the earthquake or withnoise. Because of the limited number of acquisitions, we onlyuse ascending tracks, which restricts our ability to constrainthe three-dimensional deformation field for each event. Forthe Darfield earthquake, we successfully obtained SAR pixeloffsets, which constrain displacements in the horizontal, alongtrackdirection and provide an additional component of thethree-dimensional deformation field (e.g., Fialko et al. 2001).We processed interferograms using the Caltech/JPL (JetPropulsion Laboratory) InSAR processing package ROI_PAC(Rosen et al. 2004), using a digital elevation model from theSatellite Radar Topography Mission (SRTM, Farr and Kobrick2000). PALSAR imagery from the ALOS satellite was providedby Japanese Aerospace Exploration Agency (JAXA) throughan agreement with NASA and the Alaska Satellite Facility(ASF). ENVISAT imagery was acquired through a Category-1agreement with the European Space Agency (ESA).All interferograms of the Darfield coseismic and postseismicperiod we analyze are found in Figures S1 and S3 of thismanuscript. The strong shaking, liquefaction, and high straingradient resulted in interferograms that require some manualphase unwrapping to connect coherent zones separated byregions of decorrelation. In these cases, we ensured that thephase unwrapping was consistent across spatially overlappinginterferograms by inspection and comparison to the predicteddisplacement field resulting from our inversion.816 Seismological Research Letters Volume 82, Number 6 November/December 2011

−43˚36' −43˚30'Greendale172˚06'5 kmChristchurchFigure 3bcAN100 mBNN91 mD5.8m200 mC20 m D▲ ▲ Figure 2. Examples of surface ruptures from the Darfield earthquake visible in postseismic WorldView 1 optical imagery. A) Overviewmap with surface rupture (thick black line, Quigley et al. 2010) and Global CMT solution for the Darfield event (Dziewonski et al. 1981).Roads (gray lines) and railroads (dashed lines) from http://www.diva-gis.org/. B) Example of surface rupture (arrows added by authors)and interpretive field text courtesy of local farmer (exists in field, not added by authors). C) En-echelon rupture jump of ~90 m (arrow),hedgerow offset by rupture (circle). D) Zoom view of hedge and canal offsets of ~5.8 m. Optical imagery copyright 2011 Digital Globe,provided through the NGA Commercial Imagery Program.The large number of pixels (several million) in the finalInSAR data products would be prohibitively computationallyexpensive to ingest into any inversion scheme. Therefore, we subsamplethe data using the procedure outlined in Lohman andSimons (2005) so that we retain a set of spatial averages with138 to 376 points for each interferogram (Figures S1 and S2).Because we were not able to unambiguously unwrap coherentphases across the Darfield rupture for Envisat Track 51, we treatthe regions north and south of the Darfield rupture as two separatedata sets. Peak line of sight (LOS) offsets during the Darfieldearthquake were around two meters, with horizontal pixel trackingresults of up to five meters. There were at least three distinctstrike-slip fault planes and two zones of thrust faulting activatedduring the Darfield earthquake (e.g., Quigley et al. 2010;Beavan et al. 2010; Gledhill et al. 2011). This rupture complexityis apparent in the complicated, multi-lobed deformation fieldimaged with InSAR and aftershock locations (Figure 1).The 22-Feb and 13-June Christchurch earthquakesexhibit a much simpler appearance in the InSAR observations(Figure S2), although there is a large region of decorrelationwithin the city of Christchurch, and some of the deformationoccurred offshore where it cannot be imaged with InSAR. Thesteep gradients of deformation suggest a shallow, near-surfaceslip source, as supported by our inversion described below.Peak observed LOS deformation associated with the 22-FebChristchurch earthquake is 0.52 meters. We were unable toobtain subpixel, horizontal offsets from SAR imagery for the22-Feb event, which suggests either there was no surface rupture(as confirmed by field observations) or any surface offsetswere below the noise level of subpixel offset tracking (typicallyon the order of a meter).High-resolution (~0.5 m resolution) optical imagery (Figure2) from commercial satellites was made available to scientists viathe U.S. National Geospatial Agency and the National ScienceFoundation (Table 2). We also explored the use of data fromthe Advanced Spaceborne Thermal Emission and ReflectionRadiometer (ASTER), which has lower spatial resolution but isavailable on a more consistent, global basis, particularly for preseismicimagery that may not be acquired as part of the backgroundmission for commercial satellites. Previous work usingcross-correlation of optical imagery has been primarily limitedto ASTER and SPOT imagery (e.g., Avouac et al. 2006; CrippenSeismological Research Letters Volume 82, Number 6 November/December 2011 817

TABLE 1Pairs of SAR imagery used in this study for both traditional InSAR and horizontal offsets obtained through pixel tracking(asterisks). B ⊥ is the perpendicular baseline for each pair, in meters. Dates are in GMT.Satellite Date1 Date2 Track/Path Frame/Scene B ⊥ (m)Christchurch EQ: 22 February 2011ALOS 2011Jan10 2011Feb25 335 6300 421ALOS 2010Oct27 2011Mar14 336 6290/6300 1178Darfield EQALOS 2010Mar11 2010Sep11 336 6300 1215*ALOS 2010Jan24 2010Oct27 336 6300 1893ENVI 2010Sep01 2010Oct06 323 6309 236ENVI 2010Jul09 2010Sep17 51 6309 532*Post Darfield EQALOS 2010Sep11 2010Oct27 336 6300 231ALOS 2010Sep11 2011Mar14 336 6300 1407ALOS 2010Oct27 2011Mar14 336 6300 1173Christchurch EQ: 13 June 2011ENVI 2011June08 2011July08 195 6291 14* indicates date pairs where both traditional interferometry and pixel tracking offsets were generatedAll scenes are ascending tracksTABLE 2Optical data used in this study spanning each event. Dates are in GMT. Optical imagery copyright 2010 Digital Globe,provided by the NGA Commercial Imagery Program.Satellite Date Spatial Resolution (m) Band Pre/Post-seismicChristchurch EQWorldView1 2010Sep21 0.5 Panchromatic PreWorldView1 2011Feb26 0.5 Panchromatic PostDarfield EQGeoEye 2009Oct23 0.5 Panchromatic PreWorldView2 2010Sep21 0.5 Panchromatic PostASTER 2006Feb11 15 PreASTER 2010Sep18 15 Post1992; Michel and Avouac 2002; Leprince et al. 2007; Kääb andDebella-Gilo 2010), with spatial resolutions of 15 and 2.5–10 m,respectively. For comparison, the GeoEye imagery used here hasa pixel size of 0.5 m. We performed normalized cross-correlationof imagery (e.g., Melkonian 2011) processed using the ampcorprogram contained within the ROI_PAC software package(Rosen et al. 2004). Results for the higher resolution commercialdata are described below, but we were unable to clearly resolvesubpixel offsets for either of the earthquakes based on ASTERimagery due to striping within the data.The GeoEye-1 satellite acquired pre-event high-resolutionimagery on 23 October 2009. The panchromatic15 km × 15 km scene is down-sampled from 41-cm resolutionto 50-cm resolution for civilian use. The satellite, launched inSeptember 2008, has precise pointing capabilities providingscenes that are geolocated with a circular error of probability(CEP) of about six meters without the use of ground controlpoints. We extract the radiometrically corrected JPEG2000imagery from its National Imagery Transmission Format(NTF) wrapper using the Geographic Data AbstractionLibrary (version 1.8, http://www.gdal.org/). The resulting 8.5Gb 16-bit unsigned integer geotiff is geocoded, reprojectedto Universal Transverse Mercator (UTM) coordinates andregistered and orthorectified to a 90-m SRTM digital elevationmodel (Farr and Kobrick 2000). The post-event imagerycomes from the Worldview-1 satellite. This satellite, launchedin September 2007, has a revisit time of 1.9 days and beganimaging the Canterbury region almost immediately after the818 Seismological Research Letters Volume 82, Number 6 November/December 2011

earthquake. Unfortunately clouds hampered acquisition until21 September 2010, 17 days after the earthquake. We extractedthe 17.9-km-swath-wide, half-meter panchromatic imageryusing identical procedures as with the GeoEye-1 imagery.Difficulties arise using this high-resolution imagery dueto agricultural changes in the intervening year and differentsun elevations and azimuths that result in a variable degree ofshadowing from houses and hedgerows. Much of the imagerydecorrelates over this time interval, in part because there havebeen dramatic changes in land use that are visible in the form ofradical differences in relative brightness between fields and differentplowing patterns between the two images. However, thehedgerows themselves, which are visually distorted across thefault in the postseismic images (Figures 2B–D), act as coherentfeatures that provide very strong offsets from image to image.Since the hedgerows are effectively linear and have a similarbrightness along their length, the offsets are better-resolved ina direction perpendicular to each hedgerow than they are alongtheir length. Therefore, we obtain good characterization of theE-W deflection of N-S-trending hedgerows across the fault, butpoor results for E-W motion of hedgerows and roads that trendin a near E-W direction. Since the horizontal displacements inthe E-W direction are much larger than those in the N-S directionfor this earthquake, the most useful features in the imagerypixel tracking have been the N-S-trending roads and hedgerows.Figure 3 summarizes the results of optical imagery pixeltracking for the Darfield earthquake. Colored dots (Figure 3A)indicate the magnitude of displacement in an E-W directionof a 10 × 10 pixel box that was allowed to move for 32 pixelsin any direction, posted at 5-pixel spacing. Peak displacementsacross the fault (Figure 3B) agree with what one would pick fromthe trend of the hedgerow using the postseismic imagery alone(Figure 2D). Although offsets in this example are only recoverablefrom anthropogenic features, processing images withshorter temporal baselines (days to months) produces coherentoffsets in vegetated regions, validating that this technique canbe used in remote regions if appropriate acquisitions are available.Unfortunately, the only imagery available with these shorttemporal baselines is located away from the Darfield fault trace.MODELING RESULTSSource ModelingDarfield EarthquakeFor the Darfield and Christchurch earthquakes, we invert thegeodetic observations for spatially distributed fault slip usingplanar fault geometries that we infer using a combination ofnonlinear inversion and independent data such as surface ruptures,aftershocks, etc. For the Darfield earthquake, we use foursteeply south-dipping planes (Table S1) to model the primarilyright-lateral strike-slip motion (Figures 4A and 4C) using a linearinversion for spatially distributed fault slip on a set of 328triangular dislocations (Meade 2007) with minimum momentregularization constraints. Beavan et al. (2010) demonstratedthat shallow (~4 km) thrust slip in addition to right-lateral slip isrequired to fully account for all features in the deformation field;however, our primary goal in interpreting the Darfield earthquakedeformation field is to drive modeling of Coulomb stresschange at the location of the Christchurch earthquake. At thesedistances, the effects of the shallow thrust faults are not likely tohave a strong effect on Coulomb stress change (e.g., King 2009).Our Darfield fault model location is based on mapped surfaceruptures (Quigley et al. 2010) while dips are constrained by focalmechanisms of right-lateral aftershocks. We extend our faults tothe east and west to account for significant deformation apparentin the interferograms beyond mapped surface ruptures. Ourbest-fit slip distribution and model residual is shown in Figure4A, with a moment magnitude of Mw 7.0. Slip magnitudes anddepth ranges agree well with previous inversions by Beavan et al.(2010) using InSAR and GPS observations. We are unable to fitsome features in the data near the center and easternmost end ofthe rupture (Figure S1). The misfit is influenced by a combinationof errors in model geometry, exclusion of NE-SW-dippingreverse faults, spatially correlated atmospheric noise, ionosphericperturbations, and contributions from significant postseismicdeformation evident in postseismic interferograms (Figure S3)and, therefore, likely present in varying degrees in the coseismicinterferograms used in our inversions.Figure 3B illustrates the predicted E-W horizontal offsetsfrom our best-fit model at the location of the optical imagepixel-tracking results. The predicted displacements across thefault are significantly smaller (~2.5 m compared with 5 m),which is not surprising given that there was a data gap in theInSAR imagery approaching the fault and that the regularizationplaced on our inversion tends to reduce slip in regions thathave less coverage by the data. The discrepancy may also be due,in part, to variable amounts of postseismic slip between theinterferograms and the optical imagery. Overall, the differencebetween the observed displacements and those predicted usinginversions based on InSAR data and an elastic half-space modelhighlights both the importance of using near-field data when itexists as well as the potential for issues in using elastic modelsin regions where the deformation is clearly anelastic. However,these issues are likely to primarily affect the inversion for slipin the shallow subsurface and will not contribute much to thepredicted Coulomb stress study discussed below.Christchurch EarthquakesTo obtain a fault model for the Christchurch earthquakes, weuse the Neighborhood Algorithm (Sambridge 1999) to invertALOS-PALSAR and Envisat interferograms (Table 1) for singlefault dislocations (Table S1). We then fix this best-fit geometryand extend the fault along-strike and down-dip to avoid spuriousedge effects before performing a linear inversion for distributedslip. Model trace locations are shown in Figure 4C. We use anautomated, resolution-based fault parameterization (Barnhartand Lohman 2010) that generates smaller fault patches near thesurface, where there is more constraint from data, than at depthand offshore, which allows us to more efficiently explore a rangeof potential fault geometries and constraints on slip than if weused a uniform fault patch size distribution. For the 22 FebruarySeismological Research Letters Volume 82, Number 6 November/December 2011 819

Depth (km)Coulomb Stress Change(bars)event, we obtain a distributed slip model with 182 triangulardislocations (Figure 4B), with Laplacian smoothing constraintsto regularize the inversion. We fix the slip rake direction to 64degrees, as reported by the Global Centroid Moment Tensor(GCMT) solution (http://www.globalcmt.org; Dziewonski etal. 1981). Inversions in which we allow rake to vary reveal similarsolutions. Our best-fit model strikes N57E and dips 70S beneaththe Banks Peninsula. This fault geometry agrees well with theGCMT south-dipping focal solution (N59E, 64S) and distributionsof aftershocks analyzed through the double-differencemethod (Bannister et al. 2011, page 839 of this issue). The slipmodel suggests peak slip of 2.1 m with the main rupture areaoccurring between 2 and 11 km and has a moment magnitudeMw 6.4 (Figure 4B). Some very shallow slip is observed in themodel, although this region corresponds to areas offshore whereno geodetic data is available and is probably an artifact of theinversion. Our slip model supports the ground and pixel-offsetobservations of no surface rupture during the Christchurchearthquake; however, data gaps in the InSAR observationswithin the city of Christchurch may inhibit our inversions frominferring any slip at the surface. Because only one pair of imagesis available to constrain slip during the 13-June event (Table 1,Figure S2), we do not present a distributed slip model. We showthe location of our best-fit single patch model in Figure 4C.Coulomb Stress ChangeIn order to model the potential effects of static Coulomb stresschange of the Darfield earthquake on the 22-Feb Christchurchearthquake, we use the Darfield earthquake slip distributiondescribed above (Figure 4A), which predicts a static Coulombstress change on a fault with the orientation and rake inferredfor the Christchurch earthquake as shown in Figure 5A. In ourcalculation, all slip inverted for the Christchurch earthquakeoccurs within the region of positive Coulomb stress change(Figure 5A, black curve). This suggests that static Coulombstress change from the Darfield earthquake indeed encouragedthe Christchurch earthquake. Peak calculated static Coulombstress change is 3.1 bars while the minimum is –4.5 bars.To obtain statistics describing the significance of theseinferred static Coulomb stress changes, we apply a MonteCarlo error propagation technique similar to that describedin Lohman and Barnhart (2010). We begin by simulating 500noisy data sets by adding spatially correlated noise with a spatialscale of 100 km to the predicted LOS surface displacementsfrom our best-fit slip distribution, using the same covariance aswe infer from the original Darfield data. We then invert for slipon the same four-fault geometry used above for each syntheticdata set. Lastly, we calculate the static Coulomb stress changeon the fault geometry and slip orientation inferred for theChristchurch earthquake for each realization of the syntheticdata. This method allows us to quantify errors in predictedCoulomb stress change (Figure 5B) induced by data noise, suchas correlated atmospheric water vapor. As can be seen in Figure5B, the expected variation due to these sources is far less thanthe inferred increase in stress resolved on the target fault planethat ruptured during the Christchurch earthquake. OtherDepth (km)0 West Along-StrikeLength (km) East030a9180918errors due to variations in fault plane geometry, crustal elasticstructure, or to the contribution from the rest of the aftershocksequence likely also contribute.DISCUSSION6-4Certain attributes of this earthquake sequence suggest reactivationof poorly developed faults. A particularly interestingattribute of seismicity during the 2010–2011 Canterbury earthquakesequence is the activity of steeply dipping (>50 degrees)reverse faults. First motion focal solutions for the Darfieldearthquake reveal reverse-motion rupture on a steep, eastdippingplane (Gledhill et al. 2011) before slip propagated toE-W-striking strike-slip faults. In addition, aftershock locationsand focal mechanisms located in NE-SW-trending zones at theends and center of the Darfield rupture reveal steeply dippingreverse-motion planes, and steep reverse faults are necessary tomodel geodetic observations of both the Darfield (Beavan et al.2010) and Christchurch earthquakes. Traditional Andersonianstylefaulting predicts that faults should form at angles of ~30degrees to the principal shortening direction (Anderson 1951),which results in reverse faults dipping 30 degrees with a horizontalshortening direction and normal faults dipping 60degrees with a vertical shortening direction. While Anderson’stheory predicts the angles at which faults form relative to thelocal stress field, preexisting faults can reactivate and new faults300.4StDev (bars)0.05▲ ▲ Figure 5. A) Static Coulomb stress change on the Christchurchearthquake fault plane predicted by the slip distribution inferredfor the Darfield earthquake (Figure 4A). Positive Coulomb stresschange encourages rupture, negative discourages rupture.Black outline shows extent of Christchurch earthquake slip withmagnitude > 0.7 m. B) 1σ standard deviation of static Coulombstress change, calculated using 500 realizations of the Darfieldearthquake slip distribution.bSeismological Research Letters Volume 82, Number 6 November/December 2011 821

will not be formed if it is energetically more favorable to slip onnon-optimally oriented planes (Anderson 1951). Reactivationof non-Andersonian faults is observed in numerous tectonicenvironments including Iran (e.g., Byerlee 1978) and the Aegean(e.g., Berberian 1995). The steep dip of reverse faults observed inaftershock and mainshock focal mechanisms along with geodeticallyderived fault geometries for the Darfield and Christchurchearthquakes strongly suggest Cretaceous-Oligocene faults,formed during formation of the Torlesse terrain and laterbreakup of the Rangitata Orogen (Jackson 1994), were seismicallyreactivated during the 2010–2011 Canterbury earthquakesequence. Likewise, the high stress drops, particularly for theDarfield event, calculated for each event (Fry et al. 2011, page846 of this issue) suggest reactivation of high-friction faultsunder low strain rates compared to faults in the Marlboroughfault zone or Puysegur and Hikurangi subduction zones.The lack of many aftershocks west of the Darfield earthquakein the Southern Alps (Figure 1), where thrust faultsare oriented more N-S and dip at lower angles (

Crippen, R. E. (1992). Measurement of subresolution terrain displacementsusing SPOT panchromatic imagery. Episodes 15, 56–61.DeMets, C., R. G. Gordon, D. F. Argus, and S. Stein (1994). Effect ofrecent revisions to the geomagnetic reversal time-scale on estimatesof current plate motions. Geophysical Research Letters 21 (20),2,191–2,194.Dorn, C., A. G. Green, R. Jongens, S. Carpentier, A. E. Kaiser, F.Campbell, H. Horstmeyer, J. Campbell, M. Finnemore, andJ. Pettinga (2010). High-resolution seismic images of potentiallyseismogenic structures beneath the northwest CanterburyPlains, New Zealand. Journal of Geophysical Research 115;doi:201010.1029/2010JB007459.Dziewonski, A., T. A. Chou, and J. H. Woodhouse (1981). Determinationof earthquake source parameters from waveform data for studiesof global and regional seismology. Journal of Geophysical Research286, 2,825–2,852.Farr, T., and M. Kobrick (2000). Shuttle Radar Topographic Missionproduces a wealth of data. Eos, Transactions, American GeophysicalUnion 81, 583–585.Fialko, Y., M. Simons, and D. Agnew (2001). The complete (3D) surfacedisplacement field in the epicentral area of the 1999 Mw7.1 Hector Mine Earthquake, California, from space geodeticobservations. Geophysical Research Letters 28 (16), 3,063–3,066;doi:200110.1029/2001GL013174.Fry, B., R. Benites, and A. Kaiser (2011). The character of accelerationsin the M w 6.2 Christchurch earthquake. Seismological ResearchLetters 82, 846–852.Gledhill, K., J. Ristau, M. Reyners, B. Fry, and C. Holden (2011). TheDarfield (Canterbury, New Zealand) Mw 7.1 Earthquake ofSeptember 2010: A Preliminary seismological report. SeismologicalResearch Letters 82 (3), 378–386; doi:10.1785/gssrl.82.6.378.Howard, M., A. Nicol, J. Campbell, and J. R. Pettinga (2005). Holocenepaleoearthquakes on the strike-slip Porters Pass Fault, Canterbury,New Zealand. New Zealand Journal of Geology and Geophysics 48(1), 59–74; doi:10.1080/00288306.2005.9515098.Iizuka, H., Y. Sakai, and K. Koketsu (2011). Strong ground motions anddamage conditions associated with seismic stations in the February2011 Christchurch, New Zealand, earthquake. SeismologicalResearch Letters 82, 875–822.Jackson, J. (1994). Active tectonics of the Aegean region. Annual Reviewof Earth and Planetary Sciences 22 (1), 239–271; doi:10.1146/annurev.ea.22.050194.001323.Kääb, A., and M. Debella-Gilo (2010). Sub-pixel precision image matchingfor measuring surface displacements on mass movements usingnormalized cross-correlation. Remote Sensing of Environment 115(1), 130–142; doi:10.1016/j.rse.2010.08.012.King, G. (2009). Fault interaction, earthquake stress changes, and theevolution of seismicity. In Earthquake Seismology, 225–255, H.Kanamori, ed. Elsevier.Leprince, S., S. Barbot, F. Ayoub, and J.-P. Avouac (2007). Automaticand precise orthorectification, coregistration, and subpixel correlationof satellite images, application to ground deformation measurements.IEEE Transactions on Geoscience and Remote Sensing 45(6), 1,529–1,558; doi:10.1109/TGRS.2006.888937.Lohman, R. B., and W. D. Barnhart (2010). Evaluation of earthquaketriggering during the 2005–2008 earthquake sequenceon Qeshm Island, Iran. Journal of Geophysical Research 115;doi:201010.1029/2010JB007710.Lohman, R. B., and M. Simons (2005). Some thoughts on the use ofInSAR data to constrain models of surface deformation: Noisestructure and data downsampling. Geochemistry GeophysicsGeosystems 6, doi:10.1029/2004GC000841.Mackinnon, T. C. (1983). Origin of the Torlesse terrane and coeval rocks,South Island, New Zealand. Geological Society of America Bulletin94 (8), 967–985; doi:10.1130/0016-7606(1983)942.0.CO;2.Meade, B. J. (2007). Algorithms for the calculation of exact displacements,strains, and stresses for triangular dislocation elements in auniform elastic half space. Computers & Geosciences 33 (8), 1,064–1,075; doi:10.1016/j.cageo.2006.12.003.Melkonian, A. (2011). Measuring glacier velocities and elevation changerates from ASTER data for Juneau icefield, Alaska. Master’s thesis,Cornell University, Ithaca, NY.Michel, R., and J.-P. Avouac (2002). Deformation due to the 17 August 1999Izmit, Turkey, earthquake measured from SPOT images. Journal ofGeophysical Research 107 (B4); doi:10.1029/2000JB000102; http://europa.agu.org/?view=article&uri=/journals/jb/jb0204/2000JB000102/2000JB000102.xml&t=2002. Last accessed June 24, 2011.Norris, R. J., and A. F. Cooper (2001). Late Quaternary slip rates and slippartitioning on the Alpine fault, New Zealand. Journal of StructuralGeology 23 (2–3), 507–520; doi:16/S0191-8141(00)00122-X.Pettinga, J., M. Yetton, R. Van Dissen, and G. Downes (2001).Earthquake source identification and characterization for theCanterbury region, South Island, New Zealand. Bulletin of the NewZealand Society for Earthquake Engineering 34, 282–317.Quigley, M., R. Van Dissen, P. Villamor, N. Litchfield, D. Barrell, K.Furlong et al. (2010). Surface rupture of the Greendale fault duringthe Darfield (Canterbury) earthquake, New Zealand: Initialfindings. Bulletin of the New Zealand Society for EarthquakeEngineering 43 (4), 236–242.Rosen, P. A., S. Hensley, G. Peltzer, and M. Simons (2004). UpdatedRepeat Orbit Interferometry package released. Eos, Transactions,American Geophysical Union 85 (5), 47.Sambridge, M. (1999). Geophysical inversion with a neighbourhoodalgorithm: Searching a parameter space. Geophysical JournalInternational 138 (2), 479–494.Sibson, R., F. Ghisetti, and J. Ristau (2011). Stress control of an evolvingstrike-slip fault system during the 2010–2011 Canterbury, NewZealand, earthquake sequence. Seismological Research Letters 82,824–832.Stirling, M., M. Yetton, J. Pettinga, K. R. Berryman, and G. Downes(1999). Probabilistic Hazard Assessment and Earthquake Scenariosfor the Canterbury Region, and Historic Earthquakes in Christchurch:Stage I (Part B) of Canterbury Regional Council’s Earthquake Hazardand Risk Assessment Study. Canterbury Regional Council ReportNo. U99/18.Sutherland, R., K. Berryman, and R. Norris (2006). Quaternary slip rateand geomorphology of the Alpine fault: Implications for kinematicsand seismic hazard in southwest New Zealand. Geological Societyof America Bulletin 118 (3–4), 464–474; doi:10.1130/B25627.1.Timm, C., K. Hoernle, P. Van Den Bogaard, I. Bindeman, and S. Weaver(2009). Geochemical evolution of intraplate volcanism at BanksPeninsula, New Zealand: Interaction between asthenosphericand lithospheric melts. Journal of Petrology 50 (6), 989–1,023;doi:10.1093/petrology/egp029.Wallace, L. M., J. Beavan, R. McCaffrey, K. Berryman, and P. Denys(2007). Balancing the plate motion budget in the South Island,New Zealand, using GPS, geological and seismological data.Geophysical Journal International 168 (1), 332–352; doi:10.1111/j.1365-246X.2006.03183.x.Wessel, P., and W. H. F. Smith (1998). New, improved version of GenericMapping Tools released. Eos, Transactions, American GeophysicalUnion 79, 579.Department of Earth and Atmospheric SciencesCornell University2120 Snee HallIthaca, New York 14850 U.S.A.wdb47@cornell.edu(W. D. B.)Seismological Research Letters Volume 82, Number 6 November/December 2011 823

Stress Control of an Evolving Strike-Slip FaultSystem during the 2010–2011 Canterbury, NewZealand, Earthquake SequenceRichard Sibson, Francesca Ghisetti, and John RistauRichard Sibson, 1 Francesca Ghisetti, 2 and John Ristau 3INTRODUCTIONLarge earthquakes within seismogenic crust are generallythought to require the pre-existence of large fault structures.Such fault structures appear to evolve by the progressivegrowth and amalgamation of smaller faults and fractures(Cowie and Scholz 1992). In the course of their evolution somecomponents of an evolving fault system may be inherited fromprevious tectonic episodes while others may be newly formedin the prevailing tectonic stress field. With increasing displacementand amalgamation of sub-structures, fault structurestend to become “smoother,” less complex, and perhaps weaker(Wesnousky 1988).The 2010–2011 Canterbury earthquake sequenceoccurred within the upper crust of the South Island of NewZealand around 100 km southeast from the fast-moving (20–30 mm/yr) Alpine and Hope fault strike-slip components ofthe Pacific-Australia transform fault system linking into thesouthern Hikurangi Margin subduction zone (Figure 1). Asof 15 July 2011, the sequence has included three major shocks:the M w 7.1 Darfield earthquake (3 September 2010 UTC) followedby an M w 6.2 event on 21 February 2011 UTC and an M w6.0 event on 13 June 2011 UTC, along with a rich aftershocksequence that includes 27 shocks with M w > 5.0. Rupturingoccurred on previously unrecognized faults that appear to becomponents of a highly segmented E-W structure concealedbeneath alluvial cover and/or Neogene volcanics. Some subsurfaceinformation is, however, available from seismic reflectionlines and gravity surveys (e.g., Field et al. 1989).Here we seek to demonstrate how this complex sequencehas likely arisen through reactivation under the contemporarytectonic stress field of a mixture of comparatively newly formedand older inherited fault structures.1. Department of Geology, University of Otago, P.O. Box 56, Dunedin9054, New Zealand2. Terrageologica, 129 Takamatua Bay Rd., RD1, Akaroa 7581, NewZealand3. GNS Science, Te Pu Ao, P.O. Box 30-368, Lower Hutt, New ZealandTECTONIC/GEOLOGIC SETTINGThe 2010–2011 Canterbury earthquakes occurred within30 ± 5 km thick continental crust belonging to the buoyantChatham Rise plateau contained within the Pacific plate(Eberhart-Phillips and Bannister 2002). Local geology (Figure2) comprises a basement of highly deformed Mesozoic Torlessemetagraywackes and their metamorphosed equivalents atgreater depth, unconformably overlain by a Late Cretaceous–Neogene cover sequence up to 2.5 km thick (Forsyth et al.2008). Polyphase deformation within this basement assemblageincludes accretion, folding and thrusting along the Gondwanamargin, extensional fault structures from Late Cretaceous riftingof the Zealandia microcontinent, and Neogene transpressionacross the Alpine fault system.The cover sequence consists of Late Cretaceous–Paleogeneterrestrial-marine sedimentary units (including varyingthicknesses of Late Cretaceous Mt. Somers calc-alkaline volcanicsand Eocene basalts) overlain by a regressive Miocene-Pliocene clastic sequence that contains the predominantlybasaltic Late Miocene (11–6 Ma) Banks Peninsula volcanics.Thickness variations are partly attributable to deposition asa Late Cretaceous–Paleocene syn-rift sequence accompanyingextensional rifting along the Gondwana margin, whichimposed an extensive fault fabric within the basement (Lairdand Bradshaw 2004). Neogene shortening has led to varyingreactivation of these inherited fault systems. Over the area ofthe Canterbury Plains the older units are largely obscured byPliocene and Quaternary alluvial gravels up to a few hundredmeters thick (Forsyth et al. 2008).CONTEMPORARY STRESS FIELDAvailable evidence on the contemporary regional stress fieldin the central South island (Sibson et al., forthcoming) comesfrom two principal sources summarized in Table 1: 1) stressinversions from earthquake focal mechanisms together withone breakout determination from the Galleon-1 borehole;and 2) axes of maximum contractional strain-rate derived824 Seismological Research Letters Volume 82, Number 6 November/December 2011 doi: 10.1785/gssrl.82.6.824

AlpineFaultHope Fault38 mm/yr0+ +-1.5-1.5-1-0.543 0SP0A+-1.5-1.50-1-0.50-1-1.5+0R-1-0.5H-1.5+G-2-2.5-2-1-2CHRISTCHURCH+-1.5Rangiora-2.5+0-0.5-1.5-1-0.5?-2-1.5172 0 E173 0EPegasus BayBANKS PENINSULA-2-244 0S-2.512345638 mm/yr Pacific-Australia Plateslip vectorStructural contours of top to basement (depth in km relative to s.l.)Seismic sequence from Sept. 4, 2010 to June 15, 2011(location and magnitude from GEONET, http://www.geonet.org.nz)GM6-7.1 Surface rupture of Sept. 4, 2010 earthquake+(Greendale Fault)Exploration wellM5-5.9 M4-4.9-20 to >2 km 0 to - 2 km -2 to < -2.5 km▲ ▲ Figure 1. Tectonic setting of the 2010–2011 Canterbury earthquake sequence. Mapped faults are from 1:250,000 geological maps(Forsyth 2001; Rattenbury et al. 2006; Cox and Barrell 2007; Forsyth et al. 2008). Offshore faults are from Barnes (1994), Mogg et al.(2008) and from preliminary data released from the 2011 offshore survey in the Pegasus Bay (NIWA and Otago University, http://www.gns.cri.nz/Home/News-and-Events/Media-Releases/Fault-structures-revealed). Subsurface geometry beneath the alluvial cover sequenceof the Canterbury Plains and offshore is illustrated by the structural contours of depth to basement. Top of basement is modified fromHicks (1989) and Mogg et al. (2008) and incorporates basement depths in exploration wells, depth-converted seismic reflection lines(http://www.nzpam.govt.nz), and outcrop data. A, R, and H are respectively the Ashley, Rakaia, and Hinds fault systems, interpretedas inherited normal faults. P is the Porters Pass fault system and G is the right-lateral surface break of the 4 September 2010, M w 7.1earthquake on the Greendale fault (Quigley et al. 2010). 1) right-lateral faults; 2) W-dipping reverse faults (triangles in hanging wall);3) E-dipping reverse faults (triangles in hanging wall); 4) normal faults (ticks in hanging wall); 5) inferred blind faults; 6) structurallycontrolled escarpment inferred from Bouguer gravity gradients from Bennett et al. (2000).Seismological Research Letters Volume 82, Number 6 November/December 2011 825

TMSVTTPGCRETTTQ-GTM-PPGCRETBPV▲▲Figure 2. Synoptic tectonostratigraphic column for theCanterbury Plains (not to scale). T = Torlesse basement assemblage;CRET = Late Cretaceous terrestrial sequence; MSV = Mt.Somers volcanics; PG = Paleogene marine sequence;M-P = Miocene-Pliocene marine-terrestrial sequence;BPV = Miocene Banks Peninsula volcanics; Q-G = Quaternarygravels.Tfrom geodetic retriangulation and more modern GPS studies.Congruence between the two sources suggests a remarkablyuniform regional stress field throughout the Canterburyregion with maximum compressive stress σ 1 horizontal andoriented WNW-ESE (115° ± 5°). The apparent parallelismbetween σ 1 stress trajectories and maximum contractionalstrain-rate is explicable if the latter can be treated as a measureof maximum incremental shortening subparallel to σ 1 (cf.Keiding et al. 2009). This σ 1 orientation is also consistent withpredominantly reverse-slip reactivation of structures trendingNNE-NE along the Southern Alps range front. However, thedominance of strike-slip faulting in the Canterbury earthquakesequence suggests that the regional stress field is thatof an “Andersonian” wrench (strike-slip) regime (Anderson1905, 1951) with principal compressive stresses oriented:σ 1 : 0°/115° ± 5°; σ 2 : vertical; σ 3 : 0°/025° ± 5°. Some local stressheterogeneity is, of course, to be expected, perhaps throughbuttressing by the Banks Peninsula volcanic edifice.From the combination of strike-slip and reverse faultingthroughout the region the likelihood is that the stress field isof the form σ 1 > σ v = σ 2 ~ σ 3 , with some local variance betweenσ v = σ 2 and σ v = σ 3 . Brittle structures developing within thecurrent “wrench” regime would be expected to have the orientationsshown in the Figure 3 inset. Newly forming strike-slipfaults should be vertical, lying at 25°–35° to σ 1 in accordancewith the Coulomb criterion (Anderson 1905). Note that thisalso reflects the optimal attitude for reactivating existing faultsover a broad range of sliding friction (0.8 > μ s > 0.4) (Sibson1985). Grain-scale microcracks and macroscopic extensionfractures would be statistically aligned perpendicular to σ 3with an orientation 115° ± 5°/V. Planes of maximum shearstress controlling the orientation of subvertical ductile shearzones at depth below the seismogenic zone should lie at ±45° toσ 1 along trends of 070° (dextral) and 160° (sinistral).PRINCIPAL COMPONENTS OF THE SEQUENCEThe principal components of the 2010–2011 Canterburyearthquake sequence are summarized in a seismotectonic cartoonillustrating the relationship of the different structures tothe inferred σ 1 stress trajectories (Figure 3). Regional CMTmechanisms for M w ≥ 4.0 shocks within the sequence are illustratedin Figure 4. The sequential development and structuralcharacteristics of the larger ruptures are discussed below.TABLE 1Stress Indicators in Canterbury, Central South Island (see also Sibson et al., forthcoming).Region Method σ 1 azimuth ReferenceStress DeterminationsNorth Canterbury-Marlborough Inversion of focal mechanisms 115 ± 16° Balfour et al. 2005Epicentral area 1994 Arthur’s Pass M w 6.7 Inversion of aftershock focal mechanisms 118 ± 4° Robinson and McGinty2000Central S. Alps Inversion of focal mechanisms 110 – 120° Leitner et al. 2001Offshore S. Canterbury Borehole breakouts 114 ± 9° Sibson et al. 2012Maximum Contractional Strain RatesOkarito west of Alpine F. Determination from repeated triangulation 117 ± 6° Pearson 1994Godley Valley east of Alpine F. Determination from repeated triangulation 116 ± 14° Pearson 1994Canterbury Plains NW of ChCh Determination from GPS campaigns 116 ± 9° Pearson et al. 1995Central Southern Alps Determination from GPS campaigns 105 – 115° Beavan and Haines 2001Canterbury-Otago block Rotating elastic block model 110 ± 8° Wallace et al. 2007826 Seismological Research Letters Volume 82, Number 6 November/December 2011

σ 1dσ 3Nuσ 3CM w 7.1dw 6.2u70°M w55°6.0MBPV10 kmσ 1σ 1115 ± 5°σ 3145 ± 5°▲▲Figure 3. Seismotectonic cartoon of the 2010–2011 Canterbury earthquake sequence in relation to the surface outcrop of BanksPeninsula volcanics (BPV), central Christchurch city (C), and the inferred regional stress field. Expected orientations of newlyformed structures (ellipse = extension fracture; solid lines = Coulomb shears; dashed lines = ductile shears) shown in inset at lowerleft. Epicenters of major shocks denoted by stars; thick bold line = Greendale fault surface rupture; thinner dash-dot lines with filledteeth indicating dip direction = subsurface fault models for the M w 6.2 and M w 6.0 aftershocks (Beavan et al. 2011, page 789 of thisissue); dotted line = microearthquake lineament; dark shading = area of intense aftershock activity in dilational stepover; gray-shadedbands = aftershock lineaments; hollow-toothed lines = belt of reverse-slip aftershocks (dip unconstrained).▲▲Figure 4. Regional CMT focal mechanisms for M w > 4.0 shocks within the 2010–2011 Canterbury earthquake sequence.Seismological Research Letters Volume 82, Number 6 November/December 2011 827

The 2010 M w 7.1 Darfield mainshock (16:35:46 on3 September 2010 UTC) appears to be a composite rupturethat initiated at a depth of 11 km below the Canterbury Plains,~ 6 km north of a segmented surface rupture with dextralstrike-slip < 5 m that developed during the earthquake and wasmapped for nearly 30 km west-east across the plains (Quigley etal. 2010). This rupture, on what is now termed the Greendalefault (Figures 1 and 3), occurred in an extremely low-relief areaof the plains without any prior topographic expression of thestructure. Measured dextral displacements averaging 2.5 m butranging up to ~ 5 m were consistent with pure strike-slip ona subvertical fault with an enveloping surface for the left-steppingrupture segments striking ~ 085°. While near-field focalmechanism and geodetic analyses suggest that initial rupturinginvolved reverse-slip on NE-SW striking planes (Gledhillet al. 2011; GeoNet regional moment tensor solution cataloguehttp://www.geonet.org.nz/resources/earthquake/), teleseismicanalyses from long-period waves (USGS, http://earthquake.usgs.gov; Global CMT Project, http://www.globalcmt.org/CMTsearch.html) yield moment tensor mechanisms consistentwith near-vertical dextral strike-slip on a rupture parallelingthe mapped surface trace of the Greendale fault (Figure 4).Aftershocks, largely restricted to the upper crust at depthsless than 12 km, were initially concentrated in an E-W swatharound the surface rupture trace of the Greendale fault, thougha number of subsidiary clusters and lineaments are also evident(Bannister et al. 2011, page 839 of this issue; Gledhill et al.2011).An aftershock cluster dominated by reverse-slip eventsabuts the Southern Alps foothills west and south of the westerncurving termination of the Greendale fault surface rupture(Figures 3 and 4). A strong subsidiary belt of activity with amixture of strike-slip and reverse-slip mechanisms trendsNNW from the mainshock epicenter toward the foothillsand the Porters Pass system of strike-slip faults. Of particularstructural interest is a diffuse aftershock lineament trending145°–155° that developed south of the surface rupture in thefirst two weeks of the sequence and extended out to the coastand beyond. A strong concentration of activity including fiveof the M w > 5.0 events is associated with the eastern end of theGreendale fault rupture and the area just south of it. From themixture of strike-slip and normal fault mechanisms localizedwithin this rhomboidal area of distributed activity (Gledhillet al. 2011), it appears to represent a dilational stepover to anen échelon ENE-trending aftershock lineament that extendedeastward along the Banks Peninsula volcanic rangefront.Dilational stepovers (jogs) of this kind are well known forfocusing long-continued aftershock activity with delayed sliptransfer to the en échelon fault segment (Sibson 1986).On 22 February (23:51:42 on 21 February 2011 UTC)central and eastern Christchurch were devastated by an M w6.2 aftershock located along the dominant ENE lineament followingthe northern outcrop margin of the Banks Peninsulavolcanics. No surface fault break was observed, but Beavan etal. (2011, page 789 of this issue) modeled GPS and SAR dataon surface deformation, in terms of dextral-reverse slip of up to3 m on a buried rupture with a length of 12 km and a width of7 km oriented ~060°/70° SE and extending to within 1 km ofthe surface beneath the Heathcote-Avon estuary (cf. Barnhartet al. 2011, page 815 of this issue). Beavan et al. (2011, page789 of this issue) also suggest subsidiary strike-slip rupturingon a plane oriented ~ 080°/87° S (Figure 3). Aftershock activitywas concentrated in the hanging wall of the main rupture,extending a little out to sea north and east of Banks Peninsula(Bannister et al. 2011, page 839 of this issue).A further M w 6.0 aftershock occurred on 13 June (02:20:50on 13 June 2011 UTC) with an epicenter some 5 km further tothe ENE but again without surface rupture. The regional CMTsolution yielded nodal planes suggesting either dextral-reverseslip on a fault oriented 068°/84° SSE, or sinistral-reverse slipon 161°/67° WSW with the slip vector raking 6° SSE (GeoNetcatalog; http://www.geonet.org.nz). Support for this latterrupture orientation comes from an aftershock lineament thatextended progressively along a trend of 140°–150°, subparallelto the SE-SSE trending lineament from the Greendale faultrupture (Figures 3 and 4). A “single-fault” interpretation ofsurface deformation recorded by GPS and DinSAR (J. Beavan,personal communication 2011) suggests left-reverse obliqueslip of up to 1.5 m on a rupture plane oriented 155°/55° SW thatextends from 1 km to 9 km in depth and along strike for ~ 14km, cutting across the eastern end of the 22 February rupture.BASEMENT FAULT FABRICKnown fault traces extending into the basement (includingLate Quaternary–Holocene active segments) identified by surfacemapping in the Canterbury region and from seismic profiling,both onshore and offshore (Forsyth et al. 2008; Barnes1994) (Figure 1) are clustered in two dominant groups basedon their displacement characteristics (Figure 5).1. Faults possessing dominant right-lateral components inthe rangefront and foothills of the Southern Alps are subvertical,and oriented across the azimuthal range 050°–100° (Figure 5). ENE faults cluster along the active PortersPass system, and E-W faults are mainly represented bythe new rupture trace along the Greendale fault (Figure1). Dominant orientations at 070°–100° are also shownby faults with components of normal separation, togetherwith normal fault escarpments inferred from gradients inBouguer gravity with surface traces indicating high-angledips (70°–80°) to the north and south.2. Faults dominated by reverse-slip include two groups ofopposite-facing structures: 000°–050° trending faultsdipping west, and 035°–090° trending faults dippingsoutheast (Figures 1 and 5). North-trending reverse faultsdipping at moderate angles (40°–60°) to the west are welldeveloped along the S. Canterbury rangefront (Figure 1).ENE-trending reverse faults dipping south at high angles(60°–70°) occur only in the north of the Canterburyregion (Figure 1). Both groups include Quaternary-activesegments.828 Seismological Research Letters Volume 82, Number 6 November/December 2011

Right-lateral Faults(A) Range Front and FoothillsNW23%10%5%5%10%15%20%14%Reverse FaultsEBouguer gravity anomalies (Hicks 1989; Bennett et al. 2000)together with information from exploration wells and seismiclines. These data have been used to contour the top of basementbelow the cover sequence (Figure 1). Two major depocenters(Pegasus-Rangiora basin to the north and Rakaia-Hinds basinto the south) are identified elongated in an easterly orientationand separated by an intervening structural high coincidentwith Banks Peninsula, where uplifted basement graywackesare exposed beneath the Miocene volcanics. Exploration wellshave penetrated late syn-rift sequences infilling these basins.Discontinuous E-W fault traces mapped in the Quaternarygravels of the Canterbury Plains along the Ashley, Rakaia, andHinds fault systems (Figure 1) are thus interpreted as surfacetraces of buried basement faults belonging to the structuraldomain of the Chatham Rise.STRUCTURAL ANALYSISW24%S(B) Canterbury Plains, Banks Peninsula andAdjacent Offshore40%30%20%10%NSOffshore, the northwestern edge of the Chatham Rise preservesa strong extensional fabric defined by closely spacedE-W striking, S-dipping normal faults that bound half-grabensinfilled with up to 2 km of inferred Late Cretaceous syn-riftsediments (Barnes 1994). Projection of these structures westwardbelow the Canterbury Plains is conjectural but is based on5%10%15%20%▲▲Figure 5. Rose diagrams of fault strike azimuths within theCanterbury region covered by Figure 1, weighted for mappedlength: A) Area of the Southern Alps rangefront and foothillswhere Torlesse basement is exposed; B) Area of the CanterburyPlains, Banks Peninsula, and offshore where basement is largelyconcealed. Top half of each plot is for faults where reverse-slipis dominant; bottom half is for faults with predominantly rightlateraland/or normal slip.Right-lateraland Normal Faults Reverse Faults42%EObserved slip senses on the three major ruptures within theearthquake sequence are consistent with the inferred pattern ofσ 1 stress trajectories (Figure 3). However, some stress heterogeneityis evident, especially near rupture terminations and faultintersections (Figure 4). In Anderson’s (1905, 1951) applicationof the Coulomb criterion for brittle shear failure to the initiationof faults within intact isotropic crust, strike-slip faultsforming in a wrench stress regime (σ v = σ 2 ) should be verticaland lie at ± ~30° to the σ 1 orientation. In contrast, large-displacementstrike-slip faults commonly lie at far higher angles(often >45°) to regional σ 1 trajectories and are distinctly “non-Andersonian” (Mount and Suppe 1987; Balfour et al. 2005).It follows that vertical, low-displacement strike-slip faults atAndersonian orientations are possibly newly formed structuresin the contemporary stress field, but they could also be inheritedfaults that happen to be optimally oriented for frictionalreactivation. Following the same argument, oblique-slip rupturesmost likely result from the reactivation of inherited faultsin the prevailing stress field.The principal ruptures of the Canterbury earthquakesequence can be viewed with these considerations in mind(Figure 3). First, the subvertical Greendale fault rupture lyingat 25°–35° to regional σ 1 is at optimal Andersonian orientation,implying that it is either a comparatively newly formed strikeslipfault or an inherited structure that is optimally orientedfor reactivation in the present stress field. It should be borne inmind that most of the inherited dip-slip faults within the basementare likely to have dips that are substantially less than vertical,though the Greendale rupture could occupy an amalgamof opposite-dipping structures. Note further, however, that atits western termination, the Greendale rupture trace curvesinto parallelism with the σ 1 stress trajectories (Figure 3), a propagationcharacteristic of low-displacement shear fractures andconsistent with the local existence of CMT mechanisms withcomponents of normal slip. Moreover, total strike-slip displacementon the Greendale fault appears not to be large becauseit has not been recognized to continue along strike into thebedrock geology of the Southern Alps foothills. In fact, at itsFigure 5Seismological Research Letters Volume 82, Number 6 November/December 2011 829

181614121086disallowablefrom frictionallock-updip > 75°n = 61DEXTRALallowable dextralGREENDALE FAULTfavorablyoriented45° 45°σ 1SINISTRALallowable sinistralfavorablyorienteddisallowablefrom frictionallock-up420030° 040° 050° 060° 070° 080° 090° 100° 110° 120° 130° 140° 150° 160° 170° 180° 190° 200°STRIKE AZIMUTH▲ ▲ Figure 6. Azimuthal distribution of nodal plane strikes for close-to-pure strike-slip CMT focal mechanisms (both planes dipping >75°)from the Canterbury earthquake sequence (GeoNet catalog http://www.geonet.org.nz), shown in relation to the inferred σ 1 direction.western termination the fault appears to transform into localareas of normal faulting to the north and reverse faulting to thesouth (Figure 4).While the dominant rupture in the 22 February aftershocksclearly involves dextral-reverse oblique slip, the subordinatesubvertical plane (080°/87° S) lying subparallel to theGreendale fault (Beavan et al. 2011, page 789 of this issue)is at close to the ideal Andersonian orientation for strike-slip.This part of the sequence may therefore represent competitionbetween inherited and newly formed fault segments. The twodiffuse aftershock lineaments trending 140°–155° (Figure 3)are appropriately oriented for left-lateral strike-slip on verticalfaults conjugate to the right-lateral Greendale fault with whichthey form a dihedral angle of ~ 50°–70°. Combining the CMTfocal mechanism (161°/67° WSW) with the fault model for the13 June M w 6.0 aftershock (153°/55° SW) suggests predominantlyleft-lateral strike-slip on a moderately-to-steeply dippingplane with the slip vector raking only 6°, not too dissimilar tothe ideal Andersonian relationship. However, the suggestionof a nonvertical rupture with a degree of oblique slip makes itlikely that rupturing involved the reactivation of an inheritedbasement structure. These arguments are explored further byexamining the distribution of strike azimuths, with respect tothe inferred σ 1 direction, of aftershock nodal planes for closeto-purestrike-slip CMT focal mechanisms (GeoNet catalog,http://www.geonet.org.nz) where both nodal planes dip >75°(Figure 6). Because of the ambiguity as to which nodal planerepresents the rupture plane, the distribution repeats at 90°intervals, separating potential dextral from potential sinistralstrike-slip faults. Theoretical and field studies suggest thatfaults containing the σ 2 direction undergo frictional lock-up at55°–60° to σ 1 (Collettini and Sibson 2001), reducing the allowablerange of strike-slip fault orientations. Potential strike-sliporientations are thus reduced to three categories: dark-shadedcolumns are inadmissible because of frictional lock-up; lightshadedcolumns are positively discriminated as either dextralor sinistral strike-slip ruptures; and moderate-shaded columnscould represent either dextral or sinistral strike-slip. Severalfeatures of the distribution are notable. First, despite its lengthand continuity, the Greendale fault trend is not dominant instrike-slip aftershock orientations. Moreover, a significant proportionof the positively discriminated mechanisms involvesinistral strike-slip on faults that commonly strike 135°–145°,conjugate to the dextral Greendale fault. However, by far thedominant azimuthal trend is 070° and/or 160°. Note first thatthese trends lie at ±45° to inferred σ 1 defining the orientationsof vertical planes with maximum shear stress, the expectedorientation for ductile shear zones developing in the basementbelow the brittle seismogenic crust (Figure 3). However, the070° trend also lies subparallel to the Hope fault and the presentinterplate slip vector, suggesting the possibility of somekinematic control.830 Seismological Research Letters Volume 82, Number 6 November/December 2011

DISCUSSIONThe 2010–2011 Canterbury earthquake sequence developedwithin a segmented fault system under an Andersonian wrenchstress regime (σ 1 : 0°/115° ± 5°; σ 2 : vertical; σ 3 : 0°/025° ± 5°)(Figure 3). Rupturing predominantly involved dextral strikeslipon subvertical E-W faults with varying degrees of reverseslipon differently oriented (mostly ENE-WSW) fault segments.Local normal and reverse slip faulting also occurred atstress heterogeneities at strike-slip rupture tips. Some rupturesclearly involve reactivation of inherited basement faults butother comparatively low-displacement structures may be newlyformed within the contemporary stress field.Subordinate SE-SSE trending aftershock lineamentsappear to represent a set of predominantly left-lateral strike-slipfaults conjugate to the main dextral structures (Figure 3). Theintersection angle of 50°–70° between the conjugate fault sets(±25°–35° to inferred σ 1 ) is consistent with Andersonian frictionalfault mechanics. Note that this Andersonian conjugaterelationship differs from the orthogonal relationship recognizedfor conjugate strike-slip faults in central Honshu, Japan,and southern California, which possibly reflects control of theactive brittle structures by orthogonal ductile shear zones inthe basement (Thatcher and Hill 1991).Analysis of the strike-slip focal mechanisms from theCanterbury sequence (Figure 6) suggests that the subordinateset of steep sinistral strike-slip faults may be quite widespread.This has significance for rupture segmentation because sinistraldisplacements along the conjugate faults will create contractionaljogs that impede slip along the main E-W dextral faults.In this regard, the 2010–2011 Canterbury sequence has similaritiesto the 2000 Western Tottori earthquake sequence insouthwestern Honshu. The Western Tottori M w 6.7 mainshockinvolved sinistral rupturing along a previously unrecognizedNNW-SSE strike-slip fault, but high-resolution aftershockmapping showed the mainshock lineament to be offset in aseries of contractional jogs by Andersonian conjugate dextralfaults (Fukuyama et al. 2003). As in the Canterbury sequence,such contractional jogs may act as high-strength asperitiesbecause ruptures bypassing them likely have to break throughcomparatively intact rock. It is notable that particularly highapparent stresses and ground accelerations are associated withthe intersection zone at the eastern end of the main E-W aftershockdistribution in the Canterbury sequence where the 22February M w 6.2 rupture is apparently cross-cut by the 13 JuneM w 6.0 rupture (Fry and Gerstenberger 2011, page 833 of thisissue).Overall, the fault system responsible for the Canterburyearthquake sequence appears to be controlled by the orientationof the tectonic stress field in the upper crust rather thanconforming with local plate boundary kinematics. On thisbasis the earthquakes can be regarded as intraplate eventsremote from the main Alpine-Marlborough fault systemdefining the onshore plate boundary. Continuance of conjugatefaulting has the important implication that displacementweakening leading to preferential failure has not yet reachedthe stage where one of the fault sets has become totally dominantand superseded the other. Amalgamation of inheritedand newly formed fault components of low total displacement,together with segmentation from the cross-cutting of the majorE-W dextral structures by conjugate left-lateral faults, has ledto a rough, immature fault system capable of generating highstress-drop ruptures.ACKNOWLEDGMENTSThe writers extend their thanks and appreciation to JohnBeavan, Stephen Bannister, and Martin Reyners for muchhelpful discussion and advice.REFERENCESAnderson, E. M. (1905). The dynamics of faulting. Transactions of theEdinburgh Geological Society 8, 387–402.Anderson, E. M. (1951). The Dynamics of Faulting and Dyke Formationwith Application to Britain. 2nd ed. Edinburgh: Oliver & Boyd,206 pp.Balfour, N. J., M. K. Savage, and J. Townend (2005). Stress and crustalanisotropy in Marlborough, New Zealand: Evidence for low faultstrength and structure-controlled anisotropy. Geophysical JournalInternational 163, 1,073–1,086.Bannister, S., B. Fry, M. Reyners, J. Ristau, and H. Zhang (2011).Fine-scale relocation of aftershocks of the 22 February M w 6.2Christchurch earthquake using double-difference tomography.Seismological Research Letters 82, 839–845.Barnes, P. M. (1994). Continental extension of the Pacific plate at thesouthern termination of the Hikurangi subduction zone: TheNorth Mernoo fault zone, offshore New Zealand. Tectonics 13,753–754.Barnhart, W., M. J. Willis, R. B. Lohman, and A. Melkonian (2011).InSAR and optical constraints on fault slip during the 2010–2011New Zealand earthquake sequence. Seismological Research Letters82, 815–823.Beavan, J., E. Fielding, M. Motagh, S. Samsonov, and N. Donnelly(2011). Fault location and slip distribution of the 22 February 2011M W 6.2 Christchurch, New Zealand, earthquake from geodeticdata. Seismological Research Letters 82, 789–799.Beavan, J., and J. Haines (2001). Contemporary horizontal velocityand strain-rate fields of the Pacific-Australia plate boundary zonethrough New Zealand. Journal of Geophysical Research 106, 741–770.Bennett, D., R. Brand, D. Francis, S. Langdale, C. Mills, B. Morris,and X. Tian (2000). Preliminary results of exploration in theonshore Canterbury Basin, New Zealand. New Zealand PetroleumConference Proceedings, 19–22 March 2000. Wellington, NZ:Crown Minerals, 12 pp. http://www.nzpam.govt.nz/cms/petroleum/conferences/conference-proceedings-2000.Collettini, C., and R. H. Sibson (2001). Normal faults, normal friction?Geology 29, 927–930.Cowie, P., and C. H. Scholz (1992). Growth of faults by accumulationof seismic slip. Journal of Geophysical Research 97, 11,085–11,095.Cox, S. C., and D. J. A. Barrell (2007). Geology of the Aoraki Area.Institute of Geological and Nuclear Sciences 1:250,000 geologicalmap 15, 1 sheet and 71 pp. Lower Hutt, New Zealand: GNSScience.Eberhart-Phillips, D., and S. Bannister (2002). Three-dimensionalcrustal structure in the Southern Alps region of New Zealand frominversion of local earthquake and active source data. Journal ofGeophysical Research 107; doi:10.1029/2001JB000567.Seismological Research Letters Volume 82, Number 6 November/December 2011 831

Field, B. D., G. H. Browne, B. Davy, R. H. Herzer, R. H. Hoskins, J.L. Raine, G. J. Wilson, R. J. Sewell, D. Smale, and W. A. Watters(1989). Cretaceous and Cenozoic sedimentary basins and geologicalevolution of the Canterbury region, South Island, New Zealand.New Zealand Geological Survey Basin Studies 2. Wellington, NewZealand: Department of Scientific and Industrial Research, 94 pp.+ enclosures.Forsyth, P. J. (2001). Geology of the Waitaki Area. Institute of Geologicaland Nuclear Sciences 1:250,000 geological map 19, 1 sheet and 64pp. Lower Hutt, New Zealand: GNS Science.Forsyth, P. J., D. J. A. Barrell, and R. Jongens (2008). Geology of theChristchurch Area. Institute of Geological and Nuclear Sciences1:250,000 geological map 16, 1 sheet and 67 pp. Lower Hutt, NewZealand: GNS Science.Fry, B., and M. C. Gerstenberger (2011). Large apparent stresses from theCanterbury earthquakes of 2010 and 2011. Seismological ResearchLetters 82, 833–838.Fukuyama, E., W. L. Ellsworth, F. Waldhauser, and A. Kubo (2003).Detailed fault structure of the 2000 Western Tottori, Japan, earthquakesequence. Bulletin of the Seismological Society of America 93,1,468–1,478.Gledhill, K., J. Ristau, M. Reyners, B. Fry, and C. Holden (2011). TheDarfield (Canterbury, New Zealand) M w 7.1 earthquake of 4September 2010: A preliminary seismological report. SeismologicalResearch Letters 82, 378–386.Hicks, S. R. (1989). Structure of the Canterbury Plains, New Zealand,from gravity modelling. Research Report 222, Geophysics Division,Department of Scientific and Industrial Research, Wellington,New Zealand.Keiding, M., B. Lund, and T. Árnadóttir (2009). Earthquakes, stress,and strain along an obliquely divergent plate boundary: ReykjanesPeninsula, southwest Iceland. Journal of Geophysical Research 114,B09306; doi:10.1029/2008JB006253.Laird, M. G., and J. D. Bradshaw (2004). The break-up of a long-termrelationship: The Cretaceous separation of New Zealand fromGondwana. Gondwana Research 7, 273–286.Leitner, B., D. Eberhart-Phillips, H. Anderson, and J. Nabelek (2001).A focused look at the Alpine fault, New Zealand: Seismicity,focal mechanisms and stress observations. Journal of GeophysicalResearch 106, 2,193–2,220.Mogg, W. G., K. Aurisch, R. O’Leary, and G. P. Pass (2008). TheCarrack-Caravel prospect complex: A possible sleeping giantin the deep Canterbury Basin, New Zealand. Proceedings of thePetroleum Exploration Society of Australia Eastern AustralasianBasins Symposium III, Sydney, Australia, 14–17 September 2008,369–378.Mount, V. S., and J. Suppe (1987). State of stress near the San Andreasfault. Geology 15, 1,143–1,146.Pearson, C. (1994). Geodetic strain determinations from the Okaritoand Godley-Tekapo regions, central South Island, New Zealand.New Zealand Journal of Geology and Geophysics 37, 309–318.Pearson, C., J. Beavan, D. Darby, G. H. Blick, and R. I. Walcott (1995).Strain distribution across the Australian-Pacific plate boundary inthe central South Island, New Zealand, from 1992 GPS and earlierterrestrial observations. Journal of Geophysical Research 100,22,071–22,081.Quigley, M., P. Villamor, K. Furlong, J. Beavan, R. Van Dissen, N.Litchfield, T. Stahl, B. Duffy, E. Bilderback, D. Noble, D. Barrell, R.Jongens, and S. Cox (2010). Previously unknown fault shakes NewZealand’s South Island. Eos, Transactions, American GeophysicalUnion 91, 469–472.Rattenbury, M. S., D. B. Townsend, and M. R. Johnston (2006). Geologyof the Kaikoura Area. Institute of Geological and Nuclear Sciences1:250,000 geological map 13, 1 sheet and 70 pp. Lower Hutt, NewZealand: GNS Science.Robinson, R., and P. J. McGinty (2000). The enigma of the Arthur’s Pass,New Zealand, earthquake. 2. The aftershock distribution and itsrelation to regional and induced stress fields. Journal of GeophysicalResearch 105, 16,139–16,150.Sibson, R. H. (1985). A note on fault reactivation. Journal of StructuralGeology 7, 751–754.Sibson, R. H. (1986). Rupture interaction with fault jogs. In EarthquakeSource Mechanics, ed. S. Das, J. Boatwright, and C. H. Scholz, 157–167. American Geophysical Union Monograph 37 (Maurice EwingSeries 6). Washington, DC: American Geophysical Union.Sibson, R. H., F. C. Ghisetti, and R. A. Crookbain (forthcoming).“Andersonian” wrench faulting in a regional stress field duringthe 2010–2011 Canterbury, New Zealand, earthquake sequence.In Stress Controls on Faulting, Fracturing and Igneous Intrusionin the Earth’s Crust—Commemorating the Work of Ernest MassonAnderson, ed. D. Healy et al. Geological Society of London specialpublication.Thatcher, W., and D. P. Hill (1991). Fault orientations in extensional andconjugate strike-slip environments and their implications. Geology19, 1,116–1,120.Wallace, L. M., J. Beavan, R. McCaffrey, K. Berryman, and P. Denys(2007). Balancing the plate motion budget in the South Island,New Zealand, using GPS, geological and seismological data.Geophysical Journal International 168, 332–352.Wesnousky, S. G. (1988). Seismological and structural evolution ofstrike-slip faults. Nature 335, 340–343.Department of GeologyUniversity of OtagoP.O. Box 56Dunedin 9054, New Zealandrick.sibson@otago.ac.nz(R. S.)832 Seismological Research Letters Volume 82, Number 6 November/December 2011

Large Apparent Stresses from the CanterburyEarthquakes of 2010 and 2011B. Fry and M. C. GerstenbergerB. Fry and M. C. GerstenbergerGNS ScienceINTRODUCTIONAn earthquake of Mw 6.1–6.3 1 (Beavan et al. 2011, page 789of this issue) that struck Christchurch, New Zealand, on 22February (21 February, UTC) produced recorded groundmotion acceleration over 2 g. The event caused widespread damagewith dense recordings of non-linear site behavior. Globally,dense near-field recordings of shallow intraplate earthquakesare rare. It is possible that extreme ground motions are commonwith this type of earthquake and that their rarity is merelya function of inadequate seismic sampling in the near field ofsuch low-probability, high-potency events. To better define thenature of these events, we calculate apparent stress (τ a ) of thethree largest earthquakes in the Canterbury sequence and comparethem to global and regional data. We then place recordedPGA and spectral accelerations into the context of regionaland global ground motion prediction equations and discussthe implications of high-stress events for future seismic hazardestimates for the region. For the February event, we also brieflyexplore the implications of directivity on measured groundmotions in central Christchurch.The earthquakes that occurred in the Canterbury regionof the South Island, New Zealand, from September 2010 tothe present have disproportionately large energy magnitudes(Me) to their moment magnitudes (Mw). They have producedthe largest ground motions ever measured in New Zealand.The sequence began with the Mw 7.1 earthquake that occurredabout 40 km west of the city of Christchurch on 4 September2010. The maximum recorded ground acceleration recordedduring the event was over 1.25 g, which was experienced nearthe intersection of the triggering thrust on which the rupturebegan and the strike-slip Greendale fault that carried mostof the moment in the earthquake (Gledhill et al. 2010). Peakground accelerations (PGA) in the central business districtof Christchurch averaged between about 0.2 and 0.3 g. Thesemotions were sufficient to generate liquefaction in areas of thecity. The highest recorded acceleration in the greater metropolitanarea was 0.61 g in a suburb on the southern edge of the citythat has since proved to be prone to strong site amplification.On 22 February 2011, an Mw 6.3 thrust earthquake occurred1. Mw estimates for this earthquake have ranged from 6.1 (USGS) to6.2 (Beavan et al. 2011, page 789 of this issue). To be conservative inour comparison to observed ground motions, we have used Mw 6.3 inall calculations.on a structure below the southern suburbs of the city at about7 km epicentral distance from the center of Christchurch. Thisearthquake produced extreme motions in Christchurch (Fryet al. 2011, page 846 of this issue). Maximum PGA, consideringboth horizontal and vertical components, was over 2.2g with two other recordings in the city greater than 1 g andaverage PGA in the central business district between about0.6 and 0.8 g. This intense shaking damaged many buildingsin the central business district of the city (~5–8 km epicentraldistance) and triggered widespread liquefaction (Kaiser et al.2011). On 13 June 2011, the city was again subject to intenseshaking from a nearby, shallow Mw 6.0 earthquake (Beavan etal. 2011, page 789 of this issue). Measured accelerations fromthat event were also extreme, with measured PGA over 2 g in asoutheastern suburb of the city. Taken together, this sequencehas produced widespread destruction and more than 180 fatalitiesin Christchurch.HIGH APPARENT STRESS (τ a )The faults that failed in the September 2010 Mw 7.1, theFebruary 2011 Mw 6.3, and the June 2011 Mw 6.0 earthquakeswere likely very strong, with high amounts of friction.Typically, faults in slowly deforming areas with long earthquakerecurrence intervals exhibit this attribute, as increasingdeformation typically decreases fault strength by reducingheterogeneities on the fault surface (e.g., Ben-Zion andSammis 2003). Subsequently, the radiated energies (Es) forthe three events were high for their given moments. Radiatedenergy can be determined from high-frequency velocity records(Boatwright and Choy 1986) and can be used to directly calculateMe (Choy and Boatwright 1995). Compared to theseismic moment, which is derived from displacement records,energy magnitudes are more indicative of the shaking potentialof an earthquake. Es estimates from analysis of broadband Pwaves provide Me of Me = 7.99, Me = 6.75, and Me = 6.7 forthe three events (George Choy, personal communication).Apparent stress is defined as the product of rigidity (μ) and Esper unit moment (τ a = (μ × Es)/Mo) (Wyss and Brune 1968),or the amount of stress per unit moment. There is considerabledebate regarding the scaling of τ a with earthquake moment.Aki (1957) asserts that earthquakes are self-similar, implyingthat τ a is not dependent on seismic moment. This assertionis supported by numerous other studies (e.g., Boatwright anddoi: 10.1785/gssrl.82.6.833Seismological Research Letters Volume 82, Number 6 November/December 2011 833

TABLE 1Magnitudes and Stress Calculations for the Three LargestCanterbury EventsDateMw(regional)Mw(teleseismic) Ms Meτ a(Mpa)9/03/2010 7.10 6.97 7.30 7.99 15.852/22/2011 6.30 6.12 6.30 6.75 4.106/13/2011 6.00 6.00 6.00 6.70 6.26Choy 1992; Choy and Boatwright 1995; Singh and Ordaz1994; Baltay et al. 2011). Many others have proposed a momentdependence on τ a (e.g., Abercrombie 1995; Kanamori et al.1993). τ a has important implications for earthquake hazardstudies, as increasing τ a leads to stronger ground motions. Usingteleseismically determined Es and Mo (George Choy, personalcommunication), we solve for τ a of the three largest Canterburyevents (Table 1). We assume constant bulk regional propertiesfor density and shear-wave velocity to determine rigidity. Thesequantities are informed by the region’s 3D shear-wave velocitymodel (Eberhart-Phillips et al. 2010). Values of rigiditymight be underestimated for the February and June events, asit is possible that these events occurred on surfaces that are cutby intrusions from the 6 Ma volcanic activity that resulted inthe volcanic edifice in the vicinity of the earthquakes. In thisscenario, the faults would be locally strengthened by the intrusionsand the energy release would be dominated by a subregionof the fault, greatly increasing the localized τ a .Due to the frequency dependent nature of both scatteringand attenuation, it is difficult to measure Es over a broadfrequency range (Ide et al. 2003). In this study, we utilize estimatesof Mo that are derived from teleseismic data to maintainconsistency with the teleseismically determined Es. Greaterattenuation of high-frequency energy also dictates that teleseismicestimates of Es are minimum values of actual radiatedenergy in highly attenuating regions. Ideally, energy andmoment estimates from regional data would be used to estimateτ a . However, such techniques require refined knowledgeof local attenuation structure and site responses that werepreviously only coarsely resolved. Ongoing studies are refiningthese properties (Kaiser et al. 2011) and should allow forregional estimates in the near future.We estimate the τ a of the September event to be the highestof the three earthquakes, at ~16 MPa. τ a of the February andJune events are ~5 and ~6 MPa respectively. Intraplate earthquakestypically have larger τ a than interplate events. This isalso true of the South Island, New Zealand, where recent largeevents along the Puysegur subduction zone (Fry et al. 2010)have τ a ~ 0.2MPa. We also compare the three Canterburyevents to data from four sequences of Hokkaido, Japan, eventscompiled by Baltay et al. 2011 (Figure 1) and the 2007 M 6.8event that occurred on the Hikurangi subduction zone, NorthIsland, New Zealand. The Iwate Miyagi event is a particularlyappropriate analogue as the character of the recorded near-fieldwaveforms is similar to that of the 22 February Canterburyevent (Fry et al. 2011, page 846 of this issue).The stresses calculated from the Canterbury events areremarkably high compared to global averages. The Canterburyearthquakes are also on average higher than τ a values obtainedfrom reported stress drops (most between 10 and 15 MPa) ofintraplate events in eastern North America (Atkinson andBoore 2006). For shallow subduction events, Choy et al. (2001)▲ ▲ Figure 1. Apparent stress plotted as a function of Mw for data from Japan (Baltay et al. 2011) and New Zealand. Ch 04 is the 2004Chuetsu earthquake and aftershock sequence; CO 07 is the 2007 Chuetsu-Oki earthquake and aftershock sequence; Kam is the repeatingearthquake sequence off-shore Kamaishi, Iwate; IM 08 is the 2008 Iwate-Miyagi earthquake and aftershock sequence; GB 07 isthe 2007 M 6.8 Gisborne, New Zealand, earthquake; Puy are recent earthquakes on the Puysegur subduction zone, South Island, NewZealand; Can are the largest events of the 2010–2011 Canterbury sequence.834 Seismological Research Letters Volume 82, Number 6 November/December 2011

▲ ▲ Figure 2. Observed vs. predicted ground motions from McVerry at al. (2006) for the 22 February 2011, Mw 6.3 earthquake. In allpanels, the prediction is shown with stress drop scaling (green solid line) and without (gray solid line). For both, the 16th and 84th percentilemotions are also shown in the same color. (A) PGA—both raw (black circles) and processed (open triangles) PGA are shown;(B) 0.5-s spectral accelerations; (C) 1.0-s spectral accelerations; and (D) 2.0-s spectral accelerations.find average apparent stresses to be around 0.2–0.3 MPa.Shallow events worldwide have a spread from about 6 MPa to0.04 MPa, with an average of about 0.5 MPa. While the datafor shallow, strike-slip events is limited, strike-slip mechanismsnear subduction zones show much higher τ a ; these events havean average of around 3 MPa (Choy et al. 2001). Typically, highτ a is indicative of spatially small yet strong asperities yieldingshort rise times and strong, high-frequency waves in the nearfield. Pervasive Cretaceous-age high-angle EW-trending faultzones that are hundreds of kilometers long are present offshoreof Christchurch (Wood and Herzer 1993). It is likelythat these faults continue in the crust beneath the CanterburyPlains and are capable of being reactivated as strike-slip faults.In this case, the fault strength would likely come from healingof the fault over time. Another possible fault origin for theFebruary event is reactivation of faults generated during the12 Ma–6 Ma emplacement of the Banks Peninsula volcanicrocks to the south of the city (see Figure 1 of Bannister et al.2011, page 839 of this issue). The relative immaturity of faultsin this hypothesis would contribute to the observed high τ a . Itis also possible that the faults that were active in the Februaryand June events were cross-cut by volcanic intrusions. Even ifthe faults were originally weak, cross-cutting dikes would yieldhigh stresses when broken. Precise locations of aftershocksfall along the northern and western flanks of the extinct volcano(Bannister et al. 2011, page 839 of this issue) and subsequentlydo not discriminate these competing hypotheses forthe origin of fault strength.OBSERVED GROUND MOTIONSIf the τ a of the Canterbury earthquakes is larger than the averageτ a of earthquakes used in generating the New Zealandground motion prediction equation (GMPE) (McVerry et al.2006), measured ground motions should be larger than predictedones.Figure 2 shows plots of observed, maximum vector groundmotions from the 22 February 2011 earthquake comparedto predictions for Class D, deep or soft soil (Standards NewZealand, 2004) and for an oblique mechanism; this is thestandard GMPE used in the New Zealand National SeismicHazard Model (Stirling et al. 2002). Distances are calculatedusing the closest distance to the current estimated fault plane(Beavan et al. 2011, page 789 of this issue). For comparison,we have applied a stress drop scaling term as proposed byAtkinson and Boore (2006). Stress drop is typically estimatedSeismological Research Letters Volume 82, Number 6 November/December 2011 835

September 4, 2010, Mw=7.11010A) B)SA(g)110.10.10.01PGASA(1s)0.011 10 100June 22, 2011, Mw=6.01010C) D)11SA(g)0.10.10.010.01PGA0.0011 10 100Closest Distance From Fault (km)SA(1s)0.0011 10 100Closest Distance From Fault (km)▲ ▲ Figure 3. Observed vs. predicted ground motions from McVerry at al. (2006) for the September 2010 Mw 7.1 earthquake (Panels Aand B) and the June 2011 Mw 6.0 earthquake (Panels C and D). In all panels, the prediction is shown with stress drop scaling (greensolid line) and without (gray solid line). For both, the 16th and 84th percentile motions are also shown in the same color. (A) SeptemberPGA—both raw (black circles) and processed (open triangles) PGA are shown; (B) September 1.0-s spectral acceleration; (C) JunePGA; (D) June 1-s spectral acceleration.as ~4 × τ a based on empirical calculations. The stress drop scalingis both frequency and magnitude dependent and relies onan estimate of the ratio of the reference stress drop implicit inthe GMPE to the stress drop of the target earthquake. For theMcVerry et al (2006) relationship, which is primarily based onNew Zealand earthquakes, and the Canterbury events, we havecalculated a scaler using a ratio of 1.5 (e.g., 15 MPa/10 MPa).Additionally, in Figure 2A, for PGA, we have shown boththe raw and the processed PGAs using the standard processingas done by GeoNet. The two most notable features in theplots are: (1) the sharp shift in the decrease in ground motionsafter about 10 km resulting in an overprediction of the groundmotions, and (2) the clustered increase in ground motions forthe near-field data. This effect is likely driven by a biased samplingof data in the near field toward locations subject to strongdirectivity, with that bias diminishing when including moredistant stations and sampling over a much wider area outsidethe directivity region.Figure 3 shows the PGA and 1-s spectral acceleration plotsfor the September and June earthquakes. Raw PGA values areplotted only for the September event. This event consisted ofstrike-slip displacement on the Greendale fault with smallermoments carried by surrounding thrust faults; for the distancecalculations, only the dominant Greendale fault was used(Beavan et al. 2010). The error introduced by this should beinsignificant based on the spatial distribution of stations available.Fault rupture models of the June rupture are still preliminaryand will likely introduce large (a few kilometers) errorsinto the distance calculations; we have therefore used the epicenterof the relocated mainshock (Bannister et al. 2011, page839 of this issue) projected to 1 km depth, the estimated topof the rupture. The distances are likely to be overestimated. For836 Seismological Research Letters Volume 82, Number 6 November/December 2011

10Obs SA(1s)AS1AS+SDAB 2006AB+SDSA(g)0.10.010.0011 10 100Closest Distance To Fault (km)▲ ▲ Figure 4. Observed vs. predicted ground motions from Abrahamson and Silva (2008) and Atkinson and Boore (2006) for the 22February Mw 6.3 earthquake. Both GMPEs are shown with and without stress drop scaling.both earthquakes, similar trends are seen as to the Februaryevent with a drop off and overprediction of accelerations at distancesgreater than 10 km.For comparison, using OpenSHA (Field et al. 2003), wehave plotted the median 1-s spectral acceleration curves forthe February event using Atkinson and Boore (2006) andAbrahamson and Silva (2008) GMPEs in Figure 4. Both curvesare plotted using a reverse mechanism and Vs 30 = 230 m/s. ForAbrahamson and Silva, we have used parameters based on datafrom two average sites in Christchurch. Both have been plottedwith and without the same stress drop scaling that was appliedto McVerry et al. (2006). The behavior of the models is similarto the McVerry et al. (2006) model with underpredictions inthe near field and overpredictions beyond 10 km.DISCUSSIONThe implications for probabilistic seismic hazard assessment(PSHA) for the region are not yet clear, and for any interpretationof the data it must be acknowledged that this remainsonly a sample of three earthquakes and one must be wary ofover-interpretation; this is of particular concern when quickscientific response is necessary for advising the emergency managementprocess. Due to the nature of the ongoing aftershockactivity in the Canterbury region, the hazard is likely to bedominated by earthquakes less than M 6.5 with a preliminaryPSHA (Gerstenberger et al. 2011) indicating that the 10% in50-year ground motions are dominated by smaller earthquakesat distances of 10 km or less. This indicates that more work isnecessary to understand relative contributions of both stressdrop and directivity to the near-field motions. Figures 2 and 3indicate that if directivity effects are specifically added to thecurrent stress drop scaling, possibly through increased variabilityin ground motion, we may overestimate the hazard for theregion.For future improvements to New Zealand hazard estimatesit is likely that having a catalog of Me and therefore τ awill help us to make regional refinements in ground motionestimates. This can be done by incorporating Es estimates intoroutine processing and developing a refined 3D attenuationmodel and database of static station-dependent site effects (e.g.,Boatwright et al. 2002). By calculating both regional momenttensors and radiated energy from regional data, better estimatesof strong high-frequency ground motions can be made.ACKNOWLEDGMENTSThis manuscript greatly benefited from the review of GailAtkinson. We would like to thank George Choy for providingthe estimates of radiated energy.Seismological Research Letters Volume 82, Number 6 November/December 2011 837

REFERENCESAbercrombie, R. E. (1995). Earthquake source scaling relationships from~1 to 5 M L using seismograms recorded at 2.5 km depth. Journal ofGeophysical Research 100, 24,015–24,036.Abrahamson, N. A., and W. J. Silva (2008). Summary of the Abrahamson& Silva NGA ground-motion relations. Earthquake Spectra 24 (1),67–97.Aki, K. (1967). Scaling law of seismic spectrum. Journal of GeophysicalResearch 72, 1,217–1,231.Atkinson, G. M., and D. M. Boore (2006). Earthquake ground-motionprediction equations for eastern North America. Bulletin of theSeismological Society of America 96, 2,181–2,205.Baltay, A., S. Ide, G. Prieto, and G. Beroza (2011). Variability in earthquakestress drop and apparent stress. Geophysical Research Letters38, L06303; doi:10.1029/2011GL046698.Bannister, S., B. Fry, M. Reyners, J. Ristau, and H. Zhang (2011).Fine-scale relocation of aftershocks of the 22 February M w 6.2Christchurch earthquake using double-difference tomography.Seismological Research Letters 82, 839–845.Beavan, R. J., S. Samsonov, M. Motagh, L. M. Wallace, S. M. Ellis, andN. Palmer (2010). The Darfield (Canterbury) earthquake: Geodeticobservations and preliminary source model. Bulletin of the NewZealand Society for Earthquake Engineering 43 (4), 228–235.Beavan, J., E. Fielding, M. Motagh, S. Samsonov, and N. Donnelly(2011). Fault location and slip distribution of the 22 February 2011M W 6.2 Christchurch, New Zealand, earthquake from geodeticdata. Seismological Research Letters 82, 789–799.Ben-Zion, Y., and C. G. Sammis (2003). Characterization of fault zones.Pure and Applied Geophysics 160, 677–715.Boatwright, J., and G. Choy (1986). Teleseismic estimates of the energyradiated by shallow earthquakes. Journal of Geophysical Research91, 2,095–2,112.Boatwright, J., and G. Choy (1992). Acceleration source spectra anticipatedfor large earthquakes in northeastern North America.Bulletin of the Seismological Society of America 82 (2), 660–682.Boatwright, J., G. Choy, and L. Seekins (2002). Regional estimates ofradiated seismic energy. Bulletin of the Seismological Society ofAmerica 92, 1,241–1,255.Choy, G. L., and J. L. Boatwright (1995). Global patterns of radiatedseismic energy and apparent stress. Journal of Geophysical Research100, 18,205–18,228.Choy, G. L., J. L. Boatwright, and S. Kirby (2001). The Radiated SeismicEnergy and Apparent Stress of Interplate and Intraplate Earthquakesat Subduction Zone Environments: Implications for Seismic HazardEstimation. USGS Open-File Report 01-005, 10 pp.Eberhart-Phillips, D., M. E. Reyners, S. C. Bannister, M. P. Chadwick,and S. M. Ellis (2010). Establishing a versatile 3-D seismic velocitymodel for New Zealand. Seismological Research Letters 81 (6),992–1,000; doi:10.1785/gssrl.82.6.992.Field, E. H., T. H. Jordan, and C. A. Cornell (2003). OpenSHA: Adeveloping community-modeling environment for seismic hazardanalysis. Seismological Research Letters 74 (4), 406–419.Fry, B., S. Bannister, J. Beavan, L. Bland, B. Bradley, S. Cox, J. Cousins,N. Gale, G. Hancox, C. Holden, R. Jongens, W. Power, G. Prasetya,M. Reyners, J. Ristau, R. Robinson, S. Samsonov, K. Wilson, andthe GeoNet team (2010). The M w 7.6 Dusky Sound earthquake of2009: Preliminary report. Bulletin of the New Zealand Society forEarthquake Engineering 43 (1), 24–40.Fry, B., R. Benites, and A. Kaiser (2011). The character of accelerationsin the M w 6.2 Christchurch earthquake. Seismological ResearchLetters 82, 846–852.Gerstenberger, M., M. Cubrinovski, G. McVerry, M. Stirling, D.Rhoades, B. Bradley, R. Langridge, T. Webb, B. Peng, J. Pettinga,K. Berryman, and H. Brackley (2011). Probabilistic Assessment ofLiquefaction Potential for Christchurch in the Next 50 Years. GNSScience report 2011/15, 30 pp. Lower Hutt, New Zealand.Gledhill, K., J. Ristau, M. E. Reyners, B. Fry, and C. Holden (2010).The Darfield (Canterbury, New Zealand) M w 7.1 earthquake ofSeptember 2010: A preliminary seismological report. SeismologicalResearch Letters 82 (3), 378–386; doi:10.1785/gssrl.82.6.378.Ide, S., G. C. Beroza, S. G. Prejean, and W. L. Ellsworth (2003).Apparent break in earthquake scaling due to path and site effectson deep borehole recordings. Journal of Geophysical Research 108(B5), 2,271; doi:10.1029/2001JB001617.Kaiser, A., C. Holden, J. Beavan, D. Beetham, R. Benites, A. Celentano,D. Collett et al. (2011). The February 2011 Christchurch earthquake:A preliminary report. Submitted to New Zealand Journal ofGeology and Geophysics.Kanamori, H., J. Mori, E. Hauksson, T. H. Heaton, L. K. Hutton, andL. M. Jones (1993). Determination of earthquake energy releaseand M L using TERRAscope. Bulletin of the Seismological Society ofAmerica 83, 330–346.McVerry, G. H., J. X. Zhao, N. A. Abrahamson, and P. G. Somerville(2006). Response spectral attenuation relations for crustal and subductionzone earthquakes. Bulletin for the New Zealand Society ofEarthquake Engineering 39, 1–58.Singh, S. K., and M. Ordaz (1994). Seismic energy release in Mexicansubduction zone earthquakes. Bulletin of the Seismological Societyof America 72, 2,003–2,016.Standards New Zealand (2004). Structural Design Actions, Part 5:Earthquake Actions—New Zealand. New Zealand Standard NZS1170.5:2004. Wellington, New Zealand.Stirling, M. W., G. H. McVerry, and K. R. Berryman (2002). A new seismichazard model for New Zealand. Bulletin of the SeismologicalSociety of America 92 (5), 1,878–1,903.Wood, R.A., and R. H. Herzer (1993). The Chatham Rise, New Zealand.In South Pacific Sedimentary Basins, ed. P. F. Ballance, 329–349.Vol. 2 of Sedimentary Basins of the World. Amsterdam: ElsevierScience Publishers.Wyss, M., and J. N. Brune (1968). Seismic moment, stress, and sourcedimensions for earthquakes in the California-Nevada region.Journal of Geophysical Research 73 (14), 4,681–4,694.GNS Science1 Fairway DriveAvalon, Lower Hutt, New Zealandb.fry@gns.cri.nz(B. F.)838 Seismological Research Letters Volume 82, Number 6 November/December 2011

Fine-scale Relocation of Aftershocks of the22 February M w 6.2 Christchurch EarthquakeUsing Double-difference TomographyStephen Bannister, Bill Fry, Martin Reyners, John Ristau, and Haijiang ZhangStephen Bannister, 1 Bill Fry, 1 Martin Reyners, 1 John Ristau, 1 andHaijiang Zhang 2EOnline material: Hypocenters of 2,177 earthquakes recordedduring 21 February 21–31 March 2011INTRODUCTIONOn 22 February 2011 New Zealand time (21 February UTC),the M W 6.2 Christchurch earthquake occurred just 7 kmsoutheast of the center of Christchurch city, New Zealand (Fryet al. 2011, Holden 2011, page 783 of this issue). There were181 confirmed fatalities, and the damage to Christchurch cityis estimated to be NZ$15 billion–$NZ20 billion (US$12 billion–US$16billion). The event was well-recorded by thebroadband and strong-motion national-scale GeoNet network(Petersen et al. 2011) as well as by the Canterbury regionalstrong-motion network (Avery et al. 2004). Since the 22February earthquake, more than 2,700 further aftershockshave been recorded up to 1 May 2011, including 21 events withlocal magnitude (M L ) greater than 5. Here we describe the initialrelocation analysis for these aftershocks.The M w 6.2 Christchurch earthquake is part of the largeraftershock sequence of the M w 7.1 Darfield earthquake, whichoccurred at 16:35 3 September UTC, 2010. Seismological,GPS, and InSAR data all suggest that the earthquake ruptureprocess for the M w 7.1 Darfield earthquake involved failureof multiple fault segments (Beavan et al. 2010; Gledhill et al.2011). A surface rupture for that earthquake, now termedthe Greendale fault, extending ~29.5 km and located ~4 kmsouth of the epicenter, is consistent with strike-slip faultingwith an average horizontal surface displacement of ~2.5 m(Quigley et al. 2010). The vast majority of the 7,400+ aftershocksfollowing the Darfield earthquake are shallow, at lessthan 15 km depth. Figure 1 shows that, although many of theaftershocks occurred near the surface trace of the Greendalefault, intense clusters of aftershock activity have also occurred1. GNS Science, Lower Hutt, New Zealand2. Dept. of Earth, Atmospheric, and Planetary Sciences, MassachusettsInstitute of Technology, Cambridge, MA, U.S.A.at the western and eastern ends of the Darfield fault trace, aswell as north-northwest from the Darfield epicenter.The distribution of the aftershocks (Figure 1A) indicatesa complex fault system, most of which was previously undetected;prior to the 2010–2011 activity there was negligiblerecorded seismicity in the region, as illustrated in Figure 1Bfor the time period 2000–2010. However, some faults havebeen inferred from geological mapping studies (Howard et al.2005; Pettinga et al. 2001) as well as from onshore and offshoreseismic reflection work (Dorn et al. 2010; Barnes 1995, 1996;Barnes et al. 2011; Wood et al. 1989).Following the destructive 22 February earthquake newstudies have begun to characterize the location and geometryof possible hidden faults beneath the region. This new workincludes a combination of offshore marine seismic surveys(Barnes et al. 2011), additional gravity acquisition and interpretation,new aeromagnetic surveys across sections of theCanterbury Plains (Figure 1B), some active-seismic reflectionsurveys, and relocation analysis of the existing earthquakeaftershock data (this study). Here we provide relocation analysisfor more than 2,100 aftershocks that have occurred since theM w 6.2 February earthquake. Separate analysis and processingof the aeromagnetic and reflection seismic survey data is underway.SEISMIC DATASeismic waveform data for the 22 February earthquake,and aftershocks, recorded by the New Zealand national seismographnetwork (Petersen et al. 2011) and the CanNet(Canterbury) strong-motion network (Avery et al. 2004) arepublicly available through GeoNet (http://www.geonet.org.nz).The strong-motion data coverage is excellent, with 15 strongground-motionrecorders (Figure 2) within 8 km of the top ofthe fault plane of the February M w 6.2 earthquake, 13 of theserecording vertical ground accelerations greater than 0.2 g.Many of the strong-motion recorders usually triggered foraftershocks of M L > 4 (90 events since 21 February to 1 May),doi: 10.1785/gssrl.82.6.839Seismological Research Letters Volume 82, Number 6 November/December 2011 839

(A)−43.25°0 10 20 kmM7.1−43.5°Mag 2Mag 3Mag 4Mag 5−43.75°172° 172.5° 173°(B)−43.25°Christchurch city−43.5°Canterbury Plains−43.75°Banks peninsula172° 172.5° 173°▲ ▲ Figure 1. A) Seismicity in the time period 3 September 2010 through April 30, 2011. The M w 7.1 Darfield and M w 6.2 FebruaryChristchurch earthquakes are marked as stars. Earthquakes occurring after February 21 are filled in black, earlier events are gray. B)Seismicity in the preceding 10 years, from 1 January 2000 to 1 September 2010, as cataloged by GeoNet, illustrating the low level ofseismicity before the 2010 Darfield earthquake.840 Seismological Research Letters Volume 82, Number 6 November/December 2011

−43.25°OXZ0 10 20km−43.5°−43.75°172° 172.5° 173°▲ ▲ Figure 2. Seismometer stations on the Canterbury Plains. Open triangles show the sites of temporary stations operating duringSeptember–October 2010; filled triangles show short-period and broadband seismometer sites; filled circles show strong-motion seismometersites.but also triggered for many of the 100+ M L > 3 aftershocksnear the city.Additional waveform data were also collected by a temporaryseismic array that was deployed immediately after the 22February event. This array (Figure 2), consisting of six shortperiodand three strong-motion seismometer sites, was placedto the south and north of Christchurch to provide greaterazimuthal coverage of the aftershock region and to assist withlocation and seismic tomography analysis.RELOCATION ANALYSISWe relocate aftershocks of the Christchurch M w 6.2 earthquakeusing the waveform and travel-time data currently available.We anticipate that additional waveform data currentlybeing collected (Yoshihisa Iio, personal communication 2011)will assist in further improving the regional velocity modelfor the Canterbury Plains (Figure 1B), with subsequent iterativeimprovement also expected for the aftershock locations.Additional travel-time picking by GeoNet analysts over thenext few months will also assist in iterative improvement of theaftershock locations.Initial event locations and phase picks were obtained fromGeoNet, with some additional travel time picking carried outfor the temporary stations deployed in late February. Most ofthe initial GeoNet event locations, before the relocation analysisdescribed here, have event depths at standard “fixed” depthsof 2, 5, or 12 km. Most of the larger-magnitude events thattriggered the strong-motion recorders had 20 or more phasepicks. Travel-time picking was carried out by GeoNet analysts.During phase picking it was noted that waveform data for someearthquake-station paths indicate multiple phase arrivals, inparticular to GeoNet station OXZ located approximately 40km to the west of Christchurch. This suggests strong multipatheffects from high-velocity basement or sub-basement structure.In our analysis we specifically downweight S-phase travel-timepicks from station OXZ to allow for possible ambiguities of theS-phase arrivals.We invert data from aftershocks recorded following theChristchurch M w 6.2 earthquake, using the double-differencetomography approach of Zhang et al. (2009), which buildson earlier work by Zhang and Thurber (2003). The techniqueminimizes the residuals between observed and calculatedarrival-time differences for pairs of closely located earthquakes,while also minimizing the residuals of absolute arrival times.The algorithm solves for the hypocentral parameters of theearthquakes, also allowing some modification of the P-waveand S-wave velocity structure used for travel-time calculation.Catalog-based differential times CBDT were calculatedbetween events initially separated by less than 10 km, for allSeismological Research Letters Volume 82, Number 6 November/December 2011 841

stations less than 150 km from Christchurch, using the manuallypicked P- and S-arrival times. Cross-correlation andbispectrum (BS) (Du et al. 2004) methods were then used tocalculate the waveform-based differential times (WBDT) forall event-station pairs, after pre-filtering, following the techniqueof Du et al. (2004). These derived differential times wereweighted based on the quality of the arrival time measurements.Absolute travel times and the two types of differentialtimes (CBDT and WBDT) were then combined and simultaneouslyinverted using the approach of Zhang et al. (2009)in an iterative least-squares procedure that utilizes the LSQRminimization method (Paige and Saunders 1982).In the first stage of the relocation analysis, the best 1,008events were used primarily to refine the velocity model in thevicinity of Christchurch. These events were selected fromthose earthquakes recorded in the six weeks following theDarfield M w 7.1 earthquake, as well as from the three weeksimmediately following the February M w 6.2 earthquake, whennew temporary seismometer stations were deployed south andnorth of Christchurch. Initial event selection was based onthe number of phases and initial standard errors. The datasetformed for this first stage of inversion contained 22,062 absolutephase times and 271,910 catalog-based differential times.The initial 3D velocity model was based on the most recent versionof the 3D New Zealand velocity model (Eberhart-Phillipset al. 2010). We interpolated the New Zealand–wide velocitymodel to a denser rectilinear grid centered on Christchurchcity using Delaunay triangulation and then carried out a seriesof inversion runs using the tomoDDPS approach of Zhang etal. (2009), slowly decreasing the inversion node spacing as theevent and station density allowed, with appropriate smoothingand weighting constraints (Zhang et al. 2009). The densestrong-motion array across Christchurch city, and the highnumber of events, led to high ray path coverage immediatelysouth of the city, and enabled quite a fine node spacing to beused close to the aftershocks. The velocity model after thisfirst stage had a minimum grid spacing of 1.5 km along the x-and y-axes in the section of the volume nearest Christchurchcity, and vertical nodes at 1, 2, 3, 4, 5.5, 8, 12, 24, and 30km depth, fully encompassing the aftershock volume for theChristchurch M w 6.2 earthquake. We expect further modificationand improvement of this velocity model in the next fewmonths as further waveform data is collected and analyzed(Yoshihisa Iio, personal communication 2011). Detailed resolutionanalysis, for example using checkerboard approaches,has yet to be carried out. However, we expect that the resolutionwill be quite spatially variable, reflecting the station andevent distribution, with the best resolution just south and westof Christchurch city, where the ray path coverage is highest.In a subsequent second stage we located 4,660 events,all with more than eight phase arrival observations, using thetomoDDPS software of Zhang et al (2009). This dataset consistedof 43,980 absolute phase times and 534,281 differentialtimes. Relocation was carried out using the velocity modeldeveloped in the first phase described above. Final epicentersare shown in Figure 3, and details of the relocated hypocentersfor February and March 2011 are provided in supplementarytable S1.DISCUSSIONFigure 3 shows the epicenters of the relocated events, whileFigures 4 and 5 show projections of the events onto verticalplanes AA′ (Figure 4) and BB′ (Figure 5).A feature of the relocated aftershocks is that they do notclearly define the fault plane of the 22 February earthquakeas determined from the centroid moment tensor (Sibson etal. 2011, page 824 of this issue) and geodetically (Beavan etal. 2011, page 789 of this issue); the projection of the geodeticallypreferred fault plane is overlain on cross-section AA′(Figure 4). Some vertical alignments of the aftershocks are visible,although they are not directly on the inferred M w 6.2 faultplane. One of these alignments, for example at X = 6–7 km onFigure 4, may be associated with a secondary strike-slip faultseveral kms south of the M w 6.2 earthquake, which has alsobeen inferred geodetically (Beavan et al. 2011, page 789 ofthis issue).The refined velocity model highlights P-wave velocities(V p ) greater than 6 km/s beneath part of the Banks Peninsula,as shallow as 5–8 km depth (Figures 4 and 5), at the samedepth range as the M w 6.2 February earthquake and the subsequentaftershocks. These high velocities are indicative ofschist basement, consistent with the interpretation of Wood etal. (1989). Previous passive- and active-source seismic work inthe Canterbury region by Reyners and Cowan (1993) and VanAvendonk et al. (2004) has also found V p greater than 6.0 km/sat 5–10 km depth, but with strong spatial variation. Many ofthe aftershocks appear to be located close to, or just below, thetop of this high-velocity basement.A possible reason for the lack of correspondence betweenaftershocks and the inferred fault plane for the M w 6.2 eventis that there may have been very little post-seismic slip on thefault. This explanation would be consistent with the high stressdrop of the mainshock (Fry and Gerstenberger 2011, page833 of this issue), which implies high fault friction. Lack oflarge-scale post-seismic deformation carries implications forthe mechanisms and rate of static stress transfer, and ultimatelythe longevity and the potential of the sequence to produce temporallyclustered aftershocks. This would help to explain therelatively high energy release observed for the September M w7.1 mainshock and aftershocks. In this scenario, the lack ofcontinual, shallow viscous deformation would encourage highlevels of regional stress build up, followed by large events withrelatively long aftershock sequences.ACKNOWLEDGMENTSWe thank GeoNet for providing seismological data, StaceyMartin and GeoNet staff for assisting with phase picking,and Banks Peninsula farmers for hosting temporary seismometers.We used GMT (Wessel and Smith 1998) and ObsPy(Beyreuther et al. 2010) for some figures.842 Seismological Research Letters Volume 82, Number 6 November/December 2011

A’0 km 4 8−43.5°B’B−43.6°A172.6° 172.8°Mag 2Mag 3Mag 4A’−43.5°Mag 5B’B−43.6°A172.6° 172.8°▲ ▲ Figure 3. Top: Epicenters for events occurring between 1 September 2010 and 31 January 2011. The locations of cross-sections AA’(Figure 4) and BB’ (Figure 5) are marked. Bottom: Epicenters of events occurring between 1 February 2011 and 30 April 2011, inclusive.The February M w 6.2 earthquake is marked as a solid star.Seismological Research Letters Volume 82, Number 6 November/December 2011 843

0Depth (kms)0 10Distance (kms)A A’Mag 2Mag 3Mag 4Mag 5Mag 6104 5 6km/s▲▲Figure 4. Aftershock locations for events between 1 February and 31 March 2011, projected onto the vertical plane AA’ of Figure3; only events within 2 km of the vertical plane are shown. The February M w 6.2 earthquake is at ~X = 8 km. The solid line shows theprojection of the fault plane 1 inferred from geodetic studies (Beavan et al. 2011, page 789 of this issue). Color background shows theP-wave velocity.0Distance (kms)0 10 20B B’Mag 2Depth (kms)Mag 3Mag 4Mag 5Mag 6104 5 6km/s▲ ▲ Figure 5. Aftershock locations for events between 1 February and 30 April 2011, projected onto the cross-section BB ’ of Figure 3—avertical plane with strike N79.5E degrees; the fault plane 2 of Beavan 2011, page 789 of this issue. Only events within 1.2 km of theplane are shown. Background colors show P-wave velocity; areas that have lower ray path coverage (derivative-weighted-sum lessthan 50) are grayed out. The February M w 6.2 earthquake is at ~X = 12 km. Earthquake symbol sizes are as for Figure 4.844 Seismological Research Letters Volume 82, Number 6 November/December 2011

REFERENCESAvery, H. R., J. B. Berrill, P. F. Coursey, B. L. Deam, M. B. Dewe, C. C.Francois, J. R. Pettinga, and M. D. Yetton (2004). The CanterburyUniversity strong-motion recording project. 13 th World Conferenceon Earthquake Engineering, Vancouver, B.C., Canada, 1–6 August,2004, paper 1335.[are these published proceedings? can we providename of publisher & place where published?]Barnes, P. M. (1995). High-frequency sequences deposited duringQuaternary sea-level cycles on a deforming continental shelf, northCanterbury, New Zealand. Sedimentary Geology 97, 131–156.Barnes, P. M. (1996). Active folding of Pleistocene unconformities on theedge of the Australian-Pacific plate boundary zone, offshore northCanterbury, New Zealand. Tectonics 15 (2), 623–640. Correctionto Figure 6 printed in Tectonics 15 (5), 1,110–1,111 [1996].Barnes, P., C. Castellazzi, A. Gorman and S. Wilcox (2011). SubmarineFaulting beneath Pegasus Bay, Offshore Christchurch. Short-termCanterbury Earthquake Recovery Project 2: Offshore Faults.National Institute of Water & Atmospheric Research Client reportWLG2011-28, Wellington, NZ: National Institute of Water andAtmospheric Research.Beavan, J., E. Fielding, M. Motagh, S. Samsonov, and N. Donnelly(2011). Fault location and slip distribution of the 22 February 2011M W 6.2 Christchurch, New Zealand, earthquake from geodeticdata. Seismological Research Letters 82, 789–799.Beavan, J., S. Samsonov, M. Motagh, L. Wallace, S. Ellis and N. Palmer(2010). The M w 7.1 Darfield (Canterbury) earthquake: Geodeticobservations and preliminary source model. Bulletin of the NewZealand Society for Earthquake Engineering 43, 228–235.Beyreuther, M., R. Barsch, L. Krischer, T. Megies, Y. Behr, and J.Wassermann (2010). ObsPy: A python toolbox for seismology.Seismological Research Letters 81 (3), 530–533.Dorn, C., A. G. Green, R. Jongens, S. Carpentier, A. E. Kaiser, F.Campbell, H. Horstmeyer, J. Campbell, M. Finnemore, and J.Pettinga (2010). High-resolution seismic images of potentially seismogenicstructures beneath the northwest Canterbury Plains, NewZealand. Journal of Geophysical Research, Solid Earth 115, B11303;doi:10.1029/2010JB007459.Du, W., C. H. Thurber, and D. Eberhart-Phillips (2004). Earthquakerelocation using cross-correlation time delay estimates verifiedwith the bispectrum method. Bulletin of the Seismological Society ofAmerica 94, 856–866.Eberhart-Phillips, D., M. Reyners, S. Bannister, M. Chadwick, and S.Ellis (2010). Establishing a versatile 3-D seismic velocity model forNew Zealand. Seismological Research Letters 81 (6), 992–1,000;doi:10.1785/gssrl.82.6.992.Fry, B., R. Benites, M. Reyners, C. Holden, A. Kaiser, S. Bannister, M.Gerstenberger, C. Williams, J. Ristau, and J. Beavan (2011). Verystrong shaking in New Zealand earthquakes. Eos, Transactions,American Geophysical Union.Fry, B., and M. Gerstenberger (2011). Large apparent stresses from theCanterbury earthquakes of 2010 and 2011. Seismological ResearchLetters 82, 833–838.Gledhill, K., J. Ristau, M. Reyners, B. Fry, and C. Holden (2011).The Darfield (Canterbury, New Zealand) Mw 7.1 earthquake ofSeptember 2010: A preliminary seismological report. SeismologicalResearch Letters 82, 378–386.Holden, C. (2011). Kinematic source model of the 22 February 2011M w 6.2 Christchurch earthquake using strong motion data.Seismological Research Letters 82, 783–788.Howard, M., A. Nicol, J. Campbell, and J. R. Pettinga (2005). Holocenepaleoearthquakes on the strike-slip Porters Pass Fault, Canterbury,New Zealand. New Zealand Journal of Geology and Geophysics 48(1), 59–74.Paige, C., and M. Saunders (1982). LSQR: An algorithm for sparse linearequations and sparse least squares problems. ACM Transactions onMathematical Software 8, 43–71.Petersen, T., K. Gledhill, M. Chadwick, N. Gale, and J. Ristau (2011).The New Zealand National Seismograph Network. SeismologicalResearch Letters 82, 9–20.Pettinga, J. R., M. D. Yetton, R. J. Van Dissen, and G. Downes (2001).Earthquake source identification and characterisation for theCanterbury region, South Island, New Zealand. Bulletin of the NewZealand Society for Earthquake Engineering 34, 282–317.Quigley, M., R. Van Dissen, P. Villamor, N. Litchfield, D. Barrell, K.Furlong, T. Stahl, et al. (2010). Surface rupture of the Greendalefault during the M w 7.1 Darfield (Canterbury) earthquake, NewZealand: Initial findings. Bulletin of the New Zealand Society forEarthquake Engineering 43, 236–242.Reyners, M. E., and H. Cowan (1993). The transition from subductionto continental collision: Crustal structure in the North Canterburyregion, New Zealand. Geophysical Journal International 115 (3),1,124–1,136.Sibson, R., F. Ghisetti, and J. Ristau (2011). Stress control of an evolvingstrike-slip fault system during the 2010–2011 Canterbury, NewZealand, earthquake sequence. Seismological Research Letters 82,824–832.Van Avendonk, H. J. A., W. S. Holbrook, D. Okaya, J. Austin, F. Davey,and T. Stern (2004). Continental crust under compression: A seismicrefraction study of SIGHT Transect 1, South Island, New Zealand.Journal of Geophysical Research 109; doi:10.1029/2003JB002790.Wessel, P., and W. H. F. Smith (1998). New, improved version of theGeneric Mapping Tools released. Eos, Transactions, AmericanGeophysical Union 79, 59.Wood, R. A., P. B. Andrews, and R. H. Herzer, R. A. Cook, N. de B.Hornibrook, R. H. Hoskins, A. G. Beu, et al. (1989). Cretaceousand Cenozoic Geology of the Chatham Rise Region, South Island,New Zealand. New Zealand Geological Survey Basin Studies 3,75 pp. Lower Hutt, New Zealand: New Zealand Geological Survey.Zhang, H., and C. H. Thurber (2003). Double-difference tomography:The method and its application to the Hayward fault, California.Bulletin of the Seismological Society of America 93, 1,875–1,889.Zhang, H., C. Thurber, and P. Bedrosian (2009). Joint inversion forVp, Vs, and Vp/Vs at SAFOD, Parkfield, California. Geochemistry,Geophysics, Geosystems 10 (1), Q11002.GNS Science1 Fairway Drive, AvalonLower Hutt 5040 New Zealands.bannister@gns.cri.nz(S. B.)Seismological Research Letters Volume 82, Number 6 November/December 2011 845

The Character of Accelerations in the Mw 6.2Christchurch EarthquakeB. Fry, R. Benites and A. KaiserB. Fry, R. Benites and A. KaiserGNS ScienceINTRODUCTIONThe Canterbury earthquakes of 2010 and 2011 have producedsome of the strongest ground motions ever measuredin New Zealand. Many of the highest acceleration recordingsarose from seismic stations within the city of Christchurch(population ~377,000). A dense array of strong-motion seismometerswas in place prior to the mainshock of 4 September2010. Subsequent to the mainshock, numerous rapid responseaccelerometers were installed in the Canterbury Plains, BanksPeninsula, and in the city itself (Gledhill et al. 2011; Cochranet al. 2011). Many of the strongest aftershocks were recordedby this dense amalgamation of permanent and temporaryarrays and provide a detailed record of variable ground motionthroughout the region during the aftershock sequence.The most extreme ground motions were recorded duringthe Mw 6.2 earthquake of 22 February 2011 that struck a fewkilometers to the south of Christchurch (Beavan et al. 2011,this issue; Holden et al. 2011, this issue; Bannister et al. 2011,this issue), generating severe damage throughout the city. Infact, damage to the built environment and ground liquefactionwas much more widespread in the February event than in theSeptember Mw 7.1 mainshock (Kaiser et al. 2011). This event isone of the best-recorded shallow thrust earthquakes in the nearfield. Recorded peak ground acceleration (PGA) in Februaryexceeded 2 g near the epicenter and was greater than 0.6 g overmuch of the central and eastern suburbs (Figure 1). At thesenear-source stations, vertical accelerations were generally markedlyhigher than horizontal accelerations. The large accelerationscan be reasonably well explained by the combination ofHPSCCHHCPRPCHVSCCCCC▲ ▲ Figure 1. Vertical and horizontal acceleration vectors and their location relative to the February 22 epicenter (green star). Greenhorizontal lines on each waveform show the baseline.846 Seismological Research Letters Volume 82, Number 6 November/December 2011 doi: 10.1785/gssrl.82.6.846

▲ ▲ Figure 2. A) shows three-component accelerogram from station PRPC (see Figure 1 for location) for the 22 February event, scaledto units g. B) shows their corresponding spectra.the proximity of the February event to Christchurch and theeffects of strong source directivity. However, not all featurescan be explained by source effects alone. The dense near-sourcedata from the 22 February earthquake have provided us with avaluable opportunity to study the response of the shallow subsurfaceto extreme ground motions in very fine detail.The way ground responds to an earthquake is a result ofthe earthquake rupture process, the path that the waves takebetween the source and the surface, and the response of theshallow materials below the ground. We know that the topfew meters of the ground in Christchurch played an importantrole in the shaking. This role is evident in low-frequencysignals resulting from liquefaction. Many of the poorly consolidated,low shear-wave velocity soils liquefied at shallowdepths with less than 0.1 g peak horizontal accelerations andexperienced deep liquefaction at around 0.2 to 0.3 g accelerations.The influence of the shallow subsurface is also exhibitedby the existence of energetic high-frequency signals resultingfrom the interaction of the waves with both the water table andunconsolidated soils prior to liquefaction. A marked featureof the strong-motion seismograms recorded at several nearsourcesites in Christchurch during the earthquake sequence isthe much higher frequency content of the vertical componentcompared to the corresponding horizontal recordings (Figure2). We believe that this phenomenon is due to the presence ofa shallow water table dramatically attenuating the propagationof high-frequency shear waves. A rigorous numerical demonstrationof such water effect would require the calculation ofseismic wave propagation in layered media in which one of thelayers is a porous, elastic solid containing an incompressible,inviscous fluid (Biot 1956a, Biot 1956b). However, for thiswork, we simulate such an effect in an indirect way by model-Seismological Research Letters Volume 82, Number 6 November/December 2011 847

TABLE 11-D layered model in the near field of the M 6.2 February 4,2011 Canterbury earthquake. The fault is embedded in thesecond layer. H-S denotes the half-space underneath thelayering.Thickness (km) Vp (km/s) Vs (km/s) ρ (g/cm 3 )0.02 0.3 0.165 1.50.98 4.5 2.6 1.89.0 5.6 3.2 2.4H-S 6.24 3.6 2.9ing the effect of attenuation of shear waves on the accelerationseismograms. We define a 1-D layered medium representingthe Canterbury region close to the fault that ruptured in theMw 6.2 February 2011 aftershock based on a regional subsetof the New Zealand–wide 3-D velocity model (Eberhart-Phillips et al. 2010) and shallow velocity structure determinedby microtremor analysis (Stephenson et al. 2011). The fault isrepresented by a 9-km-long and 8-km-wide rectangular thrustfault dipping 65° and striking 70° (from north), rupturing with3.2 km/s rupture velocity and 120° rake. This fault is embeddedin the second layer of a four-layered medium of elastic parameterslisted in Table 1. The modeling is accomplished with a discretewavenumber numerical scheme (Bouchon 1979) in whichattenuation is introduced by the factor exp(2π*f / Q s ) for eachplane wave in the layer (Aki and Richards 1980), where f is frequencyand Q s is the shear wave quality or attenuating factor.We apply attenuation to the top 20-m-thick soil layer to testthe hypothesis that the observed differences in frequency contentbetween the horizontal and vertical components is due tostrong shear wave attenuation in the shallow subsurface. Ideally,modeling the anelastic attenuation would include Biot’s modelof fluid-solid interaction in our wave propagation algorithm.Geli et al. (1987) successfully incorporated Biot’s model in theAki-Larner (1970) numerical technique to study the responseof smooth 2-D basins with water-saturated sediments. Theirwork prescribes a characteristic frequency ( fc) of about 5 Hzfor unconsolidated coarse sands and gravels with a permeabilityon the order of 10 –8 m 2 . This frequency depends on fluidand solid parameters such as permeability, viscosity, density,and pore density. For frequencies lower than fc, the attenuationof shear waves is stronger than for primary waves, and for frequenciesabove fc, the attenuation of primary P waves increasessignificantly with frequency due to the rise of a secondary Pwave (termed P 2 ) that results from the solid-fluid coupling.Based on an anelastic representation of attenuation in whichvelocities are estimated from the fluid and solid parameters,Geli et al. (1987) prescribe the dependency of Q on frequency,showing only weak dependency in the range 1–10 Hz for permeabilityand porosity in the range of the shallow subsurface ofChristchurch (10 –8 and 35%, respectively). We therefore applyonly shear wave attenuation (between Q s = 1 and Q s = 10 for0 Hz–7 Hz) to our range of simulations.The top panel of Figure 3 shows the three components ofdisplacement and acceleration at a station located 2 km fromthe fault, when there is no attenuation (Q s = 1,000). Although(A)(B)▲ ▲ Figure 3. Synthetic wavefield calculated with a discrete wavenumber method at a distance of 2 km from the 22 February rupture.U is east-west, V is north-south, and W is vertical. A) Results from a simulation with zero attenuation. B) Results from inclusion of aquality factor to represent the attenuating effects of the local groundwater table.848 Seismological Research Letters Volume 82, Number 6 November/December 2011

▲ ▲ Figure 4. Vertical acceleration waveforms from Figure 1. Waveforms show larger positive accelerations than negative ones. Manyof the negative acceleration troughs are also broader than the narrow positive acceleration spikes.the high acceleration may be unrealistically large (about 5 g),we recall that this is the result of a kinematic model combiningthe effect of directivity and the presence of the thin (20 m), softtop layer in the near field. The bottom panel of Figure 3 showsthe results for the case of Q s = 5. Noticeably, high frequencieshave been attenuated in the horizontal components but notin the vertical component. This is compatible with numerousobservations of this characteristic recorded during the aftershocksequence (e.g., Figure 2).VERTICAL COMPONENT ASYMMETRYAnother notable characteristic of many of the recordings fromthe February event and some recordings from other strong localevents is the occurrence of asymmetric accelerations and spikesin the vertical direction. In this observational paper, we documentan example of a newly discovered phenomenon in thevertical components of the acceleration seismograms that haspreviously been recognized in a handful of records from strongshallow earthquakes and nuclear explosions. Many accelerogramsrecorded in the Mw 6.2 earthquake exhibit maximumPGA on the vertical component (Figure 1). The asymmetricalrecordings are confined to within ~6–10 km of the epicenter,suggesting that either very strong near-field motions are necessaryto generate them or that they result from high-frequencywaves generated during source processes that subsequentlyattenuate at greater distances. Most of the high-accelerationvertical records are asymmetric with maximum accelerationsin the upward direction (>1 g) exceeding accelerations in thedownward direction (

4 Sep Mw 7.118 Oct Mw 5.013 Jun Mw 5.613 Jun Mw 6.022 Feb Mw 6.2▲ ▲ Figure 5. Vertical accelerograms from station PRPC for differentevents sorted from minimum PGA in the top panel tomaximum PGA in the bottom panel. Gray region marks twice theminimum acceleration centered about the baseline.Miyagi Nairiku earthquake of 2008 (Aoi et al., 2008; Yamadaet al., 2009). Similar asymmetric recordings from the Mw 6.9Iwate-Miyagi Nairiku earthquake of 2008 have been attributedto a “trampoline” effect (Aoi et al. 2008; Yamada et al. 2009).Aoi et al. (2008) attribute the asymmetry to the decouplingof near-surface materials during high-amplitude downwardacceleration. This occurs when the tensile forces that arise onan interface or within a granular material from downgoing particleoscillation as waves pass are larger than its tensile strength.The result is an approximate free-fall of the material. In thismodel, the high upward accelerations are caused by the compressionalresponse of the granular media to the stress of theupgoing particle oscillation. Yamada et al. (2009) suggest thelarge positive accelerations are further enhanced by “slapdown”as free-falling upper soil layers impact/interact with deeper layersthat are returning upward during the following earthquakewave cycles. Recently conducted finite element numerical modelingsupports the hypothesis that the asymmetry arises due tothe difference in response of near-surface layers to compressionand tension (Tobita et al. 2010). Figure 5 shows a series of verticalcomponent recordings from station PRPC in Christchurch.The recorded vertical acceleration from the February eventat this station was most similar to asymmetric vertical accelerationrecorded in the Iwate-Miyagi event. PRPC appears tohave experienced vertical acceleration asymmetry in numerousevents over a range of peak accelerations. Recordings from the13 June 2011 Mw 6.0 and Mw 5.6 events both exhibit slightasymmetry with PGAs as low as ~0.7 g. However, records fromthe 4 September Mw 7.1 earthquake do not appear asymmetric.Ground motions recorded at PRPC in this event reach 0.32 g.This suggests that a threshold for the non-linear effect thatcauses the asymmetry is between about 0.3 and 0.7 g at this site.In pure granular material, Tobita et al. (2010) show thatthe depth of initiation of the non-linear response is proportionalto the experienced ground acceleration. The frequencyof the response is also likely to be at least partially controlledby the depth of origin of the strong asymmetry. We will focuson station PRPC for a description of the time and frequencyof asymmetry. To quantify trace asymmetry, we normalize thedifference between the maximum (a max ) and minimum (a min )accelerations for time-windowed data (i.e., (a max – abs(a min ))/(max(abs(a max ,a min )))). Onset of dominant asymmetry occursproportionately earlier in the time series for very near-sourcestations and proportionately later for more distant stations. Atstation PRPC, onset of dominant asymmetry occurs at about3 s into the waveform (Figure 6) whereas at station HVSC,within a couple of kilometers of the rupture, onset is within thefirst second. We analyze the asymmetry of different frequenciesby applying the time-windowing to bandpass filtered data. Wedefine frequency bands in the data between 0.2–1 Hz, 1–2 Hz,2–5 Hz, 5–10 Hz, and 10–30 Hz. Asymmetry at stationPRPC is dominated by the 2–5 Hz band (Figure 6). However,we use a window half-width of 0.25 s for all considered bands.We recognize that relatively more wave cycles will contribute tothe time-window for the higher frequency bands than for thelower frequency ones. However, the asymmetry in the 2–5 Hzband is bracketed by the more symmetric 1–2 Hz and 5–10 Hzbands. This supports our suggestion that the 2–5 Hz band ofthe signal dominates the asymmetry.CONCLUSIONSThe Mw 6.2 Christchurch earthquake generated a wealth ofnear-field strong-motion data. Analysis of these data suggests850 Seismological Research Letters Volume 82, Number 6 November/December 2011

0.8PRPC Z0.40Normalized (Max PGA + Min PGA)-0.4-0.80.80.40 1 2 3 4 5 6 7 8 9 100.2 - 1 Hz1 - 2 Hz2 - 5 Hz5 - 10 Hz10 - 30 Hz0-0.4-0.80 1 2 3 4 5 6 7 8 9 10time (s)▲▲Figure 6. A) Normalized sum of maximum and minimum PGA for windows of the vertical recording at PRPC. Windows have a half-widthof 0.25 seconds and are sampled every 0.1 s. Bottom) Data windowed as above after band-pass filtering. The band between 2–-5 Hz (darkblue) is dominantly positively asymmetric between about 2.5 and 7.5 seconds (shown as shaded region).that non-linear effects in the near-surface layers such as thoseproposed by Aoi et al. (2008) and Yamada et al. (2009) are presentin many of the urban recordings. These are manifest in asymmetryin the vertical recordings. We find asymmetric behaviorin records from smaller events as well, and bracket initiation ofvertical component acceleration asymmetry at between 0.3 and0.7 g. Furthermore, a multiple-filter approach to analyzing thetime signals recorded at one of the eastern city stations suggeststhat the dominant bandwidth carrying the asymmetric accelerationsis between 2 and 5 Hz. Further analysis of these datain conjunction with detailed site response analyses and numericalmodeling has the potential to greatly increase our understandingof this phenomenon. We have noted a difference inthe frequency content of the recorded vertical and horizontalaccelerations from the earthquake. Based on numerical modeling,we propose that the lack of high-frequency energy on thehorizontal components can largely be attributed to shear waveattenuation in the shallow, water-saturated sediments. Futurework is aimed at fully implementing Biot’s theory into themodeling, explicitly describing the complete fluid interactionof the water-saturated soils in Christchurch.Many of the strong-motion recordings have vertical componentsthat are rich in high-frequency energy and horizontalcomponents that are dominated by low-frequency waves. Wemodel attenuation with a discrete wavenumber method as aproxy for this effect. The characteristic difference in energycontent appears to be the result of attenuation of high-frequencyshear waves in the shallow subsurface.REFERENCESAki, K., and K. L. Larner (1970). Surface motion of a layered mediumhaving an irregular interface due to incident plane SH waves.Journal of Geophysical Research 75, 933–954.Aki, K., and P. G. Richards (1980). Quantitative Seismology. SanFrancisco: W. H. Freeman and Company, 556 pp.Aoi, S., T. Kunugi, and H. Fujiwara (2008). Trampoline effect in extremeground motion. Science 322 (5902), 727–730, doi:10.1126/science.1163113.Seismological Research Letters Volume 82, Number 6 November/December 2011 851

Bannister, S., B. Fry, M. Reyners, J. Ristau, and H. Zhang (2011).Fine-scale relocation of aftershocks of the 22 February M w 6.2Christchurch earthquake using double-difference tomography.Seismological Research Letters, this issue.Beavan, J., E. Fielding, M. Motagh, S. Samsonov, and N. Donnelly(2011). Fault location and slip distribution of the 22 February 2011M W 6.2 Christchurch, New Zealand, earthquake from geodeticdata. Seismological Research Letters, this issue.Beavan, R. J., S. Samsonov, M. Motagh, L. M. Wallace, S. M. Ellis, andN. Palmer (2010). The Darfield (Canterbury) earthquake: Geodeticobservations and preliminary source model. Bulletin of the NewZealand Society for Earthquake Engineering 43 (4), 228–235.Biot, M. A. (1956a). Theory of propagation of elastic waves in a fluid saturatedporous solid. I. Low frequency range. Journal of the AcousticalSociety of America 28, 168–178; doi:10.1121/1.1908239.Biot, M. A. (1956b). Theory of propagation of elastic waves in a fluidsaturated porous solid. II. Higher frequency range. Journal of theAcoustical Society of America 28, 179–191; doi:10.1121/1.1908241.Bouchon, M. (1979). Discrete wave number representation of elasticwave fields in three-space dimensions. Journal of GeophysicalResearch 84, 3,609–3,614.Cochran, E., J. Lawrence, A. Kaiser, B. Fry, A. Chung, and C.Christensen (2011). Comparison between low-cost and traditionalMEMS accelerometers: A case study from the M7.1 Darfield, NewZealand aftershock deployment.Eberhart-Phillips, D., M. E. Reyners, S. C. Bannister, M. P. Chadwick,and S. M. Ellis, (2010). Establishing a versatile 3-D seismic velocitymodel for New Zealand. Seismological Research Letters 81 (6),992–1,000; doi:10.1785/gssrl.82.6.992.Geli, L., P. Bard, and D. Schmitt (1987). Seismic wave propagation in avery permeable water-saturated surface layer. Journal of GeophysicalResearch 92, 7,931–7,944.Gledhill, K., J. Ristau, M. E. Reyners, B. Fry, and C. Holden (2011).The Darfield (Canterbury, New Zealand) M w 7.1 earthquake ofSeptember 2010: A preliminary seismological report. SeismologicalResearch Letters 82 (3), 378–386; doi:10.1785/gssrl.82.6.378.Holden, C. (2011). Kinematic source model of the 22 February 2011M w 6.2 Christchurch earthquake using strong motion data.Seismological Research Letters this issue.Kaiser, A., C. Holden, J. Beavan, D. Beetham, R. Benites, A. Celentano,D. Collett et al. (2011). The February 2011 Christchurch earthquake:A preliminary report. Submitted to New Zealand Journal ofGeology and Geophysics.Stephenson, B., P. Barker, Z. Bruce, and D. Beetham (2011). ImmediateReport on the Use of Microtremors for Assessing LiquefactionPotential in the Christchurch Area. GNS Science Report 2010/30,26 pp. Lower Hutt, New Zealand: GNS Science.Tobita, T., I. Susumu, and T. Iwata (2010). Numerical analysis of nearfieldasymmetric vertical motion. Bulletin of the SeismologicalSociety of America 100, 1,456–1,469.Yamada, M., J. Mori, and T. Heaton (2009). The slapdown phase in highaccelerationrecords of large earthquakes. Seismological ResearchLetters 80, 559–564.GNS Science1 Fairway DriveLower Hutt, New Zealandb.fry@gns.cri.nz(B. F.)852 Seismological Research Letters Volume 82, Number 6 November/December 2011

Near-source Strong Ground MotionsObserved in the 22 February 2011 ChristchurchEarthquakeBrendon A. Bradley and Misko CubrinovskiBrendon A. Bradley and Misko CubrinovskiUniversity of CanterburyINTRODUCTIONOn 22 February 2011 at 12:51 p.m. local time, a momentmagnitude M w 6.3 earthquake occurred beneath the cityof Christchurch, New Zealand, causing an level of damageand human casualties unparalleled in the country’s history.Compared to the preceding 4 September 2010 M w 7.1 Darfieldearthquake, which occurred approximately 30 km to the westof Christchurch, the close proximity of the 22 February eventled to ground motions of significantly higher amplitude in thedensely populated regions of Christchurch. As a result of thesesignificantly larger ground motions, structures in general, andcommercial structures in the central business district in particular,were subjected to severe seismic demands and, combinedwith the event timing , structural collapses accounted for themajority of the 181 casualties (New Zealand Police 2011).This manuscript provides a preliminary assessment of thenear-source ground motions recorded in the Christchurchregion. Particular attention is given to the observed spatialdistribution of ground motions, which is interpreted based onsource, path, and site effects. Comparison is also made of theobserved ground motion response spectra with those of the 4September 2010 Darfield earthquake and those used in seismicdesign in order to emphasize the amplitude of the ground shakingand also elucidate the importance of local geotechnical anddeep geologic structure on surface ground motions.There are numerous identified faults in the Southern Alpsand eastern foothills (Stirling et al. 2007) and several significantearthquakes (i.e., M w > 6) have occurred in this regionin the past 150 years, most notably the M w 7.1 Darfield earthquakeon 04/09/2010 (New Zealand Society for EarthquakeEngineeering 2010). The M w 6.3 Christchurch earthquakeoccurred at 12:51 p.m. on Tuesday 22 February 2011 beneathChristchurch, New Zealand’s second-largest city, and representsthe most significant earthquake in the unfolding seismicsequence in the Canterbury region since the 4 September2010 Darfield earthquake. The 6.3 event occurred on a previouslyunrecognized steeply dipping blind fault, which trendsnortheast to southwest (the location relative to Christchurchis presented in the context of subsequently observed groundmotions). Figure 2 illustrates the inferred slip distribution onthe fault obtained by Beavan et al. (2011, page 789 of thisTECTONIC AND GEOLOGIC SETTINGNew Zealand resides on the boundary of the Pacific andAustralian plates (Figure 1) and its active tectonics are dominatedby: 1) oblique subduction of the Pacific plate beneaththe Australian plate along the Hikurangi trough in the NorthIsland; 2) oblique subduction of the Australian plate beneaththe Pacific plate along the Puysegur trench in the southwestof the South Island; and 3) oblique, right-lateral slip alongnumerous crustal faults in the axial tectonic belt, of which the650-km-long Alpine fault is inferred to accommodate approximately70–75% of the approximately 40 mm/yr plate motion(DeMets et al. 1994; Sutherland et al. 2006).▲ ▲ Figure 1. Tectonic setting of New Zealand (courtesy of J.Pettinga).doi: 10.1785/gssrl.82.6.853Seismological Research Letters Volume 82, Number 6 November/December 2011 853

▲ ▲ Figure 2. Distribution of fault slip inferred in the 22 February 2011 Christchurch earthquake (Beavan et al. 2011, this issue). Arrowsindicate the slip vector. The inferred hypocenter is indicated by a star.issue). It can be seen that slip on the fault occurred obliquelywith both significant up-dip and along-strike components(average rake, λ = 146°). The steeply dipping nature of the fault(δ = 69°), as well as the large up-dip component of slip, contributedto the large observed vertical accelerations discussed inthe next section. For the purpose of the subsequent engineeringanalysis of strong ground motion, the Beavan et al. (2011, page789 of this issue) finite fault model was “trimmed” using themethodology of Somerville et al. (1999), which resulted in theremoval of 1 km from the northeast and southwest extents ofFigure 2. The resulting “trimmed” fault therefore has dimensionsof 15 km along-strike and 8 km down-dip, giving a totalarea of 120 km 2 .Christchurch is located on the Canterbury Plains, a fandeposit resulting from the numerous rivers flowing eastwardfrom the foothills of the Southern Alps (Brown and Weeber1992). In the vicinity of Christchurch, the Canterbury Plains arecomprised of a complex sequence of gravels interbedded with silt,clay, peat, and shelly sands. The fine sediments form aquicludesand aquitards between the gravel aquifers, and with the nearbycoastline to the east, result in the majority of Christchurch havinga water table less than 5 m depth, with the majority of thearea including, and to the east of, the central business districthaving a water table less than 1 m from the surface (Brown andWeeber 1992). The postglacial Christchurch Formation createdby estuarine, lagoonal, dune, and coastal swamp deposits (containinggravel, sand, silt, clay, shell, and peat) is the predominantsurface geology layer in the Christchurch area, which outcropsup to 11 km west of the coast and has a depth of approximately40 km along the coast itself (Brown and Weeber 1992). At thesoutheast edge of Christchurch lies the extinct Banks Peninsulavolcanic complex.STRONG MOTION RECORD PROCESSINGVolume 1 ground motion records were obtained from GeoNet(http://www.geonet.org.nz/) and processed on a record-byrecordbasis. The overall processing methodology adopted iselaborated in Chiou et al. (2008, Figure 4). All ground motionswere processed with a low-pass causal Butterworth filter of50 Hz, and while the corner frequency of the high-pass filterwas record-specific, a frequency of less than 0.05 Hz providedphysically realistic Fourier spectra amplitudes and integrateddisplacement histories for all the near-source ground motions.Owing to the digital nature of all of the instruments, baselinecorrections were found to be unnecessary following theabove filtering. As a result, the processed ground motions canbe considered to provide reliable estimates of peak groundaccelerations (PGA) and spectral ordinates over the range0.01–10 seconds (Douglas and Boore 2010), which are typicallyof engineering interest. It should be noted that the aboveprocedure does not lend itself to the computation of residualdisplacements, which may be non-zero for near-source locations.However, as a result of possible instrument tilting, whichmay be significant at sites where liquefaction occurred, reliablecomputation of such residual displacements may not be possible(Graizer 2005) and is left for future study.854 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A)Magnitude, M w876542009 NZdatabase04/09/201022/02/201110 0 10 1 10 2Distance, R rup(km)(B)Number of exceedances2009 NZdatabase04/09/201022/02/2011Total10 110 00.2 0.4 0.6 0.8 1 1.2 1.4102Peak ground acceleration, PGA (g)▲ ▲ Figure 3. Significance of the 22 February 2011 Christchurch and 4 September 2010 Darfield earthquakes in relation to previouslyrecorded ground motions in New Zealand: A) magnitude-distance distribution; B) exceeded values of peak ground acceleration.While New Zealand can be considered as a region of highseismicity in a global context, prior to the 4 September 2010Darfield and 22 February 2011 Christchurch earthquakesthere was a paucity of high-amplitude recorded strong groundmotions, primarily as a result of a sparse instrumentation networkbefore the commencement of GeoNet in 2001. Figure3 illustrates the magnitude-distance distribution of recordedground motions from active shallow crustal earthquakes up to2009 as complied by Zhao and Gerstenberger (2010). Also illustratedin Figure 3A are the ground motions recorded in both the4 September 2010 Darfield and 22 February 2011 Christchurchearthquakes. The significance of the recorded ground motionsin these two earthquakes is even more apparent if the groundmotions in Figure 3A are plotted in terms of their geometricmean horizontal PGA. Figure 3B illustrates the number ofground motions exceeding specific values of PGA. It can beseen that up to 2009, the maximum PGA recorded in NewZealand was 0.39 g, with only seven observed ground motionsexceeding 0.2 g PGA. Figure 3B also illustrates the exceedancevalues observed in the Darfield and Christchurch earthquakes.With the addition of these two events (not to mention recordsobtained from numerous significant aftershocks, which arenot discussed herein) it can be seen that ground motions of upto 1.41 g have now been recorded, with 12 observed groundmotions exceeding 0.4 g and 39 exceeding 0.2 g.OBSERVED NEAR-SOURCE GROUND MOTIONSTable 1 presents a summary of the ground motions in the widerChristchurch region that were recorded within a source-to-sitedistance of R rup = 20 km, including: station site class accordingto the New Zealand loading standard, New Zealand Standards1170.5 (2004); PGA, peak ground velocity (PGV); 5–95%significant duration, (D s5–95 ) (Bommer and Martinez-Pereira1999) for geometric mean horizontal component; and peakvertical ground acceleration (PGA v ).Figures 4–6 illustrate the spatial distribution of accelerationtime histories recorded at the aforementioned strongmotion stations in the form of fault-normal, fault-parallel,and vertical components, respectively. The aforementioned“trimmed” finite fault model of Beavan et al. (2011, page 789of this issue) is also shown. The following sections discuss variousaspects of the ground motions observed in Figures 4–6.Ground Motion on Rock and Soil SitesIn interpreting the observed ground motions in Figures 4–6, itis first worth noting that only the Lyttelton Port (LPCC) stationto the southeast of Christchurch is located on engineeringbedrock (i.e., site class B). Stations HVSC and LPOC locatednear the edge of the Port Hills rock outcrop are site class C,while all remaining stations are situated on the Christchurchsedimentary basin and are predominantly site class D (theexceptions being HPSC, NNBS, PRPC, and KPOC, which aresite class E). Unfortunately at present the site characterizationof strong motion stations in the Christchurch region, and NewZealand in general, is relatively poor with the above site classesdetermined from geological maps and direct surface inspection(N. Perrin, personal communication 2011), and details suchas P- and S-wave velocity, SPT, and CPT data not available.Clearly, obtaining such information is necessary for a rigorousanalysis of the observed ground motions, and is the focus ofimmediate studies. Nonetheless, a wealth of insight can still beobtained from inspection and analysis of the observed groundmotions.A direct comparison of the effect of soil and rock site canbe made by comparing the ground motions observed at LPCCand LPOC located at Lyttelton Port approximately 1 km apart.The LPCC instrument is located on engineering bedrock, andwhile detailed information of the site conditions at LPOC arepresently unavailable, it is said to be a relatively thin (~30 m)colluvium layer comprised primarily of silt and clay (J. Berrill,personal communication 2011). In addition to a comparison ofSeismological Research Letters Volume 82, Number 6 November/December 2011 855

Station NameTABLE 1Strong Motion Stations and Near-source Recordings of the 22 February 2011 Christchurch EarthquakeCodeSiteclass * R jb†(km)R rup‡(km)PGA §(g)PGV ||(cm/s)D s5–95#(s)PGA v**(g)Canterbury Aero Club CACS D 12.7 12.8 0.21 20.0 11.8 0.19Christchurch Botanic Gardens CBGS D 4.6 4.7 0.50 46.3 10.7 0.35Christchurch Cathedral College CCCC D 2.6 2.8 0.43 56.3 9.8 0.79Christchurch Hospital CHHC D 3.7 3.8 0.37 50.9 10.3 0.62Cashmere High School CMHS D 1.0 1.4 0.37 44.4 5.1 0.85Hulverstone Dr Pumping Station HPSC E 3.8 3.9 0.22 36.7 10.0 1.03Heathcote Valley School HVSC C 1.4 4.0 1.41 81.4 5.7 2.21Kaiapoi North School KPOC E 17.3 17.4 0.20 18.9 11.3 0.06Lincoln School LINC D 13.5 13.6 0.12 12.7 12.1 0.09Lyttelton Port LPCC B 4.8 7.1 0.92 45.6 4.0 0.51Lyttelton Port Naval Point LPOC C 4.2 6.6 0.34 69.1 7.7 0.39North New Brighton School NNBS E 3.7 3.8 0.67 35.1 2.4 0.80Papanui High School PPHS D 8.6 8.6 0.21 36.7 12.8 0.21Pages Rd Pumping Station PRPC E 2.3 2.5 0.63 72.8 3.8 1.88Christchurch Resthaven REHS D 4.6 4.7 0.52 65.4 10.2 0.51Riccarton High School RHSC D 6.5 6.5 0.28 29.8 9.9 0.19Rolleston School ROLC D 19.6 19.6 0.18 8.4 10.3 0.08Shirley Library SHLC D 5.0 5.1 0.33 67.8 7.0 0.49Styx Mill Transfer Station SMTC D 10.7 10.8 0.16 27.6 13.6 0.17Templeton School TPLC D 12.5 12.5 0.11 11.3 15.3 0.16* As defined by the New Zealand Loadings Standard, NZS1170.5 (2004)† Joyner-Boore distance from surface projection of fault plane to site‡ Closest distance from fault plane to site§ Peak ground acceleration|| Peak ground velocity# Significant duration (5–95%)** Peak vertical ground acceleration. Note that with the exception of PGA v , ground motion parameters are geometric meanhorizontal definitionthe acceleration time histories in Figures 4–6, Figure 7 illustratesthe pseudo-acceleration response spectra of the geometricmean horizontal and vertical ground motion componentsat the two sites. In regard to horizontal components of groundmotion, compared to LPCC, it can be seen that the observedground motion at the LPOC site has significantly lower highfrequencyground motion amplitude (i.e. PGA LPOC = 0.34 g,PGA LPCC = 0.92 g), longer predominant period (Figure7), larger peak ground velocity (i.e., PGV LPOC = 69 cm/s,PGV LPCC = 46 cm/s), and larger significant duration (i.e., D s,LPOC = 7.7 s, D s,LPCC = 4.0 s), inferred as the result of nonlinearresponse of the surficial soils at LPOC. In contrast to thesignificant difference in horizontal ground motion, it can beseen that there is relatively little difference between the verticalground motion at LPCC and LPOC, with peak vertical accelerationsof 0.51 and 0.39 g, respectively.Evidence of LiquefactionOne of the major causes of damage in the M w 6.3 Christchurchearthquake resulted from the severity and spatial extent of liquefactionin residential, commercial, and industrial areas. Thehorizontal components of acceleration depicted in Figures 4and 5 show evidence of liquefaction phenomena in the centralbusiness district and eastern suburbs, which are locatedin the near-source region beyond the up-dip projection of thefault plane. In the central business district (i.e., REHS, CBGS,CHHC, CCCC), Cashmere (CMHS), and Shirley (SHLC),evidence of liquefaction is inferred based on the manifestedreduction in high-frequency content of ground motion followingseveral seconds of S-wave arrivals, and the subsequent acceleration“spikes,” characteristic of strain hardening deformationduring cyclic mobility. In the eastern suburbs (i.e., PRPC,HPSC, NNBS), the picture is somewhat more complex. Theground motion at Pages Road (PRPC) also has some of thecharacteristics discussed above, but in addition exhibits veryhigh accelerations in the fault-normal and vertical directions,which likely result from both surficial soil and source effects,due to its proximity to the up-dip projection of the slip asperity(i.e., Figure 2). The ground motion at North New Brighton(NNBS) exhibits several seconds of cyclic mobility before an856 Seismological Research Letters Volume 82, Number 6 November/December 2011

Scale0.5 g5 seconds▲▲Figure 4. Observed fault-normal horizontal acceleration time histories at various locations in the Christchurch region from the 22February earthquake.Scale0.5 g5 seconds▲ ▲ Figure 5. Observed fault-parallel horizontal acceleration time histories at various locations in the Christchurch region from the 22February earthquake.Seismological Research Letters Volume 82, Number 6 November/December 2011 857

Scale0.5 g5 seconds▲▲Figure 6. Observed vertical acceleration time histories at various locations in the Christchurch region from the 22 February earthquake.Spectral acc, Sa (g)10 0 Period, T (s)10 -1Eq:22/02/2011Solid:AvgHorizDashed:Verttude of ground shaking; 2) a change in surficial soil characterization;and 3) an increase in water table depth as noted previously.Given the observed spatial extent of liquefaction in the 4September 2010 Darfield earthquake (Cubrinovski et al. 2010),in which the majority of this western region was unaffected byliquefaction despite been subjected to generally stronger shakingthan the eastern regions (where liquefaction was prevalent),it can be logically concluded that the character and in situ stateof the soils in the western Christchurch region (Brown andWeeber 1992) result in a lower liquefaction susceptibility.10 -2LPCCLPOC10 -2 10 -1 10 0 10 1▲ ▲ Figure 7. Comparison of geometric mean horizontal and verticalresponse spectra observed at two stations in Lyttelton Port,one on outcropping rock (LPCC), the other on soil (LPOC).abrupt reduction in acceleration amplitude resulting in a veryshort significant duration of 2.4 seconds (Table 1). The groundmotion observed at Hulverstone Drive (HPSC) is also of interestdue to the relatively small horizontal component accelerationamplitudes compared with what might be expectedat such a near-source location (including observed shaking atnearby stations), and relative to its high vertical accelerations.No significant signs of liquefaction are evident in theground motions recorded to the west of those discussed above,which could result from three factors: 1) a reduction in ampli-Basin-generated Surface WavesAs previously mentioned, Christchurch is located on a sedimentaryfan deposit with the volcanic rock of Banks Peninsulalocated to the southeast. While specific mechanical and geometricaldetails of the predominant sedimentary basin layersare presently unknown, previous petroleum exploration hasrevealed the depth of gravel layers is in excess of 500 m, withbasement rock inferred to be at depths in excess of 2.0 km atvarious locations (Brown and Weeber 1992, Figure 1). Figure8A provides a schematic illustration of the deep geology of theregion along a plane trending southeast to northwest. Figure8A also illustrates one possible ray path from the M w 6.3 rupturein which seismic waves propagate up-dip and enter thesedimentary basin through its thickening edge. The large postcriticalincidence angles of such waves cause reflections that leadto a waveguide effect in which surface waves propagate acrossthe basin resulting in enhanced long-period ground motionamplitudes and shaking duration (Choi et al. 2005). Figure 8Billustrates the fault-normal, fault-parallel, and geometric mean858 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A)(B)Spectral Acc, Sa (g)North/West10 010 -110 -2Z 1.0=1000mChristchurchMesozoicbasement rockFault ParallelFault NormalHoriz gmTertiary VolcanicRockHypocenterZ 1.0=300mStation:CHHCR rup=3.8 km10 -2 10 -1 10 0 10 1Period, T (s)▲ ▲ Figure 8. A) Schematic illustration of waveguide effectsoccurring in the sedimentary basin underlying Christchurch (notto scale); and B) influence of basin depth on pseudo-spectralacceleration ordinates predicted empirically compared withthat observed at Christchurch Hospital (CHHC). The predictionshown is for the horizontal geometric mean and dashed linesrepresent the 16th and 84th percentiles.horizontal pseudo-response spectra at Christchurch hospital(CHHC), located at a source-to-site distance of R rup = 3.8 kmon the hanging wall. Also shown in Figure 8B is the predictedmedian, 16th, and 84th percentile response spectra for the siteusing the Bradley (2010) empirical model for two differentvalues of a proxy for basin depth. The Bradley (2010) modelis based on the Chiou and Youngs (2008) model with NewZealand–specific modifications. Basin effects are accountedfor in the model through the use of the parameter Z 1.0 , whichrepresents the depth to sediments with shear wave velocity,V s = 1.0 km/s. For site class D conditions (a nominal 30-maverage shear wave velocity of V s,30 = 250 m/s) the default valueof Z 1.0 is on the order of 300 m. Figure 8B illustrates that spectralamplitudes at CHHC for periods greater than 0.3 secondsare underpredicted using this default Z 1.0 value. The fact thatthe thickness of gravels in the Christchurch basin is known tobe greater than 500 m implies that Z 1.0 would be significantlygreater than 500 m. Figure 8B also illustrates the predictedspectral amplitudes, using a value of Z 1.0 = 1,000 m, where itcan be seen that the empirical prediction of long-period spectralamplitudes is significantly increased compared with thoseusing Z 1.0 = 300 m, in line with the observed amplitudes.The increase in amplitude of horizontal ground motionat long periods illustrated at Christchurch hospital (CHHC)was also observed at numerous other locations in the region.Significant amplitude Rayleigh surface waves are also clearlyevident in the vertical component of ground motion observedat larger source-to-site distances where body wave amplitudesare smaller (e.g., stations SMTC and CACS in Figure 6).Near-source Forward DirectivityIn the near-source region ground motions may exhibit forwarddirectivity effects due to the rupture front and direction of slipbeing aligned with the direction toward the site of interest.While the finite fault model of Beavan et al. (2011, page 789of this issue; see also Figure 2) does not provide information onthe temporal evolution of rupture, based on the central locationof the inferred hypocenter, the direction of slip is not wellaligned with an elliptically inferred rupture front. As a result, itis expected that rupture directivity effects will only be importantover a small surface area, relative to other possible rupturescenarios (Aagaard et al. 2004).Figure 9A illustrates the velocity time history at PagesRoad (PRPC), where forward directivity effects can be seen inthe fault-normal component. Figure 9B illustrates the velocitytime history at Christchurch hospital (CHHC) where avelocity pulse in the fault normal component is not clearly evident,and the large velocity amplitudes are the result of surfacewaves as previously noted. Figure 9C illustrates the observedand predicted pseudo-acceleration response spectra at CHHCwith and without the consideration of directivity effects. Thedirectivity effect was estimated empirically using the model ofShahi and Baker (2011). It can be seen that the predicted effectof forward directivity is relatively small (compared to the basindepth effect in Figure 8B) because of the small propagation distancefrom the hypocenter along the fault plane toward the site(which gives a low probability of observing a velocity pulse inthe model of Shahi and Baker 2011).The effects of near-source directivity fling step were notexamined here, both because on the near-source soil sites staticdeformations may be the result of ground failure, and alsobecause, as previously noted, the standard record processingadopted removes such long-period signals.Vertical Ground MotionFigure 6 illustrates that significant vertical ground motion amplitudeswere recorded in the near-source region in Christchurch,with peak vertical accelerations exceeding 0.6 g at sevenstrong motion stations and, in particular, values of 2.21 g and1.88 g observed at Heathcote Valley (HVSC) and Pages Road(PRPC), respectively. The vertical acceleration time histories atthese two sites also exhibit the so-called trampoline effect (Aoiet al. 2008; Yamada et al. 2009) caused by separation of surficialsoil layers in tension, limiting peak negative vertical accelerationsto approximately –1 g. Such large vertical accelerations canbe understood physically, first due to the relatively steep dip ofthe fault plane (δ = 69°), which results in a large component offault slip oriented in the vertical direction. Furthermore, at soilsites in sedimentary basins in particular, large vertical accelerationsat near-source locations can result from the conversion ofinclined SV waves to P waves at the sedimentary basin interface,Seismological Research Letters Volume 82, Number 6 November/December 2011 859

(A)Velocity, v (cm/s)100500-50-100100500-50-100Station:PRPC ; R rup=2.5 km, Fault ParallelFault Normal(B)Velocity, v (cm/s)(C)100500-50Vertical-1000 5 10 15 20 25 30 35 40Time, t (s)80400-40-8080400-40-80Station:CHHC ; R rup=3.8 kmFault ParallelFault Normal8040Vertical0-40-800 10 20 30 40 50 60Time, t (s)10 0Station:CHHCR rup=3.8 kmSpectral Acc, Sa (g)10 -110 -2Z 1.0=1000mno directivityFault ParallelFault NormalHoriz gmZ 1.0=1000mdirectivity considered10 -2 10 -1 10 0 10 1Period, T (s)▲ ▲ Figure 9. A) velocity time history recorded at Pages Road (PRPC); B) velocity time history recorded at Christchurch hospital (CHHC);and C) empirically predicted effect of directivity on spectral amplitudes at Christchurch hospital. The prediction shown is for the horizontalgeometric mean and dashed lines represent the 16th and 84th percentiles.860 Seismological Research Letters Volume 82, Number 6 November/December 2011

Vertical-to-horizontal PGA ratio54321PRPCHPSCCHHCLPOCBCDELiquefactionObservedBozorgnia andCampbell (2004)Firm SoilSoft RockLPCC00 5 10 15 20Source-to-site distance, R rup(km)▲▲Figure 10. Observed vertical-to-horizontal peak groundacceleration ratios.which are subsequently amplified and refracted toward verticalincidence due to the basin P-wave gradient (Silva 1997).Figure 10 illustrates the ratio of peak vertical accelerationand peak horizontal acceleration at the near-source strongmotion sites, as well as the empirical model of Bozorgnia andCampbell (2004). It can be seen that peak vertical-to-horizontalground acceleration ratios of up to 4.8 were observed. The peakvertical-to-horizontal ground acceleration ratios show a rapiddecay with source-to-site distance, and the observed ratios comparefavorably with the Bozorgnia and Campbell (2004) empiricalmodel for source-to-site distances beyond 5 km but significantlyunderpredict the ratios at closer distances. In Figure 10,data are also differentiated by whether liquefaction was observed(as discussed previously). Almost all strong motion records atdistances less than 5 km show liquefaction evidence (the exceptionbeing Heathcote Valley (HVSC)). At the aforementionedsites (with source-to-site distances less than 5 km), the large peakvertical-to-horizontal ground acceleration ratios observed areinterpreted to be the result of both the aforementioned steepfault dip leading to large vertical ground motions as well as significantnonlinear soil behavior (including liquefaction), whichgenerally results in more of a reduction in peak horizontal accelerationsthan peak vertical accelerations (e.g., Figure 1).(A)(B)Spectral acc, Sa (g)Spectral disp, Sd (cm)Eq:22/02/2011Solid:AvgHorizDashed:Vert10 -1 CCCCCHHCCBGSREHSNZS1170.510 -2 10 -1 10 0 10 110 0 Period, T (s)7060504030201000 2 4 6 8 10Period, T (s)▲ ▲ Figure 11. Comparison of response spectra from four strongmotionstations located in the Christchurch central business district:A) horizontal and vertical pseudo-acceleration responsespectra; and B) horizontal displacement response spectra.Ground Motion Intensity in the Central Business District(CBD)The Christchurch earthquake caused significant damage tocommercial structures in the CBD. At the time of writing, thepublic access to the majority of the 2 km 2 CBD is still prohibitedwhile an estimated 1,000 structures (of various typologies,construction materials, and age) are being demolished.The complete collapse of the Pine Gould Corporation andCanterbury Television buildings also led to the majority of the181 casualties (New Zealand Police 2011).Figures 11A and 11B illustrate the pseudo-accelerationand displacement response spectra of four strong motion stations(CCCC, CHHC, CBGS, REHS) located in the CBDregion. Despite their geographic separation distances (relativeto their respective source-to-site distances) it can be seen thatthe characteristics of the ground motion observed at theselocations is relatively similar. This is particularly the case forlong-period ground motion amplitudes, which have longerwavelengths and therefore are expected to be more coherent.On the other hand, at short vibration periods there is more ofa discrepancy in seismic intensity due to a shorter wavelengthand therefore lower wave coherency, and probably more importantlydue to the nonlinear response of significantly differentsurficial soil layers (Cubrinovski et al. 2011, page 893 of thisissue). Figure 11A, in particular, illustrates that the stronglong-period ground motion previously discussed with respectto CHHC (i.e., Figure 8B) was observed at all four CBD stationsand both A and Figures 11A and 11B illustrate that theseismic demands were above the 475-year return-period-designground motion for Christchurch site class D as specified bythe New Zealand loadings standard, New Zealand Standards1170.5 (2004). Figure 11B illustrates that for structures whosesecant period at peak displacement is in the region of 1.5 orSeismological Research Letters Volume 82, Number 6 November/December 2011 861

(A)Station:CCCCR rup=2.8 km(B)10 1Station:PRPCR rup=2.5 kmSpectral Acc, Sa (g)(C)Spectral Acc, Sa (g)(E)10 010 -110 -210 0Fault ParallelFault NormalHoriz gmVertPrediction10 -2 10 -1 10 0 10 1Period, T (s)10 -110 -210 1Fault ParallelFault NormalHoriz gmVertPredictionStation:RHSCR rup=6.5 km10 -2 10 -1 10 0 10 1Period, T (s)Station:HVSCR rup=4 kmSpectral Acc, Sa (g)(D)Spectral Acc, Sa (g)(F)10 010 -110 0Fault ParallelFault NormalHoriz gmVertPrediction10 -2 10 -1 10 0 10 1Period, T (s)10 -110 -2Fault ParallelFault NormalHoriz gmVertPredictionStation:CACSR rup=12.8 km10 -2 10 -1 10 0 10 1Period, T (s)Station:LPCCR rup=7.1 kmSpectral Acc, Sa (g)10 010 -1Fault ParallelFault NormalHoriz gmVertPrediction10 -2 10 -1 10 0 10 1Period, T (s)Spectral Acc, Sa (g)10 010 -110 -2Fault ParallelFault NormalHoriz gmVertPrediction10 -2 10 -1 10 0 10 1Period, T (s)▲ ▲ Figure 12. Horizontal and vertical pseudo-acceleration response spectra observed at various near-source strong-motion stations.The prediction shown is for the horizontal geometric mean and dashed lines represent the 16th and 84th percentiles.3.5 seconds, the displacement demands imposed by the groundmotion were in the order of two times the seismic design level.Response Spectra Observed at Various Strong MotionStationsFigure 12 illustrates the horizontal and vertical pseudoaccelerationresponse spectra observed at six locations inChristchurch as well as the empirical prediction for thegeometric mean component of Bradley (2010). No attemptis made here to rigorously assess the adequacy of empiricalground models against the observed near-source groundmotions, and the purpose of the comparisons is simply toillustrate general features of the response at the strong motionstation that depart from that which is nominally expected.Figure 12A illustrates the aforementioned strong long-periodground motion at Christchurch Cathedral College (CCCC)due to basin wave propagation. Figure 12B illustrates the forwarddirectivity polarity in the ground motion at Pages Road(PRPC), with significantly higher long-period ground motionin the fault normal component, as well as a vertical pseudospectralacceleration of 6 g for a vibration period of T = 0.1 s.Figure 12E illustrates that the high-frequency ground motionat Heathcote Valley (HVSC) is significantly above thatexpected due to its location on a colluvium wedge at the baseof the volcanic rock Port Hills. Figure 12F also illustrates thehigh-frequency ground motion on bedrock at Lyttelton Port862 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A)(B)Station:CCCCSolid:AvgHorizDashed:VertStation:PRPCSolid:AvgHorizDashed:VertSpectral acc, Sa (g)10 -1Spectral acc, Sa (g)10 010 -110 -210 0 Period, T (s)04/09/201022/02/2011NZS1170.510 -1 10 0 10 110 -210 1 Period, T (s)04/09/201022/02/2011NZS1170.510 -1 10 0 10 1(C)(D)Station:RHSCSolid:AvgHorizDashed:VertStation:CACSSolid:AvgHorizDashed:VertSpectral acc, Sa (g)10 -1Spectral acc, Sa (g)10 -110 -210 0 Period, T (s)04/09/201022/02/2011NZS1170.510 -1 10 0 10 110 -210 0 Period, T (s)04/09/201022/02/2011NZS1170.510 -1 10 0 10 1(E)Station:HVSCSolid:AvgHorizDashed:Vert(F)Station:LPCCSolid:AvgHorizDashed:VertSpectral acc, Sa (g)10 010 -1Spectral acc, Sa (g)10 -110 -210 1 Period, T (s)04/09/201022/02/2011NZS1170.510 -1 10 0 10 110 -210 0 Period, T (s)04/09/201022/02/2011NZS1170.510 -1 10 0 10 1▲ ▲ Figure 13. Comparison of geometric mean horizontal and vertical pseudo-acceleration response spectra observed in the 22 FebruaryChristchurch and 4 September Darfield earthquakes at various strong-motion stations.(LPCC) was significantly above that expected, likely due tolow attenuation through the underlying volcanic rock. Notethat the Bradley (2010) model accounts for the hanging walleffect (Abrahamson and Somerville 1996), which is not overlysignificant as a result of the steep fault dip.COMPARISON WITH GROUND MOTIONSOBSERVED IN THE 2010 DARFIELD EARTHQUAKEAND DESIGN SPECTRAThe M w 6.3 22 February 2011 Christchurch earthquake wasthe second event in approximately six months to cause significantground motion shaking in Christchurch, having beenpreceded by the 4 September 2010 Darfield earthquake (NewZealand Society for Earthquake Engineeering 2010).Figure 13 illustrates the geometric mean horizontaland vertical pseudo-acceleration response spectra of groundmotions at various strong-motion stations in Christchurchresulting from both the Christchurch and Darfield earthquakes.It can be immediately seen that for the majority ofvibration periods of engineering interest the spectral amplitudesare larger for the Christchurch earthquake. The primaryexception to the above statement is the spectral amplitudes atlong vibration periods (i.e., T > 2 s) due to both the longer durationof shaking and forward directivity effects in the Darfieldearthquake. Strong long-period spectral ordinates associatedSeismological Research Letters Volume 82, Number 6 November/December 2011 863

(A)Station:PPHSSolid:AvgHorizDashed:Vert(B)Station:SMTCSolid:AvgHorizDashed:VertSpectral acc, Sa (g)10 -1Spectral acc, Sa (g)10 -110 -210 0 Period, T (s)04/09/201022/02/2011NZS1170.510 -1 10 0 10 110 -210 0 Period, T (s)04/09/201022/02/2011NZS1170.510 -1 10 0 10 1▲ ▲ Figure 14. Similarity of response spectral shapes of horizontal and vertical ground motions observed in the Christchurch and Darfieldearthquakes at: A) Papanui (PPHS); and B) Styx Mill (SMTC).with these phenomena in the Darfield earthquake can be clearlyseen at CCCC, CACS, and HVSC stations. Figure 13A illustratesthat at Christchurch Cathedral College (CCCC), whichis located in the Christchurch CBD, spectral amplitudes in theChristchurch earthquake were approximately twice those ofthe Darfield earthquake for vibration periods less than T = 1.5s. It can also be seen that at CCCC station, spectral amplitudesresulting from the Darfield earthquake were notably below thedesign spectra for T < 2 s. Figure 13C–D also illustrate thatspectral amplitudes from the Darfield earthquake were belowthe design spectra at short periods throughout the majorityof Christchurch, with exceptions being Heathcote Valley(HVSC), Lyttelton Port (LPCC), and several western suburbs(i.e., TPLC, ROLC, LINC) not shown here.Another notable feature illustrated in Figure 13 is the similarityof the response spectral shapes at a given site from thesetwo events. In such an examination it is important to note themarkedly different source locations of the two events, withthe Christchurch earthquake occurring to the southeast andthe Darfield earthquake approximately 30 km west of centralChristchurch. Hence, the source and path effects of the groundmotion at a single site are expected to be significantly differentin both events. For example, Figure 13C and 13D illustratethe similarity of response spectral shapes, for vibration periodsless than T = 2 s, of both horizontal and vertical groundmotion components at Riccarton (RHSC) and CanterburyAero Club (CACS), while Figure 14 illustrates the similaritiesat Papanui (PPHS) and Styx Mill (SMTC). At vibration periodslarger than T = 2 s, the aforementioned source effects fromthe Darfield earthquake become significant, and the responsespectral shapes at a given site from these two events deviate.These observations clearly point to the importance of localsite effects on surface ground motions, particularly at high tomoderate vibration frequencies, and hence the benefits thatcan be obtained via site-specific response analysis as opposedto simple soil classification (recall that most of the sites in theChristchurch basin are assigned as site class D (New ZealandStandards 1170.5 2004)). It should also be noted that the foursites discussed above, while experiencing significant groundmotions, are founded on soils that did not exhibit liquefaction.CONCLUSIONSThe 22 February 2011 M w 6.3 Christchurch earthquakeimposed severe ground motion intensities, which were in excessof the current seismic design spectra and those experienced inthe 4 September 2010 Darfield earthquake, over the majorityof the Christchurch region. The severe ground motion intensitiesresulted in significant nonlinear soil behavior and severeand widespread liquefaction, which were evident in recordedacceleration time histories.The deep Christchurch sedimentary basin likely led to awaveguide effect of seismic waves entering through its thickeningedge, which resulting in increased ground motion durationsand long-period amplitudes over the majority of Christchurch.Very large vertical accelerations were also recorded at nearsourcestations, in part due to the steeply dipping fault plane,which resulted in a large component of slip oriented vertically.In contrast, forward directivity effects were not significant overa wide region, presumably related to the relatively central locationof the inferred hypocenter along-strike and down-dip andthe oblique alignment of the slip and rupture front directions.The similarity of response-spectral shapes of the groundmotion observed at a single station resulting from theChristchurch and Darfield earthquakes, for which source andpath effects were largely different, also illustrated the significanceof site-specific response for short and moderate vibrationfrequencies.ACKNOWLEDGMENTSThe ground motion records utilized in this manuscript werefreely obtained from the GeoNet project. Discussions withJohn Beavan and John Berrill are greatly appreciated.864 Seismological Research Letters Volume 82, Number 6 November/December 2011

REFERENCESAagaard, B. T., J. F. Hall, and T. H. Heaton (2004). Effects of fault dipand slip rake angles on near-source ground motions: Why rupturedirectivity was minimal in the 1999 Chi-Chi, Taiwan, earthquake.Bulletin of the Seismological Society of America 94, 155–170.Abrahamson, N. A., and P. G. Somerville (1996). Effects of the hangingwall and footwall on ground motions recorded during theNorthridge earthquake. Bulletin of the Seismological Society ofAmerica 86, S93–99.Aoi, S., T. Kunugi, and H. Fujiwara (2008). Trampoline effect in extremeground motion. Science 322, 727–730.Beavan, J., E. J. Fielding, M. Motagh, S. Samsonov, and N. Donnelly(2011). Fault location and slip distribution of the 22 February 2011M W 6.2 Christchurch, New Zealand, earthquake from geodeticdata. Seismological Research Letters 82, 789–799.Bommer, J. J., and A. Martinez-Pereira (1999). The effective duration ofearthquake strong motion. Journal of Earthquake Engineering 3,127–172.Bozorgnia, Y., and K. W. Campbell (2004). The vertical-to-horizontalresponse spectral ratio and tentative procedures for developingsimplified V/H and vertical design spectra. Journal of EarthquakeEngineering 8, 175–207.Bradley, B. A. (2010). NZ-specific Pseudo-spectral Acceleration GroundMotion Prediction Equations Based on Foreign Models. Departmentof Civil and Natural Resources Engineering, University ofCanterbury, Christchurch, New Zealand, 324 pp.Brown, L. J., and J. H. Weeber (1992). Geology of the Christchurch UrbanArea. Institute of Geological and Nuclear Sciences map. LowerHutt, New Zealand: GNS Science.Chiou, B., R. Darragh, N. Gregor, and W. J. Silva (2008). NGA projectstrong-motion database. Earthquake Spectra 24, 23–44.Chiou, B. S. J., and R. R. Youngs (2008). An NGA model for the averagehorizontal component of peak ground motion and response spectra.Earthquake Spectra 24, 173–215.Choi, Y., J. P. Stewart, and R. W. Graves (2005). Empirical model forbasin effects accounts for basin depth and source location. Bulletinof the Seismological Society of America 95, 1,412–1,427.Cubrinovski, M., J. D. Bray, M. Taylor, S. Giorgini, B. A. Bradley, L.Wotherspoon, and J. Zupan (2011). Soil liquefaction effects in thecentral business district during the February 2011 Christchurchearthquake. Seismological Research Letters 82, 893–904.Cubrinovski, M., R. A. Green, J. Allen, S. A. Ashford, E. Bowman, B. A.Bradley, B. Cox, T. C. Hutchinson, E. Kavazanjian, R. P. Orense,M. Pender, M. Quigley, and L. Wotherspoon (2010). Geotechnicalreconnaissance of the 2010 Darfield (Canterbury) earthquake.Bulletin of the New Zealand Society for Earthquake Engineering 43,243–320.DeMets, C., R. G. Gordon, D. F. Argus, and S. Stein (1994). Effect ofrecent revisions to the geomagnetic time scale on estimates of currentplate motion. Geophysical Research Letters 21, 2,191–2,194.Douglas, J., and D. Boore (2010). High-frequency filtering of strongmotionrecords. Bulletin of Earthquake Engineering 9(2): 395–409.Graizer, V. M. (2005). Effect of tilt on strong motion data processing.Soil Dynamics and Earthquake Engineering 25, 197–204.New Zealand Police (2011). Christchurch earthquake: List of deceased;http://www.police.govt.nz/list-deceased. Last accessed June 20,2011.New Zealand Society for Earthquake Engineering (NZSEE) (2010).Preliminary observations of the 2010 Darfield (Canterbury)Earthquakes. Special issue, Bulletin of the New Zealand Society forEarthquake Engineering 43, 215–439.Shahi, S. K., and J. W. Baker (2011). An empirically calibrated frameworkfor including the effects of near-fault directivity in probabilisticseismic hazard analysis. Bulletin of the Seismological Society ofAmerica 101, 742–755.Silva, W. J. (1997). Characteristics of vertical strong ground motionsfor applications to engineering design. In Proceedings of theFHWA/NCEER Workshop on the National Representation ofSeismic Ground Motion for New and Existing Highway Facilities,Burlingame, CA. Technical Report NCEER-97-0010. Buffalo, NY:National Center for Earthquake Engineering Research.Standards New Zealand Standards (2004). Structural Design Actions,Part 5: Earthquake Actions—New Zealand. Wellington, NewZealand: Standards New Zealand, 82 pp.Somerville, P. G., K. Ikikura, R. W. Graves, S. Sawada, D. Wald, N. A.Abrahamson, Y. Iwasaki, T. Kagawa, N. Smith, and A. Kowada(1999). Characterizing crustal earthquake slip models for the predictionof strong ground motion. Seismological Research Letters 70,59–80.Stirling, M. W., M. Gerstenberger, N. Litchfield, G. H. McVerry, W. D.Smith, J. R. Pettinga, and P. Barnes (2007). Updated ProbabilisticSeismic Hazard Assessment for the Canterbury Region. GNS ScienceConsultancy Report 2007/232, ECan Report Number U06/6,58 pp.Sutherland, R., K. Berryman, and R. Norris (2006). Quaternary sliprate and geomorphology of the Alpine fault: Implications for kinematicsand seismic hazard in southwest New Zealand. GeologicalSociety of America Bulletin 118, 464–474.Yamada, M., J. Mori, and T. Heaton (2009). The slapdown phase in highaccelerationrecords of large earthquakes. Seismological ResearchLetters 80, 559–564.Zhao, J. X., and M. Gerstenberger (2010). Attenuation Models for RapidPost Earthquake Assessment in New Zealand. Wellington, NewZealand: Earthquake Commission New Zealand report.Department of Civil and Natural Resources EngineeringUniversity of CanterburyPrivate Bag 4800Christchurch, New Zealandbrendon.bradley@canterbury.ac.nz(B A. B.)Seismological Research Letters Volume 82, Number 6 November/December 2011 865

Ground Motion Attenuation during M 7.1Darfield and M 6.2 Christchurch, New Zealand,Earthquakes and Performance of GlobalPredictive ModelsMargaret Segou and Erol KalkanMargaret Segou and Erol KalkanU.S. Geological SurveyEOnline material: Flat-file for both eventsINTRODUCTIONThe M 7.1 Darfield earthquake occurred 40 km west ofChristchurch (New Zealand) on 4 September 2010. Sixmonths after, the city was struck again with an M 6.2 eventon 22 February local time (21 February UTC). These eventsresulted in significant damage to infrastructure in the city andits suburbs. The purpose of this study is to evaluate the performanceof global predictive models (GMPEs) using the strongmotion data obtained from these two events to improve futureseismic hazard assessment and building code provisions for theCanterbury region.The Canterbury region is located on the boundarybetween the Pacific and Australian plates; its surface expressionis the active right lateral Alpine fault (Berryman et al.1993). Beneath the North Island and the north South Island,the Pacific plate subducts obliquely under the Australianplate, while at the southwestern part of the South Island, areverse process takes place. Although New Zealand has experiencedseveral major earthquakes in the past as a result of itscomplex seismotectonic environment (e.g., M 7.1 1888 NorthCanterbury, M 7.0 1929 Arthur’s Pass, and M 6.2 1995 Cass),there was no evidence of prior seismic activity in Christchurchand its surroundings before the September event. The Darfieldand Christchurch earthquakes occurred along the previouslyunmapped Greendale fault in the Canterbury basin, which iscovered by Quaternary alluvial deposits (Forsyth et al. 2008).In Figure 1, site conditions of the Canterbury epicentral areaare depicted on a V S30 map. This map was determined on thebasis of topographic slope calculated from a 1-km grid usingthe method of Allen and Wald (2007). Also shown are thelocations of strong motion stations.The Darfield event was generated as a result of a complexrupture mechanism; the recordings and geodetic data revealthat earthquake consists of three sub-events (Barnhart et al.2011, page 815 of this issue). The first event was due to rupturingof a blind reverse fault with M 6.2, followed by a secondevent (M 6.9), releasing the largest portion of the energyon the right-lateral Greendale fault. The third sub-event (M5.7) is due to a reverse fault with a right-lateral component(Holden et al. 2011). The Christchurch earthquake occurredon an oblique thrust fault. The comparison of spectral accelerationvalues at stations near Christchurch reveals that thesecond event produced much larger amplitudes of shakingthan the Darfield event due to its proximity to the epicenter.Both events resulted in noticeably large amplitudes ofthe vertical motion, often exceeding horizontal motion inthe near-fault area. The vertical motions, showing asymmetricacceleration traces and pulses, reached 1.26 g during theDarfield earthquake and 2.2 g during the Christchurch event.These events were recorded by more than 100 strong motionstations operated by the Institute of Geological and NuclearSciences (http://www.geonet.org.nz/). Using the processeddata from these stations, peak ground acceleration (PGA) and5%-damped spectral acceleration values at 0.3, 1, and 3 s areused for performance evaluation of the global ground motionpredictive equations (GMPEs). The selected GMPEs are theNext Generation Attenuation (NGA) models of Abrahamsonand Silva (2008), Boore and Atkinson (2008), Campbell andBozorgnia (2008), and Chiou and Youngs (2008). The Graizerand Kalkan (2007, 2009) model, which is based on the NGAproject database, is also included. These GMPEs are abbreviatedrespectively as AS08, BA08, CB08, CY08, and GK07.Because they have been used widely for seismic hazard analysisfor crustal earthquakes, their performance assessmentbecomes a critical issue especially for immediate response andrecovery planning after major events. The occurrence of aftershockssimilar to the Christchurch event will most probablycontrol seismic hazard in the broader area, as confirmed bythe recent M 6.0 event on June 13, 2011.866 Seismological Research Letters Volume 82, Number 6 November/December 2011 doi: 10.1785/gssrl.82.6.866

▲ ▲ Figure 1. Shear-wave velocity (unit = m/s) down to 30 m derived from topographic slope; the locations of strong motion stations arealso shown.PERFORMANCE EVALUATION OF GROUNDMOTION PREDICTION EQUATIONSIn order to evaluate the relative performance of the GMPEsand their ranking to be used for logic tree weighting in hazardanalysis, we used traditional residual analysis and an informationtheoretic approach. In residual analysis the predictionerror for each observation and standard deviation of the errorsfor each event are computed for each GMPE. Residuals correspondto the difference between the observations and predictionsin natural-log space; negative residuals are interpreted asoverprediction, whereas positive residuals indicate underestimationof the predictive model. The applied information theoreticapproach is based on a log-likelihood value (LLH), whichdescribes the information loss when a GMPE approximates anobservation (Scherbaum et al. 2009). The average sample loglikelihood(LLH) value of a GMPE, noted herein as g, over Nnumber of x observations, represented by a log-normal distribution,is calculated as:log( L( g x )) = − 1 N log ( g ( x i ))N∑ (1)i=1The negative average log-likelihood value is a measure of distancebetween the predictions and observations; therefore, aGMPE exhibiting a smaller absolute value of LLH, relative toother GMPEs, corresponds to a better performing model.For the Darfield event, the relative performance of GMPEswas evaluated for strike slip faulting since the greater amountof moment release occurred during the second sub-event. Forthe Christchurch event, however, the evaluation is based on thethrust fault (as discussed before). The hanging wall effects wereconsidered, although their effects are not significant because thecausative faults appear to be steep (Bradley and Cubrinovski2011, page 853 of this issue). The flat-file, listing distance metrics,V S30 for each station, and corresponding observations (PGAand spectral values), is provided for each event as an electronicsupplement (http://nsmp.wr.usgs.gov/ekalkan/NZ/index.html).Using this flat-file, predictions of GMPEs were computed foreach event. Figures 2–5 summarize the results for the M 7.1Darfield (top panels) and M 6.2 Christchurch (bottom panels)earthquakes. The plots shown in row A in each figure represent16th, 50th (median), and 84th percentile of predictions consideringan average V S30 value of 400 m/s. In these plots, observationscorrespond to the maximum value of two horizontal components.Because the NGA models predict geometric mean ofground motion, their predictions were adjusted for maximumhorizontal component by multiplying their predictions with 1.1for PGA and 1.15, 1.18, and 1.18 for spectral acceleration at 0.3 s,1 s, and 3 s, respectively. These adjustment factors were adaptedfrom Campbell and Bozorgnia (2008). The GK07 model predictsthe maximum of the two horizontal components. It shouldbe also noted that both observations and predictions are plottedagainst a distance metric specific to the model; for the BA08,the distance metric is the “Joyner-Boore distance” (R JB ), definedas the closest distance from the recording station to the surfaceprojection of the fault rupture plane (Boore et al. 1997). For theremaining models, the distance measure is the “closest fault dis-Seismological Research Letters Volume 82, Number 6 November/December 2011 867

Peak Ground Acceleration, (g)In[PGA] Actual− In[PGA] GMPE210.50.10.010.00150Graizer and Kalkan (2007, 2009)[A]Abrahamson and Silva (2008) Boore and Atkinson (2008) Campbell and Bozorgnia (2008) Chiou and Youngs (2008)M7.1 Darfield Eq.σ = 0.52 In[Y][B] σ = 0.58 In[Y]σ = 0.52 In[Y]σ = 0.63 In[Y]σ = 0.52In[Y]−58σ LLH= 0.81[C]σ LLH= 0.97σ LLH= 0.75σ LLH= 1.14σ LLH= 0.916LLH value4205 10 20 50 100 2505 10 20 50 100 2501 5 10 20 50 100 250Distance Measure (km)5 10 20 50 100 2505 10 20 50 100 250Peak Ground Acceleration, (g)In[PGA] Actual− In[PGA] GMPE210.50.10.010.00150Graizer and Kalkan (2007, 2009)[A]Abrahamson and Silva (2008) Boore and Atkinson (2008) Campbell and Bozorgnia (2008) Chiou and Youngs (2008)M6.2 Christchurch Eq.σ In[Y]= 0.60 [B] σ In[Y]= 0.61 σ In[Y]= 0.63 σ In[Y]= 0.77 σ In[Y]= 0.53−58σ LLH= 1.12[C]σ LLH= 0.93σ LLH= 1.42σ LLH= 2.82σ LLH= 0.946LLH value4205 10 20 50 100 2505 10 20 50 100 2501 5 10 20 50 100 250Distance Measure (km)5 10 20 50 100 2505 10 20 50 100 250▲ ▲ Figure 2. A) Comparison of PGA values recorded from the M 7.1 Darfield (top panels) and M 6.2 Christchurch (bottom panels) earthquakesfor 16th, 50th (median), and 84th percentile predictions from five different GMPEs. B) Residuals computed for each GMPE formedian prediction; also shown is the trend line to quantify distance bias. C) Average-log likelihood (LLH) values are to determine performanceof GMPEs; higher LLH values indicate poorer performance.868 Seismological Research Letters Volume 82, Number 6 November/December 2011

Spectral Acceleration (0.3 s), (g)In[SA(0.3 s)] Actual− In[SA(0.3 s)] GMPE210.50.10.010.00150−58Graizer and Kalkan (2007, 2009)[A]Abrahamson and Silva (2008) Boore and Atkinson (2008) Campbell and Bozorgnia (2008) Chiou and Youngs (2008)M7.1 Darfield Eq.σ In[Y]= 0.59 [B] σ In[Y]= 0.69 σ In[Y]= 0.63 σ In[Y]= 0.71 σ In[Y]= 0.63σ LLH= 0.87[C]σ LLH= 1.00σ LLH= 0.86σ LLH= 1.13σ LLH= 1.016LLH value4205 10 20 50 100 2505 10 20 50 100 2501 5 10 20 50 100 250Distance Measure (km)5 10 20 50 100 2505 10 20 50 100 250Spectral Acceleration (0.3 s), (g)In[SA(0.3 s)] Actual− In[SA(0.3 s)] GMPE210.50.10.010.00150−58Graizer and Kalkan (2007, 2009)[A]Abrahamson and Silva (2008) Boore and Atkinson (2008) Campbell and Bozorgnia (2008) Chiou and Youngs (2008)M6.2 Christchurch Eq.σ In[Y]= 0.63 [B] σ In[Y]= 0.68 σ In[Y]= 0.72 σ In[Y]= 0.80 σ In[Y]= 0.62σ = 1.03[C]σ = 1.00σ = 1.70σ = 2.29σ = 0.99LLH LLH LLH LLH LLH6LLH value4205 10 20 50 100 2505 10 20 50 100 2501 5 10 20 50 100 250Distance Measure (km)5 10 20 50 100 2505 10 20 50 100 250▲ ▲ Figure 3. A) Comparison of 5%-damped spectral acceleration values computed at 0.3 s for the M 7.1 Darfield (top panels) andM 6.2 Christchurch (bottom panels) earthquakes for 16th, 50th (median), and 84th percentile predictions from five different GMPEs.B) Residuals computed for each GMPE for median prediction; also shown is the trend line to quantify distance bias. C) Average-loglikelihood (LLH) values are to determine performance of GMPEs; higher LLH values indicate poorer performance.Seismological Research Letters Volume 82, Number 6 November/December 2011 869

Spectral Acceleration (1 s), (g)In[SA(1 s)] Actual− In[SA(1 s)] GMPE210.50.10.010.00150−54Graizer and Kalkan (2007, 2009)[A]Abrahamson and Silva (2008) Boore and Atkinson (2008) Campbell and Bozorgnia (2008) Chiou and Youngs (2008)M7.1 Darfield Eq.σ In[Y]= 0.61 [B] σ In[Y]= 0.62 σ In[Y]= 0.60 σ In[Y]= 0.60 σ In[Y]= 0.60σ LLH= 0.53[C]σ LLH= 0.52σ LLH= 0.46σ LLH= 0.53σ LLH= 0.583LLH value2105 10 20 50 100 2505 10 20 50 100 2501 5 10 20 50 100 250Distance Measure (km)5 10 20 50 100 2505 10 20 50 100 250Spectral Acceleration (1 s), (g)In[SA(1 s)] Actual− In[SA(1 s)] GMPE210.50.10.010.00150−54Graizer and Kalkan (2007, 2009)[A]Abrahamson and Silva (2008) Boore and Atkinson (2008) Campbell and Bozorgnia (2008) Chiou and Youngs (2008)M6.2 Christchurch Eq.σ In[Y]= 0.60 [B] σ In[Y]= 0.71 σ In[Y]= 0.73 σ In[Y]= 0.73 σ In[Y]= 0.64σ = 0.68[C]σ = 0.89σ = 1.13σ = 1.29σ = 0.79LLH LLH LLH LLH LLH3LLH value2105 10 20 50 100 2505 10 20 50 100 2501 5 10 20 50 100 250Distance Measure (km)5 10 20 50 100 2505 10 20 50 100 250▲ ▲ Figure 4. A) Comparison of 5%-damped spectral acceleration values computed at 1 s for the M 7.1 Darfield (top panels) andM 6.2 Christchurch (bottom panels) earthquakes for 16th, 50th (median), and 84th percentile predictions from five different GMPEs.B) Residuals computed for each GMPE for median prediction; also shown is the trend line to quantify distance bias. C) Average-loglikelihood (LLH) values are to determine performance of GMPEs; higher LLH values indicate poorer performance.870 Seismological Research Letters Volume 82, Number 6 November/December 2011

Spectral Acceleration (3 s), (g)In[SA(3 s)] Actual− In[SA(3 s)] GMPE10.50.10.010.00150−54Graizer and Kalkan (2007, 2009)[A]Abrahamson and Silva (2008) Boore and Atkinson (2008) Campbell and Bozorgnia (2008) Chiou and Youngs (2008)M7.1 Darfield Eq.σ In[Y]= 0.65 [B] σ In[Y]= 0.69 σ In[Y]= 0.71 σ In[Y]= 0.68 σ In[Y]= 0.69σ LLH= 0.45[C]σ LLH= 0.65σ LLH= 0.63σ LLH= 0.77σ LLH= 0.803LLH value2105 10 20 50 100 2505 10 20 50 100 2501 5 10 20 50 100 250Distance Measure (km)5 10 20 50 100 2505 10 20 50 100 250Spectral Acceleration (3 s), (g)In[SA(3 s)] Actual− In[SA(3 s)] GMPE10.50.10.010.00150−54Graizer and Kalkan (2007, 2009)[A]Abrahamson and Silva (2008) Boore and Atkinson (2008) Campbell and Bozorgnia (2008) Chiou and Youngs (2008)M6.2 Christchurch Eq.σ = 0.57 In[Y][B] σ = 0.65 In[Y]σ = 0.75 In[Y]σ = 0.67 In[Y]σ = 0.61In[Y]σ = 0.63[C]σ = 0.68σ = 0.83σ = 0.78σ = 0.59LLH LLH LLH LLH LLH3LLH value2105 10 20 50 100 2505 10 20 50 100 2501 5 10 20 50 100 250Distance Measure (km)5 10 20 50 100 2505 10 20 50 100 250▲ ▲ Figure 5. A) Comparison of 5%-damped spectral acceleration values computed at 3 s for the M 7.1 Darfield (top panels) andM 6.2 Christchurch (bottom panels) earthquakes for 16th, 50th (median), and 84th percentile predictions from five different GMPEs.B) Residuals computed for each GMPE for median prediction; also shown is the trend line to quantify distance bias. C) Average-loglikelihood (LLH) values are to determine performance of GMPEs; higher LLH values indicate poorer performance.Seismological Research Letters Volume 82, Number 6 November/December 2011 871

tance” (R rup ) defined as the closest distance to co-seismic ruptureplane.The next set of plots in Figures 2–5 (row B) shows thedistance distribution of residuals. Linear fit lines illustrate thedistance bias; the trend line passing through zero means thatthere is no bias in predictions. Unlike the attenuation curvesshown in row A, based on average V S30 , the residuals are computedbased on specific V S30 values at each station estimatedfrom topographic slope (Figure 1) in order to explicitly incorporatethe site effects on ground motion estimates. To quantifythe quality of fit, the standard errors of predictions (σ InY ) arecomputed based on residuals, and these values are given in eachpanel for each GMPE. The larger σ InY indicates a poorer performanceof the GMPE. For PGA, all GMPEs indicate an overallgood fit to observations up to ~100 km (Figure 2 row A); fordistances larger than 100 km, ground motion exhibits fasterattenuation, and as a result the observed peak values are lowerthan expected. This is more pronounced for the Christchurchevent. Low ground accelerations recorded at large distancesshow the effect of the anelastic attenuation due to regional lowQ (Zhao and Gerstenberger 2010).In Figure 2 (row B), residuals for the Darfield eventreveal overestimation for distances greater than 70 km for theAB08 and 100 km for the BA08 and CB08 models. On theother hand, the GK07 and CY08 fit better to the observationsbecause their trend lines fitting to residuals do not showa notable distance bias. For the AS08 and CB08, the misfit atlarger distances is more evident. In case of PGA, the GK07,BA08, and CY08 models yield the smallest σ InY of 0.52 forthe Darfield event. For the Christchurch event (Figure 2 bottompanels), the GK07, AS08, BA08, and CY08 present betterresidual fits than CB08, for which the overestimation beginsat 20 km. The smallest σ InY is due to the CY08. The same istrue for spectral acceleration at 0.3 s as shown in Figure 3 (rowB). Finally, Figure 2 (row C) shows the distance distribution ofLLH values. In these plots, trend lines identify the consistencyof the GMPE in predicting ground motion at various distances;if the slope is close to zero, then GMPE has low distance variability,meaning that it is consistent. Much higher LLH valueswith increasing distance suggest a poorer fit at far-field, whichis observed for all GMPEs for the Christchurch event.As shown in Figure 3, spectral acceleration at 0.3 s reaches1 g at 40 km for the Darfield earthquake, and 2 g at 10 kmfor the Christchurch earthquake. For the Darfield event, theAS08 and CB08 overestimate observations for distances largerthan 70 km as shown by the residual plots. The same trend isevident for the Christchurch event (Figure 3 bottom panels).The BA08 performs better for the Darfield event since there isonly a minor overestimation for distances larger than 20 km.For the Christchurch event (Figure 3 bottom panels), however,the distance trend line of LLH reveals a poorer performance ofthe BA08 and CB08.Ground motion estimates are given for spectral accelerationat 1 s in Figure 4, where the observations exceed 1 g in thenear field of both earthquakes. For Darfield, all GMPEs presentan excellent fit to the observations up to 70 km from the fault.Beyond 70 km, they slightly overestimate the observations dueto faster attenuation of ground motion at far distances. For theChristchurch event, the overestimation is evident for the BA08and CB08 over 50 km, which resulted in higher LLH values.In Figure 5 (top panels), comparisons are given for thespectral acceleration at 3 s. For both Darfield and Christchurchevents, spectral peaks reach 0.5 g. For the former event, the GK07is the best fitting model to observations with zero distance biasand with lowest LLH values. The NGA models underestimatelong period ground motions up to 100 km, whereas beyond 100km they tend to overestimate. For the Christchurch earthquake,none of the models provide an excellent fit (Figure 5 bottom panels).The GK07 overestimates observations, as opposed to underestimationof NGA models over a wide distance range.RANKING GROUND MOTION PREDICTIONEQUATIONSIn Table 1, the mean (μ LLH ) and standard deviation (σ LLH )of LLH values over the total number of observations for eachearthquake and for each GMPE are tabulated; the standarderrors (σ InY ) of predictions are also listed. This table is used forranking the GMPEs according to these three parameters. Forranking, each parameter is first normalized with the respect toits lowest value due to different GMPEs, and then the arithmetic-meanof normalized values is computed for μ LLH , σ LLH , andσ InY for each GMPE. The GMPE with the lowest arithmeticmeanis ranked as first. This exercise is repeated for each periodand for each earthquake, and the results of ranking are given inTable 2. The best performing GMPE has a combined performancevalue close to unity. The ranking results show that performanceof the GK07, BA08, and CY08 are equally the samefor different periods and events, while the CB08 and AS08show relatively poorer performance.CONCLUDING REMARKSIn this study, we examined the performance of global groundmotion prediction equations (GMPEs) for the New Zealand M7.1 Darfield and M 6.2Christchurch earthquakes, with the objective of improvingfuture seismic hazard assessment and engineering applicationsfor the Canterbury region. These events are characterized bysignificantly large ground motions at high frequencies, whichshowed faster attenuation through the crust due to low regionalQ. Amplified spectral accelerations at long periods at long distancesare attributed to Canterbury’s deep sedimentary basin.For similar shallow earthquakes in New Zealand, there is anevidence of Moho reflection, which potentially might furtheramplify long-period ground motions (Zhao and Gerstenberger2010). Comparison of predictions derived from the five differentGMPEs with observations reveal overall good performanceof these models, supporting their applicability for the region.For the purpose of selecting and weighting GMPEs in a logictree approach for regional seismic hazard analysis, we applieda simple ranking procedure based on the average LLH values872 Seismological Research Letters Volume 82, Number 6 November/December 2011

TABLE 1Mean (μ LLH ) and standard deviation (σ LLH ) of average log-likelihood (LLH) values, and standard error of predictions (σ InY ).GK07: Graizer and Kalkan (2007, 2009); AS08: Abrahamson and Silva (2008); BA08: Boore and Atkinson (2008); CB08:Campbell and Bozorgnia (2008); CY08: Chiou and Youngs (2008).μ LLH σ LLH σ InYGK07 AS08 BA08 CB08 CY08 GK07 AS08 BA08 CB08 CY08 GK07 AS08 BA08 CB08 CY08M 7.1 DarfieldPGA 0.77 0.84 0.80 0.96 0.73 0.80 0.96 0.74 1.13 0.90 0.52 0.58 0.52 0.63 0.52SA (0.3s) 0.91 1.04 0.95 1.10 0.94 0.86 0.99 0.85 1.12 1.00 0.59 0.69 0.63 0.71 0.63SA (1 s) 0.93 0.93 0.90 0.91 0.95 0.52 0.52 0.46 0.53 0.58 0.61 0.62 0.60 0.60 0.60SA (3 s) 1.02 1.06 1.09 1.11 1.23 0.45 0.65 0.63 0.76 0.80 0.65 0.69 0.71 0.68 0.69M 6.2 ChristchurchPGA 0.98 0.91 1.12 0.92 1.41 1.12 0.93 1.42 2.82 0.94 0.61 0.63 0.77 0.53 0.61SA (0.3s) 1.01 1.04 1.02 0.99 1.69 1.03 1.00 1.70 2.29 0.99 0.68 0.72 0.80 0.62 0.68SA (1 s) 1.00 1.11 0.67 0.89 1.12 0.68 0.89 1.13 1.29 0.79 0.71 0.73 0.73 0.64 0.71SA (3 s) 1.17 0.98 0.63 0.67 0.82 0.63 0.68 0.83 0.78 0.59 0.65 0.75 0.67 0.61 0.65TABLE 2Combined performance parameters of ground motion prediction equations (GMPEs; GMPE with the lowest performanceparameter can be interpreted as better performing one (shown by bold).Combined Performance ParameterGK07 AS08 BA08 CB08 CY08M 7.1 DarfieldPGA 1.05 1.19 1.04 1.35 1.07SA (0.3s) 1.00 1.16 1.04 1.24 1.09SA (1 s) 1.07 1.07 1.00 1.06 1.11SA (3 s) 1.00 1.18 1.19 1.28 1.35M 6.2 ChristchurchPGA 1.17 1.08 1.37 2.36 1.00SA (0.3s) 1.04 1.07 1.46 1.85 1.00SA (1 s) 1.01 1.21 1.39 1.49 1.08SA (3 s) 1.09 1.10 1.30 1.18 1.02considering their mean and standard deviation, as well as standarderrors (σ InY ) of predictions. The ranking results show thatperformance of GK07 (Graizer and Kalkan 2007, 2009), BA08(Boore and Atkinson 2008), and CY08 (Chiou and Youngs2008) perform equally well, while the CB08 (Campbell andBozorgnia 2008) and AS08 (Abrahamson and Silva 2008)show relatively poorer performance.ACKNOWLEDGMENTSThe authors thank Jim Cousins (GNZ) for providing parametersof strong motion recordings, and Volkan Sevilgen for preparingthe V S30 map for the Canterbury region. We also wishto thank David Boore, Tom Hanks, and Vladimir Graizer fortheir reviews.DATA AND RESOURCESFor NGA models, we used the Fortran code written by DavidBoore (Kaklamanos et al. 2010); for the Graizer and Kalkan(2007, 2009) ground motion prediction model, Matlab andFortran codes are available online at http://nsmp.wr.usgs.gov/ekalkan/PGA07/index.html. The flat-file for both Darfield andChristchurch events is also available online at http://nsmp.wr.usgs.gov/ekalkan/NZ/index.html.Seismological Research Letters Volume 82, Number 6 November/December 2011 873

REFERENCESAbrahamson, N. A., and W. J. Silva (2008). Summary of the Abrahamsonand Silva NGA ground motion relations. Earthquake Spectra 24(1), 67–98.Allen, T. I., and D. J. Wald. (2007). Topographic Slope as a Proxy forSeismic Site Conditions (V s 30 ) and Amplification around the Globe.USGS Open-File Report 2007-1357, 69 pp.; http://earthquake.usgs.gov/hazards/apps/vs30/.Barnhart, W. D., M. J. Willis, R. W. Lohman, and A. K. Melkonian(2011). InSAR and optical constraints on fault slip during the2010–2011 New Zealand earthquake sequence. SeismologicalResearch Letters 82, 815–823.Berryman, K. R., S. Beanland, A. F. Cooper, H. N. Cutten, R. J. Norris,and P. R. Wood (1993). The Alpine fault, New Zealand: Variationin Quaternary structural style and geomorphic expression. AnnalesTectonicae 6, special issue supplement, S126–163.Boore, D. M., and G. M. Atkinson (2008). Ground motion predictionequations for the average horizontal component of PGA, PGV,and 5%-damped PSA at spectral periods between 0.01 s and 10.0 s.Earthquake Spectra 24 (1), 99–138.Boore, D. M., W. B. Joyner, and T. E. Fumal (1997). Equations for estimatinghorizontal response spectra and peak acceleration fromwestern North American earthquakes: A summary of recent work.Seismological Research Letters 68, 128–153.Bradley, B. A., and M. Cubrinovski (2011). Near-source strong groundmotions observed in the 22 February 2011 Christchurch earthquake.Seismological Research Letters 82, 853–865.Campbell, K. W., and Y. Bozorgnia (2008). NGA ground motion modelfor the geometric mean horizontal component of PGA, PGV, PGD,and 5% damped linear elastic response spectra for periods rangingfrom 0.01 to 10 s. Earthquake Spectra 24 (1), 139–172.Chiou, B. S. J, and R. R. Youngs (2008). An NGA model for the averagehorizontal component of peak ground motion and response spectra.Earthquake Spectra 24 (1), 173–215.Forsyth, P. J., D. J. A. Barrell, and R. Jongens (2008). Geology of theChristchurch Area. Institute of Geological and Nuclear Sciences1;250,000 geological map 16, 1 sheet + 67 pp. Lower Hutt, NewZealand: GNS Science.Graizer, V., and E. Kalkan (2007). Ground motion attenuation modelfor peak horizontal acceleration from shallow crustal earthquakes.Earthquake Spectra 23, 585–613.Graizer V., and E. Kalkan (2009). Prediction of response spectral accelerationordinates based on PGA attenuation. Earthquake Spectra25 (1), 36–69.Holden, C., J. Beavan, B. Fry, M. Reyners, J. Ristau, R. Van Dissen, P.Villamor, and M. Quigley (2011). Preliminary source model of theMw 7.1 Darfield earthquake from geological, geodetic and seismicdata. Proceedings of the Ninth Pacific Conference on EarthquakeEngineering: Building an Earthquake-Resilient Society, 14–16April 2011, paper no. 164. Auckland, New Zealand: New ZealandSociety for Earthquake Engineering.Kaklamanos, J., D. M. Boore, E. M. Thomson, and K. W. Campbell(2010). Implementation of the Next Generation Attenuation (NGA)Ground-motion Prediction Equations in Fortran and R. USGSOpen-File Report 2010-1296, 47 pp.Scherbaum, F., E. Delavaud, and C. Riggelsen (2009). Model selectionin seismic hazard analysis: An information-theoretic perspective.Bulletin of the Seismological Society of America 99, 3,234–3,247.Zhao, J. X., and M. Gerstenberger (2010). Comparison of attenuationcharacteristics between the data from two distant regions.Proceedings of the Ninth Pacific Conference on EarthquakeEngineering: Building an Earthquake-Resilient Society, 14–16 April2011, Auckland, New Zealand. Paper no. 008.Earthquake Science CenterU.S. Geological Survey345 Middlefield RoadMenlo Park, California 94025 U.S.A.msegkou@usgs.gov(M.S.)874 Seismological Research Letters Volume 82, Number 6 November/December 2011

Strong Ground Motions and Damage ConditionsAssociated with Seismic Stations in theFebruary 2011 Christchurch, New Zealand,EarthquakeHiroaki Iizuka, Yuki Sakai, and Kazuki KoketsuHiroaki Iizuka, 1 Yuki Sakai, 1 and Kazuki Koketsu 2INTRODUCTIONThe February 2011 Christchurch, New Zealand, earthquakewas highly destructive, causing a number of buildings to collapseand killing many people. We examined the properties ofstrong ground motions in this earthquake using the recordsreleased by GeoNet (http://www.geonet.org.nz/). We alsoinvestigated the damage around the seismic stations to determinethe relationship between structural damage and strongground motions.SEISMIC GROUND MOTION INTENSITIES ANDELASTIC RESPONSE SPECTRUMThe locations of the seismic stations in our study are shown inFigure 1. Accelerograms and the elastic acceleration responsespectra, with a damping factor of 0.05 in the maximum horizontaldirection, are shown in Figures 2 and 3, respectively. Peakground accelerations (PGA) and peak ground velocities (PGV)are shown in Table 1. I j and I 1–2 are also shown in Table 1. I j isJMA (Japan Meteorological Agency) seismic intensity (Tables2, 3 and 4). It is publicly used to describe the damaging powerof seismic shaking in Japan. I 1–2 is also an index like I j . It wasdefined by Sakai, Kanno, and Koketsu (2002, 2004) based onelastic responses between 1 and 2 seconds period that wereclosely related with heavy structural damage (the subscript 1–2means between 1 and 2 seconds) and represents the damagingpower of an earthquake much better than I j .As shown in Figures 2 and 3, the records of stationsREHS, CCCC, and PRPC display pulse waves with a periodof 1–2 seconds and have high response in the region of 1–2seconds, whereas stations HVSC and LPCC with large PGAare dominated by short periods below 1 second, and theirresponses between 1 and 2 seconds are low. We comparedREHS’s spectrum, which shows the highest response in the1. Graduate School of Systems and Information Engineering,University of Tsukuba, Japan2. Earthquake Research Institute, University of Tokyo, Japan▲▲Figure 1. Locations of the seismic stations.1–2-second period, with those at Takatori, Fukiai, and JMAKobe recorded in the 1995 Kobe earthquake, which devastatedthe city of Kobe and the surrounding region (Figure4). REHS’s response in the 1–2-second period is similar tothat of JMA Kobe, but it is lower than that of Takatori andFukiai.NONLINEAR SEISMIC RESPONSE ANALYSISWe performed a nonlinear seismic response analysis with asingle-degree-of-freedom system, accounting for Japanesereinforced concrete (RC) buildings (Kumamoto and Sakai2007) and wooden houses (Sakai and Iizuka 2009) and comparedthat data with REHS’s record, which shows the highestelastic response in the period of 1–2 seconds. We adopted theTakeda Model (Figure 5A; Takeda et al. 1970) for RC build-doi: 10.1785/gssrl.82.6.875Seismological Research Letters Volume 82, Number 6 November/December 2011 875

▲▲Figure 2. Accelerograms.▲▲Figure 3. Elastic acceleration response spectrum.876 Seismological Research Letters Volume 82, Number 6 November/December 2011

ings and the Modified Takeda-Slip Model (Figure 5B; Iizukaand Sakai 2009) for wooden houses as hysteresis models. Weset these model parameters, as shown in Table 5, to assumegeneral buildings and houses and an allowable ductility factorof 6, given the heavy damage to buildings (Sakai, Koketsu andKanno 2002; Sakai and Iizuka 2009).Figure 6 shows the required strength (base shear coefficient)spectrum of REHS compared with that of Takatori,Fukiai, and JMA Kobe. The required strength figure for REHSis higher than that of the JMA Kobe in the period longer than0.4 seconds. This shows that the ground motion recorded byREHS causes heavy damage to both RC buildings and woodenhouses. The strength of RC buildings based on design standardsestablished in New Zealand in 1965 (Figure 6A, thin line;Davenport 2004) is also shown in Figure 6. These standards aremuch lower than required by the spectrum recorded by REHS.This difference explains the fact that some RC buildings builtbefore 1976 collapsed in the February 2011 Christchurch, NewZealand, earthquake.DAMAGE INVESTIGATION AROUND THE SEISMICSTATIONSWe carried out damage investigations around the seismic stationsshown in Figure 1 to examine the relationship betweenstructural damage and strong ground motions. Our investigations,which took place on March 11 and 12, 2011, encompassedall buildings within 200 m of the stations. We recordedtype of structure, number of stories, and level of damage (heavyor not) (Okada and Takai 1999; Architectural Institute ofJapan 1980) and calculated a structural damage ratio based onthe collected data.Damage conditions are shown in Table 1. Building distributionand related damage levels around the seismic stationsis shown in Figure 7. Because it was sometimes difficultto distinguish wood from masonry houses by appearance, wenoted “wood or masonry” in Figure 7. Around REHS, therewere many buildings and houses that were heavily damaged, asshown in Figure 8. Roof tiles and exterior materials on severalhouses were damaged. Around HVSC, there were some buildingswith damage to roof tiles or exterior materials, and somemasonry structures were heavily damaged. Some houses outsidethe 200-m study area were also heavily damaged. AroundPRPC, many houses displayed damaged roof tiles, but therewere no heavily damaged buildings. LPCC seismic stationis located in the port and surrounded by the sea, so there arefew structures surrounding it. There are some houses locatedon a cliff to the northeast of the station, but we assumed thattheir location would have suffered ground motion that differedfrom that of the seismic station. We were unable to determinean accurate ground position for CCCC because CatholicCathedral College, where the seismic station is located, hadbeen closed and a restricted zone had been established on itswestern side. Therefore, we investigated only the area on thenorth side of the college. Most of the buildings in the area arewarehouses or stores, with few residences. We did note, however,that the roof of the one of the college buildings was damagedand there was a heavily damaged store nearby. AroundHPSC and SHLC, the vast majority of structures displayedonly insignificant damage, and there were no buildings withheavy damage. We also observed many sand boils resultingfrom liquefaction.We observed heavily damaged buildings around REHSand CCCC, where the response between 1 and 2 seconds wasTABLE 1Seismic Intensity of Strong Ground Motions and Damage Conditions around the Seismic Station.SeismicStation I*j I 1–2 †PGA[cm/s 2 ]PGV[cm/s] Damage Situation within 200mNumber ofBuildingsDamageRatio [%]REHS 6.20 6.19 723.0 97.7 Many buildings were heavy damaged. 98 15.3HVSC 6.42 5.66 1860.0 100.4 Destruction of roofing tiles and exterior materials. 21 9.1 ‡Some masonry buildings were broken.PRPC 6.10 6.00 737.0 124.2 Many buildings were damaged at roofing tiles. 90 0LPCC 5.77 5.01 1070.2 48.0 The ground damage of cliff. 17 5.9 §CCCC 5.98 6.04 472.5 68.2 Some buildings were heavy damaged. 23 8.7 ||HPSC 5.52 5.31 324.9 49.9 Destruction of exterior materials Liquefaction. 68 0SHLC 5.76 5.59 363.6 77.1 Crack of RC structure’s pillars.113 0Liquefaction.CHHC 5.79 5.75 460.0 82.7 — — —CMHS 5.77 5.65 415.3 50.2 — — —CBGS 5.83 5.71 653.3 73.5 — — —* I j : JMA seismic intensity† I 1–2 : Seismic intensity based on elastic responses between 1 and 2 second period‡ It is the reference values because the building was few.§ It is the reference values because the damageed building was on the cliff.|| It is the reference values because the seismic station position was uncertain.Seismological Research Letters Volume 82, Number 6 November/December 2011 877

TABLE 2Explanation of JMA seismic intensity scales (http://www.jma.go.jp/) for human perception and reaction; indoor situationsand outdoor situationsSeismicIntensity I jI j < 0.5Human Perception andReaction Indoor Situation Outdoor SituationImperceptible to people, but — —recorded by seismometers.— —keeping quiet in buildings.0.5 ≤ I j < 1.5 Felt slightly by some people1.5 ≤ I j < 2.5 Felt by many people keepingquiet in buildings. Some peoplemay be awoken.2.5 ≤ I j < 3.5 Felt by most people in buildings.Felt by some people walking.Many people are awoken.3.5 ≤ I j < 4.5 Most people are startled. Feltby most people walking. Mostpeople are awoken.4.5 ≤ I j < 5.0 Many people are frightenedand feel the need to hold ontosomething stable.5.0 ≤ I j < 5.5 Many people find it hard tomove; walking is difficult withoutholding onto somethingstable.5.5≤ I j < 6.0 It is difficult to remain standing.6.0 ≤ I j < 6.5 It is impossible to remainstanding or move withoutcrawling. People may bethrown through the air.Hanging objects such as lampsswing slightly.Dishes in cupboards may rattle.Hanging objects such as lampsswing significantly, and dishes incupboards rattle. Unstable ornamentsmay fall.Hanging objects such as lampsswing violently. Dishes in cupboardsand items on bookshelvesmay fall. Many unstable ornamentsfall. Unsecured furnituremay move, and unstable furnituremay topple over.Dishes in cupboards and itemson bookshelves are more likelyto fall. TVs may fall from theirstands, and unsecured furnituremay topple over.Many unsecured furniture movesand may topple over. Doors maybecome wedged shut.Most unsecured furniture moves,and is more likely to topple over.6.5 ≤ I j Most unsecured furniture movesand topples over, or may even bethrown through the air.—Electric wires swing slightly.Electric wires swing significantly.Those driving vehicles may notice thetremor.In some cases, windows may breakand fall. People notice electricitypoles moving. Roads may sustaindamage.Windows may break and fall, unreinforcedconcrete-block walls maycollapse, poorly installed vendingmachines may topple over, automobilesmay stop due to the difficulty ofcontinued movement.Wall tiles and windows may sustaindamage and fall.Wall tiles and windows are more likelyto break and fall. Most unreinforcedconcrete-block walls collapse.Wall tiles and windows are even morelikely to break and fall. Reinforcedconcrete-block walls may collapse.TABLE 3Explanation of JMA seismic intensity scales (http://www.jma.go.jp/) for wooden housesSeismicIntensity High Earthquake Resistance Low Earthquake Resistance4.5 ≤ I j < 5.0 — Slight cracks may form in walls.5.0 ≤ I j < 5.5 — Cracks may form in walls.5.5 ≤ I j < 6.0 Slight cracks may form in walls. Cracks are more likely to form in walls.Large cracks may form in walls.Tiles may fall, and buildings may lean or collapse.6.0 ≤ I j < 6.5 Cracks may form in walls. Large cracks are more likely to form in walls.Buildings are more likely to lean or collapse.6.5 ≤ I j Cracks are more likely to form in walls.Buildings may lean in some cases.Buildings are even more likely to lean or collapse.878 Seismological Research Letters Volume 82, Number 6 November/December 2011

high, but there were no buildings damaged around HPSCand SHLC where the 1–2-second response was low. In particular,there were many heavily damaged buildings, similarto what happened in the Kobe earthquake, that correspondedto the result of the nonlinear seismic response analysis aroundREHS. Thus, our investigation confirmed that the 1–2-secondresponse bore a close relationship to heavy damage to buildings.However, there were no buildings with heavy damagearound PRPC, where the 1–2-second response was almost thesame as that of CCCC, and there were some houses with heavydamage a little away from HVSC seismic station though it wasexpected that ground motions there would not cause heavydamage to houses. We surmise that these results arise becausethe majority of the buildings around the seismic station wereranch houses and thus likely made from masonry, and theperiod corresponding to damage of such buildings is shorterthan 1–2 seconds (Sakai and Nakamura 2004).▲▲Figure 4. Elastic acceleration response spectrum comparedwith the 1995 Kobe earthquake.▲▲Figure 5. Hysteresis models: A) Takeda Model and B) Modified Takeda-Slip Model.TABLE 4Explanation of JMA seismic intensity scales (http://www.jma.go.jp/) for reinforced-concrete buildingsSeismicIntensity High Earthquake Resistance Low Earthquake Resistance5.0 ≤ I j < 5.5 — Cracks may form in walls, crossbeams and pillars.5.5 ≤ I j < 6.0 Cracks may form in walls, crossbeams and pillars. Cracks are more likely to form in walls, crossbeams andpillars.6.0 ≤ I j < 6.5 Cracks are more likely to form in walls, crossbeamsand pillars.6.5 ≤ I j Cracks are even more likely to form in walls,crossbeams and pillars.Ground level or intermediate floors may sustainsignificant damage. Buildings may lean in somecases.Slippage and X-shaped cracks may be seen in walls, crossbeamsand pillars. Pillars at ground level or on intermediatefloors may disintegrate, and buildings may collapse.Slippage and X-shaped cracks are more likely to be seen inwalls, crossbeams and pillars.Pillars at ground level or on intermediate floors are morelikely to disintegrate, and buildings are more likely to collapse.TABLE 5Parameters of the Hysteresis Models.Hysteresis Characteristics Model α y Q c / Q y α β γ δTakeda-Model 0.25 0.30 0.50 0.01 — —Modified Takeda-Slip Model 0.30 0.30 0.50 0.15 3.00 1.00Seismological Research Letters Volume 82, Number 6 November/December 2011 879

▲▲Figure 6. Required strength spectrum compared with the 1995 Kobe earthquake: A) by Takeda Model for RC buildings, and B) byModified Takeda-Slip Model for wooden houses.▲▲Figure 7. Building damage distribution surrounding seismic stations.880 Seismological Research Letters Volume 82, Number 6 November/December 2011

REFERENCES▲▲Figure 8. Heavily damaged wooden house.CONCLUSIONSWe performed a response analysis using strong ground motionsand carried out damage investigation around seismic stationslocated within the area of the February 2011 Christchurch,New Zealand, earthquake. We confirmed the relationshipbetween the 1–2-second response and heavy damage to buildings.However, some results did not correlate with the level ofdamage to buildings. We believe that these seemingly incongruousoccurrences are explained by the popularity of masonryconstruction in New Zealand, because the period correspondingto damage of masonry buildings popular in New Zealandis shorter than 1–2 seconds.ACKNOWLEDGMENTSWe thank JMA, Osaka Gas, and the Railway TechnicalResearch Institute for providing strong ground motion recordsof the Kobe earthquake. We wish to express our gratitude tothe local people who cooperated with us during our investigations.Architectural Institute of Japan (AIJ), ed. (1980). Report on the DamageInvestigation of the 1978 Miyagiken-oki Earthquake. Tokyo:Architectural Institute of Japan.Davenport, P. N. (2004). Review of seismic provisions of historic NewZealand loading codes, 2004. New Zealand Society for EarthquakeEngineering Conference, 19–21 March 2004, Rotorua, NewZealand. Paper 17, Wellington, NZ: New Zealand Society forEarthquake Engineering.Iizuka, H., and Y. Sakai (2009). Proposal of hysteresis characteristicsmodel in seismic response analysis using single-degree-of-freedomsystem for wooden house. Journal of Japan Association forEarthquake Engineering 9 (1), 113–127.Kumamoto, T., and Y. Sakai (2007). Actual strength distribution ofRC buildings considering non-structural members. Summaries ofTechnical Papers of Annual Meeting Architectural Institute of Japan2007, B-2 311–312, Tokyo: Architectural Institute of Japan.Okada, S., and N. Takai (1999). Classifications of structural types anddamage patterns of buildings for earthquake field investigation.Journal of Structural and Construction Engineering (Transactions ofAIJ) 52, 65–72.Sakai, Y., and H. Iizuka (2009). A wooden house cluster model for earthquakedamage estimation by nonlinear response analyses. Journal ofJapan Association for Earthquake Engineering 9 (1), 32–45.Sakai, Y., T. Kanno, and K. Koketsu (2002). Method of calculating seismicintensities considering structural damage and human bodysense. 11th Earthquake Engineering Symposium, CD-ROM, paperno. 4, Tokyo: 11th Japan Earthquake Engineering Symposium.Sakai, Y., T. Kanno, and K. Koketsu (2004). Proposal of instrumentalseismic intensity scale from response spectra in various periodranges. Journal of Structural and Construction Engineering(Transactions of AIJ) 585, 71–76.Sakai, Y., K. Koketsu, and T. Kanno (2002). Proposal of the destructivepower index of strong ground motion for prediction of buildingdamage ratio. Journal of Structural and Construction Engineering(Transactions of AIJ) 555, 85–91.Sakai, Y., and Y. Nakamura (2004). Investigation on destructive powerindices of strong ground motions using building damage data andstrong ground motion records by the 1994 Northridge, California,earthquake. Journal of Structural and Construction Engineering(Transactions of AIJ) 584, 59–63.Takeda, T., M. A. Sozen, and N. N. Nielsen (1970). Reinforced concreteresponse to simulated earthquakes. ASCE Journal of the StructuralDivision 96 (ST12), 2, 557–2,573.Graduate School of Systems and Information EngineeringUniversity of Tsukuba, Japan1-13-13-205, Sakura, Tsukuba-shi, 305-0003, Japaniizuka_h@edu.esys.tsukuba.ac.jp(H. I.)Seismological Research Letters Volume 82, Number 6 November/December 2011 881

Ground Motions versus Geotechnical andStructural Damage in the February 2011Christchurch EarthquakeEleni Smyrou, Panagiota Tasiopoulou, İhsan Engin Bal, and George GazetasEleni Smyrou, 1 Panagiota Tasiopoulou, 1 İhsan Engin Bal, 2 and GeorgeGazetas 1INTRODUCTIONThe M w = 6.3 earthquake of February 22 was the strongest seismicevent in a series of damaging aftershocks in and aroundChristchurch after the Darfield earthquake on 4 September2010. The source of the Darfield earthquake was in a sparselypopulated area and thus it caused no loss of life. Serious damagewas mainly due to extensive liquefaction. By contrast, theChristchurch earthquake was generated on a fault in closeproximity to the city, resulting in a death toll of 181 people.The Canterbury Plains are covered with river gravels thathide any evidence of past fault activity in this region. The newlyrevealed Greendale fault was therefore completely unknown.Only a portion of it was revealed on the ground surface duringthe Darfield earthquake. The second fault (the one thatruptured in February 2011) appears to be a continuation of thefirst, although no fault structure directly connecting the faultshas been recognized. There is a debate among seismologists atthis point whether this is a different fault from Greendale oneor not (NHRP 2011a; NHRP 2011b; Geonet 2011).Due to its magnitude, shallow depth and close proximity tothe city, the February earthquake proved particularly destructivefor the central business district (CBD) of Christchurch,where buildings suffered extensive damage. Thanks to a densenetwork of strong ground motion stations, a large number ofrecords have been obtained, which provide valuable informationon the event and offer the possibility of relating the extentof damage to actual measurements of ground shaking.Apart from the southern part of the city on the hills andthe Lyttelton port area, Christchurch is built on deep estuarinesoil, which has been shaped in the last thousands of years by theever-changing riverbed. Fine sands—the dominant soil type—and the high ground water level contributed to widespreadliquefaction in one or both earthquake events. Often accompaniedby lateral spreading, liquefaction amplified the level ofdamage, resulting in the failure of structures in the CBD andsurrounding areas, as will be explained below.1. Soil Mechanics Laboratory, School of Civil Engineering, NationalTechnical University, Athens, Greece2. Fyfe Europe S.A. Athens, GreeceThe older buildings in the city center, many of which aremade of unreinforced masonry with timber floors, were mostlybuilt in the late 19th and early 20th century, following Englisharchitectural style and construction practice and with no considerationof the high seismicity of the region. However, someof these buildings had been retrofitted in recent years. In contrast,many of the modern buildings in the CBD were designedin accordance with recent seismic codes, although their foundationsystems were not always suitable for the adverse effectsstemming from liquefaction. Thus, despite the fact that liquefiedlayers beneath the CBD restricted somewhat the amplitudeof already significantly high accelerations, the increasedvelocities and displacements due to soil softening magnified thedemands on long-period structures. Both structural and geotechnicalaspects are investigated here in an effort to broadlyexplain and quantify the observed damage.THE STRONG MOTION RECORDSThanks to a dense network of seismographs covering the broaderarea of Christchurch (Figure 1), a large number of groundmotions were recorded during the Christchurch February2011 earthquake. The CBD area includes four seismic stations:CBGS, CCCC, REHS and CHHC. The first three records aretruly free-field motions. CHHC was located near the base of atwo-story building and its motion may reflect to some degreethe effect of the structure. These ground motions may not havebeen the strongest ones recorded in terms of PGA values; however,due to certain features, their effect on structures or soilswas detrimental.There is a certain variation in the recorded accelerationtime histories (Figure 1). For instance, the range of PGA valuesvaries within a factor of 2, from 0.34 g (CHHC-NS) to 0.72 g(REHS-EW). A dominant common feature in all records is thesign of liquefaction: long-period cycles with reduced accelerationamplitudes occurring after a threshold acceleration has beenreached. Soil softening due to excess pore water pressures in combinationwith sufficient acceleration values has led to amplificationof large periods affecting a broad category of structures, asindicated by the acceleration spectra. In particular, the spectral882 Seismological Research Letters Volume 82, Number 6 November/December 2011 doi: 10.1785/gssrl.82.6.882

(A)(B)(C)(D)(E)▲▲Figure 1. A) Map of the broader Christchurch area showing the intersection of the fault plane with the ground surface (from GNSScience), the location of the accelerograph stations, the epicenter of the Christchurch 2011 earthquake, and the location with availablesoil data. B–E) Acceleration time histories and spectra of four CBD (central business district) seismic stations for NS and EWdirections.Seismological Research Letters Volume 82, Number 6 November/December 2011 883

▲ ▲ Figure 2. Observed polarity for the records in the CBD in terms of peak ground acceleration, velocity, and displacement. The contoursof PGA on the map were computed by interpolation using all records in Christchurch.amplification at periods exceeding 2 sec is attributed to the factthat once liquefaction has occurred, the overlying soil “crust”oscillates with very low frequencies, causing the bulges observedin the acceleration spectra for periods of about 3 sec (see Youdand Carter 2005 for similar observations from the then-availableliquefaction-affected acceleration spectra). In addition to structuraldamage due to high spectral accelerations, important soilrelatedfailures have directly affected houses and bridges.THE POLARITY OF THE RECORDED MOTIONSThe two orthogonal components of a record are usually alignedwith the north-south and east-west directions (Figure 1) or,ideally, if the faults were known, with fault-parallel and faultnormaldirections. Mathematically there is at least one specificangle at which a certain ground motion parameter such asPGA, PGV, or PGD reaches a maximum, indicating the governingdirection for that ground motion parameter and revealinga certain polarity of the recorded motion. Polarity plots canbe useful in determining the dominant shaking direction of anearthquake and in unveiling any directivity effects (Shabestariand Yamazaki 2003).A first index of intensity is the value of peak ground acceleration(PGA), the spatial distribution of which is depicted onthe map of Figure 2. Additionally, for the records from the fourCBD stations (CCCC, REHS, CBGS, and CHHC) the maximumpeak values of ground acceleration, velocity, and displacementare calculated trigonometrically, by varying the angle by1° between 0° and 180°, resulting in asymmetric plots of positiveand negative maxima (in absolute terms). The graphs consistentlyexhibit distinct polarity in a direction that practicallycoincides with that of the fault line. Knowing the polarity ofshaking may offer information on the rupture mechanism andinsight into the dominant damage observed in the CBD area.TYPICAL SOIL PROFILE, LIQUEFACTION,ANALYSISThe Christchurch urban area, extending from Riccarton inthe west to Bexley in the east and reaching Heathcote Valleyand the Port Hills in the south, is located on the CanterburyPlains. Its dominant geomorphic feature is the river floodplains.In particular, the Avon (primarily) and Heathcote (secondarily)rivers, originating from various springs in westernChristchurch, form endless meanders through the city and theeastern suburbs as they head to the estuary near the sea.As depicted in Figure 3A, the subsoil in the CBD systematicallyconsists of profiles with random variations in layeringin the upper 15–25 m (Cubrinovski et al. 2010; Toshinawa etal. 1997). The volcanic bedrock is located at an approximate884 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A) (B) (C)▲ ▲ Figure 3. A) Typical in-depth soil profile in the CBD. B) Accelerograms and response spectra of the LPCC record used as excitation(applied in outcrop), and at two different depths obtained from the analyses. C) Polarity plots of LPCC record and output of analysis onthe ground surface.depth of 400 m and emerges on the surface at the southern borderof the Canterbury Plains, forming the Port Hills of BanksPeninsula. Thick layers of gravel formations overlay the bedrock(Brown and Weeber 1992). The surficial sediments havean average thickness of about 25 m and consist of alternatinglayers of alluvial sand, silt, and gravel. They have been depositedby overbank flooding (Eidinger et al. 2010)—hence, theirloose disposition. In the CBD, especially, sand and non-plasticsilt with low content of fines are the dominant soil types (Rees2010). The latter feature combined with the high ground waterlevel (from 0 to 3 m) below the center of the city explains thesensitivity to liquefaction.There is significant variability of soil deposits within shortdistances that can differentiate the ground motion characteristics.For example, Toshinawa et al. (1997) describe the soil profilesof two characteristic sites 1.2 km distant, one consistingof only sandy gravels and sand close to CBGS seismic station(Figure 1), and the other composed of silt and peat depositsto a depth of 7 m close to REHS seismic station. Accordingto Toshinawa et al. (1997), during a 1994 distant earthquakegreater amplification was observed at the second site, close toREHS, in agreement with the records of February 2011 (Figure1). This seems quite reasonable in cases of strong earthquakes,where the response of such soft, mostly sandy soils is expectedto be dominated by the effects of severe liquefaction. However,both sites belong to the same broader classification of soft soils(class D) for structural design purposes in the New Zealanddesign standards (New Zealand Standards 1170.5 2004).To investigate the soil response in the CBD urban areawhile accounting for liquefaction effects, we chose a typical“generic” soil profile (Figure 3A). Soil properties were obtainedfrom boreholes conducted close to the Fitzgerald Bridge, situatedat the eastern part of the CBD (see the star on the mapin Figure 1). Standard penetration test (SPT) values wereobtained from Bradley et al. (2009) and Rees (2010). Shearwave velocity, V s , values were based on empirical correlationswith SPT (Dikmen 2009).With the “generic” soil profile defined, dynamic effectivestressanalyses were conducted in order to capture the excesspore water pressure rise and dissipation, using the finite differencecode FLAC (Itasca Consulting Group 2005). Groundmotion recorded at station LPCC on the volcanic outcropSeismological Research Letters Volume 82, Number 6 November/December 2011 885

at Lyttelton Port was selected as the (outcrop) input motionreferred to the base of the gravel formations (Figure 3B). Thepresumption that this rock motion (the only one on [soft] rockin the area) is a suitable candidate for the base of the CBD isonly a crude approximation, because although the LPCC andCBD stations have the same distance from the about 65°-dippingfault, LPCC lies on the hanging wall and the CBD onthe footwall of this partly thrust and partly strike-slip fault.The NS and EW components of the LPPC record excited thesoil column in two different one-dimensional wave propagationanalyses. The numerical simulation involves the constitutivelaw of Byrne (1991) for pore pressure generation, which isincorporated in the standard Mohr-Coulomb plasticity model.In general, as one would expect, the results of the analysisin terms of acceleration time histories and acceleration spectrafor the two components (Figure 3B) demonstrate that as theshear waves propagate from the base of volcanic rock, the soilde-amplifies the low-period components of motion and amplifiesthose of high period. Moreover, the computed responseon top of the dense gravel formation indicates that there is nosubstantial influence of the gravel layer in altering the inputmotion, other than de-amplifying the values in the high-frequencyrange (above 5 Hz) and slightly amplifying lower frequencies.In addition, the peak ground acceleration values donot change.In contrast to the minor effect of gravel on the soil response,the surficial soil layers play a dominant role in defining theground motion characteristics—hardly a surprise. These layersbehave as a filter cutting off the high frequency spikes, whilethe duration of motion cycles is lengthened. As a result, thepeak accelerations have diminished to 0.35 g approximatelyin both directions. Moreover, in terms of spectral accelerationvalues, there is considerable spectral amplification to 1 g in thehigher period range of up to 1.8 sec. Overall, both componentsshow similar response, with certain disparities in the frequencycontent, e.g., N-S output is richer in higher periods.Polarity plots have also been constructed for LPCCmotion and the computed ground surface motion. They areportrayed in Figure 3C. Evidently, there is no single (common)dominant direction for all PGA, PGV, and PGD values, contraryto the consistency in polarity of the CBD records (Figure2). The PGA principal direction is normal (rather than parallel)to the fault. This discrepancy with CBD polarity might beattributed to the fact that Lyttelton is on the hanging wall sidewhile the CBD lies on the footwall. For the “thrust” componentof faulting this difference may indeed have an effect, butthis is an issue that needs further investigation and is beyondthe scope of this paper. The polarity of the output diverges onlyslightly from the polarity of LPCC. The comparison of polarityplots demonstrates clearly the cut-off of PGA values in alldirections and increase of PGV and PGD values. Evidently, theliquefied layers play the role of a seismic isolator, reducing theacceleration amplitude of the wave components propagatingthrough them.The occurrence of liquefaction is visible in the pattern ofrecorded ground acceleration time histories and is captured(with engineering accuracy) in our analysis: pore water pressureincreases during shaking, reaching the initial effective overburdenstress, σ′ νο . At that point onward the soil loses most of itsstrength and begins to behave as a heavy liquid mass filteringout the high-frequency components, cutting off the accelerationvalues, and allowing only (long–period) oscillations of thedry cover layer that is “floating” on the top of the liquefied layer.The ratio of the earthquake-generated (excess) pore waterpressure, Δu, normalized by the initial vertical effective overburdenstress, σ′ νo ,r u =uσ νo,approaches value of 1 at the onset of liquefaction. Values of r uabove 0.8 indicate that already large excess pore water pressureshave taken place and the soil has softened significantly. Figures4B and 4C depict the distribution of computed peak values ofr u (t) with depth and the time histories of r u (t) at five characteristicdepths. In detail, Figure 3B shows that liquefaction didoccur from 2.5 m to 17 m (r u > 0.8) throughout the silty sandlayer. The dense sand layer experienced some excess pore pressuresfrom flow from the overlying layer, but r u values were toolow for liquefaction, as depicted in the time history of r u at 18m (Figure 4C). In addition, according to the time histories of r uin Figure 4C, liquefaction occurs early, just 3 to 4 sec after thebeginning of shaking, which is close to the cut-off of accelerationsin the time histories shown in Figure 3B.To validate the analysis, the authors attempted a comparisonbetween real records and numerical results. The recordselected for the comparison, CBGS, is depicted on the map ofFigure 1. The CBGS station is located in the Botanic Gardensand the recorder is housed in a very light kiosk (Figure 5A). Thesigns of liquefaction sand boils are visible, although they hadbeen cleaned following the earthquake (the picture was takenby our research team in April 2011 [Tasiopoulou et al. 2011]).No other facilities exist in the surroundings, ensuring free fieldconditions. Moreover, the soil profile described by Toshinawaet al. (1997) is appropriate for this location.As already discussed, LPCC and the four CBD recordshave different polarity. However, LPCC was the only optionin the search for a rock outcrop motion to be used as excitationin our analysis. That is why the comparison of spectra has beenconducted in the direction of polarity of the CBGS record. Forexample, the strong PGA and PGV direction (polarity) forCBGS is approximately S56W and its PGD polarity is S51W.The acceleration time histories of the CBGS recorded and computedmotions in the direction of S56W are depicted in Figure5B. Although these time histories seem to differ, especially interms of PGA values, a closer look reveals that they have certaincommon features, better depicted in Figure 5C after filteringout components with frequency above 4 Hz. Notice in particularthat the main pulse at 4 sec exists in both time histories.The response spectral SA, SV, and SD are compared in Figure5D. The agreement of analysis with reality confirms that theanalysis achieves a realistic insight of the mechanisms of soilresponse during the Christchurch earthquake.886 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A) (B) (C)▲ ▲ Figure 4. A) Typical surficial soil deposit: layers and properties. B) Distribution of computed excess pore water pressure ratio r uwith depth. C) Computed time histories of r u at several depths.BUILDING CATEGORIES AND THE OBSERVEDDAMAGEBuilding Exposure in ChristchurchStructures in New Zealand exhibit great variety. Timber andmasonry buildings constitute around 80% of the building stock(Uma et al. 2008). Christchurch in particular has many oneortwo-story timber and masonry residential buildings outsidethe CBD and very few modern reinforced concrete (RC) highrisebuildings. The building composition in the CBD is different,with medium-rise modern steel and RC structures as wellas mid-rise unreinforced masonry (URM) and timber dwellingsand office buildings, some of which date from the late 19thand early 20th centuries. The one-story timber houses are commonlyfound in the suburbs surrounding the city, especiallythose along the Avon and Heathcote rivers.It can be said, roughly, that the area outside the CBD consistsof relatively low-rise and light structures, while long-periodstructures are more abundant in the CBD. This might be one ofthe reasons for the high concentration of damage in the CBDarea during the February 2011 earthquake. A preliminary studypresented below investigates the spatial distribution of the driftdemands of the recorded strong motions for a range of periods.Observed DamageMost of the casualties in the CBD were due to the collapse oftwo older mid-rise RC structures, called CTV and PGC, thefailure conditions of which are presently being investigated. Atthe time of writing the final statistics regarding the buildingsafety evaluation are not yet available. However, as of 18 March2011, the data by Civil Defence (Kam et al. 2011) referred to3,621 buildings checked within the CBD, out of which 1,933were posted red (needs to be demolished), 862 were posted yellow(has serious damage requiring extensive repair), and 826were posted green (needs minor repair in order to be usable).More specifically, 19% of the reinforced concrete structures,14% of the timber, and 7% of the steel buildings checked wereevaluated as red, while the equivalent percentages for reinforcedand unreinforced masonry structures were 16% and62%, respectively, reconfirming the poor behavior of URMstructures. Insufficient detailing and bad construction techniques,mostly related to non-structural elements, aggravatedthe damage. Although the aforementioned data have come upbefore the completion of the second phase of building safetyassessment and thus reflect the situation in CBD one monthafter the earthquake, they offer a representative picture of theextent and severity of damage in the CBD.Reasons behind the Extended Structural DamageThe demand imposed by the Christchurch earthquake on differentstructures is assessed in terms of maximum inter-storydrift demands (median values of all simulations done for themaximum values of all possible recording directions considered)in an attempt to broadly correlate ground motion featureswith the spatial distribution of damage. To this end, somecharacteristic buildings have been selected as representative ofSeismological Research Letters Volume 82, Number 6 November/December 2011 887

(A)(D)(B)(C)▲ ▲ Figure 5. A) The CBGS seismic station, with the remnants of liquefaction sand boils seen as scars on the grass. B) Accelerationtime histories of CBGS: record and analysis. C) Comparison of the above acceleration time histories after filtering them at 4 Hz. D)Comparison of 5% damped spectra between CBGS record and analysis.the building stock in the CBD, representing short-, medium-,and long-period structures. Our goal is not, obviously, to studyin detail certain structures but to “reconcile” earthquake damagewith ground motions. One- and two-story timber residentialhouses and two RC frame structures of different height,one six stories and one 17 stories, have been examined “generically”as described below in detail. Another case study couldselect URM buildings, a fairly representative typology in CBD,which suffered much from out-of-plane wall failures.The selected buildings are treated as reference structuresfor their category, while the variability in the structural characteristicswithin each structural category is assumed to follow astatistical distribution simulated through a Monte-Carlo algorithm.This approach is a necessity since at this stage detailedstructural data are not available. The parameters of the statisticaldistribution, i.e., mean value, coefficient of variation, typeof distribution, etc., are either taken from the available literatureor estimated using engineering judgment guided by the(macroscopic) visual inspection. The assumed values, as well asthe relative references for each parameter and structural categoryexamined, are summarized in Table 1.Having created a large number of simulated buildings, weapplied the displacement-based assessment procedure establishedby Priestley et al. (2007) to evaluate the demand on eachbuilding. This is then translated to displacement demands foreach floor and to inter-story drifts, utilizing the displacementprofiles proposed in Priestley et al. (2007). The method is basedon the substitute-structure theory, first suggested by Gülkan andSözen (1974) and Shibata and Sözen (1976), according to whichan inelastic multi-degree-of-freedom (MDOF) system can berepresented by an equivalent inelastic single-degree-of-freedomsystem (SDOF). The only aspect of our methodology that, outof necessity, deviates from the Priestley et al. (2007) is that the“yield period” of each structural category is based on literaturesuggestions rather than an initial estimate of stiffness and themass of each specific building. The “yield period” refers to thestiffness at the point of yielding, which is the limit beyond whichsubstantial inelastic response begins that eventually may lead to888 Seismological Research Letters Volume 82, Number 6 November/December 2011

Case StudyKey ParametersTABLE 1Key Parameters Used in the Representative Analyses1- and 2-story timber Displacement limit statesEquivalent viscous damping equationRatio of the first yield to the base shear coefficientStory height6-story RC frame17-story RC frameBeam depthBeam lengthRebar yield strengthYield-period equationEquivalent viscous damping equationBeam depthBeam lengthRebar yield strengthYield-period equationEquivalent viscous damping equationRanges Used in MonteCarlo Simulations *µ = 8 mm, CoV = 0.15, [N]Deterministica = 0.5, b = 0.8, [U]a = 2.8 m, b = 3.1 m, [U]µ = 0.8 m, CoV = 0.15, [N]µ = 7.0 m, CoV = 0.15, [N]µ = 330 MPa, CoV = 0.15, [N]DeterministicDeterministicµ = 0.6 m, CoV = 0.15, [N]µ = 5.0 m, CoV = 0.15, [N]µ = 330 MPa, CoV = 0.15, [N]DeterministicDeterministicReference for the KeyParametersUma et al. 2008NZSEE 2006ATC 1996Field dataTasiopoulou et al. 2011Tasiopoulou et al. 2011Uma et al. 2008Crowley et al. 2004Priestley et al. 2007Galloway et al. 2011Galloway et al. 2011Uma et al. 2008Crowley et al. 2004Priestley et al. 2007* µ: mean, CoV: coefficient of variation, [N]: Normal distribution, [U]: Uniform distribution, a and b: limits of the uniform distribution.significant damage. The yield period has been successfully usedas a key parameter in performance assessment by Crowley et al.(2004) and Bal et al. (2010). References for the parameters usedfor each category of buildings are given in Table 1.To ensure that the maximum displacement demand is estimated,the components for each record have been rotated inincrements of 1° degree from 0° to 180°, thus creating a newset of 180 records and the corresponding response spectra. Thecontours of the maps presented in the paper (see Figures 6 to8) have been derived after assessing each simulated buildingfor a total of 180 response spectra. Note that the inter-storydrift demands have been calculated only at the position of therecording stations, as shown on the maps in Figures 6 to 8. Thevalues presented between the stations are only the result of linearinterpolation among several “anchor” points. Obviously,the interpolation in these figures is bound by the coastline andcannot be extended to Kaiapoi and to Lyttelton Port stations.Short-period structures are mostly timber buildings.Such two-story houses are found in the CBD, while one-storyhouses are outside the CBD and in the suburbs. They are nonengineeredbuildings, with few if any exceptions; local regulationsallow simple timber houses to be constructed without anapproved design. Both groups were significantly damaged, butonly a few collapsed. However, the damage to such houses dueto liquefaction-induced ground differential settlements andhorizontal displacements was unprecedented. A generic buildinghas been used in this study as a reference structure. Theproperties of this generic structure are taken from the workby Uma et al. (2008), in which the story drift limits are givenas 0.3%, 0.6%, 1.2%, and 1.6% for slight, moderate, significantdamage, and collapse limit states. Details of the assumedparameters can be found in Table 1.There are several commercial buildings in the CBD, mostof which are mid-rise RC structures designed and built in the1970s and 1980s when the developed modern design conceptshad only partially (at best) been incorporated in codes. A specificbuilding from Kilmore Street (Markham’s Building),shown in Figure 9, is used for generating an ensemble of similarbuildings for moderately long-period structures. Despitewidespread liquefaction in the area, its pile foundation helpedto limit the damage; thus, the results presented below refer tosimilar buildings founded on stable upper soil layers. The finalcase study is a real building in Worchester Street, known asClarendon Tower, which has been reported to have undergonesignificant but repairable damage in the February earthquake.It is a regular moment-resisting frame structure, the details ofwhich are given in Galloway et al. (2011).The spatial distribution of the mean values of inter-storydrift demands in Christchurch for two-story timber structures(Figure 6), computed using the approach described above,clearly suggests that there must have been concentration of theinter-story demand in and near the Heathcote Valley wherethe strongest recorded shaking (HVSC) in terms of PGAand low-period SA and SD took place. On the contrary, damagein the area of the CBD must have been somewhat lighter,apparently due to the smaller low-period SD in the CBD. Suchdifferences can be attributed to the somewhat larger distancefrom the source and the fact that the soft soils de-amplified theshort-period seismic waves. But still, the median inter-storydrift demands in the CBD are computed to have been in theorder of 1.0%–1.5% for two-story timber structures, a level ofdemand definitely sufficient to induce substantial structuraldamage. Indeed, observations from different parts of the CBDon a variety of timber two-story structures confirm this theo-Seismological Research Letters Volume 82, Number 6 November/December 2011 889

▲▲Figure 6. Computed median inter-story drift demands (%) on two-story timber structures:maximum of all possible recording directionsis considered (coefficient of variation of the results: 31%).▲▲Figure 7. Computed median inter-story drift demands (%) on six-story RC frame structures:maximum of all possible recording directionsis considered (coefficient of variation of the results: 19%).▲ ▲ Figure 8. Computed median inter-story drift demands (%) on 17-story RC frame structures:maximum of all possible recording directionsis considered (coefficient of variation of the results: 19%).890 Seismological Research Letters Volume 82, Number 6 November/December 2011

▲ ▲ Figure 9. Representative buildings for the considered categories: timber residential house (left), mid-rise RC structure (middle), andtall RC structure (right).retical finding. Analyses on one-story timber structures withsimilar assumptions showed that the expected drift demandswere quite limited, thus explaining the low damage ratio ofone-story houses.The computed spatial distribution of the median inter-storydrift demands for six-story RC frames is portrayed in Figure 7,which shows that the highest demands on such structures, inthe order of 3.0–3.5% inter-story drifts, are concentrated in theCBD—which explains the damage on mid-rise RC structuresduring the February 22 earthquake. Interestingly, the computeddamage potential for such structures specifically reaches its climaxin the CBD. Readers are reminded that both the CTVand PGC buildings, which fatally collapsed from the Februaryshaking, were mid-rise RC structures constructed on soft soil.Tall RC frame structures in the CBD, though limited innumber, also experienced some extent of damage with the mostcharacteristic case being that of the Grand Chancellor Hotel, a26-story wall-frame structure that was rendered unusable dueto large residual displacement. The building we chose to lookat, the Clarendon Tower, underwent significant but repairabledamage. Nevertheless, the building will be demolished as notmeeting insurance standards. The findings illustrated in Figure8 exhibit 0.6% to 0.8% median inter-story drift demands forsimilar 17-story structures, a drift level that certainly translatesinto damage but remains below the unrepairable drift limits inline with field observations (Tasiopoulou et al. 2011).The inter-story drift demands reach their peak in the CBDas shown in Figure 8, a fact that could arguably be attributedto the characteristic bulges appearing in the response spectra(Figure 1) in the range of long periods. The elastic fundamentalperiod of such structures is estimated around 2 sec, whilethe yield period (beyond which significant damage arises), isof the order of 5 sec (Crowley et al. 2004). The effective secantperiod, for example, is expected to elongate up to about 7 secin the case of an overall ductility equal to 2, as computed withthe approximate expression for effective period of Priestley andKowalsky (2000).CONCLUSIONSDamaging earthquakes feature large variations in spatial distributionof the strong ground motion parameters, a fact that ismostly attributed to the complexity of source mechanism, radiationpattern, and site conditions. The M w = 6.3 Christchurchearthquake was a surprising and unusual event which occurredin an unknown fault that had already been awakened by theSeptember 2010 stronger earthquake, and it had a strong thrustcomponent and a steeply dipping plane.This paper has attempted to identify quantifiable parametersthat could provide better insight to seismologists and engineerswho try to systematically investigate the reasons behindthe structural and soil failures that occurred in the Februaryshaking. The study focuses on connecting the basic features ofthe recorded strong motions to the nonlinear behavior of thesoil layers. Liquefaction, a phenomenon that played a majorand devastating role, has been examined through a “generic”downtown soil profile and dynamic effective stress analysis.The LPCC record was applied as the base excitation, as it wasthe only available rock outcrop motion. Despite several uncertainties,the output spectra obtained from the liquefactionanalyses and the one recorded in the free field in the BotanicGardens have shown quite a satisfactory match provided thatthe compared spectra are aligned with the strong direction ofthe recorded motion. The dominant direction of the CBGSrecord is consistently almost parallel to the fault plane whilethe Lyttelton record exhibits more inconsistencies, somethingthat may be related to the effects of the hanging wall and thesteep thrust-fault plane. The governing direction of each recordhas been found by simply turning the record in every possibledirection with one-degree intervals and re-recording the strongmotion parameters sought—a venerable procedure to uncoverthe dominant direction of the shaking of a given site.The paper concludes with an effort to better explain thereasons why some particular structural types showed bad performancein the CBD area. Short, medium, and long periodSeismological Research Letters Volume 82, Number 6 November/December 2011 891

structures have been examined adopting a displacement-basedprocedure. Results show that the inter-story drift demands inthe CBD were particularly damaging for all types of structuresbut especially catastrophic for mid-rise RC buildings on shallowfoundations. This is an important finding that may contribute tounderstanding why the CTV and PGC buildings collapsed.ACKNOWLEDGMENTSFinancial support for the expedition to the earthquakestrickenarea and the work outlined in this paper has beenprovided under the research project “DARE,” funded throughthe “IDEAS” Programme of the European Research Council(ERC) under contract number ERC-2-9-AdG228254-DARE.The authors would like to thank Professors John Berrill, MiskoCubrinovski, Stefano Pampanin, and Dr. Umut Akgüzel forproviding data and assisting the authors during their reconnaissancevisit in Christchurch in April 2011.REFERENCESApplied Technology Council (ATC) (1996). Seismic Evaluation andRetrofit of Concrete Buildings. ATC-40 Report, vols. 1 and 2.Redwood City, CA: Applied Technology Council.Bal, İ. E., J. J. Bommer, P. J. Stafford, H. Crowley, and R. Pinho (2010).The Influence of Geographical Resolution of Urban Exposure Datain an Earthquake Loss Model for Istanbul, Earthquake Spectra 26(3), 619–634.Bradley, B. A., M. Cubrinovski, R. P. Dhakal, and G. A. MacRae (2009).Probabilistic seismic performance assessment of a bridge-foundation-soilsystem. Soil Dynamics and Earthquake Engineering 30,395–411.Brown L. J., and J. H. Weeber (1992). Geology of the Christchurch UrbanArea. Institute of Geological and Nuclear Sciences, Scale 1:25,000,Geological Map 1, New Zealand. Lower Hutt, New Zealand: GNSScience.Byrne, P. (1991). A cyclic shear-volume coupling and pore-pressure modelfor sand. Proceedings of the Second International Conference onRecent Advances in Geotechnical Earthquake Engineering and SoilDynamics, St. Louis, Missouri, 47–55.Crowley, H., and R. Pinho (2004). Period-height relationship for existingEuropean reinforced concrete buildings. Journal of EarthquakeEngineering 8 (S1), 305–332.Cubrinovski, M., R. Green, J. Allen, S. Ashford, E. Bowman, B. Bradley,B. Cox, T. Hutchinson, E. Kavazanjian, R. Orense, M. Pender, M.Quigley, T. Wilson, and L. Wotherspoon (2010). Geotechnicalreconnaissance of the 2010 Darfield (New Zealand) earthquake.Bulletin of the New Zealand Society for Earthquake Engineering 43,243–320.Dikmen, Ü. (2009). Statistical correlations of shear wave velocityand penetration resistance for soils. Journal of Geophysics andEngineering 6, 61–72.Eidinger, J., A. Tang, and Thomas O’Rourke (2010). Technical Councilon Lifeline Earthquake Engineering (TCLEE), Report of the 4September 2010 Mw 7.1 Canterbury (Darfield), New ZealandEarthquake. Reston, VA: American Society of Civil Engineers.Galloway, B. D., H. J. Hare, and D. K. Bull (2011). Performance ofmulti-storey reinforced concrete buildings in the Darfield earthquake.Proceedings of the Ninth Pacific Conference on EarthquakeEngineering—Building an Earthquake-Resilient Society, 14–16April, 2011, Auckland, New Zealand, paper no. 168.Geonet (2011). Christchurch badly damaged by magnitude 6.3 earthquake(22 February 2011), http://www.geonet.org.nz.Gülkan, P., and M. Sözen (1974). Inelastic response of reinforced concretestructures to earthquake motions. ACI Journal 71 (12), 604–610.Itasca Consulting Group (2005). Fast Lagrangian Analysis of Continua.Minneapolis, MN: Itasca Consulting Group Inc.Kam, W. Y., U. Akguzel, and S. Pampanin (2011). 4 Weeks on:Preliminary Reconnaissance Report from the Christchurch 22Feb 2011 6.3M w Earthquake. Report, New Zealand Society forEarthquake Engineering Library, Wellington, New Zealand.Natural Hazards Research Platform (NHRP) (2011a). Why the 2011Christchurch earthquake is considered an aftershock, http://www.naturalhazards.org.nz.Natural Hazards Research Platform (NHRP) (2011b). Magnitude 6.3earthquake not on Greendale Fault, http://www.naturalhazards.org.nz.New Zealand Society for Earthquake Engineering (NZSEE) (2006).Assessment and Improvement of the Structural Performance ofBuildings in Earthquakes, New Zealand Society for EarthquakeEngineering.New Zealand Standards 1170.5 (2004). Structural Design Actions, Part5: Earthquake Actions—New Zealand. Wellington, New Zealand:Standards New Zealand, 82 pp.Priestley, M. J. N., G. M. Calvi, and M. J. Kowalsky (2007). DisplacementbasedSeismic Design of Structures. Pavia, Italy: IUSS Press.Priestley, M. J. N., and M. J. Kowalsky (2000). Direct displacement-basedseismic design of concrete buildings. Bulletin of the New ZealandNational Society for Earthquake Engineering 33 (4), 421–444.Rees, S. D. (2010). Effects of fines on the un-drained behavior ofChristchurch sandy soils. PhD thesis, Civil and Natural ResourcesEngineering, University of Canterbury, Christchurch, New Zealand.Shabestari, K. T., and F. Yamazaki (2003). Near-fault spatial variationin strong ground motion due to rupture directivity and hangingwall effects from the Chi-Chi, Taiwan earthquake. EarthquakeEngineering and Structural Dynamics 32, 2,197–2,219.Shibata, A., and M. Sözen (1976). Substitute structure method for seismicdesign in reinforced concrete. ASCE Journal of the StructuralDivision 102 (ST1), 1–8.Tasiopoulou, P., E. Smyrou, İ. E. Bal, G. Gazetas, and E. Vintzileou (2011).Geotechnical and Structural Field Observations from Christchurch,New Zealand, Earthquakes. Research Report, National TechnicalUniversity of Athens, Greece.Toshinawa, T., J. J. Taber, and J. B. Berrill (1997). Distribution ofground-motion intensity inferred from questionnaire survey, earthquakerecordings, and microtremor measurements: A case study inChristchurch, New Zealand, during the 1994 Arthurs Pass earthquake.Bulletin of the Seismological Society of America 87, 356–369.Uma, S. R., J. Bothara, R. Jury, and A. King (2008). Performance assessmentof existing buildings in New Zealand. Proceedings of the NewZealand Society for Earthquake Engineering Conference, Wairakei,New Zealand, 11–13 April, paper no. 45.Youd, T. L., and B. L. Carter (2005). Influence on soil softening and liquefactionon spectral acceleration. ASCE Journal of Geotechnicaland Geoenvironmental Engineering 131 (7), 811–825.Soil Mechanics LaboratorySchool of Civil EngineeringNational Technical UniversityHeroon Polytechneiou 9Zografou CampusAthens 15780 Greecesmiroulena@gmail.com(E. S.)892 Seismological Research Letters Volume 82, Number 6 November/December 2011

Soil Liquefaction Effects in the CentralBusiness District during the February 2011Christchurch EarthquakeMisko Cubrinovski, Jonathan D. Bray, Merrick Taylor, Simona Giorgini, Brendon Bradley, Liam Wotherspoon, and Joshua ZupanMisko Cubrinovski, 1 Jonathan D. Bray, 2 Merrick Taylor, 1 SimonaGiorgini, 1 Brendon Bradley, 1 Liam Wotherspoon, 3 and Joshua Zupan 2INTRODUCTIONDuring the period between September 2010 and June 2011,the city of Christchurch was strongly shaken by a series ofearthquakes that included the 4 September 2010 (M w = 7.1),26 December 2010 (M w = 4.8), 22 February 2011 (M w = 6.2),and 13 June 2011 (M w = 5.3 and M w = 6.0) earthquakes. Themoment magnitude (M w ) values adopted in this paper are takenfrom GNS Science, New Zealand (http://www.geonet.org.nz);they are 0.1 units higher than the corresponding M w valuesreported by the U.S. Geological Survey (http://earthquake.usgs.gov/earthquakes/eqinthenews/2011/usb0001igm/). Theseearthquakes produced strong ground motions within the centralbusiness district (CBD) of Christchurch, which is the centralheart of the city just east of Hagley Park and encompassesapproximately 200 ha. Some of the recorded ground motionshad 5% damped spectral accelerations that surpassed the 475-year return-period design motions by a factor of two. Groundshaking caused substantial damage to a large number of buildingsand significant ground failure in areas with liquefiablesoils. The 22 February earthquake was the most devastating.It caused 181 fatalities and widespread liquefaction and lateralspreading in the suburbs to the east of the CBD and in areaswithin the CBD, particularly along the stretch of the AvonRiver that runs through the city. There were pockets of heavydamage in the CBD, including the collapse of two multistoryreinforced concrete buildings, as well as the collapse and partialcollapse of many unreinforced masonry structures includingthe historic Christchurch Cathedral in the center of the CBD.Soil liquefaction in a substantial part of the CBD adverselyaffected the performance of many multistory buildings, resultingin global and differential settlements, lateral movement offoundations, tilt of buildings, and bearing failures.The M w = 6.2, 22 February 2011 earthquake is especiallymeaningful for earthquake professionals because it occurredjust five months after the M w = 7.1, 4 September 2010 Darfield1. University of Canterbury, Christchurch, New Zealand2. University of California, Berkeley, California, U.S.A.3. University of Auckland, Auckland, New Zealandearthquake, the epicenter of which was approximately 40 kmfrom the Christchurch CBD. Whereas the 22 February eventkilled almost two hundred people, the 4 September eventresulted in no deaths. Although the September event causedwidespread liquefaction-induced damage in the Christchurcharea, it did not cause significant liquefaction-induced damagewithin the CBD. There is much to learn from comparingthe different levels of soil liquefaction, differing magnitudesand seismic source distances, and variable performanceof buildings, lifelines, and engineered systems during thesetwo earthquakes. It is rare to have the opportunity to documentthe effects of one significant earthquake on a moderncity with good building codes. It is extremely rare to have theopportunity to learn how the same ground and infrastructureresponded to two significant earthquakes.This paper summarizes the key field observations madefollowing the 22 February 2011 Christchurch earthquakeregarding the effects of soil liquefaction on building performancein the CBD. Other papers in this special issue provideinformation on earthquake ground motions and the geotechnicaleffects of this event outside the CBD. Additionally, theeffects of the 4 September 2010 Darfield earthquake were documentedpreviously (e.g., Cubrinovski et al. 2010). After a briefoverview of the CBD, we describe the typical soil conditions inthe CBD, followed by a summary of recorded ground motionsin the CBD. There are several cases of buildings with differentfoundation types (e.g., isolated spread footings, spread footingswith grade beams, raft foundations, and pile foundations) thatperformed differently in liquefied ground. Representative casesof building performance on liquefied ground are described toprovide insights regarding the effects of soil liquefaction onurban areas with modern construction.CHRISTCHURCH CENTRAL BUSINESS DISTRICTChristchurch is situated in the middle part of the east coastof the South Island of New Zealand. It has a population ofabout 350,000 (the second-largest city in New Zealand). Itsurban area covers approximately 450 km 2 . It is sparsely developedwith approximately 150,000 dwellings (predominantlydoi: 10.1785/gssrl.82.6.893Seismological Research Letters Volume 82, Number 6 November/December 2011 893

Ground motion recording station with geo-meanpeak horizontal ground accelerations from 4 Sept.2010 event (left) and 22 Feb. 2011 event (right).Christchurch CathedralFigure 2 Section LineChristchurch CBDFinite fault model for 22 Feb2011 M w = 6.2 event. Referto Beavan et al. (this issue)and Bradley and Cubrinovski(this issue) for more detail.▲▲Figure 1. Christchurch CBD relative to subsurface fault rupture of 22 February 2011 event.single-story timber-framed houses with a smaller number oftwo-story houses) spread across a large area with many parks,natural reserves, and recreation grounds. The CBD is the areaencompassed by the four avenues, Rolleston to the west, Bealeyto the north, Fitzgerald to the east, and Moorhouse to thesouth. The CBD is more densely developed, including multistorybuildings in its central area, a relatively large number ofhistoric masonry buildings dating from the late 19th and early20th century, residential buildings (typically two- to five-storystructures located north of Kilmore Street), and some industrialbuildings in the south and southeastern parts of the CBD.In total, about 3,000 buildings of various heights, constructionage, and structural systems were within the CBD boundariesbefore the 2010–2011 earthquake series. Latest estimatesindicate that about 1,000 of these buildings will have to bedemolished because of excessive earthquake damage. Figure 1outlines the boundaries of the CBD and the approximate locationof the causative fault of the 22 February 2011 earthquake.LOCAL GEOLOGYThe city of Christchurch is located on Holocene deposits ofthe Canterbury Plains, except for its southern edge, which islocated on the slopes of the Port Hills of Banks Peninsula,the eroded remnant of the extinct Lyttelton Volcano, composedof weathered volcanic rock (basalt) and thick depositsof Pleistocene loess. The river floodplain, Pacific coastline,and the Port Hills are the dominant geomorphic featuresof the Christchurch urban area. The Canterbury Plains arecomplex fans deposited by eastward-flowing rivers from theSouthern Alps, a NS-trending mountain range, into PegasusBay on the Pacific coast. The fan surfaces cover an area 50 kmwide × 160 km long. At Christchurch, surface postglacialsediments have a thickness between 15 and 40 m and overlie300–400-m-thick interlayered formations of gravels and fineto very fine grained sediments, representing deposition duringepisodic glacial and interglacial periods, and together comprisea series of ground water aquifers (Brown and Weeber 1992).As shown in Figure 2, the surface sediments are made up offluvial gravels, sands, and silts (Springston Formation, with amaximum thickness of 20 m to the west of Christchurch) orestuarine, lagoon, beach, dune, and coastal swamp deposits ofsand, silt, clay, and peat (Christchurch Formation, with a maximumthickness of 40 m on the coast at New Brighton, east ofthe CBD). The shallow soil deposits (i.e., depths of up to 15–20m) vary significantly within short distances, both horizontallyand vertically.Brown and Weeber (1992) describe the original site conditionsand development of Christchurch as follows:Originally the site of Christchurch was mainly swamplying behind beach dune sand; estuaries and lagoons,and gravel, sand and silt of river channel and flooddeposits of the coastal Waimakariri River flood plain.The Waimakariri River regularly flooded Christchurchprior to stopbank construction and river realignment.Since European settlement in the 1850s, extensivedrainage and infilling of swamps has been undertaken.894 Seismological Research Letters Volume 82, Number 6 November/December 2011

▲ ▲ Figure 2. Representative subsurface cross-section of Christchurch CBD along Hereford Street (modified from I. McCahon, personalcommunication, 19 July 2011).Brown and Weeber also state that surface deposits are activelyaccumulating and that the present-day river channel depositsare excluded from the above-mentioned Christchurch andSpringston formations.Canterbury has an abundant water supply through riversand streams and rich aquifers. The dominant features ofpresent-day Christchurch are the Avon and Heathcote rivers,which originate from springs in western Christchurch, meanderthrough the city, and feed the estuary at the southeast endof the city. As illustrated in Figure 3, the Avon River meandersthrough the CBD while the Heathcote River flows south of theCBD. The groundwater table is deepest at the west end of thecity (i.e., about 5 m depth), becoming progressively shallowereastward (i.e., within 1.0–1.5 m of the ground surface for mostof the city east of the CBD), and approaching the ground surfacenear the coastline. The water table is generally within 1.5to 2.0 m of the ground surface within the CBD.GROUND MOTION CHARACTERISTICSThe 4 September 2010 M w = 7.1 Darfield earthquake was causedby a complex rupture of several fault segments, the largest andnearest to Christchurch being on the Greendale fault about20 km west of the CBD. A maximum horizontal peak groundacceleration (PGA) of 0.24 g was recorded in the CBD, and thePGA decreased generally with distance downstream along theAvon River. The M w = 6.2, 22 February 2011 Christchurchearthquake was less than 10 km from the CBD along the southeasternperimeter of the city in the Port Hills (Figure 1). Theclose proximity of this event caused higher-intensity shaking inthe CBD as compared to the Darfield earthquake. Several ofthe recordings exhibited forward-directivity significant velocitypulses. In the CBD, horizontal PGAs of between 0.37 gand 0.52 g were recorded. There are four strong motion stationslocated within or very close to the CBD (Figure 1). The recordedPGAs at these four stations are summarized in Table 1 for thefive earthquakes producing highest accelerations (Bradley andCubrinovski 2011, page 853 this issue).For the shallow part of a deposit, the variation in therecorded PGA values corresponds closely with variations in thecyclic stress ratio (CSR) for each of these events. Magnitudescaling factors can then be applied to adjust each calculated CSRvalue to an equivalent value for an M w = 7.5 event (CSR M7.5 ) assummarized in Table 1 for the geometric mean horizontal valuesof the PGA (Bradley and Cubrinovski 2011, page 853 ofthis issue). The data show that in addition to the high PGAs duringthe 22 February 2011 earthquake (PGA = 0.37–0.52 g), theCBD buildings were subjected to significant PGAs in the rangeof 0.16–0.27 g in four additional events. The highest adjustedCSR 7.5 values of 0.14–0.20 were obtained for the M w = 6.2, 22February 2011 earthquake, which were about 1.6 times the correspondingCSR-values from the M w = 7.1, 4 September 2010Darfield earthquake. At many sites in Christchurch liquefactionre-occurred during these earthquakes, which in conjunctionwith the numerous smaller aftershock records providesinvaluable data both in terms of thresholds for liquefactiontriggering and CSR levels responsible for producing damagingliquefaction. In the CBD itself, only isolated areas liquefied duringmultiple events. Instead, widespread liquefaction occurredonly in the CBD during the 22 February 2011 earthquake.LIQUEFACTION IN THE CBDImmediately after the 22 February 2011 earthquake (i.e., from23 February to 1 March) an extensive drive-through reconnaissancewas conducted to map liquefaction and to document theSeismological Research Letters Volume 82, Number 6 November/December 2011 895

Christchurch CBDAvon RiverHeathcote RiverChristchurch Cathedral▲▲Figure 3. Liquefaction documentation map of eastern Christchurch from drive-through reconnaissance.TABLE 1Geometric mean PGAs and adjusted cyclic stress ratios to M w = 7.5 earthquake (CSR 7.5 ) for four strong motion stationswithin/close to CBD, for five earthquakes in the period September 2010–June 2011Geometric Mean PGA (g) Cyclic Stress Ratio CSR*7.5Magnitude ScalingEventCBGS CCCC CHHC REHS CBGS CCCC CHHC REHS Factor MSF †4 SEP 10, M w = 7.1 0.158 0.224 0.173 0.252 0.089 0.127 0.098 0.142 1.1526 DEC 10, M w = 4.8 0.270 0.227 0.162 0.245 0.097 0.082 0.058 0.088 1.8022 FEB 11, M w = 6.2 0.501 0.429 0.366 0.522 0.199 0.170 0.145 0.208 1.6313 JUN 11, M w = 5.3 0.183 – 0.199 0.188 0.066 – 0.072 0.068 1.8013 JUN 11, M w = 6.0 0.163 – 0.215 0.264 0.060 – 0.079 0.097 1.77* CSR 7.5 = 0.65 (PGA/g)/MSF (at depth of groundwater)† MSF = 10 2.24 / M2.56w ≤ 1.8 (corresponding to the lower bound range recommended in Youd et al. 2001, with a cap of 1.8)]severity of its manifestation across Christchurch. The drivethroughsurvey aimed at capturing surface evidence of liquefactionas quickly as possible and quantifying its severity in aconsistent and systematic manner. The resulting liquefactiondocumentation map is shown in Figure 3. Three areas of differentliquefaction severity are indicated: A) moderate to severeliquefaction (black zone), B) low to moderate liquefaction (darkgray zone), and C) liquefaction predominantly on roads withsome on properties (light gray zone). Traces of liquefactionwere also observed in other areas. The suburbs to the east of theCBD along the Avon River (Avonside, Dallington, Avondale,Burwood, and Bexley) were most severely affected by liquefaction.About 5,000 residential properties in these suburbs willbe abandoned (New Zealand Government 2011). There wasalso substantial damage in areas of the CBD, and many heavilydamaged structures will require retrofit or demolition.Ten days after the earthquake, after the urban search andrescue efforts had largely finished, the authors initiated a comprehensiveground survey within the CBD to document liquefactioneffects in this area. Figure 4 shows a preliminary liquefactiondocumentation map for the CBD. This paper focusesprimarily on key observations of building performance withinthe principal liquefied zone stretching west-east through theCBD, from Hagley Park to the west, along the Avon River tothe northeast boundary of the CBD at the Fitzgerald Bridge.This zone is of particular interest because many high-risebuildings on shallow foundations and deep foundations wereaffected by the liquefaction in different ways.Even though the map shown in Figure 4 distinguishes thezone most significantly affected by liquefaction, the severity ofliquefaction within this zone was not uniform. The manifestationof liquefaction was primarily of moderate intensity with896 Seismological Research Letters Volume 82, Number 6 November/December 2011

Location of structures illustrated insubsequent figures.Geomorphic featureModerate to severeliquefaction zone indicatedwith black shading.Avon RiverChristchurch Cathedral▲▲Figure 4. Preliminary liquefaction map indicating zones of weakness and locations of buildings discussed in the paper.(A)(B)▲▲Figure 5. Representative areas of: A) moderate liquefaction (7 March 2011; S43.52791 E172.63653), and B) severe liquefaction withinthe CBD principal liquefaction zone (4 March 2011; S43.52604 E172.63839).relatively extensive areas and volumes of sediment ejecta (Figure5). There were also areas of low manifestation or only traces ofliquefaction, but also pockets of severe liquefaction with verypronounced ground distortion, fissures, large settlements, andsubstantial lateral ground movements. This non-uniformityin liquefaction manifestation reflects the complex and highlyvariable soil conditions even within the CBD principal liquefactionzone. Survey maps of Christchurch dating back to thetime of early European settlement (1850s) show a network ofstreams and swamps scattered across this area (Archives NewZealand 2011).The north extent of the zone, which is shown by the thicksolid line in Figure 4, is a clearly defined geomorphic boundaryrunning east-west that was delineated by a slight change in elevationof about 1 m to 1.5 m over an approximately 2 m to 10 mwide zone before the earthquakes. After the 22 February event,Seismological Research Letters Volume 82, Number 6 November/December 2011 897

(A)(B)Liquefaction inducedsediment ejecta.▲ ▲ Figure 6. Apartment complex: A) looking south from northern building showing tilt of southern building, and B) looking north atliquefaction feature at edge of southern building (7 March 2011; S43.52434 E172.64432).it was further characterized by ground fissuring or distortionassociated with localized spreading, as well as gentle slumpingof the ground surface on the downslope side. Ground cracks,fissures, and a distorted pavement surface marked this feature,which runs continuously through properties and affected anumber of buildings causing cracks in both the foundationsand superstructures. Liquefaction and associated grounddeformation were pronounced and extensive on the downslopeside between the identified geomorphic feature and the AvonRiver, but noticeably absent on the slightly higher elevationto the north (upslope side away from the river). This featureis thought to delineate the extent of a geologically recent rivermeander loop characterized by deposition of loose sand depositsunder low-velocity conditions. A similar geomorphic featurewas observed delineating the boundary between liquefactiondamage and unaffected ground within a current meander loopof the river to the east of this area (Oxford Terrace betweenBarbadoes Street and Fitzgerald Avenue).EFFECTS OF LIQUEFACTION IN THE CBDSome of the most important observations of the effects of soilliquefaction on structures in the CBD of Christchurch aredescribed in this paper. These are the types of case histories thatare required to glean important findings from and to add tothe empirical database of the seismic performance of buildingsat sites that have liquefied. There are several important cases ofbuildings with different foundation types (e.g., isolated spreadfootings, spread footings with grade beams, raft foundations,and pile foundations) that performed differently in liquefiedground, as we will describe in the following sections of thispaper.Preliminary geotechnical zoning based on existing dataindicates several different areas within the CBD that are dominatedeither by gravelly layers, thick liquefiable sands or sandysiltmixtures, and peat in the top 8–10 m of the deposits. Thesoil profiles and thicknesses of these layers are highly variableeven within a single zone, thus imposing difficult foundationconditions and sometimes resulting in unconventional orhybrid types of foundations being adopted for buildings. Thegravelly soils, even though relatively more competent foundationsoils, typically show medium standard penetration test(SPT) N values of about 15 to 25 blow counts, whereas the liquefiableloose sands and silt-sand mixtures have low resistanceof less than N = 12 or cone penetration test (CPT) q c valuesless than 3–6 MPa. Another influencing feature in the designof deep foundations is the presence of gravel aquifers, with artesianwater pressures, at a depth of approximately 22 m, whichhas imposed additional restrictions in terms of the cost and useof deep foundations. Additionally, the upward gradient haspotentially adversely affected the liquefaction resistance of theoverlying soils by increasing the pore water pressures in thesesoils during the 2010–2011 earthquakes.Ground Failure Effects on Nearly IdenticalStructures—East Salisbury AreaA mini-complex of three nearly identical buildings (with onesmall but important difference) is shown in Figure 6. Thebuildings are three-story structures with a garage at the groundfloor, constructed on shallow foundations. This case clearlyillustrates the impact of liquefaction, as the nearly identicalstructures have been built across the EW-trending geomorphicfeature identified previously in Figure 4, with one building898 Seismological Research Letters Volume 82, Number 6 November/December 2011

Shading indicates areaover which pronouncedgrade change occursApartment buildingsshown in Figure 6Duplex homes;center structure isshown in Figure 8Tilted structureshown in Figure 6bN▲ ▲ Figure 7. Location of geomorphic feature in area of apartment and duplex complexes north of Salisbury Street in CBD. Darkenedband is the area of pronounced grade change.located on the higher level to the north suffering no damage,and the buildings located below the crest suffering progressivelyhigher amounts of damage. This geomorphic feature, which isexpressed here by a significant change in grade of the pavementbetween the northern and middle buildings, is shown in Figure7. The northern building that sits on the higher ground showedno evidence of cracking and distortion of the pavement surface.Conversely, large sediment ejecta were found along the perimeterof the southern building indicating severe liquefaction inits foundation soils (Figure 6B). Liquefaction features were alsoobserved near the middle building, but the resulting distress ofthis building was significantly less than that of the southernbuilding. The southern building had a shortened end wall witha column at its southwest corner, which appeared to produceadditional settlement at the location of the column’s concentratedload. It suffered differential settlement of about 40 cmand more than 3 degrees of tilt toward the west-southwest,which is visible in Figure 6A.Adjacent to these buildings is another complex of threeidentical but structurally different buildings from the formerset. Their locations relative to the abovementioned geomorphicfeature is identical, but these buildings are two-story duplexesthat are apparently supported on different foundations. Figure8A shows the middle building with clear evidence of pavementdistortion, cracking, and settlement of the surroundingground. The settlement of the building was likely not significant,but the ground settled about 20 cm, exposing the top ofthe foundation at the southwest corner (Figure 8B).Another apartment complex, constructed on a single levelbasement that extends almost the full length of the complexand provides off-street parking for the development, lies tothe west of the two case histories discussed previously. It alsocrosses the geomorphic feature. Noticeable settlement of theground at the southern end of the complex of the order of15–20 cm occurred and compression features in the pavementsuggest that it displaced laterally toward the street. The concretebasement floor and structure appeared to have undergonenegligible distortion, which indicates an overall rigid responsedespite the differential ground movements across the site.Punching Settlement—Madras-Salisbury-Peterborough AreaSeveral buildings with shallow foundations located within theliquefied zone underwent punching settlements with someundergoing significant differential settlements and bearingcapacity failures. An example of such performance is shownin Figure 9 for a two-story industrial building located 200 msouthwest of the buildings discussed previously. There are clearmarks of the mud-water ejecta on the walls of the building,indicating about 25-cm-thick layer of water and ejected soilsdue to the severe liquefaction. Note the continuous sand ejectaaround the perimeter of the footing and signs of punchingshear failure mechanism in Figure 9. At the front entrance ofthe building large ground distortion and sinkholes were createddue to excessive pore water pressure and upward flow of water.Settlement of the building around its perimeter was evidentand appeared substantially larger than that of the surroundingsoil, that was unaffected by the building. The building settledapproximately 25 cm relative to a fence at its southeast cornerand settled 10–20 cm relative to the ground at its northwestcorner. The ground floor at the entrance was uplifted and blis-Seismological Research Letters Volume 82, Number 6 November/December 2011 899

(A)(B)Focus area of Figure 8bPavement levelprior to 22 February2011 eventExposedfoundation▲▲Figure 8. Duplex housing complex: A) looking north at center building, and B) close-up of ground settlement next to center building(16 March 2011; S43.52399 E172.64417).Structure shownin Figure 11Structure shownin Figure 12Observed liquefaction features▲▲Figure 9. Two-story building that underwent liquefactioninducedpunching movements (7 March 2011; S43.52506E172.64176).tered, which is consistent with the pronounced settlementbeneath the walls along the perimeter of the building.Differential Settlement and Sliding—Armagh-MadrasAreaFarther to the south, at the intersection of Madras and Armaghstreets, several buildings were affected by severe liquefactionthat induced significant differential settlements or lateralmovements. At this location, the liquefaction was manifestedby a well-defined, narrow zone of surface cracks, fissures, anddepression of the ground surface about 50 m wide, as well aswater and sand ejecta (Figure 10). This zone stretches from theAvon River to the north toward the buildings affected by thisliquefaction feature, shown in the background of Figure 10 tothe south. Traces of liquefaction were evident further to thesouth of these buildings.▲ ▲ Figure 10. Relatively narrow liquefaction-induced featurethat traverses parking lot northeast of the intersection of Madrasand Armagh streets (24 March 2011; S43.52842 E172.64308).The building shown in Figure 11 is a three-story structureon shallow foundations that settled substantially at its front,resulting in large differential settlements that tilted the buildingabout 2 degrees. The building was also uniformly displacedlaterally approximately 15 cm toward the area of significantliquefaction near the front of the building (i.e., to the north).There was a large volume of sand ejecta at the front part of thebuilding with ground tension cracks propagating east of thebuilding and in the rear car-park that were consistent with thelateral movement of the building toward the north.The building shown in Figure 12 is immediately across thestreet to the north. It is a six-story building on isolated footingswith tie beams and perimeter grade beam. The isolated footingsare 2.4 m × 2.4 m and 0.6 m deep. Figure 12 shows the view ofthe building looking toward the west and indicates total settlementsmeasured relative to the building to its north, which did900 Seismological Research Letters Volume 82, Number 6 November/December 2011

1.8 deg15 cmGround cracking due to lateral displacements▲▲Figure 11. Liquefaction-induced differential settlement and sliding of building in the CBD (24 March 2011; S43.52878 E172.64252).NLiquefaction-inducedsediment ejecta29 cm 18 9 6 5 4 0▲ ▲ Figure 12. Building undergoing significant liquefaction-induced differential settlement due to part of it being founded on the liquefactionfeature in this area (24 March 2011; S43.52878 E172.64252).not appear to settle. Starting from its northern edge and proceedingsouth, the differential settlement is 1 cm for the firstspan, 1 cm for the second span, then 3 cm, 9 cm, and 11 cm,respectively, for the final three spans. This results in an overalldifferential settlement across the structure of 25 cm, with20 cm of it occurring across the two southernmost spans. Astrong tie beam 0.6 m wide × 1.2 m deep was used between thefootings for the first two northernmost spans, whereas the tiebeams between the footings for the remaining spans were only0.3 m wide × 0.45 m deep. This foundation detail, togetherwith the fact that the observations of liquefaction were mostsevere at the southeast corner of the building and diminishedacross the footprint of the building toward the north, led tosubstantial differential settlements and pronounced structuraldistortion and cracking. Both buildings were considereduneconomical to repair after the 22 February 2011 earthquake.The building shown in Figure 11 has been demolished and thebuilding shown in Figure 12 will be demolished.Performance of Adjacent Structures—Town Hall AreaThe Christchurch Town Hall for Performing Arts, designed bySir Miles Warren and Maurice Mahoney and opened in 1972,is located within the northwest quadrant of the CBD, with themeandering Avon River to its immediate south. It is a complexfacility comprising a main auditorium (seating 2,500) withadjoining entrance lobby, ticketing, and café areas. Furtherextensions provide a second, smaller auditorium, the JamesHay Theatre (seating 1,000), and a variety of function roomsand a restaurant. The structures are supported on shallow foundationsexcept for the kitchen facility, which was added later.Seismological Research Letters Volume 82, Number 6 November/December 2011 901

(A) (B) (C)Crack▲▲Figure 13. Town Hall auditorium and adjacent dining facility undergoing significant liquefaction-induced differential settlement andlateral movements (24 March 2011; S43.52727 E172.63521).(A)(B)(C)▲ ▲ Figure 14. Building in area of significant liquefaction that displays negligible to minor differential settlement or punching settlement(24 March 2011; S43.526508 E172.634646).Air bridges connect the complex to the Crowne Plaza, a majorhotel, and to the Christchurch Convention Centre (opened1997) to the north. Tiled paved steps lead from the southernside of the complex down to the river’s edge, with fountainsand views across to Victoria Park.The facility suffered extensive damage caused primarilyby liquefaction-induced ground failure. Differential settlements,caused by punching shear beneath the building’s maininternal columns that surround the auditorium and carry thelargest dead loads to shallow foundations and a second ringof exterior columns (Figure 13A) that are connected to theinner ring via beams (Figure 13B), caused distortion to thestructure. The cracked beam shown in Figure 13B underwentan angular distortion of 1/70 across its span. The seating forthe auditorium has been tilted and dragged backward due tothe settlement of the surrounding columns. Additionally,the floor of the auditorium is now domed due to differentialuplift relative to the columns. The air bridge connecting themain auditorium to the Christchurch Convention Centre tothe north (away from river) has separated from the building.With no significant deformations of the ground as the obvioussource of this lengthening between the two buildings, theexplanation appears to be that distortions to the auditoriumstructure have pulled the outer walls in toward the building,creating this separation. The entire complex appears to havemoved laterally toward the river (albeit by a barely perceptibleamount on the northern side) with parts of the complex closestto the river undergoing increasingly larger movements (Figure13C). These sections have settled and moved laterally towardthe river more than the remainder of the building, leading tosignificant structural deformations where the extension andoriginal structures are joined.Contrary to the liquefaction-induced punching settlementof buildings into the surrounding ground that wasobserved at the Town Hall and in other parts of the CBD,the seven-story building shown in Figure 14A did not punchsignificantly into the liquefied ground nor undergo significantdifferential settlement. As shown in Figure 14B therewere significant amounts of sand ejecta observed in this area.However, there was no obvious evidence of significant differentialground or building movement (Figure 14C). The differentialsettlement measured between adjacent columns wastypically negligible, but differential settlements of up to 3.5cm were measured at a few locations. This building is acrossthe street and slightly to the west of the Town Hall. It is a case902 Seismological Research Letters Volume 82, Number 6 November/December 2011

30 cm17 cm30 cmFoundation beam▲ ▲ Figure 15. Building on pile foundations in area of severe liquefaction showing large settlement of the surrounding soils relative tothe foundation beams (4-6 March 2011; S43.526575 E172.638668).of liquefaction without significant differential settlement andbuilding damage.Contrasting Performance of a Pile-SupportedStructure—Kilmore AreaSeveral pile-supported structures were identified in areasof severe liquefaction. Although significant ground failureoccurred and the ground surrounding the structures settled,the buildings supported on piles typically suffered less damage.However, there are cases where pile-supported structureswere damaged in areas that underwent lateral spreading nearthe Avon River. In other cases, such as the building shown inFigure 15, which is located approximately 200 m to the eastfrom the Town Hall, the ground-floor garage pavement washeavily damaged in combination with surrounding grounddeformation and disruption of buried utilities. The settlementof the surrounding soils was substantial, with about 30 cm ofground settlement on the north side of the building and up to17 cm on its south side. The first-story structural frame of thebuilding that was supported by the pile foundation with strongtie-beams did not show significant damage from these liquefaction-inducedground settlements.Across from this building to the north is a seven-storyreinforced concrete building on shallow spread footing foundationsthat suffered damage to the columns at the ground level.This building tilted toward the southeast as a result of approximately10-cm differential settlement caused by the more severeand extensive liquefaction at the south-southeast part of thesite. It is interesting to note that in the vicinity of this building,the site contained areas that liquefied during the 4 September2010 earthquake. Following the extensive liquefaction in the22 February 2011 event, there was also significant liquefactionin some areas during the 13 June 2011 earthquakes.Having all of this in mind, these two buildings provideinvaluable information on the performance of shallow foundationsand pile foundations in an area of moderate to severe liquefactionthat induced uneven ground settlements. Extensivefield investigations are planned to document the ground conditionsin detail at these sites.Presence of Shallow Gravelly Soils—Victoria SquareNear Victoria Square, the liquefied zone was composed predominantlyof relatively deep loose sand deposits that transitionedrelatively sharply into a zone where gravelly soil layersreach close to the ground surface. Shallow foundations (spreadfootings and rafts) for many of the high-rise buildings in thislatter area are supported on these competent gravelly soils.However, the ground conditions are quite complex in the transitionzone, which resulted in permanent lateral movements,settlements, and tilt of buildings either on shallow foundationsor hybrid foundation systems, as illustrated in Figure 16.Immediately to the north of these buildings, the liquefactionwas severe with massive sand ejecta; however, approximately100 m and further to the south where the gravels predominate,there was neither evidence of liquefaction on the groundsurface nor visible distress of the pavement surface. Again, itappears that the ground and foundation conditions have playeda key role in the performance of these buildings, which thereforehave been selected for in-depth field investigations.Lateral Spreading—Avon RiverLiquefaction-induced lateral spreading was evident within theCBD along the Avon River in the liquefied zone, and the horizontalstretching of the ground adversely affected several buildings.Detailed measurements by ground surveying conductedat about 10 transects on the Avon River within the CBD afterthe 22 February earthquake indicated that at several locationsthe maximum spreading displacements at the banks of theAvon River reached about 50–70 cm, whereas at most of theother locations the spreading was on the order of 10 cm to 20Seismological Research Letters Volume 82, Number 6 November/December 2011 903

documented observations, critically important case histories ofsoil-foundation-structure-interaction can be developed. Whencompleted, these well-documented case histories of buildingperformance in liquefied ground can be used to evaluate andcalibrate computational software with advanced geotechnicalsoil models and provide empirical data for developing designprocedures for evaluating the effects of liquefaction on buildingperformance.ACKNOWLEDGMENTSThe primary support for the New Zealand GEER team memberswas provided by the Earthquake Commission New Zealand(EQC) and University of Canterbury. The primary supportfor the U.S. GEER team members was provided by grantsfrom the U.S. National Science Foundation (NSF) as part ofthe Geotechnical Extreme Events Reconnaissance (GEER)Association activity through CMMI-0825734 and CMMI-1137977. Any opinions, findings, and conclusions or recommendationsexpressed in this material are those of the authorsand do not necessarily reflect the views of the National ScienceFoundation, EQC, or the host institutions of the authors.We would also like to acknowledge the assistance of all NewZealand and U.S. GEER team members who participated in thereconnaissance of these events. Their contributions are noted atthe GEER Web site (http://www.geerassociation.org/).▲ ▲ Figure 16. Buildings on shallow and hybrid foundations intransition area from moderate liquefaction to low/no liquefaction;arrows indicate direction of tilt of the buildings (7 March2011; S43.52878 E172.63528).cm. There were many smaller buildings suffering serious damageto the foundations due to spreading as well as clear signs ofthe effects of spreading on some larger buildings both at thefoundations and through the superstructure.CONCLUSIONSDocumenting and learning from observations after designlevelearthquakes are vital to advancing the state-of-practicein earthquake engineering. Surveying the re-occurrence ofliquefaction, documenting cases of liquefaction-inducedground movements, and evaluating the effects of liquefactionon buildings and lifelines provide invaluable information thatwill serve as benchmarks to the profession’s understandingof the effects of earthquakes. The series of earthquakes thatshook Christchurch in 2010 and 2011 provides insights anddata more valuable than that which can be developed throughexperiments due to the problems of model scaling. These earthquakes,in particular, represent important earthquake scenariosworldwide. Each of the documented building responses inthe CBD provides critical insights regarding the performanceof structures and foundations sited on ground that couldpotentially liquefy. Site investigations are planned to documentfully the ground conditions at these sites, so that with theseREFERENCESArchives New Zealand (2011). Black Map of Christchurch, March 1850.http://archives.govt.nz/gallery/v/Online+Regional+Exhibitions/Chregionalofficegallery/sss/Black+Map+of+Christchurch/. Lastaccessed July 18, 2011.Bradley, B. A., and M. Cubrinovski (2011). Near-source strong groundmotions observed in the 22 February 2011 Christchurch earthquake.Seismological Research Letters 82,853–865.Brown, L. J., and J. H. Weeber (1992). Geology of the Christchurch UrbanArea. Institute of Geological and Nuclear Sciences. Lower Hutt,New Zealand: GNS Science.Cubrinovski, M., R. Green, J. Allen, S. Ashford, E. Bowman, B. Bradley,B. Cox, T. Hutchinson, E. Kavazanjian, R. Orense, M. Pender, M.Quigley, and L. Wotherspoon (2010). Geotechnical reconnaissanceof the 2010 Darfield (Canterbury) earthquake. Bulletin of the NewZealand Society for Earthquake Engineering 43 (4), 243–320.New Zealand Government (2011). http://www.beehive.govt.nz/release/govt-outlines-next-steps-people-canterbury. Last accessed 18 July2011.Youd, T. L., I. M. Idriss, R. D. Andrus, I. Arango, G. Castro, J. T.Christian, R. Dobry, et al. (2001). Liquefaction resistance of soils:Summary report from the 1996 NCEER and 1998 NCEER/NSFworkshops on evaluation of liquefaction resistance of soils. ASCEJournal of Geotechnical & Geoenvironmental Engineering 127 (10),817–833.Department of Civil and Natural Resources EngineeringUniversity of CanterburyPrivate Bag 4800Christchurch 8140 New Zealandmisko.cubrinovski@canterbury.ac.nz(M. C.)904 Seismological Research Letters Volume 82, Number 6 November/December 2011

Comparison of Liquefaction FeaturesObserved during the 2010 and 2011 CanterburyEarthquakesR. P. Orense, T. Kiyota, S. Yamada, M. Cubrinovski, Y. Hosono, M. Okamura, and S. YasudaR. P. Orense, 1 T. Kiyota, 2 S. Yamada, 3 M. Cubrinovski, 4 Y. Hosono, 5M. Okamura, 6 and S. Yasuda 7INTRODUCTION1. Department of Civil and Environmental Engineering, University ofAuckland, New Zealand2. Institute of Industrial Science, University of Tokyo, Japan3. Department of Civil Engineering, University of Tokyo, Japan4. Department of Civil and Natural Resources Engineering, Universityof Canterbury, New Zealand5. Department of Architecture and Civil Engineering, ToyohashiUniversity of Technology, Japan6. Department of Civil and Environmental Engineering, EhimeUniversity, Japan7. Department of Civil and Environmental Engineering, Tokyo DenkiUniversity, Tokyo, JapanOn 4 September 2010, a magnitude M = 7.1 earthquake struckthe Canterbury region on the South Island of New Zealand.The epicenter of the earthquake was located near Darfield,about 40 km west of the central business district (CBD) of thecity of Christchurch and at a depth of about 10 km. Extensivedamage was inflicted on lifelines and residential houses dueto widespread liquefaction and lateral spreading in areas closeto major streams, rivers, and wetlands throughout the city ofChristchurch and the town of Kaiapoi. In the months followingthe Darfield M 7.1 earthquake, numerous aftershocks werefelt across the city.Almost six months after the Darfield mainshock, on 22February 2011, the Canterbury region was hit by a magnitudeM = 6.3 earthquake. The epicenter was located near Lyttelton,only 6 km to the southeast of the Christchurch CBD and at adepth of 5 km. In spite of its smaller magnitude, this earthquakeresulted in more damage to pipeline networks, transport facilities,residential houses/properties, and multistory buildings inthe CBD than the September 2010 event, mainly because ofthe short distance to the city and the shallower depth.Although there were no casualties after the 2010 Darfieldearthquake, which is sort of a miracle considering the magnitudeof the earthquake, the 2011 Christchurch earthquakeresulted in a significant number of casualties due to the collapseof multistory buildings and unreinforced masonry structuresin the Christchurch city center. As of 1 June 2011, 181 casualtieswere reported (New Zealand Police; http://www.police.govt.nz/list-deceased).While it is extremely regrettable that the 2011Christchurch earthquake resulted in significant casualties,engineers and seismologists now have a hard-to-find opportunityto learn the response of ground and structures to twolarge-scale earthquakes that occurred less than six monthsapart. From a geotechnical engineering point of view, it is interestingto look at the widespread liquefaction in natural sediments,re-liquefaction of ground occurring over a short periodof time, and further damage to earth structures that had beendamaged as a result of the first earthquake.Following the two earthquake events, detailed geotechnicalinvestigations were conducted by the authors as part of theJapanese Geotechnical Society (JGS) earthquake reconnaissanceteams. The reconnaissance was a collaboration betweenthe society’s New Zealand-based members and researchers dispatchedfrom Japan for this purpose. The first visit was made12–15 September 2010, while the second one was 27 February–3March 2011. This paper attempts to present a comparison ofthe two events based on the observations made by the authorsfollowing these reconnaissance trips, with emphasis on the geotechnicalimplications of liquefaction-observed damage in theaffected areas.It is worth mentioning that a series of aftershocks, the largestof which were M 5.6 and M 6.3, rattled the city on 13 June2011. These aftershocks again caused extensive liquefactionin many parts of Christchurch. As we write this paper, reconnaissancework is underway to shed more light on the damagecaused by re-liquefaction.GEOLOGIC SETTINGThe Canterbury Plains, about 180 km long and of varyingwidth, are New Zealand’s largest areas of flat land. They havebeen formed by the overlapping fans of glacier-fed rivers issuingfrom the Southern Alps, the mountain range of the SouthIsland. The plains are often described as fertile, but the soils arevariable. Most are derived from the greywacke of the mountainsor from loess (fine sediment blown from riverbeds). Indoi: 10.1785/gssrl.82.6.905Seismological Research Letters Volume 82, Number 6 November/December 2011 905

43°21′35″2010 EQ0 27 km2011 EQKaiapoiWaimakariri RiverAvon RiverCity CentreBexleyHeathcote RiverPort Hills2011 EQEpicentreBanks Peninsula43°39′18″172° 26′23″172°55′40″▲▲Figure 1. Map of Christchurch region highlighting relevant geographical information.addition, clay and volcanic rock are present near Christchurchfrom the Port Hills slopes of Banks Peninsula.The city of Christchurch is located at the east coast of theCanterbury Plains adjacent to an extinct volcanic complexforming Banks Peninsula. A map of Christchurch CBD andits environs is shown in Figure 1. Most of the city was mainlyswamp behind sand dunes, and estuaries and lagoons that havenow been drained (Brown et al. 1995). The surface geology ofthe greater Christchurch area consists of predominantly recentHolocene (

(A)(B)▲▲Figure 2. A) Simplified geology of Christchurch region. B) Simplified soil strata along cross-section A-A’ (modified from Brown andWeeber 1992).43°27′36″SWSDallington - 1CBGSCHHCREHSCCCCSHLCPRPCSWSDallington - 2HPSCSWSBexley - 1SWSBexley - 2CMHSHVSCLPCC43°37′01″172° 30′10″172°47′41″▲ ▲ Figure 3. Location of strong-motion sites near the city of Christchurch (after GeoNet Web site, http://magma.geonet.org.nz/delta/app?service=page/Home). Also shown are the locations of Swedish weight sounding (SWS) tests.Seismological Research Letters Volume 82, Number 6 November/December 2011 907

TABLE 1Comparison of peak ground accelerations recorded at strong-motion sites near the city during the 2010 Darfield earthquakeand 2011 Christchurch earthquake. Data are from GeoNet strong motion FTP site. The unit of acceleration is g (1 g = 9.80 m/s 2 ).SeismicStationsSite Name2010 Darfield Earthquake 2011 Christchurch EarthquakeEp. Dist.(km) Vert Hor-1 Hor-2MaxHorEp. Dist.(km) Vert Hor-1 Hor-2HVSC Heathcote Valley Primary School 43 0.28 0.56 0.62 0.66 1 1.47 1.46 1.19 1.50LPCC Lyttelton Port Company 45 0.16 0.33 0.23 0.37 4 0.41 0.78 0.88 1.00CCCC Chch Cathedral College 38 0.16 0.23 0.20 0.24 6 0.69 0.48 0.37 0.49CMHS ChCh Cashmere High School 36 0.25 0.25 0.24 0.26 6 0.80 0.35 0.38 0.42PRPC Pages Road Pumping Station 41 0.31 0.20 0.23 0.23 6 1.63 0.66 0.59 0.73CHHC Christchurch Hospital 36 0.16 0.20 0.15 0.20 8 0.51 0.34 0.36 0.46REHS Christchurch Resthaven 37 0.21 0.24 0.25 0.33 8 0.53 0.71 0.37 0.73CBGS Christchurch Botanic Gardens 36 0.11 0.15 0.17 0.18 9 0.27 0.53 0.43 0.64HPSC Hulverstone Dr Pumping Station 43 0.13 0.16 0.11 0.16 9 0.86 0.15 0.24 0.25SHLC Shirley Library 39 0.12 0.18 0.18 0.19 9 0.50 0.31 0.34 0.34Note: Ep. Dist – Epicentral distance; Vert – vertical acceleration; Hor-1 and Hor-2 – horizontal components of acceleration;Max. Hor – calculated maximum resultant acceleration of horizontal components. Unit of acceleration is g (1 g = 980 cm/s 2 ). Source:GeoNet 2011.MaxHormotion FTP site, the maximum recorded acceleration was onthe order of 0.95 g near the earthquake epicenter (GeoNet2010); however, no serious damage was reported in the area.In the city of Christchurch, the recorded peak ground accelerations(PGA) were on the order of 0.15–0.30 g, as shown inTable 1. The seismic stations indicated in the table correspondto those shown in Figure 3. Figure 4 shows a typical accelerationrecord obtained during the 2010 earthquake. Note thatthe duration of significant shaking at Christchurch Hospital(CHHC), located at the southwest edge of the CBD, is on theorder of about 25–30 sec.2011 Christchurch EarthquakeThe 2011 Christchurch earthquake occurred more than fivemonths after the 2010 Darfield earthquake, with an epicenterlocated on an unmapped fault different from the Greendalefault. Yet it is considered an aftershock because it was causedby a fault rupture within the zone of aftershocks that followedthe September 2010 mainshock (National Hazards ResearchPlatform 2011). Because the M 6.3 aftershock was much closerto the Christchurch CBD than the M 7.1 mainshock, theground accelerations experienced in the CBD as a result of the2011 earthquake were three to four times greater than during2010 event (see Table 1); in the eastern suburbs, they wereabout five times greater. The vertical PGA recorded was 1.47g at Heathcote Valley primary school (about midway betweenthe CBD and the epicenter) while in the CBD the PGA was0.5–0.7 g and in the eastern suburbs the maximum recordedvertical PGA was 1.63 g (GeoNet 2011). A feature of this earthquakewas the very strong vertical component of PGA, whichin general was greater than the horizontal components. Figure4 also illustrates the time histories of acceleration recorded atChristchurch Hospital on 22 February 2011. Because of theshorter distance to the epicenter, the acceleration records inthis earthquake have a higher frequency and shorter durationtime, as well as larger amplitude, in comparison with the onesrecorded on 4 September 2010.COMPARISON OF OBSERVED LIQUEFACTIONDAMAGEAlthough structural failure of commercial buildings led to thegreatest casualties in the M 6.3 Christchurch earthquake, byfar the most significant damage to residential buildings andlifelines in both Canterbury earthquakes was the result of liquefactionand associated ground deformations. Liquefactionoccurred in areas that are known to have high potential forliquefaction—former river channels, abandoned meanders,wetlands, and ponds. Immediately following some of the largestaftershocks from the M 7.1 earthquake, liquefaction reoccurredin some of these areas. During the M 6.3 earthquake,liquefaction was more widespread and vents continued to surgeduring the aftershocks immediately following this event. Theimpact of sand boils and cracks caused by lateral spreading wasthat parts of the eastern suburbs were inundated with sand andsilt—in places there were layers of ejected soil that were manytens of centimeters thick.As mentioned earlier, two large aftershocks, measuringM 5.6 and M 6.3, shook the city on 13 June 2011 and causedextensive re-liquefaction in many parts of the city. Streets wereagain flooded with water and ejected sands, reminiscent ofwhat happened immediately after the 22 February 2011 earthquake.Such reoccurrence of liquefaction indicates that the soildeposits in the area were still loose even after the intense shakingthey had been subjected to over the last nine months.908 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A)Acceleration (Gal)6004002000-200-400-6006004002000-200-400-6006004002000-200-400-600Christchurch Hospital, 4 Sep 2010, MainshockN01W (Horizontal 1)S89W (Horizontal 2)Vertical0 10 20 30 40 50 60Time (sec)(B)Acceleration (Gal)6004002000-200-400-6006004002000-200-400-6006004002000-200-400Christchurch Hospital, 22 Feb 2011, MainshockN01W (Horizontal 1)S89W (Horizontal 2)Vertical-6000 10 20 30 40 50 60Time (sec)▲▲Figure 4. Acceleration time histories recorded at Christchurch hospital during the 2010 Darfield earthquake and 2011 Christchurchearthquake (data from GeoNet strong motion FTP Web site). Note: 1 g = 980 Gal.43°31′18″43°32′16″172°36′43″ 172°39′14″▲ ▲ Figure 5. Location of Avon River as well as wetlands and streams in 1850 superposed to the present-day map of the ChristchurchCBD.Eastern SuburbsLiquefaction and lateral spreading were extensive in areas adjacentto the Avon River, which follows a meandering coursethrough Christchurch from its source in the west throughthe CBD, then toward the east passing through Avonside,Dallington, Avondale, and Aranui, and finally flowing to thePacific Ocean via the Avon-Heathcote estuary.Figure 5 shows the locations of Avon River and otherstreams based on a 1850 map of the city superposed on thepresent map of Christchurch CBD. The meandering nature ofthe Avon is conspicuous as it flows from the west toward theeast. Also, it can be seen that several wetlands and streams crisscrossedthe future city center, some of which were later artificiallyreclaimed as the city grew. The locations of these formerriver channels had a significant effect on the damage observedfollowing the M 6.3 earthquake. Details of liquefactioninduceddamage observed in the central business district arepresented by Cubrinovski et al. (2011, page 893 of this issue).After the 2010 Darfield earthquake, Swedish weightsounding (SWS) tests were performed by the JGS-University ofCanterbury reconnaissance teams at numerous locations affectedby liquefaction and lateral spreading. The SWS test is a simplemanually operated penetration test under a dead-load of 100 kgin which the number of half-rotations required for a 25-cm penetrationof a rod (screw point) is recorded (Japanese StandardsAssociation 1995). As a result of the SWS test, the correspondingstandard penetration test (SPT) N-value can be obtainedthrough the following empirical equation (Inada 1960).Seismological Research Letters Volume 82, Number 6 November/December 2011 909

0000111122223333Depth Depth (m) (m)45Depth (m) (m)45Depth (m) (m)45Depth (m) (m)456666777788889109(e) Spot No.5(d) Spot No.4(b) Spot No.2(c) Spot No.31010100 10 20 30 400 10 20 30 400 10 20 30 400 10 20 30 40Conversion N-valueConversion N-valueConversion N-valueConversion N-valueConverted N-value Converted N-value Converted N-value Converted N-value9Bexley Bexley Dallington Dallington9▲ ▲ Figure 6. Converted SPT N-value profiles from Swedish weight sounding tests conducted by the University of Canterbury and JGSreconnaissance teams in September 2010.N = 0.002W SW + 0.067N SW , (1)where W SW (kg) is the weight less than 100 kg and N SW is thenumber of half-rotations for every meter of penetration. W SWis counted when penetration occurs with dead-load less than100 kg. Note that this equation is applicable to gravel, sand,and sandy soils.Typical results of SWS tests in Christchurch are shown inFigure 6. The locations of the test sites are indicated in Figure3. It can be seen from the strength-depth profiles that in theseareas, layers of about 5 m or thicker exist with high potential toliquefy (very loose silt/sand layer with SPT N-value < 5). Thepresence of loose sandy deposits in many areas in Christchurchhas also been confirmed through dynamic cone penetrometer(DCP) tests and spectral analysis of surface waves(SASW) tests conducted by the Geotechnical Extreme EventsReconnaissance (GEER) team (Green et al. 2011, page 927of this issue).Immediately following the two earthquakes, reconnaissancework was performed to investigate the extent and featuresof the damage. Figures 7A and 7B show the distributionof liquefaction observed in the suburbs east of the CBD followingthe September 2010 and February 2011 earthquakes,respectively. These maps, which were constructed from on-footinvestigations and drive-through surveys with the help of theUniversity of Canterbury reconnaissance team, may be incompletedue to the limited time spent by the team in the areafollowing each earthquake. As mentioned earlier, the shorterdistance to the city and the shallower depth of the February2011 earthquake resulted in more significant and more widespreaddamage in these areas than the September 2010 earthquake.The circle and square data points plotted in the figurescorrespond to the maximum distance from the Avon River atwhich lateral spreading was observed in the north and southbanks, respectively, based on ground inspection. It is worthnoting that while major liquefied sites in the September 2010earthquake were concentrated along the Avon River, liquefactionwas observed in the 2011 earthquake across a wider area,i.e., not only in the eastern suburbs but in the north and in theCBD as well.Considering the short time interval between the two largeearthquakes, the 2011 earthquake induced additional damageto many facilities that suffered liquefaction-induced damageafter the 2010 earthquake and had not been repaired. Figures8A and 8B show the condition of a river embankment adjacentto the Avon Rowing Club (east bank of Avon River) after the2010 earthquake and 2011 earthquake, respectively. The widthof crack openings on the shoulder and the settlement of thecrown became larger due to the re-liquefaction of the foundationground of the embankment. Extensive re-liquefaction wasobserved in the entire Porritt Park in 2011, where almost halfof the green grassy area was covered by sand boils, similar tothat observed after the 2010 earthquake, as shown in Figure 9.The southern portion of the Bexley suburb was formerlya swamp and formed part of the Bexley wetlands. It had beenreclaimed in the late 1990s by filling the area, and the subdivisionwas built over it. Interviews with homeowners indicatethat the area was fairly new, with some houses built as recent asfive years ago (Orense et al. 2011). The September 2010 earth-910 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A)(B)▲▲Figure 7. Distribution of liquefaction-induced damage in the eastern suburbs: A) September 2010 earthquake; B) February 2011earthquake.Seismological Research Letters Volume 82, Number 6 November/December 2011 911

(A)(B)▲▲Figure 8. Damage to a river embankment near the Avon Rowing Club: A) 2010 earthquake; and B) 2011 earthquake.(A)(B)▲▲Figure 9. Cracks observed in Porritt Park: A) 2010 earthquake; and B) 2011 earthquake.quake triggered liquefaction of the loose uncompacted fill,resulting in ground settlement and lateral spreading. Ejectedsands filled up the whole neighborhood, as thick as 30 cm insome areas (Figure 10A). Following the 2011 earthquake,Bexley was again one of the worst-hit areas in terms of liquefaction-induceddamage. Massive amount of sands were againejected and deposited around houses (Figure 10B). The massivesand boils ejected from underground caused differentialground settlements, resulting in the tilting of many houses.Sand boils were also observed in the swamps of Bexley wetlands,indicating that the ground below the swamp also underwentliquefaction.KaiapoiThe township of Kaiapoi is located in the northeastern end ofthe Canterbury Plains, about 20 km north of Christchurch(Figure 1). The Kaiapoi River, which cuts through the centerof the town, joins the Waimakariri River on the eastern edge ofthe town and flows to the sea. In terms of liquefaction duringthe 2010 earthquake, Kaiapoi was probably the worst-hit area,with many residential houses, several commercial buildings,and other infrastructure facilities suffering damage due to lateralspreading, ground subsidence, and differential settlement.Investigations of old maps by Wotherspoon et al. (2010)showed that that much of the most significant liquefactiondamage in and around Kaiapoi during the 2010 Darfield eventoccurred in areas where river channels had been reclaimed orin old channels that have had flow diverted away. The highlymodified nature of the Waimakariri River and its proximity toKaiapoi meant that some of these reclaimed areas overlappedregions that have since been developed as the town has grown,as shown in Figure 11.Following the 2011 earthquake, re-liquefaction occurredin Kaiapoi; however, because of the farther epicentral distance,the impact of liquefaction was minor compared to thatobserved in September 2010. The peak ground accelerationsrecorded in Kaiapoi during the 2010 and 2011 earthquakeswere 0.36 g and 0.20 g, respectively. A comparison of the distributionof liquefaction in Kaiapoi during the two events showedsmaller liquefaction zones in 2011. Areas where re-liquefaction912 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A)(B)▲▲Figure 10. Damage to a residential house due to liquefaction in Seabreeze Close, Bexley: A) 2010 earthquake; and B) 2011 earthquake.43°22′11″43°24′52″172°37′56″ 172°42′32″▲▲Figure 11. Map of river channels in 1865 superposed on the present-day map of Kaiapoi (courtesy of L. Wotherspoon).was observed include some areas adjacent to the stopbanks(levees) in northern Kaiapoi and in the filled-up sections ofCourtenay Drive in southern Kaiapoi.Figure 12 shows a residential house that suffered severedamage due to lateral spreading following the September 2010earthquake. This house was standing on ground that movedtoward the Waimakariri River, resulting in tilting of the houseand formation of a 1.6-m-wide crack between the house andthe adjacent ground. Following the 2011 earthquake, the JGSteam revisited the same house. Sand boils were observed onlyin the ground cracks adjacent to the house, with the width ofthe crack increasing to 1.9 m. It is unknown, however, whetherthe increase in the crack opening was caused by the 2011earthquake alone. Creep deformation in this area due to theSeismological Research Letters Volume 82, Number 6 November/December 2011 913

(A)(B)▲▲Figure 12. A) Damage to a residential house in south Kaiapoi that underwent foundation failure due to lateral spreading and liquefactionin the 2010 Darfield earthquake. B) Additional damage after the 2011 Christchurch earthquake.(A)(B)▲ ▲ Figure 13. A) Liquefaction at a park adjacent to Courtenay Lake, south Kaiapoi, following the September 2010 earthquake. B)Re-liquefaction during the February 2011 earthquake resulted in sand boils being ejected through existing cracks, but ground distortionwas minor.aftershocks of the 2010 Darfield earthquake has been reported(Cubrinovski and Orense 2010). Therefore, there is a possibilitythat the width of the crack was more than 1.6 m before theFebruary 2011 earthquake and the impact of the earthquake tothis area was minor. No other remarkable additional damageto residential houses/properties was observed in south Kaiapoi.After February 2011, most of the sand boils in areas closeto the waterways were observed at existing/repaired crackscaused by the 2010 earthquake. Aside from the lower intensityof ground shaking in Kaiapoi, it is possible that the excess porewater pressure generated by the earthquake motion could havebeen dissipated easily through the existing cracks and thereforethis earthquake did not induce significant ground deformation(Figure 13).On the other hand, a more pronounced liquefaction wasobserved in residential houses/properties in north Kaiapoi,although relatively minor compared to that after the 2010Darfield earthquake. Figure 14 compares the settlement of atwo-story house due to liquefaction. The slope in front of thegarage was originally uphill, but it became downhill after the2010 Darfield earthquake as a result of more than 50 cm ofground subsidence. This house suffered an additional 15 cmsubsidence following the 2011 earthquake. The narrowinggap between the roof and the head of a member of the reconnaissanceteam can be recognized from the figure. Althoughthe team members appearing in the photos are different, theirheights are almost the same.914 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A)(B)▲ ▲ Figure 14. A) Subsidence of a two-story house adjacent to the stopbank in north Kaiapoi caused by the 2010 earthquake. B) Additionalsubsidence of the same house after the 2011 earthquake.LIQUEFACTION OBSERVED ADJACENT TOHEATHCOTE RIVERHeathcote River, located in the southern boundary ofChristchurch, meanders around the base of the Port Hills fromwest to southeast. It drains into the Avon-Heathcote estuarybefore draining into Pegasus Bay. Earlier studies have indicatedthat aside from parts of the eastern suburbs, the areas aroundthe Heathcote River are underlain by loose saturated sand andsilt, which have high potential to liquefy (Christchurch CityCouncil 2005).Following the September 2010 earthquake, a quick drivethroughinvestigation was conducted along the HeathcoteRiver, specifically targeting areas that were denoted as havinghigh potential for liquefaction-induced damage. However,there was very little evidence of ground distortion and liquefactionin this area, with only a few sand boils found in aperiod of about two hours of drive-through and on-foot surveys(Cubrinovski et al. 2010).On the other hand, after the February 2011 earthquake,significant ground distortions due to liquefactionwere observed adjacent to Heathcote River. In flat areas withshallow ground-water tables (e.g., St. Martins, Opawa, andWoolston), a number of structures such as stopbanks, bridges,and residential properties suffered severe damage due to liquefaction.Figure 15 shows the distribution of liquefaction alongthe Heathcote River observed during a walk-through investigationconducted two weeks after the 2011 earthquake. Again,the circular and square dots in the figure correspond to themaximum distance from the Heathcote River at which lateralspreading was observed. It is clear from the figure that severeliquefaction occurred at limited areas along the HeathcoteRiver—considerably smaller than the ones observed adjacentto Avon River (see Figure 7B).The lower areas along Wilson Road in St. Martins may bethe worst-hit areas near the Heathcote River. Figure 16 showsthe St. Martins library, a brick building whose collapse wascaused by the differential subsidence of the foundation grounddue to liquefaction. Ejected sands, with thickness on the orderof 20 cm, were deposited around the right half of the foundation.Additionally, a tilted power pole can be seen in the rightside of the figure, indicating that liquefaction occurred at shallowdepth in this area.Heathcote River winds along the foot of Port Hills, andtherefore the topography around the river is full of ups anddowns. From geological information, loess deposits are presentin the subsurface at the base of Port Hills (Brown and Weeber1992). No liquefaction was observed at the ground that is consideredto be loess.Figure 17 shows a trench in a residential property underconstruction in Eastern Terrace just beside the river. Thetrench depth was greater than 2 m and yet no groundwater wasobserved. Soil samples were collected from the trench at depthsof 1 and 2 m from the ground surface, and their grain sizes wereanalyzed. Figure 18 shows the grain size distribution curves ofsoils taken from the trench in comparison with those of soilscollected in other parts of Christchurch and Kaiapoi. The opendots correspond to the ejected sand collected from sites adjacentto Avon River and Kaiapoi, while the solid dots representsoils collected at a slope in Port Hills and near the HeathcoteRiver where liquefaction was not observed. It can be seen thatthe sand boils have similar grain size distributions, regardless ofthe location where they were collected (in Kaiapoi or adjacentto the Avon or Heathcote rivers). They have fines content lessthan 25%. On the other hand, the subsurface soil adjacent tothe Heathcote River, which did not liquefy, contained >90%fines with clay content >20%. Therefore, it is possible that thepresence of unliquefiable soil at subsurface is a major reasonSeismological Research Letters Volume 82, Number 6 November/December 2011 915

▲▲Figure 15. Distribution of liquefaction-induced damage adjacent to the Heathcote River after the February 2011 earthquake.▲▲Figure 16. Damage to a brick structure in St. Martins (near the Heathcote River) induced by soil liquefaction.916 Seismological Research Letters Volume 82, Number 6 November/December 2011

▲▲Figure 17. A trench in a residential property in Heathcote. Groundwater table was not evident even up to a depth of 2 m below theground surface.100Percent finer by weight (%)806040Huntsbury(Port Hills)Trench 2m (Eastern Tce)Trench 1m (Eastern Tce)Porritt Park20Seabreeze(Bexley)North KaiapoiWilsons st(St. Martins)00.001 0.01 0.1 1 10Grain size (mm)▲ ▲ Figure 18. Comparison of grain size distribution curves of ejected sands from different sites in Christchurch and soils from unliquefiedsubsurface sites near the Heathcote River.why the extent of liquefaction along the Heathcote River wasminor compared to that near the Avon River.CONCLUDING REMARKSAlthough the M 7.1 Darfield earthquake caused liquefactionin Christchurch and adjacent areas, the M 6.3 Christchurchearthquake induced more widespread liquefaction and causedmore serious damage to infrastructure. Liquefaction and reliquefactionwere observed in areas with high potential to liquefy,such as natural deposits close to major streams, rivers, andwetlands as well as loose or uncompacted fill. Experiences fromcase histories all over the world have highlighted the effect ofliquefaction on buildings and buried structures, but the scale ofdamage experienced in Christchurch following the 2010 and2011 events was unprecedented and may be the greatest everobserved in an urban area. Moreover, the short time intervalbetween the two large earthquakes presented a very rare oppor-Seismological Research Letters Volume 82, Number 6 November/December 2011 917

tunity to investigate the liquefaction mechanism in naturaldeposits. Finally, the re-liquefaction experienced by the city asa result of the recent aftershocks on June 2011 highlights thehigh susceptibility of soil deposits in Christchurch to liquefactionand presents a very challenging problem not only to thelocal residents but to the geotechnical engineering professionas well.ACKNOWLEDGMENTSThe authors would like to acknowledge the other members ofthe NZ-JGS reconnaissance team: Kohji Tokimatsu (TokyoInstitute of Technology, Japan), Ryosuke Uzuoka (TokushimaUniversity, Japan), and Hirofumi Toyota (Nagaoka Universityof Technology, Japan). The insights provided by Michael Pender,Tam Larkin, and Liam Wotherspoon, all of the Universityof Auckland, as well as the assistance of many postgraduatestudents from the University of Auckland and University ofCanterbury, are gratefully acknowledged. Finally, we acknowledgethe New Zealand GeoNet project and its sponsors EQC,GNS Science, and Land Information New Zealand for providingdata used in this paper.REFERENCESBerill, J., H. Avery, M. Dewe, A. Chanerley, N. Alexander, C. Dyer,C. Holden, and B. Fry (2011). The Canterbury AccelerographNetwork (CanNet) and some results from the September 2010M 7.1 Darfield earthquake. In Proceedings of the Ninth PacificConference on Earthquake Engineering, paper no. 181 (CD-ROM).Auckland: New Zealand Society for Earthquake EngineeringBrown, L. J., R. D. Beetham, B. R. Paterson, and J. H. Weeber (1995).Geology of Christchurch, New Zealand. Environmental &Engineering Geoscience 1 (4), 427–488.Brown, L. J., and J. H. Weeber (1992). Geology of the Christchurch UrbanArea. Lower Hutt, New Zealand: Institute of Geological andNuclear Sciences.Christchurch City Council (CCC) (2005). 3.4.5 Earthquake Risk.City Plan Online; http://www.cityplan.ccc.govt.nz (updated 14November 2005).Cubrinovski, M., J. D. Bray, M. Taylor, S. Giorgini, B. Bradley, L.Wotherspoon, and J. Zupan (2011). Soil liquefaction effects in thecentral business district during the February 2011 Christchurchearthquake. Seismological Research Letters 82, 893–904.Cubrinovski, M., R. Green, J. Allen, S. Ashford, E. Bowman, B. Bradley,B. Cox, T. Hutchinson, E. Kavazanjian, R. Orense, M. Pender, M.Quigley, and L. Wotherspoon (2010). Geotechnical reconnaissanceof the 2010 Darfield (Canterbury) earthquake. Bulletin of the NewZealand Society for Earthquake Engineering 43 (4), 243–320.Cubrinovski, M., and R. Orense (2010). 2010 Darfield (New Zealand)earthquake—Impacts of liquefaction and lateral spreading. Bulletinof the International Society for Soil Mechanics and GeotechnicalEngineering 4 (4), 15–23.GeoNet (2010). Strong motion FTP site; ftp://ftp.geonet.org.nz/strong/processed/Proc/2010/09_Darfield_mainshock_extended_pass_band/.GeoNet (2011). Strong motion FTP site; ftp://ftp.geonet.org.nz/strong/processed/Proc/2011/02_Christchurch_mainshock_extended_pass_band/.Green, R. A., C. Wood, B. Cox, M. Cubrinovski, L. Wotherspoon,B. Bradley, T. Algie, J. Allen, A. Bradshaw, and G. Rix (2011).Use of DCP and SASW tests to evaluate liquefaction potential:Predictions vs. observations during the recent New Zealand earthquakes.Seismological Research Letters 82, 927–938.Inada, M. (1960). Interpretation of Swedish weight sounding. Tsuchi-to-Kiso [monthly magazine of the Japanese Geotechnical Society] 8(1), 13–18 (in Japanese).Japanese Standards Association (JSA) (1975). Japanese IndustrialStandards: Method of Swedish Weight Sounding—JIS A 1221(1975), 1995 revision.Natural Hazards Research Platform (NHRP) (2011). Why the 2011Christchurch Earthquake is Considered an Aftershock; http://www.naturalhazards.org.nz.Orense, R., M. Pender, L. Wotherspoon, and M. Cubrinovski (2011).Geotechnical aspects of the 2010 Darfield (New Zealand) earthquake.Invited lecture, Eighth International Conference on UrbanEarthquake Engineering, Tokyo (Japan) (7–8 March 2011).Wotherspoon, L. M., M. J. Pender, and R. P. Orense (2010). Relationshipbetween observed liquefaction at Kaiapoi following the 2010Darfield earthquake and old channels of the Waimakariri River.Submitted to Engineering Geology.Department of Civil and Environmental EngineeringUniversity of AucklandPrivate Bag 92019Auckland 1142 New Zealandr.orense@auckland.ac.nz(R. P. O.)918 Seismological Research Letters Volume 82, Number 6 November/December 2011

Ambient Noise Measurements following the2011 Christchurch Earthquake: Relationshipswith Previous Microzonation Studies,Liquefaction, and NonlinearityMarco MucciarelliMarco MucciarelliBasilicata UniversityINTRODUCTIONFollowing the Christchurch 2011 earthquake, the BasilicataUniversity (Potenza, Italy) organized a field trip to NewZealand mainly to examine structural engineering issues butalso to investigate the similarity between this event and theL’Aquila 2009 quake that struck central Italy. In both casesan event with magnitude slightly above 6 occurred on a blindfault underlying an area inhabited by a population of the orderof hundreds of thousands, killing a few hundred people andseverely damaging the city center, and in both cases a site amplificationstudy was available before the event. At the same timethere were striking differences between the two earthquakesin maximum recorded acceleration, the nonlinear behavior ofsoils, and the occurrence of liquefaction.It was also an opportunity to look at some issues related tothe use of microtremor measurements, in particular:1. to verify if the soil frequencies estimated more than 15years ago by Toshinawa et al. (1997) are a persisting featureor if there were changes following the strong motionsin 2010 and 2011;2. to verify the usefulness of the soil vulnerability index proposedby Nakamura (1996) as a proxy of liquefaction susceptibility;and3. to compare the strong-motion recordings with elasticlimit soil behavior derived from ambient noise, looking forhints of hardening nonlinearity as proposed by Bonilla etal. (2005) and similarity with the observations in L’Aquila(Puglia et al. 2011).PREVIOUS STUDIES AND DATA COLLECTIONIn 1994 the Arthurs Pass Earthquake (M l = 6.6) occurredabout 100 km northwest of Christchurch. The macroseismicintensity was estimated for the city together with local siteamplifications inferred from seismic recordings and microtremors(Toshinawa et al. 1997). The authors found a satisfactorycorrelation among the results of the different techniquesand prepared a microzonation map.As for horizontal-to-vertical spectral ratio (HVSR) analysisof microtremors, Toshinawa et al. (1997) collected three setsof 40-sec-long samples at each site on a 1 by 1 km grid. Theyfound that the H/V spectral ratio of microtremors was wellcorrelated to the ground motion characteristics during earthquakesrecorded at a seismic array deployed within the city andalso correlated with the local geology. The outcropping lithologyof the Christchurch area (Brown and Weeber 1992) iscomposed of:1. Volcanic rock, in the southern part of the city.2. Holocene marine dunes, in the vicinity of the coast.3. Alluvial sand and silt deposits from the estuarine area tothe center of the city, where swamps and lagoons weredrained to reclaim land.4. Alluvial gravel area, in the westernmost part of the city.5. Transition area, with alternating deposits, located betweenthe estuarine and gravel areas.The soil fundamental frequency had higher values in the westerngravel area, starting from 5 Hz and decreasing down to 1Hz proceeding eastward in the transition area and in the sandand silt alluvium beneath the city center. The volcanic rock atthe south returned a flat response and no clear peak was identifiedin the dune area near the ocean coastline.During our field trip, we devoted three days to microtremormeasurements. It was possible to perform 43 measurements, aslisted in Table 1. We also collected 12 recordings as close aspossible to accelerometric stations that recorded the February2011 event, while the others were taken in the most damagedareas but with an effort to obtain good spatial coverage (Figure1). The data were sampled at 128 Hz using a digital threecomponenttromometer (Micromed Tromino) for an acquisitionlength of at least 12 minutes. The data were then filtered,tapered, transformed in frequency domain, smoothed with atriangular filter (n = 5), and finally averaged between horizontalcomponents and among 20-sec subsets using the geometricmean. Almost all the recordings returned HVSR peaks thatpassed the ensemble of tests proposed by the SESAME project(Chatelain et al. 2008).doi: 10.1785/gssrl.82.6.919Seismological Research Letters Volume 82, Number 6 November/December 2011 919

TABLE 1List of the HVSR measurements, their geographiccoordinates, the frequency of the highest HVSR peak andits amplitude. The last column is the K g coefficient asintroduced by Nakamura (1996).Site Lon Lat Fmax Amax KgCACS 172.5275768 –43.4826609 5.88 2.8 1.3RSMC 172.5628001 –43.5355676 0.94 3.2 10.9N13 172.5705568 –43.5437258 2.00 5.0 12.5N26 172.5777560 –43.5108726 3.88 3.7 3.5N27 172.5818618 –43.5213617 3.19 2.5PPHS 172.6064083 –43.4936198 2.06 5.9 16.9N28 172.6125849 –43.5122464 1.81 3.2 5.7N25 172.6167294 –43.5493591 0.50 3.1 19.2CBGS 172.6199070 –43.5292657 1.38 4.0 11.6CHHC 172.6234361 –43.5368474 1.44 3.5 8.5CMHS 172.6273183 –43.5675024 3.75 4.5 5.4N3 172.6306775 –43.5275045 2.06 4.5 9.8N2 172.6306950 –43.5285730 2.13 5.0 11.7N1 172.6313143 –43.5311002 1.46 4.0 11.0N12 172.6346267 –43.5806096 0.81 3.8 17.8RHES 172.6354074 –43.5235211 2.05 6.0 17.6N30 172.6355683 –43.5310515 1.50 3.9 10.1N32 172.6355845 –43.5284343 1.50 5.0 16.7N31 172.6413124 –43.5309137 1.75 6.0 20.6N14 172.6416583 –43.5213395 2.25 5.0 11.1N24 172.6429684 –43.5476508 0.69 2.5 9.1N4-N5 172.6458412 –43.5257455 1.75 4.3 10.6CCCC 172.6461493 –43.5373261 1.50 4.9 16.0N29 172.6473579 –43.4943542 1.56 5.1 16.7N15 172.6519704 –43.5319409 1.63 3.7 8.4N10 172.6596168 –43.5500570 1.13 4.4 17.1N16 172.6642247 –43.5319177 1.38 2.2 3.5SHLC 172.6651345 –43.5040313 1.50 2.5 4.2N11 172.6752989 –43.5624248 2.63 4.0 6.1PRPC 172.6823280 –43.5281150 1.25 3.4 9.2N17 172.6874653 –43.5401185 1.63 2.8 4.8N6 172.6879189 –43.5015612 1.80 2.2 2.7HPSC 172.7019825 –43.5024466 1.25 1.5 1.8N18 172.7049053 –43.5361145 1.94 2.8 4.0N21 172.7071643 –43.5575285 1.55 5.6 20.2HVSC 172.7082133 –43.5807977 3.56 4.2 5.0N9 172.7228573 –43.6046841 7.44 6.2 5.2N8 172.7233046 –43.6033945 6.06 3.4 1.9N19 172.7291321 –43.5239186 0.38 1.7 7.6N7 172.7313212 –43.5077893 0.31 2.2 15.6N23 172.7360934 –43.5611169 2.00 4.5 10.1N20 172.7433333 –43.5401993 0.50 3.8 28.9N22 172.7604590 –43.5769734 1.25 9.8 76.8RESULTSWe compared our results with the results of Toshinawa et al.(1997) by preparing two maps (Figure 2 and Figure 3) thatshow the frequency of the highest peak in the HVSR curveand the relevant amplitude (Fmax and Amax). It is worth notingthat this highest peak does not always coincide with thesoil fundamental frequency but is closer to the parameterschosen by Toshinawa et al. (1997), for a reason that is evidentif one compares the frequency maps. As mentioned before,Toshinawa et al. did not find any resonance peak in the coastaldune area, while our measurements return values below 1 Hz.It is probable that the older instrumentation, coupled with ashorter-duration sampling time, could not detect low frequencysoil resonance due to a deeper structure. This low frequency isalso often visible moving westward, providing a peak whoseamplitude is generally lower than those in the range above 1Hz. The peaks at frequency >1 Hz appear to be in good agreementwith those reported by Toshinawa et al. (1997), passingfrom 1 Hz below the city center up to 5–6 Hz in the westernmostpart of Christchurch. Our results present quite a differentpicture when the amplitude is taken into account. While theToshinawa et al. (1997) values never exceed 6 and below thecity center HVSR maximum amplitudes are in the range of 2to 3, our measurements top a value of 10 and under the citycenter (or central business district, CBD, as it is often called)the HVSR amplitudes range from 3.5 to 7. It is well known thatHVSR amplitude could be rather sensitive to processing techniques,and a more precise comparison was not possible due tolack of further information from Toshinawa et al. (1997).More detail on our results is given in Figures 4 and 5.Figure 4 shows the clear change in fundamental frequencymoving along longitude, with an increase moving westward.The second plot (Figure 5) shows a site with HVSR that has asingle, low-frequency peak while another site has a similar peakplus another one at higher frequency. It is known that HVSRhas a problem in showing modes higher than the fundamentalone (Parolai and Richwalski 2004), so the two peaks are mostlikely related to a complex stratigraphic situation with tworesonant strata separated by a velocity inversion (Di Giacomoet al. 2005).The second check we performed was the reliability ofthe soil damage index or vulnerability index introduced byNakamura (1996) for use in estimating earthquake damage onthe ground surface. Its distribution should delineate weak areason the ground. It is given as:K g = A g 2 / F g ,where A g is the amplitude corresponding to soil resonance frequencyF g .This coefficient was used in studies aimed at mapping thesoil susceptibility to liquefaction prior to an earthquake (see, e.g.,Beroya et al. 2009 for Laoag City, Philippines), but to our knowledgewas never tested after an earthquake. The fact that the soilfrequencies measured after the 2011 quake are in good agree-920 Seismological Research Letters Volume 82, Number 6 November/December 2011

▲▲Figure 1. Location of the noise measurements taken in Christchurch overlaid on a satellite view from Google Maps. Accelerometricstation CACS and the relevant noise measurement are located northwest of Christchurch airport, out of the upper left corner of theimage.▲▲Figure 2. Map of the frequency of the higher peak in HVSR curves.`Seismological Research Letters Volume 82, Number 6 November/December 2011 921

▲▲Figure 3. Map of the amplitude of the higher peak in HVSR curves.▲▲Figure 4. Shift in soil fundamental frequency moving from the easternmost site N10 to the westernmost site N27.922 Seismological Research Letters Volume 82, Number 6 November/December 2011

▲▲Figure 5. Site N20 showing HVSR as a single, low frequency peak and site N25 with a similar peak plus another one at higher frequency.ment with those determined 17 years earlier provides an opportunityto test the K g coefficient on soil that has not changed itscharacteristics as assessed with microtremor measurements.The map of this coefficient is provided in Figure 6. Thedotted area in the figure indicates the extent of the most severeliquefaction episodes as reported by Cubrinovski and Taylor(2011). The lack of correlation is evident in the most affectedarea, where both high and low values of K g are present, andthe same happens in those zones with no evidence of liquefaction.More detailed analyses are needed, however, before rulingout the K g parameter, since the original formulation is basedon the idea that just one peak is visible in the HVSR curve,corresponding to the fundamental frequency. As stated above,our measurement sometimes returned two peaks correspondingto two resonant strata at different depths, so it would beimportant to have more data about the depth of the layer whereliquefaction occurred.Our third activity was about the possible difference insoil behavior between the elastic, weak-motion domain andthe strong-motion signals. In this paper I report preliminaryresults, since a comprehensive study on all the recording stationsis still underway.To check for variation of soil fundamental frequency thatcould be taken as a possible indication of nonlinear behavior,our research group has recently developed a technique that wasfirst applied to the recordings of the L’Aquila, 2009 earthquake(Puglia et al. 2011). The basic idea is the comparison betweenthe soil fundamental frequency estimated using the HVSRtechnique and the time-frequency behavior during strongmotion estimated using the S-transform technique (Stockwellet al.1996). This transform is particularly useful for analyzing asystem that changes its dynamic characteristics over time sinceit provides information about the local spectrum of a genericsignal overcoming the limitations derived from the assumptionsof the stationarity of a signal (as is the case for the shorttimeFourier transform).Using the same frequency scale it is possible to comparethe S-transform with the HVSR to verify whether the fundamentalfrequency obtained from the noise recording remainsconstant during the S-waves phase, until the coda-waves andto the end of the signal. The S-transform is normalized to themaximum of each time step. The normalized S-transformallows us to better identify the site fundamental frequency,while the standard S-transform is more useful for identifyingthe instant-by-instant frequency content of the signal, itschange over time, and the related energy.The example given here is for station CBGS at ChristchurchBotanical Gardens (Figure 7). The dashed red line highlights thecracks in the ground and the sand left by liquefaction, still clearlyvisible two months after the February 2011 quake. The acceler-Seismological Research Letters Volume 82, Number 6 November/December 2011 923

▲▲Figure 6. Map of the K g parameter and its relation to liquefaction.▲ ▲ Figure 7. The building hosting the accelerometric station CBGS at Christchurch Botanical Gardens. The dashed red line highlightsthe cracks in the ground and the sand left by liquefaction. There are a few centimeters of settlement to the left of this line.924 Seismological Research Letters Volume 82, Number 6 November/December 2011

▲▲Figure 8. Comparison between normalized S-transform and HVSR at CBGS (see text for details).ometer is located inside the building, directly on the concretefoundation slab visible outside the door and under the bench. Inaddition to cracks and liquefaction, a settlement of the stationoccurred with a few centimeters displacement. This place is theideal candidate for strong nonlinear effects, which are indeedvisible in Figure 8. Figure 8 includes information derived fromaccelerometric and noise recordings, showing the accelerometricrecording on the top left, the S-transform on the bottom left,and the HVSR on the bottom right panel; most of the energyof the largest horizontal component of motion is at frequencieslower than the fundamental one determined by HVSR. Itis also worth noting that between 15 to 20 s, the time-domaintrace and the S-transform show high-frequency accelerationevidence of hardening nonlinearity of the kind first describedby Bonilla et al. (2005), due to hysteretic dilatant behavior ofnon-cohesive, partially saturated soils (for more details on accelerometricrecordings and liquefaction studies, see Bradley andCubrinovski (2011, page 853 of this issue) and Smyrou et al.(2011, page 882 of this issue).CONCLUSIONSDuring a quick field survey in Christchurch after the February2011 earthquake it was possible to collect microtremor measurementsclose to accelerometric stations that recorded theevent in the most damaged areas and with overall good spatialcoverage. This allowed us to compare current measurementswith the microzonation map produced by Toshinawa et al.(1997), and in particular with their map of HVSR frequenciesand peak values. There is a general agreement in the frequencymap, with frequency decreasing from the western part of thecity moving toward the estuarine area. The main differencewith the previous study is that we were able to identify frequenciesbelow 1 Hz that returned the maximum HVSR amplitudein the ocean coastline area but were also visible in other partsof the city. Another difference is that the HVSR amplitudemap always provided higher values in this study with respect toToshinawa et al. (1997).Christchurch was affected by severe and widespread soilliquefaction. The fact that the soil frequency remained stablewith respect to measurements performed before the ongoingearthquake sequence prompted us to check the reliability ofthe soil vulnerability index K g introduced by Nakamura (1996)for estimating earthquake damage on the ground surface andpreviously used to map the soil susceptibility to liquefactionbut never tested after an earthquake. No clear correlation hasbeen found in this study between K g and the occurrence of liquefaction.Possible explanations for this are (1) the NakamuraSeismological Research Letters Volume 82, Number 6 November/December 2011 925

(1996) formulation is based on just the fundamental frequencypeak, while HVSR in Christchurch often returns two peakscorresponding to resonant strata at different depths; and (2)the K g parameter is dependent on the square of HVSR amplitude,which is quite unstable as discussed before.Finally, the same technique based on S-transform, whichafter the L’Aquila 2009 earthquake did not point out significantevidence of nonlinearity, here shows clear signs of energyat frequencies lower than the fundamental one in the elasticdomain (softening nonlinearity) in the coda of accelerogramsfrom CBGS; at the same time it is possible to recognize hintsof hardening nonlinearity due to hysteretic dilatant behaviorof soils.Future research will include a second, more detailed mappingof soil frequency using HVSR and comparison betweenelastic and nonlinear behavior at all the accelerometric stations,including the recordings of the September 2010 Darfieldearthquake and the June 2011 Christchurch earthquake.ACKNOWLEDGMENTSMany thanks are due to the staff of Canterbury University(Christchurch) who helped with logistical assistance, insightfulfield trips, and stimulating discussions, and in particular toStefano Pampanin, Misko Cubrinovski, Tobias Smith, WengKam, and Umut Akguzel. Thanks to Rocco Ditommaso forthe S-transform calculations. The paper was prepared duringa stay at GFZ–Helmholtz Zentrum, Potsdam, and benefitedfrom comments from colleagues after a seminar presentation.REFERENCESBeroya, M. A. A., A. Aydin, R. Tiglao, and M. Lasala (2009). Use ofmicrotremor in liquefaction hazard mapping. Engineering Geology107, 140–153.Bonilla, L. F., R. J. Archuleta, and D. Lavallée (2005). Histeretic anddilatant behavior of cohesionless soils and their effects on nonlinearsite response: Field data observation and modeling. Bulletin of theSeismological Society of America 95, 2,373– 2,395.Bradley, B. A., and M. Cubrinovski (2011). Near-source strong groundmotions observed in the 22 February 2011 Christchurch earthquake.Seismological Research Letters 82, 853–865.Brown, L. J., and J. H. Weeber (1992). Geology of the Christchurch UrbanArea. Institute of Geological and Nuclear Sciences, Map 1, 1 sheet+ 104 pp. Lower Hutt, New Zealand: GNS Science.Chatelain, J.-L., B. Guillier, F. Cara, A.-M. Duval, K. Atakan, and theSESAME Working Group (2008). Evaluation of the influenceof experimental conditions on H/V results from ambient noiserecordings. Bulletin of Earthquake Engineering 6 (1), 33–74.Cubrinovski, M., and M. Taylor (2011). Liquefaction Map V.1.0 22Feb. 2001 earthquake, http://db.nzsee.org.nz:8080/en/web/chch_2011/geotechnical/-/blogs/liquefaction-map-drive-throughreconnaissance.Last accessed 19 September 2011.Di Giacomo, D., M. R. Gallipoli, M. Mucciarelli, S. Parolai, and S. M.Richwalski (2005). Analysis and modeling of HVSR in the presenceof a velocity inversion: The case of Venosa, Italy. Bulletin of theSeismological Society of America 95, 2,364–2,372.Nakamura, Y. (1996). Real-time information systems for hazard mitigation.In Proceedings of the 10th World Conference in EarthquakeEngineering, paper # 2134.Parolai, S., and S. M. Richwalski (2004). The importance of convertedwaves in comparing H/V and RSM site response estimates. Bulletinof the Seismological Society of America 94 (1), 304–313.Puglia, R., R. Ditommaso, F. Pacor, M. Mucciarelli, L. Luzi, and M.Bianca (2011). Frequency variation in site response as observedfrom strong motion data of the L’Aquila, 2009 seismic sequence.Bulletin of Earthquake Engineering 9 (3), 869–892; doi:10.1007/s10518-011-9266-2.Smyrou, E., P. Tasiopoulou, İ. E. Bal, and G. Gazetas (2011). Groundmotions versus geotechnical and structural damage in the February2011 Christchurch earthquake. Seismological Research Letters 82,882–892.Stockwell, R. G., L. Mansinha, and R. P. Lowe (1996). Localization ofthe complex spectrum: The S transform. IEEE Transactions onSignal Processing 44, 998–1,001.Toshinawa, T., J. J. Taber, and J. B. Berrill (1997). Distribution ofground-motion intensity inferred from questionnaire survey, earthquakerecordings, and microtremor measurements—A case studyin Christchurch, New Zealand, during the 1994 Arthurs Passearthquake, Bulletin of the Seismological Society of America 87 (2),356–369.Department of Structural Engineering, GeotechnicalEngineering, Engineering GeologyBasilicata UniversityViale dell’Ateneo Lucano, 1085100 Potenza Italymarco.mucciarelli@unibas.it926 Seismological Research Letters Volume 82, Number 6 November/December 2011

Use of DCP and SASW Tests to EvaluateLiquefaction Potential: Predictions vs.Observations during the Recent New ZealandEarthquakesRussell A. Green, Clint Wood, Brady Cox, Misko Cubrinovski, Liam Wotherspoon, Brendon Bradley, Thomas Algie, John Allen, Aaron Bradshaw, and Glenn RixRussell A. Green, 1 Clint Wood, 2 Brady Cox, 2 Misko Cubrinovski, 3Liam Wotherspoon, 4 Brendon Bradley, 3 Thomas Algie, 5 John Allen, 6Aaron Bradshaw, 7 and Glenn Rix 8INTRODUCTIONFollowing both the 4 September 2010 M w 7.1 Darfield and22 February 2011 M w 6.2 Christchurch, New Zealand, earthquakes,Geotechnical Extreme Events Reconnaissance (GEER)team members from the United States and New Zealand visitedthe affected areas to assess geotechnical related damage(e.g., Allen et al. 2010a, b). As shown in Figure 1, liquefactionwas pervasive in large portions of the region after both earthquakes.The widespread liquefaction caused extensive damageto residential properties, water and wastewater networks,high-rise buildings, and bridges. For example, nearly 15,000residential houses and properties were severely damaged fromliquefaction and lateral spreading. More than 50% of thesehouses were damaged beyond economic repair. Also, portionsof the central business district (CBD) were severely damagedby liquefaction during the Christchurch earthquake. It is estimatedthat approximately 30% of the buildings in the CBDwere damaged beyond repair, although not all of the damageresulted from liquefaction.Among the field tests performed by the GEER teamwere the dynamic cone penetrometer (DCP) test (Sowers andHedges 1966) and spectral analysis of surface waves (SASW)test (Stokoe et al. 1994). Both of these tests can provide informationabout the liquefaction susceptibility of soil and are relativelyportable, making them suitable for rapid post-earthquakereconnaissance field studies. The objective of this paper is to1. Department of Civil and Environmental Engineering, VirginiaTech, Blacksburg, Virginia U.S.A.2. University of Arkansas, Fayetteville, Arkansas U.S.A.3. University of Canterbury, Christchurch, New Zealand4. University of Auckland, Auckland, New Zealand5. Partners in Performance, Sydney, Australia6. TRI Environmental, Inc., Duluth Minnesota, U.S.A.7. University of Rhode Island, Kingston, Rhode Island, U.S.A.8. Georgia Tech, Atlanta, Georgia, U.S.A.provide an overview of DCP and SASW tests performed acrossthe Christchurch region and to summarize the comparison ofthe observed versus predicted liquefaction occurrence duringboth the Darfield and Christchurch earthquakes.BACKGROUNDAt 4:35 a.m. on 4 September 2010 NZ Standard Time, thepreviously unmapped Greendale fault ruptured, producing theM w 7.1 Darfield earthquake. The epicenter for this event wasapproximately 40 km west of the center of Christchurch, butthe closest distance from the fault rupture to the western suburbsof Christchurch (e.g., Hornby) was only about 10 km (e.g.,Allen et al. 2010a). As shown in Figure 2, representative geometricmeans of the recorded horizontal peak ground accelerations(PGAs) were 0.71 g in the epicentral region, 0.20 g in theCBD, 0.32 g in Kaiapoi (north of Christchurch), and 0.27 gin Lyttelton (south of Christchurch) (e.g., Allen et al. 2010b).The M w 6.2 Christchurch earthquake occurred at 12:51p.m. on 22 February 2011 NZ Standard Time. As with theDarfield earthquake, the Christchurch earthquake occurred ona previously unmapped fault, the Port Hills fault, located in thePort Hills south of Christchurch. The distance from the epicenterto the center of Christchurch was about 8 km, but the ruptureplane was directly beneath some of the southern neighborhoodsof Christchurch (e.g., Heathcote Valley) and Lyttelton.As shown in Figure 2, representative geometric means of therecorded PGAs were 1.31 g in the epicentral region, 0.42 g inthe CBD, 0.20 in Kaiapoi (north of Christchurch), and 0.11 gin Templeton (west of Christchurch).Much of Christchurch and its environs were originallyswampland, beach dune sand, estuaries, and lagoons that weredrained as part of European settlement (Brown et al. 1995).Consequently, in large areas the near-surface soil stratigraphyis characterized by inter-bedded, loose Holocene aged silt,sand, and gravel that are highly susceptible to liquefactiondoi: 10.1785/gssrl.82.6.927Seismological Research Letters Volume 82, Number 6 November/December 2011 927

▲ ▲ Figure 1. Aerial image of the Canterbury Plains region. Black bordered areas are those that liquefied during the 4 September 2010M w 7.1 Darfield earthquake, while the white shaded areas are those that liquefied during the 22 February 2011 M w 6.2 Christchurchearthquake.(Environment Canterbury [ECan] 2004). This is especially thecase in the eastern portion of Christchurch where the groundwater table is only one to two meters below the ground surface.Unfortunately, the intense shaking of the Darfield andChristchurch earthquakes proved correct ECan’s (2004) findingsregarding the high liquefaction susceptibility of these soils.As shown in Figure 1, widespread liquefaction occurred inthe eastern part of Christchurch and in Kaiapoi during boththe Darfield and Christchurch earthquakes (also see Orense etal. 2011, page 905 of this issue). However, the Christchurchearthquake caused more widespread liquefaction in highlydeveloped areas than did the Darfield earthquake due to therelative close proximity of the fault rupture. For example,liquefaction occurred in large portions of the CBD duringthe Christchurch earthquake that did not liquefy during theDarfield earthquake (Cubrinovski et al. 2011, page 893 ofthis issue). The areas most severely affected by liquefactionwere the suburbs along the Avon River to the east of the CBD(Avonside, Dallington, Avondale, Burwood, and Bexley). Thesoils in these suburbs are predominantly loose fluvial depositsof clean fine sands and sands with non-plastic silts, with the top5–6 m in a very loose state (Gerstenberger et al. 2011). Also,the town of Kaiapoi was significantly impacted by liquefactionduring both events, especially portions of the town that werebuilt on abandoned river channels and fill (Wotherspoon et al.2011).Figure 3 illustrates the typical manifestation of liquefactionon the streets of Christchurch. The severity of the liquefactionled to large settlements of many houses including differentialsettlements that caused foundation and structuraldamage. The largest damage to land and built environmentwas caused by liquefaction-induced lateral spreading along theAvon River, streams, and wetlands in the eastern Christchurchsuburbs. As mentioned above, lateral spreading displacementsranged anywhere from a few tens of centimeters up to 2 m at theriver banks and extended inland as far as 200–300 m from thewaterway, severely damaging roads, pipe networks, and residentialproperties within the affected zone (Robinson et al. 2011).IN-SITU TESTSFollowing both the Darfield and Christchurch earthquakes,GEER team members from the United States and NewZealand performed in-situ tests using the DCP and SASW.Both tests are portable and provide information about thesubsurface properties, making them suitable for immediate928 Seismological Research Letters Volume 82, Number 6 November/December 2011

▲ ▲ Figure 2. Aerial image of Christchurch and its environs. Bullets are the location of the strong motion instrument stations, with thegeometric mean of the peak horizontal accelerations listed for the M w 7.1 Darfield earthquake (first number) and the M w 6.2 Christchurchearthquake (second number).post-earthquake reconnaissance investigations. Of particularinterest to the team were the properties of the soils thatliquefied in either, or both, the Darfield and Christchurchearthquakes. In the following, the DCP and SASW equipment,tests performed, and data reduction are described inmore detail.Dynamic Cone Penetrometer (DCP)The dynamic cone penetrometer (DCP) used for this studywas designed by Professor George Sowers (Sowers and Hedges1966) and is shown in Figure 4. This system utilizes a 6.8-kgmass (15-lb drop weight) on an E-rod slide drive to penetratean oversized 45° apex angle cone. The cone is oversized toreduce rod friction behind the tip. At sites that liquefied, theDCP tests were performed in hand-augered holes that werebored to the top of the layer that liquefied, as determined bycomparing the liquefaction ejecta to the auger tailings. At thesites tested that did not liquefy, the augered holes were boredto the top of the potentially liquefiable layer (i.e., sand layerbelow the ground water table), if such a layer was found. Theaugered holes minimized rod friction and allowed collectionof samples of the liquefiable soil. Experience has shown thatthe DCP can be used effectively in augered holes to depths upto 4.6 to 6.1 m.The DCP tests consist of counting the number of drops ofthe 6.8-kg mass that is required to advance the cone ~4.5 cm(1.75 inches), with the number of drops, or blow count, referredto as the DCP N-value or N DCPT . N DCPT is approximatelyequal to the standard penetration test (SPT) blow count upto an N-value of about 10 (Sowers and Hedges 1966; Green etal. forthcoming). However, beyond an N-value of 10, the relationshipbecomes non-linear. Figure 5A shows the relationshipbetween SPT and DCP N-values that was used in this study,which is a slightly modified version of the one proposed bySowers and Hedges (1966). The modifications to the Sowersand Hedges (1966) relationship are specific to the soils in theCanterbury region and are based on comparing the N DCPTvalues to SPT N-values, cone penetration test (CPT) tip resistance,and shear wave velocity measurements made near theDCP test sites.Following the procedure outlined in Olson et al. (forthcoming),the SPT equivalent N-values (N SPTequiv ) values werenormalized for effective overburden stress and hammer energyusing the following relationship:Seismological Research Letters Volume 82, Number 6 November/December 2011 929

(A)(B)(C)▲▲Figure 3. Typical manifestation of liquefaction during the M w 6.2 Christchurch earthquake (similar manifestation occurred during theM w 7.1 Darfield earthquake): A) A street in Hoon Hay after initial clean up (piling up) of sand and silt ejecta (25 February 2011); B) pilesof sand ejecta from residential properties at Burwood (26 February 2011); and C) massive sand boils in recently developed residentialareas of Burwood; this area did not liquefy in the 4 September quake (26 February 2011). All photos by M. Cubrinovski.930 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A)(B)▲▲Figure 4. DCP test being performed adjacent to a house inBexley after the 4 September 2010 M w 7.1 Darfield earthquake.Photo by R. Green on 15 September 2010.PN 1,60−SPTequiv ≈ N SPTequiv ( N DCPT )⋅⎛ a⎝ ′σ vo⎞0.5 ER⎠ 60% , (1)where N SPT equiv (N DCPT ) is the functional relationship betweenN SPT and N DCPT shown in Figure 5A, P a is atmospheric pressure(i.e., 101.3 kPa), σ′ vo is initial vertical effective stress (in thesame units as P a ), and ER is energy ratio. This relationship usesthe effective stress and hammer energy normalization schemesoutlined in Youd et al. (2001).Although the energy ratio for the system was not measured,the DCP hammer is similar to the donut hammer usedfor the SPT. Skempton (1986) and Seed et al. (1984) suggestedthat the energy ratio for an SPT donut hammer system rangesfrom about 30 to 60%. However, because the DCP system does▲▲Figure 5. A) Relationship between DCP test and SPT N-valuesfor an energy ratio of 60%, and B) comparison of N DCPT and theequivalent N 1,60cs (N 1,60cs-SPTequiv ) for a site in an eastern neighborhoodof Bexley.not have pulleys, a cathead, etc., we anticipate that the energyratio for the DCP is likely to be near the upper end of this range.Therefore, we assumed an ER = 60% for our calculations. Inaddition to the effective stress and hammer energy corrections,the N SPT equiv values were also corrected for fines content followingthe procedure proposed in Youd et al. (2001). Figure5B shows a plot of N DCPT and N 1,60cs–SPTequiv for a test sitein the eastern Christchurch neighborhood of Bexley, whichSeismological Research Letters Volume 82, Number 6 November/December 2011 931

▲ ▲ Figure 6. Aerial image of Christchurch and its environs. Superimposed on the image are locations where DCP tests were performedafter either the Darfield or the Christchurch earthquake.experienced severe liquefaction during both the Darfield andChristchurch earthquakes.In total, 30 DCP tests were performed across Christchurchand its environs after the Darfield and Christchurch earthquakes.Figure 6 shows the locations of these test sites. In additionto the Darfield and Christchurch earthquakes, the DCPhas been used on several other recent post-earthquake investigationsto evaluate deposits that liquefied (e.g., the 2008 M w 6.3Olfus, Iceland, earthquake; the 2010 M w 7.0 Haiti earthquake;the 2010 M w 8.8 Maule, Chile, earthquake; and the 2011 M w5.8 Central Virginia, U.S.A. earthquake).Spectral Analysis of Surface Waves (SASW)The spectral analysis of surface waves (SASW) method is usedto determine the shear wave velocity (V S ) profile at sites tested.The SASW method is widely accepted and has been used tocharacterize the subsurface shear stiffness of soil and rock sitesfor the past 20-plus years (e.g., Nazarian and Stokoe 1984;Stokoe et al. 1994, 2003, 2004; Cox and Wood 2010, 2011;Wong et al. 2011). In particular, the SASW method has oftenbeen applied in geotechnical earthquake engineering to characterizematerials for near-surface site response analyses (e.g.,Rosenblad et al. 2001; Wong and Silva 2006) and soil liquefactionanalyses (e.g., Andrus and Stokoe 2001). The SASW test isa non-intrusive, active source seismic method that utilizes thedispersive nature of Rayleigh-type surface waves propagatingthrough a layered material to infer the subsurface V S profile ofa site.The SASW field measurements in this study were madeusing three 4.5-Hz geophones, a “pocket-portable” dynamicsignal analyzer, and a sledge hammer. Figure 7 shows the testsetup at a site in south Kaiapoi. The geophones were modelGSC-11Ds manufactured by Geo Space Technologies, whilethe analyzer was a Quattro system manufactured by DataPhysics Corporation. The Quattro is a USB-powered, fourinputchannel, two-output channel dynamic signal analyzerwith 205-kHz simultaneous sampling rate, 24-bit ADC,110-dB dynamic range, and 100-dB anti-alias filters. It is controlledwith a flexible, Windows-based software package (DataPhysics Signal Calc) that has measurement capabilities in boththe time and frequency domains. The compact, highly portablenature of this setup is ideal for earthquake reconnaissanceefforts where shallow V S profiles are desired. At most locations,receiver spacings of approximately 0.61, 1.22, 2.44, 4.88, 6.10,and 12.20 m were used to collect surface wave data. These teststook less than 45 minutes per location and typically enabled V Sprofiles to be generated down to 6.1–9.1 m below the surface.In total, 36 SASW tests were performed across Christchurch932 Seismological Research Letters Volume 82, Number 6 November/December 2011

▲▲Figure 7. SASW setup at a site in south Kaiapoi. Photo by B. Cox on 12 Sept 2010.and its environs after the Darfield and Christchurch earthquakes.Figure 8 shows the locations of the SASW test sites.Spectral analysis was used to separate the measured surfacewaves by frequency and wavelength to determine theexperimental (“field”) dispersion curve for the sites via phaseunwrapping. An effective/superposed-mode inversion thattakes into account ground motions induced by fundamentaland higher-mode surface waves as well as body waves (i.e., a fullwavefield solution) was then used to match theoretically thefield dispersion curve with a one-dimensional (1D) layered systemof varying layer stiffnesses and thicknesses (Roesset et al.1991; Joh 1996). The 1D V S profile that generated a dispersioncurve that best matched the field dispersion curve was selectedas the site profile. Per Youd et al. (2001), the V S profiles werethen normalized for effective overburden stress using the followingrelationship:V S1 = V⎛ P aS⎝ ′σ vo0.25⎞ , (2)⎠where V S1 is the shear wave velocity normalized to 1 atm effectivestress, P a is atmospheric pressure (i.e., 101.3 kPa), andσ′ vo is initial vertical effective stress (in the same units as P a ).Figure 9 shows a plot of V S and V S1 for a test site in the easternChristchurch neighborhood of Bexley, which experiencedsevere liquefaction during both the Darfield and Christchurchearthquakes. Also plotted in this figure is the empirically determinedupper-bound V S1 for liquefiable soils (i.e., soils havingV S1 > V* S1 will not liquefy regardless of the intensity of shakingimposed on them).ESTIMATION OF PGAs AT DCP AND SASW TESTSITESAs discussed in the next section, the in-situ test data describedabove correlates to the ability of the soil to resist liquefaction(i.e., capacity). However, to evaluate liquefaction potential,both the soil’s ability to resist liquefaction and the demandimposed on the soil by the earthquake needs to be known. Forthe approach used herein to evaluate liquefaction potential(i.e., stress-based simplified procedure), the amplitude of cyclicSeismological Research Letters Volume 82, Number 6 November/December 2011 933

▲▲Figure 8. Aerial image of Christchurch and its environs. Superimposed on the image are locations where SASW tests were performedafter either the Darfield or the Christchurch earthquake.loading correlates to the PGA at the ground surface and theduration correlates to earthquake magnitude. Accordingly,the PGAs at the sites where DCP and SASW tests were performedneeded to be estimated for both the Darfield andthe Christchurch earthquakes. As outlined below, the PGAsrecorded at the strong motion stations (refer to Figure 2) wereused to compute the conditional PGA distribution at the testsites.The PGA at a strong motion station i can be expressed as:ln PGA i = ln PGA i (Site, Rup) + η + ε i , (3)▲ ▲ Figure 9. Measured (V S ) and corrected (V S1 ) shear wavevelocity profiles for a test site in the eastern Christchurch neighborhoodof Bexley. Also shown is the theoretical limiting upperboundvalue of V S1 for liquefaction triggering (V* S1 ) for soil havingFC = 9%.where ln(PGA i ) is the natural logarithm of the observed PGAat station i; ln PGA i (Site, Rup) is the predicted median naturallogarithm of PGA at the same station by an empirical groundmotion prediction equation (GMPE), which is a function ofthe site and earthquake rupture; η is the inter-event residual;and ε i is the intra-event residual. Based on Equation 3, empiricalGMPEs provide the distribution (unconditional) of PGAshaking as:( ) , (4)ln(PGA i )~ N ln PGA i , ση2 + σε2934 Seismological Research Letters Volume 82, Number 6 November/December 2011

where X ~ N(μ X , σ X 2 ) is shorthand notation for X having a normaldistribution with mean μ X and variance σ X 2 .By definition, all recorded PGAs from a single earthquakehave the same inter-event residual, η. On the other hand, theintra-event residual, ε i , varies from site to site, but is correlatedspatially due to similarities of path and site effects among variouslocations. Accordingly, use can be made of recorded PGAsat strong motion stations (e.g., Figure 2) to compute a conditionaldistribution of PGAs at the DCP and SASW test sites.First, we used the empirical GMPE proposed by Bradley(2010) to compute the unconditional distribution of PGAs atthe strong motion stations. A mixed-effects regression was thenused to determine the inter-event residual, η, and the intra-eventresiduals, ε i ’s, for each strong motion station (Abrahamson andYoungs 1992; Pinheiro et al. 2008).Second, the covariance matrix of intra-event residuals wascomputed by accounting for the spatial correlation betweenall of the strong motion stations and a test site of interest. Thejoint distribution of intra-event residuals at a test site of interestand the strong motion stations is given as:εsiteε = N 0 σ ε Σ0 , site 12, (5)Σ 21 Σ 22SMstation2where X ~ N(μ X , Σ) is shorthand notation for X having amultivariate normal distribution with mean μ X and covariancematrix Σ (i.e., as before, but in vector form); and σ 2 ε siteisthe variance in the intra-event residual at the site of interest.In Equation 5, the covariance matrix has been expressed in apartitioned fashion to elucidate the subsequent computation ofthe conditional distribution of ε site . The individual elements ofthe covariance matrix were computed from:Σ (i, j) = ρ i,j σ εiσ εj, (6)where ρ i,j is the spatial correlation of intra-event residualsbetween the two locations i and j; and σ εiand σ εjare the standarddeviations of the intra-event residual at locations i and j.Based on the joint distribution of intra-event residuals given byEquation 5, the conditional distribution of ε site was computedfrom Johnson et al. (2007):site SMstation1 SMstation 2[ ε ε ] = N ( Σ 12 Σ 22 ε , σεΣsite 12 Σ22 1 Σ21)= N με ε , 2(site SMstation σεsite ε ) (7)SMstationUsing the conditional distribution of the intra-event residualat a test site of interest given by Equation 7 and substitutinginto Equation 4, the conditional distribution of the PGA i wascomputed from:[ ln PGA site ln PGA SMstation ]=2( )N ln PGA site + η + μ ε σsite ε SMstation , ε site ε (8)SMstationIt should be noted that in cases where the test site of interestwas located far from any strong motion station, the conditionaldistribution was similar to the unconditional distribution, andfor test sites of interest located very close to a strong motionstation the conditional distribution approached the valueobserved at the strong motion station.To estimate the PGAs at the DCP and SASW test sites,the unconditional PGAs were estimated using the empiricalGMPE proposed by Bradley (2010) and the conditional PGAswere estimated following the approach outlined above whereinthe spatial correlation model of Goda and Hong (2008) wasused.LIQUEFACTION EVALUATIONUsing the PGAs determined as described above, the cyclicstress ratios (CSRs) at the DCP test sites, for both the Darfieldand Christchurch earthquakes, were calculated following themethodology outlined in Youd et al. (2001). The average of therecommended range of magnitude scaling factors (MSFs) proposedin Youd et al. (2001) was used to compute CSR M7.5 atthe sites.As outlined previously, equivalent SPT N 1,60 values weredetermined from the N DCPT values using Equation 1. Thesevalues were then corrected for fines content (FC) using the procedureproposed in Youd et al. (2001). For many of the sites,samples of the liquefiable soil were collected and analyzed inthe laboratory to determine the FC. However, for sites whereno samples were collected, FC = 12% was assumed, which isrepresentative of the approximate fines content of soils at thesites sampled. Once the N 1,60cs-SPTequiv were determined, thecorrelation proposed by Youd et al. (2001) was used to estimatethe cyclic resistance ratio (CRR) for an M w 7.5 event (i.e.,CRR M7.5 ). Comparisons of the computed CSR M7.5 for boththe Darfield and Christchurch earthquakes and CRR M7.5 fora test site in the eastern Christchurch suburb of Bexley areshown in Figure 10A. As shown in this figure, liquefactionis predicted to have occurred during both earthquakes (i.e.,CSR M7.5 > CRR M7.5 ). However, the factor of safety againstliquefaction (FS) is lower for the Christchurch earthquakethan the Darfield earthquake, where FS = CRR M7.5 /CSR M7.5 . Thelower factor of safety indicates increased severity of liquefaction.These predictions are consistent with field observationsin Bexley made shortly after the two earthquakes (i.e., liquefactionoccurred during both earthquakes, but was more severeduring the Christchurch earthquake).To compare the predicted versus observed liquefaction atall the DCP test sites, each of the DCP logs was analyzed forquality, and critical depths for liquefaction/thickness of thecritical layers were selected. Logs where refusal was met within~0.3 to 0.5 m of the start of the test were removed from thedatabase, where refusal was taken as N DCPT > ~35 for morethan two 4.5-cm drives. The reason for this is that too little ofthe profile was tested in these cases to make a meaningful interpretation.The thicknesses of the critical layers were selectedbased on how liquefaction manifested at the ground surface.Seismological Research Letters Volume 82, Number 6 November/December 2011 935

(A)(A)(B)(B)▲▲Figure 10. Comparison of CSR M7.5 for the Darfield andChristchurch earthquakes with CRR M7.5 for a site in Bexley(FC = 9%): A) profiles for DCP test; and B) profiles for SASWtest.In general, the selected critical layer thickness was thinnest forcases of lateral spreading with no ejecta, intermediate for lateralspreading with ejecta, and thickest for large sand boils withno associated lateral spreading. For example, the profile shownin Figure 10A laterally spread (see Figure 4) and there was asignificant amount of ejecta that vented to the ground surfacenearby. Using this information, and trends in the N DCPT ,shown in Figure 5B, the selected critical layer was ~2 m thick,as indicated in Figure 10A. Once the critical layers were determinedfor each test site, the N 1,60cs-SPTequiv values, CSR M7.5 ,and CRR M7.5 were averaged over these depths. The results wereplotted along with Youd et al. (2001) SPT CRR M7.5 curve inFigure 11A.A similar procedure as that outlined above was used tocompute the CSR M7.5 for the SASW test sites. However, theMSF proposed by Andrus and Stokoe (2000) was used instead▲ ▲ Figure 11. Comparison of predicted versus observed liquefaction:A) DCP test; and B) SASW test.of the average of the recommended range proposed by Youdet al. (2001). The reason for using slightly different MSFs wasto be consistent with how the respective cyclic resistance ratiocurves were developed from the observational data. Usingthe computed V S1 , the CRR M7.5 for the test site profiles werecalculated following the Andrus and Stokoe (2000) procedure;this procedure is also outlined in Youd et al. (2001).Comparisons of the computed CSR M7.5 for both the Darfieldand Christchurch earthquakes and CRR M7.5 for a test site inthe eastern Christchurch suburb of Bexley are shown in Figure10B. Consistent with the DCP test results, liquefaction ispredicted to occur at this site during both the Darfield andChristchurch earthquakes, with the liquefaction predicted tobe more severe during the Christchurch earthquake. Again,these predictions are in line with the post-earthquake observations.936 Seismological Research Letters Volume 82, Number 6 November/December 2011

Using the same critical layers as selected for DCP testliquefaction evaluations, V S1 , CSR M7.5 , and CRR M7.5 wereaveraged over the critical depths for each test site profile. Theresults were plotted along with the Andrus and Stokoe (2000)CRR M7.5 curves in Figure 11B.DISCUSSIONAs shown in Figure 11, the liquefaction predictions made usingboth the DCP and SASW test data reasonably match fieldobservations. This is particularly significant for the DCP databecause a correlation was first required to convert the measuredN DCPT to SPT N-values (shown in Figure 5A), and undoubtedly,this correlation is inherently uncertain. Also, the DCPwas only able to test down to a depth of ~6 m at a maximumand usually less than about 4.5 m. Below this depth, N DCPTbecame large because of the presence of a dense layer and/orbecause of the increase in effective confining stress. Becausethe DCP is manually operated, performing tests beyond ~5 mdepths becomes very laborious even in relatively loose sanddeposits. The SASW test was able to test to deeper depths thatthe DCP, but was still limited to depths of ~6 to 9 m with thesledge hammer source. These depth limits are true shortcomingsof both tests because at a few DCP and SASW test sites,available cone penetration test (CPT) soundings indicated thepresence of potentially liquefiable layers deeper in the profiles.As a result, our selected critical layer may only be one of multiplecritical layers in the profile and may not be the most critical.Also from Figure 11, it can be noted that most of the DCPand SASW tests were performed at sites that liquefied, with apaucity of data from sites that did not liquefy. The reason forthis is the manifestation of liquefaction at the ground surfaceis a definite indication that liquefiable soils are present. Severalno-liquefaction sites were investigated, especially ones adjacentto sites that liquefied. However, in the majority of these caseswe were not able to find a sandy stratum below the groundwater table in the upper ~5 m of these sites using the handauger.As a result, DCP tests were not performed at these sites,and the sites were not included in the DCP database.CONCLUSIONSThe U.S. and New Zealand members of the GEER team performedDCP and SASW tests after the 4 September 2010 M w7.1 Darfield earthquake and the 22 February 2011 M w 6.2Christchurch earthquake. Both tests are relatively portable,making them suitable for rapid, post-earthquake investigations.Of particular interest to the team were characterizing sites thatliquefied during either one or both of the earthquakes. Usingthe in-situ test data in combination with estimated PGAs,the liquefaction potential at the test sites was evaluated andcompared with post-earthquake observations. Despite someshortcomings of the tests, they did a relatively good job in correctlypredicting the occurrence/non-occurrence of liquefaction,proving the value of these tests for rapid, post-earthquakeinvestigations.ACKNOWLEDGMENTSThe primary support for the US GEER Team memberswas provided by grants from the U.S. National ScienceFoundation (NSF) as part of the Geotechnical Extreme EventReconnaissance (GEER) Association activity through CMMI-00323914 and NSF RAPID grant CMMI-1137977. Also,Dr. Wotherspoon’s position at the University of Aucklandis funded by the Earthquake Commission (EQC). However,any opinions, findings, and conclusions or recommendationsexpressed in this material are those of the authors and do notnecessarily reflect the views of the National Science Foundationor the EQC.REFERENCESAbrahamson, N. A. and R. R. Youngs (1992). A stable algorithm forregression analyses using the random effects model. Bulletin of theSeismological Society of America 82 (1), 505–510.Allen, J., S. Ashford, E. Bowman, B. Bradley, B. Cox, M. Cubrinovski,R. Green, T. Hutchinson, E. Kavazanjian, R. Orense, M. Pender,M. Quigley, and L. Wotherspoon (2010a). Geotechnical reconnaissanceof the 2010 Darfield (Canterbury) earthquake. Bulletin of theNew Zealand Society for Earthquake Engineering 43 (4), 243–320.Allen, J., S. Ashford, E. Bowman, B. Bradley, B. Cox, M. Cubrinovski, R.Green, T. Hutchinson, E. Kavazanjian, R. Orense, M. Pender, M.Quigley, and L. Wotherspoon (2010b). Geotechnical Reconnaissanceof the 2010 Darfield (New Zealand) Earthquake. GEER AssociationReport No. GEER-024, ed. R. A. Green and M. Cubrinovski.Andrus, R. D., and K. H. Stokoe II (2000). Liquefaction resistance ofsoils from shear wave velocity. ASCE Journal of Geotechnical &Geoenvironmental Engineering 126 (11), 1,015–1,025.Bradley, B. A. (2010). NZ-specific Pseudo-spectral Acceleration GroundMotion Prediction Equations Based on Foreign Models. Departmentof Civil and Natural Resources Engineering, University ofCanterbury, Christchurch, New Zealand, 324 pp.Brown L. J., R. D. Beetham, B. R. Paterson, and J. H. Weeber (1995).Geology of Christchurch, New Zealand. Environmental &Engineering Geoscience 1 (4), 427–488.Cox, B. R., and C. M. Wood (2010). A comparison of linear-array surfacewave methods at a soft soil site in the Mississippi Embayment. InGeoFlorida 2010: Advances in Analysis, Modeling, and Design, ed.D. O. Fratta et al., 1,369–1,378. Reston, VA: American Society ofCivil Engineers.Cox, B. R., and C. M. Wood (2011). Surface wave benchmarking exercise:Methodologies, results and uncertainties. In GeoRisk 2011:Geotechnical Risk Assessment and Management, ed. C. H. Juang etal., 845–852. Reston, VA: American Society of Civil Engineers.Cubrinovski, M., J. D. Bray, M. Taylor, S. Giorgini, B. Bradley, L.Wotherspoon, and J. Zupan (2011). Soil liquefaction effects in thecentral business district during the February 2011 Christchurchearthquake. Seismological Research Letters 82, 893–904.Environment Canterbury (ECan) (2004). Solid Facts on ChristchurchLiquefaction. Environment Canterbury, Christchurch, NewZealand; http://ecan.govt.nz/publications/General/solid-factschristchurch-liquefaction.pdf.Gerstenberger, M., M. Cubrinovski, G. McVerry, M. Stirling, D.Rhoades, B. Bradley, R. Langridge, T. Webb, B. Peng, J. Pettinga,K. Berryman, and H. Brackley (2011). Probabilistic Assessment ofLiquefaction Potential for Christchurch in the Next 50 Years. GNSScience Report 2011/15, 30 pp.Goda, K., and H. P. Hong (2008). Spatial correlation of peak groundmotions and response spectra. Bulletin of the Seismological Societyof America 98 (1), 354–465.Seismological Research Letters Volume 82, Number 6 November/December 2011 937

Green, R. A., S. M. Olson, B. R. Cox, G. J. Rix, E. Rathje, J. Bachhuber,J. French, S. Lasley, and N. Martin (forthcoming). Geotechnicalaspects of failures at Port-au-Prince seaport during the 12 January2010 Haiti earthquake. Earthquake Spectra.Joh, S. H. (1996). Advances in interpretation and analysis techniquesfor spectral-analysis-of-surface-waves (SASW) measurements.PhD diss., Dept. of Civil, Architectural, and EnvironmentalEngineering, University of Texas, Austin, TX, 240 pp.Johnson, R. A., and D. W. Wichern (2007). Applied MultivariateStatistical Analysis. Upper Saddle River, NJ: Pearson Prentice-Hall.Nazarian, S., and K. H. Stokoe II. (1984). In situ shear wave velocitiesfrom spectral analysis of surface wave tests. Proceedings ofthe Eighth World Conference on Earthquake Engineering, SanFrancisco, California, 21–28 July 1984. International Associationfor Earthquake Engineering (IAEE), 31–38.Olson, S. M., R. A. Green, S. Lasley, N. Martin, B. R. Cox, E, Rathje,J. Bachhuber, and J. French (forthcoming). Documenting liquefactionand lateral spreading triggered by the 12 January 2010 Haitiearthquake. Earthquake Spectra.Orense, R. P., T. Kiyota, S. Yamada, Y. Hosono, M. Okamura, and S.Yasuda (2011). Comparison of liquefaction features observed duringthe 2010 and 2011 Canterbury earthquakes. SeismologicalResearch Letters 82, 905–918.Pinheiro, J., D. M. Bates, S. DebRoy, D. Sarkar, and the R Core Team(2008). nlme: Linear and Nonlinear Mixed Effects Models. R packageversion 3.1, 89 pp.Robinson, K., M. Cubrinovski, and P. Kailey (2011). Field measurements oflateral spreading following the 2010 Darfield earthquake. Proceedingsof the Ninth Pacific Conference on Earthquake Engineering, 14–16April 2011, Auckland, New Zealand, paper no. 52.Roesset, J. M., D. W. Chang, and K. H. Stokoe II (1991). Comparison of2-D and 3-D models for analysis of surface wave tests. Proceedingsof the Fifth International Conference on Soil Dynamics andEarthquake Engineering, vol. 1, 111–126. International Society forSoil Mechanics and Geotechnical Engineering.Rosenblad, B. L., K. H. Stokoe II, E. M. Rathje, and M. B. Darendeli(2001). Characterization of Strong Motion Stations Shaken by theKocaeli and Duzce Earthquake in Turkey. Geotechnical EngineeringReport GR01-1, Geotechnical Engineering Center, University ofTexas at Austin.Seed, H. B., K. Tokimatsu, L. F. Harder, and R. Chung (1984). The Influenceof SPT Procedures on Soil Liquefaction Resistance Evaluations.Report no. UCB\EERC-84/15, Earthquake Engineering ResearchCenter, University of California, Berkeley, CA.Skempton, A. W. (1986). Standard penetration test procedures and theeffects in sands of overburden pressure, relative density, particlesize, aging and overconsolidation. Geotechnique 36 (3), 425–447.Sowers, G. F., and C. S. Hedges (1966). Dynamic cone for shallow in-situpenetration testing, vane shear and cone penetration resistance testingof in-situ soils. American Society of Testing Materials (ASTM)Select Technical Paper 399, Philadelphia, PA: American Society ofTesting Materials.Stokoe, K. H. II, G. W. Wright, A. B. James, and M. R. Jose (1994).Characterization of geotechnical sites by SASW method, inGeophysical Characterization of Sites, ed. R. D. Woods, 15–25. NewDelhi: Oxford Publishers.Stokoe, K. H. II, S. H. Joh, and R. D. Woods (2004). Some contributionsof in situ geophysical measurements to solving geotechnicalengineering problems, in Geotechnical and GeophysicalSite Characterization, Proceedings of the International SiteCharacterization ISC’2 Porto, eds. A. Viana da Fonseca, A. and P.W. Mayne, 97–132. Rotterdam: Millpress.Stokoe, K. H. II, B. L. Rosenblad, J. A. Bay, B. Redpath, J. G. Diehl.,R. A. Steller, I. G. Wong, P. A. Thomas, and M. Luebbers (2003).Comparison of VS Profiles from Three Seismic Methods at YuccaMountain. Proceedings of Soil and Rock America 2003 1, 22–25June 2003, Cambridge MA, 299–306.Wong, I., and W. Silva (2006). The importance of in-situ shear-wavevelocity measurements in developing urban and regional earthquakehazard maps. Proceedings of the 19th Annual Symposium onthe Application of Geophysics to Engineering and EnvironmentalProblems, 2–6 April 2006, 1,304–1,315. Environmental &Engineering Geophysical Society, CD-ROM.Wong, I., K. H. Stokoe II, B. R. Cox, Y.-C. Lin, and F.-Y. Menq (2011).Shear-wave velocity profiling of strong motion sites that recordedthe 2001 Nisqually, Washington, earthquake. Earthquake Spectra27 (1), 183–212.Wotherspoon, L. M., M. J. Pender, and R. P. Orense (2011). Relationshipbetween observed liquefaction at Kaiapoi following the 2010Darfield earthquake and former channels of the WaimakaririRiver. Submitted to Engineering Geology.Youd, T. L., I. M. Idriss, R. D. Andrus, I. Arango, G. Castro, J. T.Christian, R. Dobry, et al. (2001). Liquefaction resistance of soils:Summary report from the 1996 NCEER and 1998 NCEER/NSFworkshops on evaluation of liquefaction resistance of soils. ACSEJournal of Geotechnical and Geoenvironmental Engineering 127(10), 817–833.Department of Civil and Environmental EngineeringVirginia Tech120B Patton HallBlacksburg, Virginia 24061 U.S.A.(R. A. G.)938 Seismological Research Letters Volume 82, Number 6 November/December 2011

Performance of Levees (Stopbanks) duringthe 4 September 2010 M w 7.1 Darfield and22 February 2011 M w 6.2 Christchurch,New Zealand, EarthquakesRussell A. Green, John Allen, Liam Wotherspoon, Misko Cubrinovski, Brendon Bradley, Aaron Bradshaw, Brady Cox, and Thomas AlgieRussell A. Green, 1 John Allen, 2 Liam Wotherspoon, 3Misko Cubrinovski, 4 Brendon Bradley, 4 Aaron Bradshaw, 5 Brady Cox, 6and Thomas Algie 7INTRODUCTIONThe objective of this paper is to summarize the performance ofthe levees (or stopbanks) along the Waimakariri and Kaiapoirivers during the 4 September 2010 M w 7.1 Darfield and 22February 2011 M w 6.2 Christchurch, New Zealand, earthquakes.Shortly after their arrival in the Canterbury area in themid-nineteenth century European settlers started constructingdrainage systems and levees along rivers (Larned et al. 2008).In particular, flooding of the Waimakariri River and its tributariesposed a constant threat to the Christchurch and Kaiapoiareas. The current levee system is a culmination of severalcoordinated efforts that started in earnest in the 1930s and iscomposed of both primary and secondary levee systems. Theprimary levee system is designed for a 450-year flood. Damageestimates for scenarios where the flood protection system isbreached have been assessed at approximately NZ$5 billion(van Kalken et al. 2007). As a result, the performance of thelevee system during seismic events is of critical importance tothe flood hazard in Christchurch and surrounding areas.During the 2010 Darfield and 2011 Christchurch earthquakes,stretches of levees were subjected to motions with peakhorizontal ground accelerations (PGAs) of approximately 0.32g and 0.20 g, respectively. Consequently, in areas where thelevees were founded on loose, saturated fluvial sandy deposits,liquefaction-related damage occurred (i.e., lateral spreading,slumping, and settlement). The performance summary presentedherein is the result of field observations and analysis of1. Department of Civil and Environmental Engineering, VirginiaTech, Blacksburg, Virginia U.S.A.2. TRI Environmental, Duluth, Minnesota U.S.A.3. University of Auckland, Auckland, New Zealand4. University of Canterbury, Christchurch, New Zealand5. University of Rhode Island, Kingston, Rhode Island, U.S.A.6. University of Arkansas, Fayetteville, Arkansas, U.S.A.7. Partners in Performance, Sydney, Australiaaerial images (New Zealand Aerial Mapping 2010, 2011), withparticular focus on the performance of the levees along theeastern reach of the Waimakariri River and along the KaiapoiRiver.In the sections that follow, we first present backgroundinformation about the levee system. This is followed by anoverview of the performance of the levees during the Darfieldand Christchurch earthquakes. Next, we discuss the relationshipbetween the severity of damage to the levees along thedowntown stretch of the Kaiapoi River and the response of thefoundation soils. Finally, we present a summary of the findingsand draw conclusions.BACKGROUND OF THE LEVEE SYSTEMThe Waimakariri River flows from the Southern Alps acrossthe Canterbury Plains between Christchurch, to the south,and Kaiapoi, to the north, and empties into Pegasus Bay inthe east (Figure 1). The river drains a mountainous catchmentarea of 3,566 km 2 and poses the most significant floodhazard in New Zealand (van Kalken et al. 2007). Early effortsby European settlers to realign and contain the river withinits banks were piecemeal and only partially successful (e.g.,Wotherspoon et al. 2011). To better coordinate the efforts andto ensure equal flood protection to both Christchurch andKaiapoi, the Waimakariri River Trust was established in 1923(Griffiths 1979). In response to the 1926 floods (Figure 2),the Trust implemented a major river improvement scheme in1930, known as the Hays No. 2 Scheme. Among other things,the scheme entailed an overall improvement of the levee systemalong the Waimakiriri River. However, these improvementswere unable to prevent the major floods in 1940, 1950,and 1957. These floods prompted a further river improvementscheme in 1960, which entailed benching existing levees andconstructing new levees.doi: 10.1785/gssrl.82.6.939Seismological Research Letters Volume 82, Number 6 November/December 2011 939

▲▲Figure 1. Canterbury region of New Zealand.▲ ▲ Figure 2. 1926 photograph of the Waimakariri River overflowingits banks in Christchurch. (Te Ara Encyclopedia of NewZealand 2010.)The mean annual flow of the Waimakariri River is120 m 3 /s. However, in 1957 the largest flood on recordoccurred, with an estimated peak discharge of 4,248 m 3 /s(Griffiths 1979). This flood initially was estimated to have a100-year return period (Griffiths 1979) but was later revised toan approximately 450-year return period (Ian Heslop, personalcommunication 2010) and essentially served as the design basisflood for the levee improvement scheme implemented in the1960s. Flood protection now includes approximately 100 kmof levees.A typical levee cross-section in the Canterbury regionhas 3:1 horizontal to vertical slopes on both the river and landsides (Figure 3). They range in height from 3 to 5 m above thesubgrade and have a 4-m-wide top, which also serves as anaccess road. A flood event originating in the headwaters ofthe Waimakariri River takes approximately 1.5 days to traveldownstream before it reaches the levee system that protectsChristchurch and surrounding areas. At its crest, the 450-yearevent would leave 90 cm of freeboard, but may only last for fourhours (Heslop, personal communication 2010).The levees were often constructed by pushing up river gravelsand silts. A typical cross-section is made up of a gravel corewith 1-m-thick silt cap, which extends from the river side acrossthe top (Figure 3). The levees typically sit on sandy soils at ornear the ground water level. A toe filter was also constructedon the land side of the levee to prevent piping of sand during ahigh-water event. During the 1960 river improvement scheme,some new levees were constructed and benches were addedto some of the existing levees, both of which were compactedusing vibrating rollers (Tony Boyle, personal communication2010). However, no compaction control or foundation analysiswas conducted (Heslop, personal communication 2010).940 Seismological Research Letters Volume 82, Number 6 November/December 2011

▲▲Figure 3. Typical geometry and soil composition of levee cross-section.SEISMIC PERFORMANCE OF THE LEVEESThe authors performed damage surveys along stretches of theprimary and secondary levees for the Waimakariri River andalong the primary levees for the Kaiapoi River following boththe 2010 Darfield and 2011 Christchurch earthquakes. Thesurveys were performed on foot, in an automobile, and from ahelicopter. Additionally, the authors used high-resolution aerialimages to aid in the damage survey and corresponded withEnvironment Canterbury (ECan) personnel (Ian Heslop andTony Boyle). Heslop and Boyle oversaw the post-earthquakedamage assessments performed by a local consulting firm andcontinue to oversee the repairs to the sections of the damagedlevees. Below is a summary of the levee performance along theeastern reach of the Waimakariri River and along the KaiapoiRiver.As a result of the damage caused by the Darfield earthquake,ECan estimated that the flood capacity of theWaimakariri levee system had been reduced from a 450-yearevent (4,730 m 3 /s) to approximately a 15-year event (1,500m 3 /s). Subsequently, concerns were raised when river flows roseto approximately 1,000 m 3 /s in the days following the Darfieldearthquake (personal communications with Boyle and Heslop2010). ECan proceeded with repairs to the most severely damagedsections of levees within weeks of the Darfield earthquake.Repairs progressed from severely damaged areas tothose with only minor damage. By December 2010, the reconstructionhad increased the flood protection capacity of thesystem to 2,500 m 3 /s or 20-year flood event (Heslop, personalcommunication 2011). These reconstruction costs, as a resultof the Darfield earthquake, were approximately $NZ4 million,which was at the upper bound of the estimate providedby ECan shortly after the earthquake (Heslop, personal communication2010).Damage repair to the levee system on the WaimakaririRiver was nearly complete at the time of the February 2011Christchurch earthquake, with the system having beenreturned to a 3,000 m 3 /s or 1-in-30-year event in December2010. The Christchurch earthquake reduced the flood protectioncapacity to 2,500 m 3 /s or a 20-year flood event. As of July2011, the restoration work has nearly been completed, increasingthe capacity to 4,000 m 3 /s or a 100-year flood event. ECanestimates that an additional $NZ 2 million in damages to thelevees were caused by the Christchurch earthquake (Heslop,personal communication 2011). Total restoration to pre-Darfield earthquake flood capacity is expected by end of 2011.There was minor damage to the levee system caused by the 13June 2011 M w 6.0 aftershock, but it did not result in a reductionin flood capacity.The majority of the damage to the levees resulting fromboth the M w 7.1 Darfield and M w 6.2 Christchurch earthquakesoccurred east of SH1 as depicted in Figure 4 (note, SH1is shown in Figure 7). In Figure 4, damage severity is categorizedusing the scale developed by Riley Consultants (2010,2011). The scale has five grades that range from No Damage toSevere Damage, as summarized in Table 1. As may be observedfrom Figure 4, the damage patterns to the levees following bothearthquakes are very similar, but are in general less severe for theChristchurch earthquake compared to the Darfield earthquake.Note that some portions of the levees were already under repairby the time the authors were able to inspect them following theChristchurch earthquake. In these cases, the authors supplementedtheir field observations, to the extent possible, withobservations both from high resolution aerial images taken theday after the Christchurch earthquake and field observationsmade by ECan consultants (Riley Consultants 2011).The majority of the damage to the levees was a consequenceof liquefaction in the foundation soils that resulted inlateral spreading, slumping, and/or settlement. The damagemostly manifested as longitudinal cracks running along thecrest of the levees (Figure 5A). Although not desirable, moderatecrack widths for this mode of damage are not believed to becritical to the functionality of the levees because they do notprovide a direct seepage path from one side of the levee to theother. However, there is always the potential for these longitu-Seismological Research Letters Volume 82, Number 6 November/December 2011 941

(A)(B)▲▲Figure 4. Observed damage to levees following the A) 4 September 2010 M w 7.1 Darfield earthquake and B) 22 February 2011 M w 6.2Christchurch earthquake. Adapted from Riley Consultants 2010, 2011.942 Seismological Research Letters Volume 82, Number 6 November/December 2011

TABLE 1Damage severity categories (Riley Consultants 2010, 2011)CategoryNo DamageMinor DamageModerate DamageMajor DamageSevere DamageDescriptionNo observed damageCracks up to 5 mm wide and/or 300 mm deep. Negligible settlement of crest.Cracks up to 1 m deep. Some settlement of crest.Cracks greater than 1 m deep. Evidence of deep seated movement and/or settlement.Severe damage or collapse. Gross lateral spread and/or settlement, cracks showing deformation of500 mm or more.dinal cracks to connect undetected transverse cracks or flawsthat only penetrate partway through opposite sides of the levee.Such a tortuous seepage path could potentially enlarge rapidlydue to internal erosion and piping at high river levels.Transverse cracks in the levees were less commonlyobserved than longitudinal cracks and were often associatedwith sharp bends along the length of the levees and/or slumpingof the embankment (Figure 5B). Because these cracks provide adirect seepage path from one side of the levee to the other, theycan severely impact the functionality of the levees. Even transversecracks of minor width could potentially enlarge rapidlydue to internal erosion and piping at high river levels and leadto the failure of that section of the levee.Settlement of levee sections resulted from both post-liquefactionconsolidation in the foundation soils and bearingcapacity failures due to the reduced strength of the liquefiedfoundation soil. In addition to the degradation in levee functionalitydue to cracking associated with the settlement (similarto that discussed above), settlement also reduces the amountof freeboard at high river levels. The significance of this loss isdependent on the magnitude of the settlement, but in generalit is not thought to be a significant issue with the levee system.Another liquefaction-related mode of degradation to thelevees’ capacity is where liquefaction and/or lateral spreadingformed on both sides of the levee. In these cases a potentialflow path is formed down through the vertical feeder dike onthe river side of the levee, laterally through the source stratumunder the levee, and up through the vertical feeder dike on theother side of the levee. Extensive liquefaction was observed onboth sides of the levee along an approximately 0.5-km stretchof the Waimakariri River on Coutts Island Road (Figure 6).From interviews with local land owners and review of maps ofthe area from 1865, this area was part of an old river channel(Wotherspoon et al. 2011). Additionally, borings performedby the authors using a hand auger showed a deep sand depositalong this 0.5-km stretch of levee and buried sticks and logs onboth ends, consistent with an old river channel and river channelbanks.SEVERITY OF DAMAGE AND FOUNDATION SOILSTo examine the relationship between the severity of theinduced damage to the levees and the liquefaction response ofthe foundation soil, a stretch of levees along the Kaiapoi Riverwas examined in more depth. As shown in Figure 4, theselevees sustained damage ranging from No Damage to SevereDamage (Table 1). Following the Darfield earthquake, theNew Zealand Earthquake Commission (EQC) contracted alocal firm to perform a series of cone penetration tests (CPTs),among other in-situ tests, throughout north and south Kaiapoi(Tonkin and Taylor 2010). The locations of the CPT soundingsperformed on, or adjacent to, the levees along the KaiapoiRiver are shown in Figure 7.Representative CPT soundings from the north and southbanks of the Kaiapoi River are presented in Figures 8A and 9A.From interpreting 27 such CPT logs, as well as available boreholedata (Tonkin and Taylor 2011), the soil profile along thenorth bank of the Kaiapoi River east of the Williams StreetBridge is characterized by approximately 4 m of very loose toloose sand overlying approximately 4 m of loose to mediumdense gravelly sand. Below approximately 8 m, the sand andgravel layers tend to be significantly denser than the overlyinglayers. The depth to the ground water table varies, but isapproximately 1.5 m deep. Samples of the liquefiable soilstaken adjacent to the levees on the north bank had fines contentsaround 15%, with the fines being non-plastic. On thesouth bank of the Kaiapoi River east of the Williams StreetBridge, the soil profile is characterized by approximately 6 m ofvery loose to loose silty sand/sand overlying an approximately2-m-thick layer of loose to medium dense sand/gravelly sand.Below approximately 8 m the sand and gravel layers tend to besignificantly denser than the overlying layers. The ground watertable is approximately 2 m deep.Using the 27 Kaiapoi levee CPT soundings and twoadditional soundings performed adjacent to the levees alongthe southern bank of the Waimakariri River in Kainga andBrooklands, the authors analyzed liquefaction of the foundationsoils following the procedures outlined in Youd et al.(2001). The strong motion seismograph station KPOC, locatedin north Kaiapoi (Figure 7), recorded motions from both earthquakes.The geometric mean of the horizontal peak groundaccelerations (PGAs) of the motions recorded during theDarfield and Christchurch earthquakes were 0.32 g and 0.20 g,respectively. The distance from the strong motion station to theCPT sounding locations ranges from approximately 0.7 to 3.7km, with the majority of the soundings being located less than1 km from the station. Because of this close proximity, 0.32 gand 0.20 g PGAs were used to compute the cyclic stress ratios(CSRs) imposed on the soil at all the sounding locations duringthe Darfield and Christchurch earthquakes, respectively.Seismological Research Letters Volume 82, Number 6 November/December 2011 943

(A)(B)▲▲Figure 5. Cracks in levee: A) Example of longitudinal cracks running along the crest of the levee; and B) example of transverse (oroblique) crack in levee.944 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A)(B)▲▲Figure 6. A) Large sand boil on landside of the levee on Coutts Island Rd. B) Large sand boil on river side of the same section of thelevee.Seismological Research Letters Volume 82, Number 6 November/December 2011 945

▲▲Figure 7. Locations of CPT soundings on or adjacent to levees.Figures 8B and 9B show the results from the liquefactionevaluation for the two representative CPT soundingsmentioned above. In these figures, the cyclic resistance ratio(CRR) for each profile and the CSRs for both events are plottedtogether, where both the CRR and CSR are adjusted to anM w 7.5 earthquake. For liquefiable soils (i.e., gravels, sands, andcohesionless silts), liquefaction is predicted to have occurredat depths where the CSR M7.5 > CRR M7.5 . Accordingly, forboth profiles, liquefaction is predicted to have occurred duringthe Darfield earthquake for almost the entire depth fromthe ground water table to the top of the dense gravel/sand layer(i.e., to ~7.5 m and ~11 m for the north and south river banks,respectively). However, during the Christchurch earthquake,marginal liquefaction is predicted to occur at a few isolateddepths within both profiles.In an attempt to relate the severity of the observed leveedamage to the liquefaction response of the foundation soil,plots of factor of safety against liquefaction versus damageindex (i.e., FS Liq vs. DI) and thickness of the liquefied layer versusdamage index (i.e., T vs. DI) were made for the 29 CPTsoundings analyzed. Note, damage index corresponds to thedamage categories proposed by Riley Consultants (2010,2011): 1 = No Damage, 2 = Minor Damage, 3 = ModerateDamage, 4 = Major Damage, and 5 = Severe Damage. Weperformed linear regressions on the data, where first the datafrom the two earthquakes were kept separate (Figure 10) andthen they were combined (Figure 11). In developing these plots,the sections of the levees that were under repair at the time ofthe authors’ field inspections were assumed to have DI = 4.The basis for this is that these sections were given high priorityfor repair, which implies that the sustained damage was significant.However, because the intensity of shaking during theChristchurch earthquake at these locations was significantlyless than that during the Darfield earthquake, it is likely thatthe levels of damage induced by the Christchurch earthquakewere less severe than those from the Darfield earthquake.Expected trends can be identified in all plots (i.e., the damageindex increases as the factor of safety against liquefactiondecreases and as the thickness of the liquefied layer increases).However, the strength of the trends, as indicated by the correlationcoefficients (r 2 ), varies between the two earthquakes whenthe data is treated separately. For example, for the Darfieldearthquake, the lowest correlation coefficient (r 2 = 0.147) is forT vs. DI, but T vs. DI has the highest correlation coefficient(r 2 = 0.625) for the Christchurch earthquake. In contrast, thecorrelation coefficients for FS Liq vs. DI are relatively consistentfor both the Darfield and Christchurch earthquakes (i.e.,r 2 = 0.562 and r 2 = 0.595, respectively). When the data fromthe two earthquakes are combined, r 2 = 0.348 and r 2 = 0.578for T vs. DI and FS Liq vs. DI, respectively.From the correlation coefficients, the factor of safetyagainst liquefaction appears to be a better index for damageseverity than the thickness of the liquefied layer. Thisis not altogether surprising given that a lot of the damage to946 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A)(B)▲▲Figure 8. A) Representative CPT sounding for the north bank of the Kaiapoi River; B) liquefaction evaluation of the site for both theDarfield and Christchurch earthquakes.(A)(B)▲ ▲ Figure 9. A) Representative CPT sounding for the south bank of the Kaiapoi River; B) liquefaction evaluation of the site for both theDarfield and Christchurch earthquakes.the levees resulted from lateral spreading, more so than deepseated slumping and settlement/bearing capacity failures. Ofthese three failure modes, lateral spreading can occur even if arelatively thin layer liquefies, while deep seated slumping andsettlement/bearing capacity failures require a thicker layer toliquefy. This is likely the reason for the disparity between the r 2values for the T vs. DI plots for the Darfield and Christchurchevents. In the case of the Darfield earthquake, the levees weresubjected to relatively intense shaking and the thickness of theliquefied layer was large. However, because lateral spreadingcan occur on even a thin liquefied layer, the r 2 value for the Tvs. DI plot was very low (i.e., r 2 = 0.147). In contrast, the leveeswere subjected to less shaking during the Christchurch earthquakeand the liquefied layers were relatively thin where liquefactionoccurred. However, even these relatively thin liquefiedlayers were thick enough for lateral spreading to occur, whichresulted in damage to the levees and a relatively high value of r 2for the T vs. DI plot (i.e., r 2 = 0.625). The implication of this isthat liquefaction severity indices that account for both the factorof safety against liquefaction and thickness of the liquefiedlayer, such as the liquefaction potential index (LPI) (Iwasakiet al. 1982), may not be appropriate for evaluating the risk ofdamage from liquefaction where lateral spreading is the primaryfailure mode.SUMMARY AND CONCLUSIONSThe seismic stability of the levees in the Christchurch, NewZealand, area is critically important to the flood protectionfor the region. Overall, the levee system performed well duringboth the M w 7.1 Darfield and M w 6.2 Christchurch earthquakes.However, portions of the levees along the easternSeismological Research Letters Volume 82, Number 6 November/December 2011 947

(A)(B)(C)(D)▲▲Figure 10. Correlations relating factor of safety against liquefaction and damage index and thickness of the liquefied layer for thedata from the Darfield and Christchurch earthquakes regressed separately. (Note that the low r 2 value in (B) indicates an extremelyweak correlation between the thickness of the liquefied layer and damage index for the Darfield earthquake; hence a dotted line is usedto show the results of the regression.)(A)(B)▲ ▲ Figure 11. Correlations relating factor of safety against liquefaction and damage index and thickness of the liquefied layer for combinedDarfield and Christchurch earthquake data.948 Seismological Research Letters Volume 82, Number 6 November/December 2011

each of the Waimakariri River and along the Kaiapoi Riversustained varying levels of damage during both events. In allcases observed by the authors, damage was related to liquefactionin the foundation soils. In an attempt to relate the severityof the observed levee damage to the liquefaction response ofthe foundation soil, damage severity was correlated to factorof safety against liquefaction and to thickness of the liquefiedlayer. While damage severity correlated to both of these measures,the factor of safety against liquefaction appears to be thebetter index of damage severity when lateral spreading is theprimary failure mode.DATA AND RESOURCESPersonal communications with Tony Boyle took place 27September 2010. Personal communications with Ian Hesloptook place October and November 2010 and May and July2011.ACKNOWLEDGMENTSThe primary support for the U.S. GEER team memberswas provided by grants from the U.S. National ScienceFoundation (NSF) as part of the Geotechnical Extreme EventsReconnaissance (GEER) Association activity through CMMI-00323914 and NSF RAPID grant CMMI-1137977. Also,Dr. Wotherspoon’s position at the University of Auckland isfunded by the New Zealand Earthquake Commission (EQC).However, any opinions, findings, and conclusions or recommendationsexpressed in this material are those of the authorsand do not necessarily reflect the views of the National ScienceFoundation or the EQC.REFERENCESGriffiths, G. A. (1979). Recent sedimentation history of the WaimakaririRiver, New Zealand. Journal of Hydrology (New Zealand) 18, 6–28.Iwasaki, T., K. Tokida, F. Tatsuoka, S. Watanabe, S. Yasuda, and H. Sato(1982). Microzonation for soil liquefaction potential using simplifiedmethods. Proceedings of the Third International Conference onEarthquake Microzonation, Seattle WA, 1,319–1,330.Larned, S. T., D. M. Hicks, J. Schmidt, A. J. H. Davey, K. Dey, M.Scarsbrook, D. B. Arscott, and R. A. Woods (2008). The SelwynRiver of New Zealand: A benchmark system for alluvial plain rivers.River Research and Applications 24, 1–21.New Zealand Aerial Mapping (2010). Kaiapoi (air photo). Wellington,New Zealand.New Zealand Aerial Mapping (2011). Kaiapoi (air photo). Wellington,New Zealand.Riley Consultants (2010). Waimakariri and Kaiapoi River StopbanksPost Earthquake Condition Assessment. Report 10820-B preparedby Riley Consultants for Environment Canterbury (ECan). RileyConsultants, Christchurch, New Zealand.Riley Consultants (2011). Waimakariri and Kaiapoi River StopbanksFindings of Condition Assessment Post 22 February 2011 Earthquake.Letter Report 10820/2-A from Riley Consultants to EnvironmentCanterbury (ECan). Riley Consultants, Christchurch, NewZealand.Te Ara Encyclopedia of New Zealand (2010). http://www.teara.govt.nz/en/floods/6/7.Tonkin and Taylor (2011). Darfield Earthquake Recovery GeotechnicalFactual Report—Kaiapoi North. Report EP-KAN-FAC preparedby Tonkin and Taylor, Ltd. for the New Zealand EarthquakeCommission. Tonkin and Taylor, Christchurch, New Zealand.van Kalken, T., T. Oliver, I. Heslop, and T. Boyle (2007). Impacts ofsecondary flood embankments on the Waimakariri floodplain,New Zealand. Proceedings of the 32nd Congress of the InternationalAssociation of Hydraulic Engineering and Researchers, July 1–6,2007, Venice, Italy.Wotherspoon, L. M., M. J. Pender, and R. P. Orense (2011). Relationshipbetween observed liquefaction at Kaiapoi following the 2010Darfield earthquake and former channels of the WaimakaririRiver. Submitted to Engineering Geology.Youd, T. L., I. M. Idriss, R. D. Andrus, I. Arango, G. Castro, J. T.Christian, R. Dobry, et al. (2001). Liquefaction resistance of soils:Summary report from the 1996 NCEER and 1998 NCEER/NSFworkshops on evaluation of liquefaction resistance of soils. ASCEJournal of Geotechnical & Geoenvironmental Engineering 127 (10),817–833.Department of Civil and Environmental EngineeringVirginia Tech120B Patton HallBlacksburg, Virginia 24061 U.S.A.rugreen@vt.edu(R. A. G.)Seismological Research Letters Volume 82, Number 6 November/December 2011 949

Performance of Bridges during the 2010Darfield and 2011 Christchurch EarthquakesLiam Wotherspoon, Aaron Bradshaw, Russell Green, Clinton Wood, Alessandro Palermo, Misko Cubrinovski, and Brendon BradleyLiam Wotherspoon, 1 Aaron Bradshaw, 2 Russell Green, 3 Clinton Wood, 4Alessandro Palermo, 5 Misko Cubrinovski, 5 and Brendon Bradley 5INTRODUCTIONThe region in and around Christchurch, encompassingChristchurch city and the Selwyn and Waimakariri districts,contains more than 800 road, rail, and pedestrian bridges.Most of these bridges are reinforced concrete, symmetric, andhave small to moderate spans (15–25 m). The 22 February2011 moment magnitude (M w ) 6.2 Christchurch earthquakeinduced high levels of localized ground shaking (Bradley andCubrinovski 2011, page 853 of this issue; Guidotti et al. 2011,page 767 of this issue; Smyrou et al. 2011, page 882 of thisissue), with damage to bridges mainly confined to the centraland eastern parts of Christchurch. Liquefaction was evidentover much of this part of the city, with lateral spreading affectingbridges spanning both the Avon and Heathcote rivers.The majority of bridge damage was a result of liquefaction-inducedlateral spreading, with only four bridges sufferingsignificant damage on non-liquefiable sites. Abutments,approaches, and piers suffered varying levels of damage, withvery little damage observed in the bridge superstructure.However, bridges suffered only a moderate amount of damagecompared to other structural systems. Because some bridgescritical to the city infrastructure network sustained substantialdamage, extensive traffic disruption occurred immediately followingthe event.This paper presents a summary of field observations andsubsequent analyses on the damage to some of the bridges inthe Canterbury region as a result of the Christchurch earthquake.Reference is also made to the performance of bridgesfollowing the 4 September 2010 M w 7.1 Darfield earthquake(Gledhill et al. 2011), and details of damage progression are presentedwhere applicable. The ground motion characteristics forboth events and the regional soil conditions are first described.We provide descriptions of the damage at each selected bridgesite and compare observations of liquefaction with predictedresponse using in situ test data.1. University of Auckland, New Zealand2 . University of Rhode Island, Kingston, Rhode Island, U.S.A.3. Virginia Tech, Blacksburg, Virginia, U.S.A.4. University of Arkansas, Fayetteville, Arkansas, U.S.A.5. University of Canterbury, Christchurch, New ZealandLOCAL GEOLOGYThe city of Christchurch, shown in Figure 1, is located alongthe central coast of the Canterbury Plains, an approximately50-km-wide and 160-km-long region created by the overlappingalluvial fans of rivers flowing east from the SouthernAlps. Interbedded marine and terrestrial sediments up to 40m deep overlie 300 to 400 m of late Pleistocene sands andgravels (Brown and Weeber 1992). Much of the city was originallyswampland, beach dune sand, estuaries, and lagoons,which were drained as part of the settlement and expansion ofthe city (Brown et al. 1995). A high water table, one to twometers below the ground surface in the east of the city, graduallyincreases in depth moving across the city to the west. Tothe south of the city are the Port Hills, formed from volcanicactivity (Brown and Weeber 1992).Two spring-fed meandering rivers, the Avon and theHeathcote, cut through Christchurch (Figure 2). The AvonRiver passes through the city from west of the Christchurch▲ ▲ Figure 1. Overview of Christchurch city and its surroundings,with the epicenters of the Darfield and Christchurch earthquakesshown by stars. Boundaries of moderate bridge damageduring the Darfield earthquake at Lincoln and Kaiapoi arerepresented by circles. The region of interest for Christchurchbridges presented in Figure 2 is bounded by the dashed whiterectangle (Google Inc. 2011).950 Seismological Research Letters Volume 82, Number 6 November/December 2011 doi: 10.1785/gssrl.82.6.950

▲ ▲ Figure 2. Horizontal peak ground accelerations recorded at strong motion sites in Christchurch during the Darfield and Christchurchearthquakes, and the locations of bridges highlighted in this paper (Google Inc. 2011).central business district (CBD), through the CBD, and to theeastern edge of the city where it enters the Avon-Heathcoteestuary. East of the CBD the Avon River widens as it nears theestuary. The Heathcote is a smaller river and runs from west toeast in the southern part of the city before entering the estuary.GROUND MOTION CHARACTERISTICSOn 4 September 2010, the M w 7.1 Darfield earthquake struck40 km west of the Christchurch CBD at a focal depth of 11 km(Gledhill et al. 2011). The highest recorded ground motionswere near the epicenter, having a maximum horizontal PGA of0.76 g (geometric mean of the horizontal components, appliesto all horizontal PGAs stated herein) and a maximum verticalPGA of 1.26 g. These large vertical accelerations are typicalof the near-source strong motion recordings for this event. Amaximum horizontal PGA of 0.25 g and maximum verticalPGA of 0.22 g were recorded in the Christchurch CBD, andthe PGA generally decreased with distance downstream alongthe Avon River. The largest vertical PGA in the central andeastern areas of Christchurch was 0.32 g at Pages Road pumpingstation.The M w 6.2 2011 Christchurch earthquake was centeredless than 10 km from the Christchurch CBD along the southeasternperimeter of the city in the Port Hills (Figure 1). Theclose proximity and shallow depth of this event caused higherintensity shaking in Christchurch as compared to the Darfieldearthquake. In the city, ground motions were characterized bylarge vertical accelerations resulting from the close proximity tothe fault plane, steeply dipping oblique thrust faulting mecha-Seismological Research Letters Volume 82, Number 6 November/December 2011 951

(A)(B)▲▲Figure 3. Response spectra of the geometric mean of the horizontalaccelerations at strong motion station recordings in centraland eastern Christchurch compared to NZS1170.5 designresponse spectrum for Christchurch, site subsoil class D for500-year return period. A) Darfield earthquake, B) Christchurchearthquake. Four-letter symbols represent different strongmotion stations, positions of which are indicated in Figure 2.nism, and deep alluvial deposits (Beavan et al. 2011, page 789of this issue; Bradley and Cubrinovski 2011, page 853 of thisissue). The highest recorded ground motions were near the epicenterat the Heathcote Valley primary school, with the horizontaland vertical PGAs 1.41 g and 2.21 g, respectively. In theCBD, horizontal PGAs of between 0.37 g and 0.52 g and verticalPGAs of 0.35 g to 0.79 g were recorded. Horizontal PGAsranging from 0.22 g to 0.67 g and vertical PGAs from 0.49 g to1.88 g were recorded in the vicinity of the Avon River (Bradleyand Cubrinovski 2011, page 853 of this issue).The horizontal PGAs for the Darfield and Christchurchearthquakes at the strong motion stations in central and easternChristchurch are summarized in Figure 2. It is clear fromthe data in Figure 2 that the Christchurch event producedmuch higher ground motions than the Darfield event in theCBD and along the Avon and Heathcote rivers. While notshown in this figure, the same can be said for the level of verticalaccelerations experienced in these areas.The horizontal acceleration response spectra from five ofthe strong motions stations in Figure 2 for the Darfield andChristchurch events are compared to the NZS1170.5 designresponse spectrum for a 500-year return period event inChristchurch (hazard factor Z = 0.22) on a site subsoil class D(Standards New Zealand 2004) in Figure 3.Because the bridges in the region are typically short to midspan, the natural period can reasonably be assumed as less than0.8 seconds. Figure 3A shows that during the Darfield event,the spectral acceleration values in this range were generallyless than the values that a bridge would have been designed forusing current standards (although most bridges were designedaccording to older standards with lower design levels). Onlythe spectral accelerations of the ground motion recorded atHeathcote Valley primary school (HVSC) are above the designcode values in this range, likely a result of basin wedge effectsgiven its position at the head of the Heathcote Valley in thePort Hills.In general, the ground motion response spectra from theChristchurch earthquake in Figure 3B were higher than the500-year-return-period design spectrum over the entire vibrationperiod range. The periods of highest spectral response correspondto the expected natural periods of the bridge structuresin the region. Even though bridges likely experienced shakinglevels at or above their design levels throughout this region, themajority sustained minimal damage as a result of ground shakingalone. This can be attributed to the sturdy designs typicalof bridges constructed in the 1950s and 1960s, which was aperiod of extensive bridge replacement in Christchurch.OVERVIEW OF CANTERBURY BRIDGEPERFORMANCEAlthough liquefaction was widespread in central and easternChristchurch, only five bridges suffered severe damage andten developed moderate damage in the 22 February 2011Christchurch earthquake. Most bridges were reopened withina week of the earthquake, with only one closed for a longerperiod of time. Because of the location of the earthquake on thesoutheastern edge of the city, most of the bridge damage wasconfined to central and eastern regions, where ground shakingwas strongest and soil conditions weakest. This paper focuseson the performance of ten of these bridges, the locations ofwhich are indicated in Figure 2. The majority of bridge damagewas a result of lateral spreading of river banks, with only fourbridges damaged on sites that did not experience liquefaction(locations 1, 8, 9, and 10 in Figure 2). The largest distance fromthe fault rupture to an affected bridge was 17 km (correspondingto the moderately damaged Chaney’s Overpass). Eleven ofthe 14 bridges along the Avon River within the CBD suffered952 Seismological Research Letters Volume 82, Number 6 November/December 2011

only minor damage, mostly to their approaches. Outside theCBD, the two remaining bridges along the Avon that did notsuffer moderate or severe damage had only minor approachdamage. Compared to the Avon River, bridges crossing theHeathcote River sustained much less damage despite beingclose to the fault rupture, primarily due to the larger seismicresistance of the foundation soils of these bridges. Apart fromthree cases, all bridges along the Heathcote River were eitherundamaged or developed only minor approach damage.Eight road bridges suffered moderate damage followingthe 4 September 2010 Darfield earthquake, with five of theseclosed for five days or longer. Traffic weight limitations and/orrestricted lane access was in place for a more extended period,all but one of which instances was due to approach damage as aresult of lateral spreading. The Darfield earthquake had a largermagnitude, and thus resulting ground motions affected a muchlarger region, with bridge damage occurring from Lincoln, 15km south of central Christchurch, to Kaiapoi, 16 km north.The most distant bridge damage, at the Williams Street Bridgein Kaiapoi due to lateral spreading, was approximately 30 kmfrom the rupture of the Greendale fault. Within Christchurchcity itself, Gayhurst Road Bridge and South Brighton Bridgeboth experienced moderate damage, principally as a consequenceof lateral spreading (Allen et al. 2010; Palermo et al.2010).DAMAGE ASSOCIATED WITH LIQUEFACTIONBridges along both the Avon and Heathcote rivers sufferedvarying levels of damage from lateral spreading due to theDarfield and Christchurch earthquakes, with ground conditionsand distance from the epicenter influencing this responseas described previously. Even at a given bridge location the levelof damage varied significantly from one end of the bridge to theother, with more damage observed on the inner banks of thelocal river bends, likely a result of the low-energy depositionalenvironment, as compared to the outer banks. In this sectionof the paper, we present an overview on the heavily damagedFerrymead Bridge at the mouth of the Heathcote River andon the most affected bridges along the Avon River from theChristchurch earthquake.The type of bridge damage along the Avon was fairly consistent:settlement and lateral spreading of approaches, backrotationand cracking of the abutments, and some pier damage.In most cases bridge decks restrained movement of the top ofthe abutment, resulting in their back-rotation. There was littlebridge superstructure damage, with only minor crushing andspalling as a result of pounding and relative movement. Unlessotherwise noted, simply supported bridges discussed hereindid not have any bearings. All the damaged bridges previouslymentioned had pile foundations, with lateral spreading forcesplacing large demands on the abutment piles and likely resultingin plastic hinging below grade. The approach fill of severalbridges subsided by up to a meter, resulting in the bridges beingclosed up to a week. In most cases, settlement and spreading ofthe approaches impacted bridge serviceability.The Christchurch CBD bridges crossing the Avon Rivergenerally performed well, with the most common damagebeing minor lateral spreading, compression or slight slumpingof approach material, and minor cracking in abutments.All bridges were single span and were passable to recoveryvehicles in the cordon soon after the event. (The cordon isthe restricted-access area of the CBD, put in place due to thewidespread earthquake damage in the area.) Compared to theAvon River, bridges crossing the Heathcote suffered muchless damage. Apart from the Ferrymead Bridge at the mouthof the Heathcote, all bridges were either undamaged or experiencedonly minor damage. As previously noted, we inferthat this is the result of more resistant foundation soils alongthe Heathcote River relative to the Avon River. Typical damagewas minor approach settlement, with little impact on thebridge abutments and superstructure.Detailed descriptions of the bridges shown in Figure 4with the most severe liquefaction-induced damage and theanalyses of in situ test data at these sites follow. The PGAsused in the liquefaction evaluations were estimated using theground motion prediction equations of Bradley (2010) andthe spatial correlation model of Goda and Hong (2008). Theestimated PGAs at the bridge sites are the geometric mean ofthe two horizontal components for site class D (Standards NewZealand 2004) and are summarized in Table 1. Further informationon the calculation of these PGA values can be found inGreen et al. (2011, page 927 of this issue).Ferrymead BridgeThe Ferrymead Bridge (Figure 4A) was constructed in 1967,runs in the east-west direction, and spans the mouth ofHeathcote River (Figure 2). The bridge is a three-span reinforcedconcrete bridge supported by wall abutments withwingwalls and two four-column bents connected to pile caps.The west abutment and bents are supported by floating pilefoundations, while the eastern bent is supported by end-bearingpile foundations to bedrock, and the east abutment on shallowfoundations on bedrock.Although the Ferrymead Bridge performed well duringthe 2010 Darfield earthquake, at the time of the Christchurchearthquake it was undergoing a major upgrade to includewidening and underpinning of the deck with two reinforcedconcrete girders supported on two drilled shaft foundations.These upgrades had been planned before the occurrence of theDarfield earthquake. One of the girders at the east abutmenthad been completed and the girder at the west abutment waspartly completed when the Christchurch earthquake struck.Also, to allow access for construction cranes and equipment,two temporary steel bridges were erected on both sides of thebridge and were in place at the time of the Christchurch earthquake.Each abutment consisted of two separate sections, one infront of the other (i.e., one section supporting the superstructureand the other abutment block behind it). Lateral spreadingoccurred at the east abutment, with the material overlying thebedrock moving both perpendicular and parallel to the bridgeSeismological Research Letters Volume 82, Number 6 November/December 2011 953

(A)(B)(C)(D)(E)(F)▲▲Figure 4. Bridges damaged primarily as a result of liquefaction: A) Ferrymead, B) South Brighton, C) ANZAC Drive, D) Avondale Road,E) Gayhurst Road, F) Fitzgerald Avenue.TABLE 1Estimates of peak ground accelerations during Darfield and Christchurch earthquakes in the absence of liquefaction atbridges presented in Figure 2.Darfield EarthquakeChristchurch EarthquakeBridge NameConditional MedianPGA (g)Conditional StandardDeviation (ln PGA)Conditional MedianPGA (g)Conditional StandardDeviation(ln PGA)Moorhouse Ave Bridge 0.208 0.259 0.412 0.284Fitzgerald Ave Bridge 0.214 0.293 0.448 0.323Gayhurst Rd Bridge 0.206 0.293 0.495 0.319Avondale Rd Bridge 0.183 0.360 0.344 0.339ANZAC Dr Bridge 0.180 0.379 0.276 0.168South Brighton Bridge 0.188 0.392 0.618 0.404Ferrymead Bridge 0.247 0.371 0.673 0.400Port Hills Overbridge 0.284 0.350 0.677 0.379Horotane Overbridge 0.292 0.344 0.682 0.373Railway Bridge 3 0.364 0.266 0.814 0.288954 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A)(C)(B)▲▲Figure 5. Ferrymead Bridge field investigation data: A) shearwave velocity (V s ) profile; B) liquefaction assessment using V sdata, comparing the cyclic resistance ratio CRR 7.5 for the site tothe Darfield (CSR 7.5 DAR) and Christchurch (CSR 7.5 CHC) cyclicstress ratios. C) Damage to western abutment and temporarystabilization works.axis. The lateral spreading caused permanent rotation andcracking of the abutments and a number of the piers. Extensiveflexural cracking was evident at the base of the piers at theirconnection to the pile cap. The rear section of the east abutmentback-rotated 2.5° and the section supporting the bridgedeck back-rotated 5°. Additionally, surveys showed the eastabutment moved vertically upward 10 cm, but there was negligiblemovement of the eastern pier.Approximately 8-cm-wide lateral cracks were observedin the vicinity of the drilled shaft supporting the new girder,with the cracks running in both the longitudinal and transversedirections. This caused the top of the new concrete bridgegirder to rotate about 2° toward the river and caused approximately30 cm of ground settlement, measured relative to thebottom surface of the new girder, which was originally cast ongrade.Severe liquefaction, as evidenced by significant volumes ofejecta, and lateral spreading occurred in the area leading up tothe west abutment. Surveys showed that the west abutment andpier had settled 20 cm and shifted horizontally 20 cm towardthe river. The soil in front of the abutment settled approximately80 cm, but no appreciable rotation of the abutment wasobserved. The foundations supporting the west bridge pier inFigure 5C had shifted to the east, causing the support columnsto be out of plumb. Remedial efforts have been completed totie back the foundations supporting the western pier that experiencedsignificant tilting to the west abutment using highstrengthsteel rods.Following the Christchurch event, Spectral Analysis ofSurface Waves (SASW) was performed at a location 60 m tothe west of the west abutment. The shear wave velocity (V s ) profilefor the west end of the bridge is shown in Figure 5A. The V sprofile shows a soft soil layer between 1.5 and 4 m depth, overlyinga much stiffer layer, and the water table at 1.75 m depth.Using the PGAs listed in Table 1, the cyclic stress ratios (CSRs)for both the Darfield and Christchurch earthquakes were calculatedfollowing the methodology outlined in Youd et al.(2001). The magnitude scaling factors (MSF) recommended byAndrus and Stokoe (2000) were used to scale the CSRs to anM w 7.5 event (i.e., CSR 7.5 ). Using the shear wave velocity datashown in Figure 5A, the cyclic resistance ratio (CRR 7.5 ) for theprofile was calculated following the Andrus and Stokoe (2000)procedure, also outlined in Youd et al. (2001). The overburdencorrection factor, K σ , was further used to modify the CRR 7.5values (Hynes and Olsen 1999). This method allows for thedirect comparison of the CSR 7.5 induced by the two earthquakeswith the CRR 7.5 for the profile, as shown in Figure 5B.As may be observed from this figure, liquefaction is predictedto have occurred from ~1.5 to 4 m during both the Darfield andChristchurch earthquakes (i.e., CSR 7.5 > CRR 7.5 ), with thefactor of safety against liquefaction being significantly lowerduring the Christchurch event. While evidence of severe liquefactionwas observed following the Christchurch earthquake,no liquefaction was evident following the Darfield earthquake.South Brighton BridgeThe South Brighton Bridge (Figure 4B) was constructed in1980, runs in the east-west direction, and spans the Avon Riverjust north of where the river empties into the Avon-Heathcoteestuary (Figure 2). The bridge is a three-span skewed reinforcedconcrete structure with seat-type abutments on rubberbearings and single piers, all of which are supported by rakedoctagonal precast, prestressed concrete piles. The abutmentrubber bearings were removed due to the permanent movementsthat developed during the Darfield earthquake (Palermoet al. 2010) and were replaced with temporary hardwood packers.The bridge site was a wide wetland prior to the bridgeconstruction. To construct the bridge, two approach embankmentsapproximately 4 m in height were extended out into thewetlands, with the bridge structure spanning the river channel.These embankments were constructed of uncontrolled fillmaterial.Significant cracking of the approach embankments onboth sides of the bridge occurred during the Darfield earth-Seismological Research Letters Volume 82, Number 6 November/December 2011 955

spalling of the bottom flange of the deck. These displacementsrepresent the cumulative effect of both seismic events. Minorflexural cracking at the base of the central pier from transversemovement due to ground shaking was evident following theDarfield event, with minimal additional damage following theChristchurch event.▲ ▲ Figure 6. Settlement of approach material, exposure of rakedpiles, and cracking of the western abutment of South BrightonBridge. Note the rotation of the abutment in relation to the girderand cracking at rear of abutment seat.quake. Slumping of the material adjacent to the abutmentsdeveloped as a result of movement toward the river, whilethe approaches developed lateral spreading perpendicular tothe river (parallel to the sides of the approach embankment).Liquefaction ejecta was evident in the area surrounding theapproaches, with lateral spreading parallel to the river extendingto the north and south of the approaches on both sides.Similar damage occurred as a result of the Christchurch earthquake,with further severe lateral spreading.Lateral spreading due to both the Darfield andChristchurch earthquakes caused the east abutment to backrotateby approximately 7°, with spreading of the underlyingsoils exposing the abutment piles. The piles rotated along withthe abutment structure, with evidence of plastic hinge developmentin both the front and rear rows of piles. The gabionmat used for erosion protection in front of the abutments hadmoved away from the abutment as the underlying soil spread.These soil movements were larger than those observed in theDarfield event.The west abutment back-rotated by approximately 5° andlight cracking was observed on the tension face of the abutmentpiles after that event. The damage to the piles supporting thewest abutment caused by the Darfield earthquake was exacerbatedduring the February earthquake where the abutment hadback-rotated by an additional 3° for a total rotation of approximately8° (Figure 6) and plastic hinging was clearly visible onthe abutment piles. Soil beneath the abutment had settled significantly,exposing 80 cm of the supporting piles. Comparedto the post-Darfield conditions, there had also been a significantincrease in settlement and spreading at this abutment.Differential movement of the abutments relative to thebridge deck was evident, with the east abutment moving about22 cm along the line of skew to the north and settled about 3 to4.5 cm. The west abutment moved 20 cm along the line of skewto the south and settled 8.5 to 9.5 cm, with minor crushing andANZAC Drive BridgeANZAC Drive Bridge was constructed in 2000, runs in thenorth-south direction on State Highway 74 and spans theAvon River (Figure 2). Shown in Figure 4C, the bridge is atriple-span precast concrete girder structure (hollow core deck)that is supported by two four-column bents and concrete abutmentwalls with wingwalls. The south approach and abutmentwere constructed on an embankment fill, while the north endof the bridge was constructed at surrounding grade.The bridge site experienced marginal liquefaction andminor lateral spreading during the Darfield earthquake,but the bridge and its functionality were not affected by thisevent. However, the bridge was damaged by the Christchurchearthquake, yet remained functional after regrading theapproaches. Severe liquefaction, as evidenced by the large volumesof ejecta, and significant lateral spreading occurred inthe areas surrounding the north and south abutments duringthe latter earthquake, with evidence of liquefaction beingmore pronounced on the south end of the bridge. There werea significant number of sand boils and ejecta observed in thelow-lying areas adjacent to the embankment on the south end.Additionally, lateral spreading was observed on both the sidesof the embankment with the cracks running parallel to theroadway and having widths of about 8 to 18 cm. A short sectionof the south approach roadway was repaved and showed anabrupt elevation change due to ground settlement in the vicinityof the bridge abutment.Liquefaction and lateral spreading were less evident onthe north end of the bridge. However, a roundabout directlynorth of the approach possibly obscured some of the evidence.Cracking parallel to the river developed across the roadwayleading up to the north abutment and extended to both sidesof the bridge. The higher elevation of the area around the northapproach likely resulted in smaller volumes of liquefactionejecta as compared to the south approach area.The south abutment back-rotated 6°, as shown in Figure7, and lateral spreading at the base of the abutment resulted ina 30 to 40 cm gap between the concrete abutment and backfill.Also, a large horizontal gap formed between the abutmentand the edge of a walkway running along the riverbank, withthe bridge superstructure restraining the horizontal abutmentmovement. The rotation of the south abutment exposed a rowof steel H-piles supporting the abutment, which also appearedto have rotated along with the abutment. Numerous rubbertires were also exposed that had been placed between the abutmentand a walkway running along the riverbank. These tireswere designed to act as a lateral spreading buffer for the walkway.The north abutment showed similar rotational movementsbut had less rotation, of 3.5–4°. The lateral spreading along the956 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A)(B)▲▲Figure 7. Damage to southern abutment of the ANZAC DriveBridge, with back-rotation of approximately 6° and spreadingbetween abutment and adjacent walkway.(C)(D)▲▲Figure 8. ANZAC Drive Bridge pier damage, with crackingand spalling of cover concrete.base of the abutment was also less, resulting in an 18 to 24 cmgap between the abutment and the backfill. Additionally, thehorizontal gap between the abutment and a walkway runningalong the riverbank was much less relative to the south end.Both of the bridge piers suffered extensive but superficialcracking to the concrete columns and bent as well as the beamcolumnjoint region, with up to 2° of rotation (Figure 8). Whilethe damage first appeared extensive, with apparent shear cracking,further inspection showed that in reality these cracks werelimited to the concrete cover. Spalling of the cover concreteappeared to be primarily the result of rotation of the piles,causing stresses to be concentrated at the edges of the members.These rotations can be attributed to horizontal movement ofthe pile foundations toward the center of the river due to lateralspreading.Following the Christchurch event, SASW tests andDynamic Cone Penetration tests (DCPTs) were carried out▲▲Figure 9. ANZAC Drive Bridge field investigation: A) shearwave velocity (V s ) profile; B) liquefaction assessment using V s ,comparing the cyclic resistance ratio CRR 7.5 for the site to theDarfield (CSR 7.5 DAR) and Christchurch (CSR 7.5 CHC) cyclic resistanceratios; C) dynamic cone penetration test (DCPT) profile (i.e.,N DCPT and equivalent N 1,60cs ); D) liquefaction assessment usingequivalent N 1,60cs , comparing the CRR 7.5 for the site to the CSR 7.5DAR and CSR 7.5 CHC.50 m southwest of the south abutment. The DCPT N-values(N DCPT ) were converted to equivalent standard penetrationtest (SPT) N-values using a modified relationship to that proposedby Sowers and Hedges (1966). Then, the N-values werefurther corrected for rod length, hammer energy, effectiveconfining stress, and fines content following the proceduresoutlined in Youd et al. (2001). The resulting profiles from theSASW tests and DCPTs are shown in Figures 9A and 9C. TheV s data from the SASW test indicates a soft soil layer betweendepths of 1 and 6 m, and the water table at a depth of 1.5 m.The CRR 7.5 profiles for the site were determined usingboth the SASW and DCPT data, per Youd et al. (2001) andas outlined previously for the SASW. Using the PGAs listedin Table 1, the cyclic stress ratios (CSRs) for both the Darfieldand Christchurch earthquakes were calculated following theSeismological Research Letters Volume 82, Number 6 November/December 2011 957

methodology in Youd et al. (2001) and as outlined previously.Comparisons of the CRR 7.5 and CSR 7.5 for the Darfield andChristchurch earthquakes are presented in Figures 9B and 9Dfor the SASW and DCP tests, respectively.As may be observed from Figure 9B (V s data), liquefactionis predicted to have occurred from ~1.75 to ~6 m during boththe Darfield and Christchurch earthquakes, with the factorof safety against liquefaction being slightly lower during theChristchurch event. Similar trends are predicted in Figure 9D(DCPT data), but liquefaction is predicted to have occurredduring both earthquakes in a slightly thinner layer, ~2.5 to~3.25 m. These predictions are consistent with field observations(i.e., liquefaction occurred at the site during both earthquakes,with the liquefaction being more severe during theChristchurch earthquake).Avondale Road BridgeAvondale Road Bridge (Figure 4D) was constructed in 1962,runs approximately in the north-south direction, and spansthe Avon River (Figure 2). The bridge consists of three spans ofprecast reinforced concrete girders that are supported on twothree-column bents and seat-type abutment walls with wingwalls.Since its construction, the bridge has been seismicallyretrofitted using steel brackets, which are bolted to tie the elementsof the bridge together.The bridge was not damaged during the Darfield earthquake,with the region north of the bridge showing no signs ofliquefaction damage. However, just south of the bridge, alongthe inner bank of the river, there were minor to moderate levelsof liquefaction ejecta, with the volume increasing toward thesouthwest. Liquefaction and lateral spreading were more severeduring the Christchurch earthquake, with larger volumes ofejecta and significant lateral spreading adjacent to both sidesof the south abutment. To the north, there was also increasedvolume of ejecta, and moderate spreading 30 m to the west.There was minimal roadway damage adjacent to the northabutment; however the north abutment back-rotated approximately3°. At the south, the abutment has back-rotated 7°, withmoderate settlement of the approach and damage to roadwayand services (Figure 10C). Large lateral spreading cracksextended out from both sides of the abutment, transitioningfrom perpendicular to the riverbanks to parallel over a distanceof approximately 15 m. The superstructure and piers showed nosigns of damage after either earthquake.A cone penetration test (CPT) was performed after theDarfield earthquake, approximately 15 m to the west of thesouth abutment, with the results shown in Figure 10A (Tonkinand Taylor 2011a). The CRR 7.5 profile for the site was determinedusing the CPT data, per Youd et al. (2001). Using thePGAs listed in Table 1, the cyclic stress ratios (CSRs) for boththe Darfield and Christchurch earthquakes were calculatedfollowing the methodology outlined in Youd et al. (2001).Comparisons of the CRR 7.5 and CSR 7.5 for the Darfield andChristchurch earthquakes are presented in Figure 10B. As maybe observed from this figure, the site is predicted to marginallyliquefy during the Darfield earthquake, with the severity(A)(C)(B)▲▲Figure 10. Avondale Road Bridge field investigation: A) CPTprofile; B) liquefaction assessment using CPT data, comparingthe cyclic resistance ratio CRR 7.5 to the Darfield (CSR 7.5 DAR)and Christchurch (CSR 7.5 CHC) cyclic resistance ratio; C) damageto southern abutment and approach.of the liquefaction increased for the Christchurch earthquake.These predictions are consistent with field observations (i.e.,liquefaction occurred at the site during both earthquakes, withthe liquefaction being more severe during the Christchurchearthquake).Gayhurst Road BridgeGayhurst Road Bridge was constructed in 1954, runs inapproximately the north-south direction, and spans the AvonRiver (Figure 2). This integral bridge, shown in Figure 4E, consistsof three-spans of precast reinforced concrete girders supportedby wall piers that were cast in place within the deck andseat-type concrete abutments with wingwalls. Both the piersand the abutments are founded on reinforced concrete piles.Prior to the earthquakes, both approaches were approximatelylevel with the bridge deck as part of the natural level of the riverbanks.Severe liquefaction occurred during the Darfield earthquake,indicated by the significant volume of ejecta to the north958 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A)(B)▲▲Figure 11. Settlement of the region surrounding the northernabutment of the Gayhurst Road Bridge.of the bridge on the inner bank of the river, with lateral spreadingand large settlements developing throughout the entirearea The effects were more severe following the Christchurchearthquake, with an increased volume of ejecta and further lateralspreading and settlement (Figure 11). To the south on theouter bend of the river there was minimal spreading on eitherside of the bridge and only moderate ejecta volumes after theChristchurch earthquake.Both earthquakes caused significant damage to the northapproach and abutment of this bridge, with their combinedeffects resulting in approximately one meter of settlementof the approach adjacent to the abutment. The wingwalls onboth sides of the north abutment displaced toward the river byabout 90 cm at their top and moved laterally between 10 and15 cm away from the abutment perpendicular to the bridgeaxis (Figure 12C). Extensive cracking was evident, with totalexposure of the reinforcement connecting the wingwalls tothe abutment. The north abutment developed 5° of back-rotation,with a fraction of this being initiated during the Darfieldearthquake, as were the wingwall movements.The base of the northern pier rotated toward the center ofthe river, with one face of the pier cracking horizontally alongits length, approximately one meter from the deck soffit. Thiswas initiated in the Darfield event, with crack widening andfurther rotation during the Christchurch earthquake. Lateralspreading was the cause of this damage, with the lateral forceon the pier base developing a large moment at the stiff pier-deckinterface and cracking the pier.At the south abutment there was little indication of settlementof the approach. The wingwalls did not show any appreciabledisplacement, nor did the abutment show any measureablerotation. The southern pier also did not show any obvioussigns of distress.A CPT was performed after the Darfield earthquake,approximately 5 m east of the north abutment, with the resultsshown in Figure 12A (Tonkin and Taylor 2011b). Followingthe procedure outlined in the previous section, CRR 7.5 and(C)▲▲Figure 12. Gayhurst Road Bridge field investigation: A) CPTprofile; B) liquefaction assessment using CPT data, comparingthe cyclic resistance ratio CRR 7.5 to the Darfield (CSR 7.5 DAR)and Christchurch (CSR 7.5 CHC) cyclic resistance ratio; C) damageto northern approach, with approximately 1 m of slumping ofthe approach, with damage and movement of wingwalls.CSR 7.5 were developed for the Darfield and Christchurchearthquakes and are compared in Figure 12B. As may beobserved from this figure, the site is predicted to marginallyliquefy during the Darfield earthquake, with the severity ofliquefaction increased for the Christchurch earthquake. Whilethe prediction for the Christchurch earthquake is consistentwith field observations (i.e., severe liquefaction), the predictionunderestimated the severity of the liquefaction observed duringthe Darfield earthquake.Fitzgerald Avenue BridgesFitzgerald Avenue Bridge, constructed in 1964, runs in thenorth-south direction and spans the Avon River (Figure 2). Thebridge, shown in Figure 4F, consists of two structures supportingsouthbound traffic and northbound traffic, respectively.Each bridge consists of double-span precast concrete girderswith a single wall pier and pile-supported concrete wall abutments.Retrofit had recently been carried out, involving steelSeismological Research Letters Volume 82, Number 6 November/December 2011 959

ackets linking the piers and abutments to the deck. At thelocation of the bridge, the river undergoes a significant changeof direction with the north abutment on the inner bank andthe south abutment on the outer bank. This bridge was on theedge of the central city cordon setup after the Christchurchearthquake, and consequently, was inaccessible to the generalpublic and was used only by vehicles with cordon access.As discussed in Bradley et al. (2010), the soil profileunderlying the north end of the bridge can be approximated asfour distinct layers: (1) sand, ~4.5 m thick, N 1 = 10, V s = 130m/s; (2) sand with fines, ~6.5 m thick, N 1 =15, V s = 160 m/s;(3) sand, ~6.5 m thick, N 1 = 10, V s = 130 m/s; and (4) sand,N 1 = 30, V s = 220 m/s. The soil profile underlying the southend of the bridge is similar to the north end, minus layer (3).The bridge was undamaged by the Darfield earthquake,with no evidence of liquefaction on either side of the bridge.However, during the Christchurch earthquake, significantlateral spreading developed on the east side of the north abutment,with cracks running parallel to the riverbank and materialmoving south toward the river. The north abutment ofthe western bridge was very near the bend in the river, witha free face both perpendicular and parallel to the bridge.Lateral spreading was noted with movement occurring both tothe south and west. Settlements of approximately 0.5 m wereobserved on the north approach as well.Both north abutments showed back-rotation, which—combined with settlement of the river banks at the base ofthe abutments—exposed the abutment piles. The abutmentrotation caused the easternmost pile on the north abutment(Figure 13A) to fail in tension, with the tension face openingup and crack widths measured up to 10 mm. Spalling of thecover concrete on the bottom flange of the deck girder (Figure13B) developed as a result of relative movement of the superstructureand abutment. Minimal settlement of the approachwas observed at the south abutments. Large cracks were noted,however, in the abutment and wingwalls.This bridge had been previously identified as critical tothe bridge network, with an extensive field testing programperformed in the late 1990s. The program included multipleCPTs and standard penetration tests (SPTs) performed at theabutments of both the twin bridges. The subsequent analysesshowed that the north abutment of the eastern bridge was mostvulnerable to liquefaction and structural damage (Bowen andCubrinovski 2008a, 2008b; Bradley et al. 2010), with liquefactionpredicted in the relatively loose sandy soil between2.5 m and 17.5 m. These predictions are very consistent withthe observed response on the bridge during the Christchurchearthquake.DAMAGE NOT ASSOCIATED WITH LIQUEFACTIONMoving away from the Avon and Heathcote rivers, whereliquefaction-induced lateral spreading was the main cause ofdamage, four bridges suffered damage not related to the effectsof liquefaction. One bridge, Railway Bridge 3, was damageddue to the seismically induced lateral earth pressures acting on(A)(B)▲▲Figure 13. Fitzgerald Avenue Bridge damage: A) tension failureof abutment pile and exposure of reinforcement, B) spallingof bottom flange of deck girder.the abutments. Two bridges, Moorhouse Overbridge and PortHills Overbridge, were damaged due to shaking effects thatactivated the transverse response of the structure. The finalbridge, Horotane Overbridge, sustained damage as a result ofshaking and slope stability issues. The final three bridges didnot develop any significant superstructure damage in any of theearthquakes.Railway Bridge 3The Railway Bridge 3 was constructed in 1950 and consistsof a timber deck with brick masonry wingwall abutments,spanning a roadway between built-up railway embankmentsapproximately 3 m in height (Figure 2).The bridge was not damaged by the Darfield earthquake,but extreme shaking during the Christchurch earthquakeresulted in severe cracking and movement of the abutments.This caused deformation in the track ballast and tracks, result-960 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A)(C)(B)▲▲Figure 14. Railway Bridge 3 field investigation: A) shear wavevelocity (V s ) profile; B) liquefaction assessment using V s data,comparing the cyclic resistance ratio CRR 7.5 to the Darfield(CSR 7.5 DAR) and Christchurch (CSR 7.5 CHC) cyclic resistanceratio; C) remediation of bridge using steel frame structure toprop abutments and tracks.ing in a train derailment soon after the event. The bridge wastemporarily remediated in the days following the event usinga steel frame between the abutments and stabilizing walls infront of the wingwalls as shown in Figure 14C.An SASW test was performed 20 m to the west of thebridge following the Christchurch event; the shear wave velocityprofile at this site is shown in Figure 14A. As indicated bythis V s data, the profile consists of ~4 m of a medium denselayer overlying a denser stratum. Using this shear wave velocitydata and the PGAs listed in Table 1, the CRR 7.5 for the siteand the CSR 7.5 induced during the Darfield and Christchurchearthquakes were calculated as outlined above for the otherbridges. The results are plotted in Figure 14B. As may beobserved from this figure, liquefaction is not predicted to occurduring either the Darfield and Christchurch earthquakes (i.e.,CSR 7.5 < CRR 7.5 ). These predictions are consistent with fieldobservations.Moorhouse Avenue OverbridgeThe Moorhouse Avenue Overbridge (Figure 15A) was constructedin 1960 and runs in the east-west direction (Figure 2).The bridge consists of 11 spans of T-girders that are supportedby dual reinforced concrete column bents.This bridge was not damaged during the Darfield earthquake.Following the Christchurch earthquake, the bridge wasout of service to all traffic for about a month due to damage sustainedduring the event. The damage was primarily to a singlecolumn where a deck expansion joint is located. There were nolinkages between sections of the deck as this position, while thecorresponding pier on the opposite side of the bridge did havethese linkages in place. The expansion joint detail extendedinto the column, increasing the slenderness of the piers (i.e.,these columns were of a size comparable to the other columnsalong the span, except that they were split down the middle bythe expansion joint). The columns also had widely spaced transversereinforcement. The damage was likely caused by a combinationof the high accelerations (estimated PGA = 0.41g) anda large velocity pulse, exciting the transverse response of thebridge and resulting in the flexural-buckling failure of the columnsshown in Figure 16.Upon first inspection the bridge had only suffered shearcracking in both columns, but several hours later the bridgewas inspected again and it was observed that the damaged columnshad started to buckle, putting the central span at risk ofcollapse. Temporary props were then put in place to providegravity support for the span until a rehabilitation plan could beimplemented. There was also evidence of concrete spalling andbar buckling at the abutment-deck interface.Port Hills OverbridgeThe Port Hills Overbridge was constructed in 1970 and runsin approximately the northwest-southeast direction (Figure2). The bridge, shown in Figure 15B, consists of a dual six-spanreinforced concrete voided-slab bridge supported by single pierbents and seat-type abutments. This highway bridge had beenrecently retrofitted, with spans and abutments linked togetherwith steel brackets and rods to form an integral system. Soilhad also been excavated from around the end piers so that theheight of the piers would be uniform along the structure.The bridge was not damaged by the Darfield earthquake,but was approximately 1.5 km from the epicenter of theChristchurch earthquake, with an estimated PGA of 0.68 g atthe site. Column damage developed during this event, with thecenter pier forming a plastic hinge at its base and two of thecorner reinforcing bars buckling over a length of 150 mm. Thisdamage was induced by ground shaking, which activated thetransverse response of the bridge. The retrofitted links betweenthe spans and the bolts connecting the span to the abutmenthad elongated, but the bridge was still able to service trafficwith the damage it sustained.Horotane OverbridgeThe Horotane Overbridge, constructed in 1970, is on StateHighway 74 and runs in approximately the northwest-south-Seismological Research Letters Volume 82, Number 6 November/December 2011 961

(A)(B)(C)▲▲Figure 15. Bridges with damage not associated with liquefaction: A) Moorhouse Avenue, B) Port Hills, C) Horotane.east direction (Figure 2). Shown in Figure 15C, the bridge is adual three-span reinforced concrete bridge supported by singlepier bents. The end spans are supported by seat-type abutments,with the structure spanning between two large builtupembankments approximately 9 m high. The embankmentslopes beneath the abutments and parallel to the roadway havean angle of about 33° relative to the horizontal (i.e., 1.5H:1Vslope). This bridge is ~200 m from the Port Hills Overbridge,and had also been recently retrofitted using a similar approach,with abutment seat extensions and linkages between the bridgeelements.This bridge did not suffer any damage during the Darfieldearthquake. During the Christchurch earthquake, the retrofitmethod appeared to have worked well in terms of protectingthe structure. The ties between spans and at the abutmentselongated and pulled out, as they had in the Port HillsOverbridge. Additionally, 60% of the bolts that attached thesoffit of the precast concrete beams to the abutment seat extensionhad sheared off.The northwest abutment back-rotated 1° and a transversecrack developed at the top of the northwest slope near theconcrete abutment; however, this was not continuous acrossthe slope. The southeast abutment back-rotated by 3.4°and atransverse crack developed at the bottom of the slope on thenortheast side. A significant transverse crack 10 cm wide and60 cm deep opened up at the top of the southeast slope and wascontinuous across the width of the bridge. A transverse scarpalso developed near the toe of the 13-m-long slope, extendingbetween the southern piers of the two bridges, suggesting thata slope failure had been initiated in the embankment fill butdid not become unstable. Movement was also evident perpendicularto the bridge axis, with cracking in the slope extendingthrough the abutment, resulting in wide cracking and lateralmovement of the abutment and superstructure (Figure 17).CONCLUSIONS▲▲Figure 16. Moorhouse Avenue Overbridge pier flexural bucklingfailure.Overall, the bridges in Christchurch and the Canterburyregion performed well during the Darfield and Christchurchearthquakes, given the magnitude of the observed groundmotions. Of those bridges that were damaged, the majoritywere as a result of liquefaction-induced lateral spreading, withonly four bridges suffering damage not related to liquefactioneffects. Even though the larger-magnitude Darfield eventaffected a much wider region, the location of the Christchurchevent resulted in more significant damage due to the intensityof shaking in a region of the city with many bridges and highliquefaction susceptibility.As a result of the significant lateral spreading, the mostaffected components of the bridges were the approaches, abutments,piers, and foundation system. Bridges were able to resistthe inertial forces due to shaking, while the compressive lateralspreading forces resulted in abutment rotation and foundationdamage. For almost all cases the predicted and observed lique-962 Seismological Research Letters Volume 82, Number 6 November/December 2011

Events Reconnaissance (GEER) Association activity througha CMMI-00323914 and NSF RAPID grant CMMI-1137977.However, any opinions, findings, and conclusions or recommendationsexpressed in this material are those of the authorsand do not necessarily reflect the views of the National ScienceFoundation. We acknowledge the New Zealand GeoNet projectand its sponsors EQC, GNS Science, and Land InformationNew Zealand for providing ground motion records used in thisstudy.REFERENCES▲ ▲ Figure 17. Abutment damage and superstructure movementof the southeastern abutment of the Horotane Overbridge.faction occurrences were in close accord, independent of whetherthe liquefaction was evaluated using the V s , DCPT, or CPT.Settlement of bridge approaches affected the serviceabilityof many of the affected bridges, and bridges critical to the networkwere seriously damaged, causing significant traffic disruptionimmediately following the event. Nevertheless, the overallnetwork performed well, with only the Moorhouse AvenueOverbridge closed for an extended period of time. This goodperformance is attributed to the fact that most ChristchurchCity Council road bridges built in the 1950s and 1960s wererobust integral bridges. For the recently constructed bridges,good performance was a result of the significant improvementin bridge seismic safety in New Zealand and retrofittingefforts in the past decade. Additionally, the regular configuration,limited span length, and effective restraining methodswere important factors in the reduced vulnerability of theChristchurch bridge network.ACKNOWLEDGMENTSDr. Wotherspoon’s position at the University of Aucklandis funded by the New Zealand Earthquake Commission(EQC). The primary support for the U.S. GEER team memberswas provided by grants from the U.S. National ScienceFoundation (NSF) as part of the Geotechnical ExtremeAllen, J., S. Ashford, E. Bowman, B. Bradley, B. Cox, M. Cubrinovski,R. Green, T. Hutchinson, E. Kavazanjian, R. Orense, M. Pender,M. Quigley, and L. Wotherspoon (2010). Geotechnical reconnaissanceof the 2010 Darfield (Canterbury) earthquake. Bulletin of theNew Zealand Society for Earthquake Engineering 43 (4), 243–320.Andrus, R. D., and K. H. Stokoe II (2000). Liquefaction resistance ofsoils from shear-wave velocity. ASCE Journal of Geotechnical andGeoenvironmental Engineering 126 (11), 1,015–1,025.Beavan, J., E. Fielding, M. Motagh, S. Samsonov, and N. Donnelly(2011). Fault location and slip distribution of the 22 February 2011M W 6.2 Christchurch, New Zealand, earthquake from geodeticdata. Seismological Research Letters 82, 789–799.Bowen, H. J., and M. Cubrinovski (2008a). Psuedo-static analysis ofpiles in liquefiable soils: Parametric evaluation of liquefied layerproperties. Bulletin of the New Zealand Society for EarthquakeEngineering 41 (4), 234–246.Bowen, H. J., and M. Cubrinovski (2008b). Effective stress analysis ofpiles in liquefiable soil: A case study of a bridge foundation. Bulletinof the New Zealand Society for Earthquake Engineering 41 (4), 247–262.Bradley, B. A. (2010). NZ-Specific Pseudo-spectral Acceleration GroundMotion Prediction Equations based on Foreign Models. University ofCanterbury, Department of Civil Engineering, 319 pp.Bradley, B. A., and M. Cubrinovski (2011). Near-source strong groundmotions observed in the 22 February 2011 Christchurch earthquake.Seismological Research Letters 82, 853–865.Bradley, B. A., M. Cubrinovski, R. P. Dhakal, and G. A. MacRae(2010). Probabilistic seismic performance and loss assessment ofa bridge-foundation-soil system. Soil Dynamics and EarthquakeEngineering 30 (5), 395–411.Brown, L. J., R. D. Beetham, B. R. Paterson, and J. H. Weeber (1995).Geology of Christchurch, New Zealand. Environmental &Engineering Geoscience 1 (4), 427–488.Brown, L. J., and J. H. Weeber (1992). Geology of the Christchurch UrbanArea. Institute of Geological and Nuclear Sciences. Lower Hutt,New Zealand: GNS Science.Gledhill, K., J. Ristau, M. Reyners, B. Fry, and C. Holden (2011).The Darfield (Canterbury, New Zealand) M w 7.1 earthquake ofSeptember 2010: A preliminary seismological report. SeismologicalResearch Letters 82 (3), 378–386.Goda, K., and H. P. Hong (2008). Estimation of seismic loss for spatiallydistributed buildings. Earthquake Spectra 24, 889–910.Green, R. A., C. Wood, B. Cox, M. Cubrinovski, L. Wotherspoon,B. Bradley, T. Algie, J. Allen, A. Bradshaw, and G. Rix (2011).Use of DCP and SASW tests to evaluate liquefaction potential:Predictions vs. observations during the recent New Zealand earthquakes.Seismological Research Letters 82, 927–938.Guidotti, R., M. Stupazzini, C. Smerzini, R. Paolucci, and P. Rameri(2011). Numerical study on the role of basin geometry and kinematicseismic source in 3D ground motion simulation of the 22February 2011 M W 6.2 Christchurch earthquake. SeismologicalResearch Letters 82, 767–782.Seismological Research Letters Volume 82, Number 6 November/December 2011 963

Hynes, M. E., and R. S. Olsen (1999). Influence of confining stress onliquefaction resistance. Proceedings of the International Workshopon Physics and Mechanics of Soil Liquefaction, Baltimore, Maryland,USA, 10–11 September 1998Rotterdam, the Netherlands: Balkema,145–152.Palermo, A., M. LeHeux, M. Bruneau, M. Anagnostopoulou, L.Wotherspoon, and L. Hogan (2010). Preliminary findings on performanceof bridges in the 2010 Darfield earthquake. Bulletin of theNew Zealand Society for Earthquake Engineering 43 (4), 412–420.Smyrou, E., P. Tasiopoulou, I. Engin Bal, and G. Gazetas (2011). Groundmotions versus geotechnical and structural damage in the February2011 Christchurch earthquake. Seismological Research Letters 82,882–892.Sowers, G. F., and C. S. Hedges (1966). Dynamic cone for shallow in-situpenetration testing. In Vane shear and cone penetration resistancetesting of in-situ soils. ASTM STP 399. Philadelphia, PA: AmericanSociety of Testing Materials, 29–37.Standards New Zealand (2004). Structural Design Actions, Part 5:Earthquake Actions—New Zealand. Wellington, New Zealand:Standards New Zealand, 82 pp.Tonkin and Taylor (2011a). Darfield Earthquake Recovery GeotechnicalFactual Report—Avondale. Report REP-AVD-FAC preparedby Tonkin and Taylor, Ltd for the New Zealand EarthquakeCommission. Christchurch, New Zealand: Tonkin and Taylor.Tonkin and Taylor (2011b). Darfield Earthquake Recovery GeotechnicalFactual Report—Dallington Lower. Report REP-DAL-FAC preparedby Tonkin and Taylor, Ltd for the New Zealand EarthquakeCommission. Christchurch, New Zealand: Tonkin and Taylor.Youd, T. L., I. M. Idriss, R. D. Andrus, I. Arango, G. Castro, J. T.Christian, R. Dobry, et al. (2001). Liquefaction resistance of soils:Summary report from the 1996 NCEER and 1998 NCEER/NSFworkshops on evaluation of liquefaction resistance of soils. ASCEJournal of Geotechnical and Geoenvironmental Engineering 127(10), 817–833.University of AucklandPrivate Bag 92019Auckland 1142 New Zealandl.wotherspoon@auckland.ac.nz(L. W.)964 Seismological Research Letters Volume 82, Number 6 November/December 2011

New PublicationsCanGeoRefThe American Geosciences Institute (AGI) and the CanadianFederation of Earth Sciences (CFES) have launchedCanGeoRef (www.cangeoref.org), a bibliographic database coveringthe Canadian geoscience literature since the early 1800s.CanGeoRef is the result of a cooperative arrangement betweenCFES and AGI to expand access for smaller companies andindividuals focused on Canadian geoscience to GeoRef, AGI’sglobal bibliographic database for the geosciences. Data havealready been added for Alberta and Manitoba and Ontariois near completion; Newfoundland/Labrador and BritishColumbia will follow shortly.Operational Earthquake Forecasting ReportThe final report of the International Commission onEarthquake Forecasting for Civil Protection, entitled“Operational Earthquake Forecasting: State of Knowledgeand Guidelines for Implementation,” has been published inAnnals of Geophysics (vol. 54, no. 4, 315–391, doi: 10.4401/ag 5350). The complete report can be freely downloaded fromhttp://www.annalsofgeophysics.eu/index.php/annals/article/view/5350.The report has been accepted by the ItalianDepartment of Civil Protection, which commissioned thestudy immediately following the L’Aquila earthquake of 6April 2009. Although written in response to this request, theCommission intends for the report will be useful to othercountries developing operational forecasting procedures andprotocols.Rapid Observation of Vulnerability and Estimationof RiskThe Federal Emergency Management Agency (FEMA) hasreleased Rapid Observation of Vulnerability and Estimationof Risk (ROVER), free, mobile software for pre- and postearthquakebuilding safety screening. ROVER is availableon CD-ROM (FEMA P-154 ROVER CD) from the FEMAPublications Warehouse or via online download. ROVERautomates two de facto international standard paper-based seismicsafety screening procedures: FEMA P-154, Rapid VisualScreening (RVS) of Buildings for Potential Seismic Hazards, andATC-20, Postearthquake Safety Evaluation of Buildings. Thepre-earthquake module is used by field inspectors to quicklycompile an electronic inventory of buildings, record importantseismic features of a building, and generate an automaticestimate of the need for detailed seismic evaluation. The postearthquakemodule is used to quickly perform and manage thesafety tagging (red, yellow, and green tags) almost universallyapplied to buildings after earthquakes. To order FEMA P-154ROVER on CD from Publications Warehouse, phone 800-480-2520 or fax 240-699-0525; to download the free software,visit www.atc-rover.org.doi: 10.1785/gssrl.82.6.965Seismological Research Letters Volume 82, Number 6 November/December 2011 965

SSA 2012 Annual Meeting AnnouncementSeismological Society of AmericaTechnical Sessions17–19 April 2012 (Tuesday–Thursday)San Diego, CaliforniaIMPORTANT DATESTravel Grant Deadline 30 November 2011Abstract Submission Deadline 11 January 2012Program w/Abstracts Online 24 February 2012Meeting Pre-registration Deadline 9 March 2012Hotel Reservation Cut-Off 24 March 2012Online Registration Cut-Off 6 April 2012SSA Annual Meeting AnnouncementPROGRAM COMMITTEEMeeting ContactsTechnical Program Co-ChairsDavid Oglesby and Raul Castro2012Program@seismosoc.orgAbstract SubmissionsJoy TroyerSeismological Society of America510.559.1784joy@seismosoc.orgRegistrationSissy StoneSeismological Society of America510.559.1780sissy@seismosoc.orgExhibitsKatie KadasSeismological Society of America510.559.1783katie@seismosoc.orgPress RelationsNan BroadbentSeismological Society of America408-431-9885nan0604@msn.comCALL FOR PAPERSAbstract Deadline: Tuesday, 11 January 2012. Electronic submissionsrequired. Instructions will be available on the SSAWeb site at www.seismosoc.org/meetings/ on 1 December2011.MEETING INFORMATIONRegistrationRegistration information will be published in the January andMarch issues of SRL and will be available online beginning 1January 2012.Preliminary SchedulePlease note that this year the technical sessions start on Tuesdayrather than on Wednesday as in recent years and the sessions endon Thursday. The field trip will be held on Friday. Events will beheld at the Town and Country Resort and Convention Centerin the Mission Valley area of San Diego, California.Monday, 16 AprilBoard of Directors Meeting (9am–5pm)Registration (3pm–8pm)Icebreaker (6pm–8pm)Tuesday, 17 AprilTechnical Sessions (8am–6pm)Annual Luncheon (12pm–2pm)Town Hall Meeting (7pm–9 pm)966 Seismological Research Letters Volume 82, Number 6 November/December 2011 doi: 10.1785/gssrl.82.6.966

Wednesday, 18 AprilTechnical Sessions (8am–5pm)Lunch (12pm–1pm)Joyner Lecture & Reception (6pm–8pm)Thursday, 19 AprilTechnical Sessions (8am–5pm)Lunch (12pm–1pm)Friday, 20 AprilField TripsThis schedule is subject to change.HOTEL INFORMATIONTown and Country Resort and Convention CenterThe Town and Country, a beautiful garden-filled resort,includes five restaurants, two swimming pools and a day spa.It is next door to a championship golf course and just minutesfrom downtown, Old Town, and other attractions via trolley.By making your reservation from this URL, you will beinsured the SSA conference rate. https://resweb.passkey.com/Resweb.do?mode=welcome_ei_new&eventID=3640472Reasons to Stay at the Town and CountryWhen you stay at the conference hotel, you not only stay atthe most convenient location for the meeting, you help theSeismological Society. The hotel gives us a significant discounton our meeting rooms and food if we book a certain number ofguest rooms. We base the price of meeting registration on makingthat number. If we don’t meet that projection, SSA losesmoney on the meeting and our ability to serve our membersis reduced. This year we have negotiated an affordable roomrate and complimentary in-room Internet. We have blockedrooms in the nicest part of the hotel. Of course, you can stay atother hotels nearby, but you will have to pay for parking at theTown and Country (around $20/day), your room will not beas convenient, and you won’t have the satisfaction of knowingthat your room reservation is helping make the SSA AnnualMeeting a financial success.TRAVEL GRANTSModest travel grants are available to help defray some of thecosts for international SSA members and student SSA memberswho wish to attend the annual meeting in San Diego. Theapplication deadline is 30 November 2011.EligibilityAll grant recipients must be SSA members current through2012.• The International Travel Grant is available to SSA memberswho must travel from outside the US to attend theSSA meeting.• The Student Travel Grant is available to SSA student memberswho must travel more than 500 km (310 miles) toattend the meeting.• The ESC/SSA Travel Grant is available to any student travelingfrom a member-state of the European SeismologicalCommission. This grant is provided under a cooperativeagreement between SSA and ESC.All grant recipients must present their work in either an oral orposter session at the meeting.ApplicationTo apply for any of these awards, please submit an applicationelectronically via the online form at http://www.seismosoc.org/meetings/2012/travel_grants/. The application shouldstate which award you are applying for, provide reasons whyyou are a good candidate for the award, and include the text ofthe abstract for your presentation.DeadlineApplications must be received by SSA no later than 30 November2011. Applicants will be notified about their award status by4 January 2012, one week before the SSA deadline for abstractsubmission. The awards will be announced and the checks presentedat the Annual Luncheon during the SSA meeting.MembershipTo ensure that your membership is current, please renew yourmembership by logging into the members area of the SSA website(http://www.seismosoc.org/members). For more information,contact meetings@seismosoc.org.Seismological Research Letters Volume 82, Number 6 November/December 2011 967

Coming in BSSAIssue 101:6 of the Bulletin of the Seismological Society of America, expected publication December 2011, will present the followingarticles. indicates that online material will be available at the SSA Web site.ARTICLESOn the Probability of Detecting PicoseismicityKatrin Plenkers, Danijel Schorlemmer, Grzegorz Kwiatek, andthe JAQUARS groupSource Parameters of Picoseismicity Recorded atMponeng Deep Gold Mine, South Africa: Implications forScaling RelationsG. Kwiatek, K. Plenkers, G. Dresen, and the JAGUARSResearch GroupMonitoring the Earthquake Source Process in NorthAmericaR.B. Herrmann, H. Benz, and C. J. AmmonInvestigating the Distributions of Differences betweenMainshock and Foreshock MagnitudesChristine Smyth, Jim Mori, and Masumi YamadaResolution of Seismic-Moment Tensor Inversions from aSingle Array of ReceiversIsmael Vera Rodriguez, Yu J. Gu, and Mauricio D. SacchiMagnitude-Scaling Rate in Ground Motion PredictionEquations for Response Spectra from Large, ShallowCrustal EarthquakesJohn Zhao and Ming LuStatistical Analysis of the 2002 M w 7.9 Denali EarthquakeAftershock SequencePathikrit Bhattacharya, Mary Phan, and Robert ShcherbakovMoment-Constrained Finite-Fault Analysis UsingTeleseismic P waves: Mexico Subduction ZoneC. Mendoza, S. Castro Torres, and J. M. Gomez GonzalezCalifornia Integrated Seismic Network (CISN) LocalMagnitude Determination in California and VicinityR. A. Uhrhammer, M. Hellweg, K. Hutton, P. Lombard, A. W.Walters, E. Hauksson, and D. OppenheimerQuantifying a Potential Bias in Probabilistic SeismicHazard Assessment: Seismotectonic Zonation WithFractal PropertiesMatteo Spada, Stefan Wiemer, and Eduard KisslingEpistemic Uncertainty in the Location and Magnitude ofEarthquakes in Italy from Macroseismic DataW. H. Bakun, A. Gómez Capera, and M. StucchiPreliminary Probabilistic Seismic Hazard Analysis of theCO2CRC Otway Project Site, Victoria, AustraliaMark Stirling, Nicola Litchfield, Matthew Gerstenberger, DanClark, Brendon Bradley, John Beavan, Graeme McVerry, RussVan Dissen, Andy Nicol, Laura Wallace, and Robert BuxtonImprovements to Seismic Monitoring of the EuropeanArctic Using Three-Component Array Processing at SPITSS. J. Gibbons, J. Schweitzer, F. Ringdal, T. Kværna, S. Mykkeltveit,and B. PaulsenRecurrent Morphogenic Earthquakes in the PastMillennium along the Strike-Slip Yushu Fault, CentralTibetan PlateauAiming Lin, Dong Jia, Gang Rao, Bing Yan, Xiaojun Wu, andZhikun RenIntegration of Paleoseismic Data from Multiple Sites toDevelop an Objective Earthquake Chronology:Application to the Weber Segment of the Wasatch FaultZone, UtahChristopher B. DuRoss, Stephen F. Personius, Anthony J.Crone, Susan S. Olig, and William R. LundThe Crustal and Upper-Mantle Structures beneath theNortheastern Margin of TibetXuzhang Shen, Xiuping Mei, and Yuansheng ZhangCrustal Structure in the Southern Appalachians:A Comparison of Results Obtained from Broadband Dataand Three-Component, Wide-Angle P and S ReflectionDataM. Scott Baker and Robert B. HawmanStripping Analysis of Ps-Converted Wave PolarizationAnisotropyHitoshi OdaUpper-Crust Shear-Wave Velocity of South KoreaConstrained by Explosion and Earthquake DataHeeok Jung, Yong-seok Jang, and Bong Gon JoThe Green’s Functions Constructed from 17 Years ofAmbient Seismic Noise Recorded at Ten Stations of theGerman Regional Seismic NetworkDanuta Garus and Ulrich WeglerScattered P′P′ Waves Observed at Short DistancesPaul S. Earle, Sebastian Rost, Peter M. Shearer, and ChristineThomas968 Seismological Research Letters Volume 82, Number 6 November/December 2011 doi: 10.1785/gssrl.82.6.968

Verification of a Spectral-Element Method Code for theSouthern California Earthquake Center LOH.3Viscoelastic CaseFlorent De MartinApplication of the Nearly Perfectly Matched Layer toSeismic-Wave Propagation Modeling in ElasticAnisotropic MediaJingyi Chen and Jianguo ZhaoComparison of the Empirical Green’s Spatial DerivativeMethod and Empirical Green’s Function MethodMichihiro Ohori and Yoshiaki HisadaApplication of the Multichannel Wiener Filter toRegional Event Detection Using NORSAR Seismic-ArrayDataJ. Wang, J. Schweitzer, F. Tilmann, R. S. White, and H. SoosaluScattering and Attenuation of Seismic Waves inNortheastern North AmericaR.D. Cicerone, C.G. Doll Jr., and M. N. ToksözReal-Time Strong-Motion Broadband Displacementsfrom Collocated GPS and AccelerometersYehuda Bock, Diego Melgar, and Brendan W. CrowellAnalysis of the Origins of κ (Kappa) to Compute HardRock to Rock Adjustment Factors for GMPEsChris Van Houtte, Stéphane Drouet, and Fabrice CottonComparison of Site Periods Derived from DifferentEvaluation MethodsD. Motazedian, K. Khaheshi Banab, J. A. Hunter, S. Sivathayalan,H. Crow, and G. BrooksA Stochastic Approach for Evaluating the NonlinearDynamics of Vertical Motion Recorded at the IWTH25Site for the 2008 M w 6.9 Iwate–Miyagi Inland EarthquakeShigeo KinoshitaComparison of Nonlinear Structural Responses forAccelerograms Simulated from the Stochastic Finite-Fault Approach versus the Hybrid Broadband ApproachGail M. Atkinson, Katsuichiro Goda, and Karen AssatouriansNear-Field Response of a 1D-Structure Alluvial SiteDenis Sandron, Livio Sirovich, and Franco PettenatiA Predictive Equation for the Vertical-to-Horizontal Ratioof Ground Motion at Rock Sites Based on Shear-WaveVelocity Profiles from Japan and SwitzerlandBenjamin Edwards, Valerio Poggi, and Donat FähEmpirical Distance Attenuation and the Local MagnitudeScale for Northwest IranMehdi Rezapour and Reza RezaeiForearc versus Backarc Attenuation of EarthquakeGround MotionHadi Ghofrani and Gail AtkinsonRegional Correlations of V S30 and Velocities AveragedOver Depths Less Than and Greater Than 30 MetersDavid M. Boore, Eric M. Thompson, and Héloïse CadetSHORT NOTESNear-Surface Expression of Early-to-Late HoloceneDisplacement along the Northeastern Himalayan FrontalThrust at Marbang Korong Creek, Arunachal Pradesh,IndiaR. Jayangondaperumal, Steven G. Wesnousky, and Barun K.ChoudhuriThe 16 May 1909 Northern Great Plains EarthquakeW. H. Bakun, M. C. Stickney, and G. C. RogersLocation of Aftershocks of the 4 April, 2010 M w 7.2El Mayor–Cucapah Earthquake of Baja California,MexicoRaúl R. Castro, José G. Acosta, Víctor M. Wong, Arturo Pérez-Vertti, Antonio Mendoza, and Luis InzunzaNonvolcanic Tremor in the Aleutian ArcC. L. Peterson, S. R. McNutt, and D. H. ChristensenA New Empirical Magnitude Scaling Relation forSwitzerlandBettina P. Goertz-Allmann, Benjamin Edwards, Falko Bethmann,Nicholas Deichmann, John Clinton, Donat Fäh, andDomenico GiardiniDetermination of Love- and Rayleigh-Wave Magnitudesfor Earthquakes and ExplosionsJessie L. Bonner, Anastasia Stroujkova, and Dale AndersonInversion of Ground-Motion Data from a SeismometerArray for Rotation Using a Modification of Jeager’sMethodWu-Cheng Chi, W. H. K. Lee, J. A. D. Aston, C. J. Lin, andC. C. LiuSite Effects in Unstable Rock Slopes: Dynamic Behaviorof the Randa Instability (Switzerland)Jeffrey R. Moore, Valentin Gischig, Jan Burjanek, Simon Loew,and Donat FähSeismological Research Letters Volume 82, Number 6 November/December 2011 969

COMMENTS AND REPLIESComment on “Evidence that the 2008 M w 7.9 WenchuanEarthquake Could Not Have Been Induced by the ZipingpuReservoir” by Kai Deng, Shiyong Zhou, Rui Wang, RussellRobinson, Cuiping Zhao, and Wanzheng ChengShemin GeReply to “Comment on ‘Evidence that the 2008 M w 7.9Wenchuan Earthquake Could Not Have Been Induced bythe Zipingpu Reservoir’ by Kai Deng, Shiyong Zhou, RuiWang, Russell Robinson, Cuiping Zhao, and WanzhengCheng” by Shemin GeShiyong Zhou and Kai Deng970 Seismological Research Letters Volume 82, Number 6 November/December 2011

EASTERN SECTIONRESEARCH LETTERSReassessment of Stable Continental Regionsof Southeast AsiaRussell L. WheelerRussell L. WheelerU. S. Geological SurveyABSTRACTProbabilistic seismic-hazard assessments of the central andeastern United States (CEUS) require estimates of the size ofthe largest possible earthquake (Mmax). In most of the CEUS,sparse historical seismicity does not provide a record of moderateand large earthquakes that is sufficient to constrain Mmax.One remedy for the insufficient catalog is to combine thecatalog of moderate to large CEUS earthquakes with catalogsfrom other regions worldwide that are tectonically analogousto the CEUS (stable continental regions, or SCRs). After theNorth America SCR, the largest contribution of earthquakesto this global SCR catalog comes from a Southeast Asian SCRthat extends from Indochina to southeasternmost Russia.Integration and interpretation of recently published geologicaland geophysical results show that most of these SoutheastAsian earthquakes occurred in areas exposing abundant alkalineigneous rocks and extensional faults, both of Neogene age(last 23 million years). The implied Neogene extension precludesclassification of the areas as SCR crust. The extensionalso reduces the number of moderate and large Southeast Asianhistorical earthquakes that are available to constrain CEUSMmax by 86 percent, from 43 to six.INTRODUCTIONMost probabilistic seismic-hazard assessments of the centraland eastern United States (CEUS: east of the RockyMountains) and elsewhere worldwide require an estimateof Mmax, the moment magnitude of the largest earthquakethat is thought to be possible within a specified area (Wheeler2009a,b). Wheeler (2009a) cited example assessments, includingRisk Engineering Inc. et al. (1986), Johnston et al. (1994),and Petersen et al. (2008). The value of Mmax is important inprobabilistic computations for building codes and for design ofcritical structures such as nuclear power plants (Mueller 2010).Accurate estimates of Mmax are more important for nuclearreactors than building codes because reactor designs requireconsideration of smaller annual probabilities of unexpectedlystrong ground motions (Petersen et al. 2008; Office of NuclearRegulatory Research 2007).The historical record of the CEUS contains earthquakes ofmoment magnitude M 7.0 or larger only at the seismic zones ofNew Madrid, Missouri; Charleston, South Carolina; and perhapsCharlevoix, Quebec (Ebel 1996, 2011; Johnston 1996c;Hough et al. 2000; Bakun and Hopper 2004). Elsewhere inthe CEUS, sparse seismicity suggests that large earthquakesmay have recurrence intervals longer than the historical record,which is generally two to four centuries long. Wherever sufficientpaleoseismic work has been done in the CEUS outsidethe New Madrid, Charleston, and Charlevoix zones, findingsdocument occurrences of prehistoric earthquakes larger thanany in the historical record that occurred at intervals longerthan the historical record (Madole 1988; Crone and Luza1990; Crone, Machette, and Bowman 1997; Crone, Machette,Bradley et al. 1997; Obermeier 1998; McNulty and Obermeier1999; Tuttle et al. 2006; Cox et al. 2010). If recurrence intervalsare that long, then earthquakes larger than any observedhistorically are possible. If such an earthquake is not in the historicalrecord, then Mmax may not have been observed and itmust be estimated by other means. Indirect methods based onphysics, statistics, or the geologic properties of small areas havegenerally given Mmax estimates that lack strong supportingevidence (Chinnery 1979; Coppersmith et al. 1987; Wheeler2009a). Another approach was needed and the next sectionsummarizes it.Stable Continental RegionsA recent workshop on CEUS Mmax concluded that identificationand study of global tectonic analogs of the CEUS andtheir seismicity is the preferred approach to the problems arisingfrom short historical records (Wheeler 2009b, 141–143).doi: 10.1785/gssrl.82.6.971Seismological Research Letters Volume 82, Number 6 November/December 2011 971

TABLE 1Criteria for Identifying Stable Continental Regions (SCRs)Time Interval * SCR Identification Criteria †Neogene Period(0–23 Ma)1. No rifting or major extensionor transtension after Paleogene2. No deformation of orogenicforeland after Early Cretaceous3. No orogenic activity afterEarly CretaceousPaleogene Period Allowed Not allowed Not allowed(23–65.5 Ma)Late Cretaceous Epoch Allowed Not alowed Not allowed(65.5–99.6 Ma)Early Cretaceous Epoch Allowed Allowed Allowed(99.6–145.5 Ma)* In parentheses, age range of the interval of geologic time, from Gradstein et al. (2004). Ma, millions of years ago.† After Kanter (1994). Allowed: deformation of this age and kind does not disqualify an area from being an SCR. Kanter listeda fourth criterion of no major anorogenic intrusions younger than Early Cretaceous. The criterion is not necessary forevaluation of the Southeast Asian SCRs.Coppersmith et al. (1987) and Coppersmith (1994) suggestedthat the CEUS and other areas worldwide that are tectonicallyanalogous to it may have similar values of Mmax. Coppersmithet al. (1987) suggested combining the historical earthquakesof geologically similar regions into a dataset large enough topotentially provide robust lower bounds on Mmax, or perhapsto be candidates for Mmax itself. Accordingly, Kanter (1994)expressed “tectonically analogous” in terms of four criteria thatbroadly characterize the tectonics of the CEUS and centraland eastern Canada (Table 1). Johnston et al. (1994) used theterm stable continental region (SCR) for an area that meets allfour criteria. Other continental areas are not considered to betectonic analogs of the CEUS and are classified as active continentalcrust (ACR).Participants in the CEUS Mmax workshop were acutelyaware that the geologic variables that control the value ofCEUS Mmax are poorly known (see discussions throughoutWheeler 2009b). Furthermore, the distinction between SCRsand ACRs is not clear in all continental areas. For example, tectonismyoung enough to classify an area as active crust accordingto the criteria of Table 1 may be sparse or unrecognized.Alternatively, the tectonism might not be clearly rifting, orogenicactivity, or deformation of an orogenic foreland. In caseswhere the distinction between stable and active crust is enigmatic,focusing attention on the brittle upper crust can helpto make the distinction. In other cases argument by geologicor tectonic analogy can clarify the distinction. Later sectionsdescribe illustrative cases in and around eastern Mongolia andin Indochina, respectively.With these uncertainties in mind, Kanter (1994) utilizedher criteria to define eight SCRs. Each continent contains atleast one SCR. The CEUS forms the southern half of the NorthAmerica SCR. Johnston et al. (1994) compiled geological andseismological information on SCR earthquakes worldwide.As already mentioned, the recent Mmax workshop produceda recommendation that future estimates of CEUS Mmax forseismic-hazard analyses should utilize the global SCR catalogof Johnston et al. (1994) (Wheeler 2009b, 141–143). Theglobal catalog shows that, after North America, the muchsmaller China SCR in Southeast Asia has the most historicalearthquakes of M 6.0 or larger (Figure 1). Consequently the1994 China SCR and its three parts as shown in Figure 1 areimportant tectonic analogs in estimating CEUS Mmax.PurposeSince the definition and delineation of SCRs in 1994, manypapers on the geophysics and tectonics of Southeast Asia haveappeared in English-language Western journals, for exampleYin (2010) and papers cited there. My purpose is to reassessthe 1994 China SCR and its earthquakes in light of the newinformation presented in these papers, in order to improve estimatesof CEUS Mmax. The 1994 China SCR of Kanter (1994)includes two thin bands of active continental crust that are centeredon large, active, strike-slip fault systems. The thin bandsdivide the 1994 China SCR into three parts that are labeledMO, CH, and IO in Figure 1. The rest of this paper utilizesthe new information and the criteria in Table 1 to update theMongolia, 2011 China, and Indochina SCRs of Figure 1. Theupdate will result in reclassifying most of the 1994 China SCRof Figure 1 as active crust. Nearly all of the Mongolia SCR willretain its classification as SCR crust, as will the southwesternpart of the 2011 China SCR.SOUTHEAST ASIAN SCRsContinental Extension and Alkaline Igneous RocksInformation published since the early 1990s (for example, Yin2010) shows that much of Southeast Asia is undergoing horizontalextension. Reassessing the SCR with the new informationrequires determining which parts of the 1994 ChinaSCR have undergone Neogene extension (Table 1). Geodeticand geophysical data and mapped extensional faults and continentalrifts provide well-known indicators of continentalextension.It may be less well known that dated alkaline igneous rocks,when combined with geologic field relations showing relativeages of faulting, eruption, and intrusion, can determine boththe occurrence of continental rifting and its age. Worldwide,972 Seismological Research Letters Volume 82, Number 6 November/December 2011

(A)70°N100°W 60°W70°N(B)100°E140°E50°N50°NMO140°ECH30°N*30°N10°NIO10°N100°E0 2000Kilometers100°W0 2000Kilometers60°WEarthquake epicenterMainland and coastline of continent*Boundary of stable continental regionLarge earthquakes of New Madrid seismic zone▲ ▲ Figure 1. Comparison of sizes and seismicities of North America (A) and China (B) stable continental regions (SCRs). Coastlinesand SCR boundaries after Kanter (1994) and Broadbent and Allan Cartography (1994). Epicenters are of earthquakes of magnitude 6.0or larger on any magnitude scale (Wheeler, in preparation). Both maps use Lambert azimuthal equal area projections and the samescale to aid visual comparison (Broadbent and Allan Cartography 1994). The North America SCR covers 24,007,000 km 2 and generated35 reported earthquakes of magnitude 6.0 or larger over the four-century historical record, whereas the smaller China SCR covers7,118,000 km 2 and has generated 27 such earthquakes over its 15-century historical record (Johnston et al. 1994). Chinese earthquakesappear more numerous in the figure because more of the North American epicenters overlap one another at the scale of the figure. Forease of discussion, I will refer to the single large SCR of part B as the 1994 China SCR, and to its three components as the Mongolia SCR(MO), the 2011 China SCR (CH), and the Indochina SCR (IO).igneous rocks of alkaline compositions are spatially associatedwith continental rifts (Bailey 1974; Neumann and Ramberg1978; Keller and Hoover 1988; McKenzie and Bickle 1988;Wilson 1989). The spatial association is generally accepted asimplying that extension produces alkaline melts (Wilson 1989).Furthermore, common crustal rocks have melting temperatureswell below those of basalts. This implies that melting inWheeler, Figure 1the mantle generates alkaline basalts. Petrological modelingcalculations of McKenzie and Bickle (1988) and of Barry et al.(2003) imply that the melting of peridotite, a common mantlerock, yields alkaline basaltic melts at depths exceeding 70 km. Inlaboratory experiments, the initial melting of peridotite undermantle pressures and temperatures produces small amounts ofalkaline basaltic melts (Jaques and Green 1980; Olafsson andEggler 1983; Takahashi and Kushiro 1983). Both the meltingexperiments and the petrological calculations show that additionalmelting shifts the composition of basaltic melts away fromalkaline toward less-alkaline basaltic compositions. Most riftrelatedalkaline igneous rocks are alkaline basalts. In addition,some rifts also contain alkaline volcanic and intrusive rocks ofgranitic compositions. Alkaline igneous rocks are known in severalparts of a rift that contains the New Madrid seismic zone,the most active seismic zone in the CEUS (Figure 1) (see summaryof these alkaline rocks in Wheeler 1997). Importantly forthe present study, basaltic rocks dominate in Southeast Asia(Whitford-Stark 1987; Yin 2010), for example in the Baikalrift system of the study area (Wilson 1989; Figure 2 this paper).Thus, volcanic rocks of alkaline basaltic composition imply asmall amount of extension within the upper mantle.Exposed or shallow normal or transtensional faults demonstrateextension of the upper crust and its seismogenic zone.The larger the extensional fault slips, the more likely it is thatbrittle extension penetrates into or spans the seismogenic zone.Where both rifts and alkaline basaltic volcanic rocks are presentand are of similar ages, they indicate that extension affectsthe upper crust, upper mantle, and therefore perhaps the middleand lower crust as well. Rifts without known alkaline volcanicrocks demonstrate brittle extension of at least the uppermostcrust, perhaps including the seismogenic zone. However,alkaline volcanic rocks without recognized, coeval extensionalfaults are more problematic. Absent known extensional faults,alkaline rocks might indicate that incipient extension in theupper mantle has not extended far enough upward to affect theseismogenic zone.Seismological Research Letters Volume 82, Number 6 November/December 2011 973

80°50°80°90°EXPLANATIONBorder of stable continentalregion (SCR) (Kanter, 1994)Borders of China and Mongolia100°SiberiaKESTKFS110°120°BRSRussiaMongoliaChina130°SFZ140°50°140°Large crustal block (see text)HESMongoliaSCRArea of Neogene volcanic rocksChina40°40°30°Epicenter in activecontinental crustEpicenter in stablecontinental crustNormal fault (hachure ondowndropped wall)Reverse fault (teeth onupthrown wall)Left lateral strike-slip faultXXFSLMSFWNORTHCHINABLOCKSBFSOUTHCHINABLOCKQLFZTLFECCMNorthKoreaYellowSea2011ChinaSCREastChinaSeaTaiwan30°130°EChinaBurma20°Right lateral strike-slip faultASRRLaosChinaVietnamECCMHainan Island20°IndianOceanSFLPSFBIndo-ChinaSCRSouthChinaSea10°10°0 1,000KilometersASRROutlines of islands otherthan Hainan and Taiwanare omitted for clarity.90°100°110°120°▲▲Figure 2. Selected Wheeler, Figure tectonic 2 elements of Southeast Asia. Stable continental regions (SCRs) after Broadbent and Allan Cartography(1994) and Kanter (1994) (See Figure 1 for new SCR names introduced here.) North China block and the adjoining part of the Koreanpeninsula after Zhang et al. (1984), Zhao et al. (1998), and Kwon et al. (2009). South China block after Ren et al. (2002), Liu et al. (2007),J. Adams (see Wheeler 2009b, 83), Yin (2010), and analyses in this paper. Locations of areas of Neogene volcanic rocks are from thecontinent-scale map in Figure 1 of Yin (2010). For legibility here I generalized the locations. Each solid square represents the center of agroup of approximately five of the locations shown by Yin (2010). Epicenters are of reported earthquakes of magnitude 6.0 or larger onany magnitude scale (Wheeler, in preparation). W, epicenter of Wenchuan earthquake (12 May 2008; M 7.9; 31.00°N, 103.32°E; http://earthquake.usgs.gov/). Faults shown have Neogene movement. Their locations and age assignments are after Peizhen et al. (1991),Ren et al. (2002), Jia et al. (2006), Zhu et al. (2010), and Yin (2010). ECCM, eastern China continental margin. Fault names: ASRR, AilaoShan–Red River shear zone; BRS, Baikal rift system; HES, Hangay extensional system; KES, Khubsugul extensional system; LMSF,Longmen Shan thrust fault; LPSFB, Lanping-Simao fold belt; QLFZ, Qinling fault zone; SBF, Sichuan Basin fault; SF, Sangaing fault; SFZ,Stanovoy fault zone; TKFS, Tunka fault system; TLF, Tanlu fault; XXFS, Xiangshuihe-Xiaojiang fault system. Lambert azimuthal equalarea projection centered at 25°N, 100°E.974 Seismological Research Letters Volume 82, Number 6 November/December 2011

Mongolia SCRThe presence or absence of Neogene alkaline igneous rocks andrifting provides a guide to whether the Mongolia SCR shouldbe classified as an SCR or an ACR. The Mongolia SCR includesthe eastern half of Mongolia, most of northeastern China, andadjacent areas of Russia (Figure 2). The Mongolia SCR is a comparativelystable region. ACR crust surrounds this SCR; latersections will summarize the active nature of the North Chinablock on the south, a region of active faults and volcanism onthe west and north, and the eastern China continental marginon the east. The Tanlu fault separates the Mongolia and 2011China SCRs. The Mongolia SCR is moving eastward withrespect to a fixed Eurasia (Liu et al. 2007; Wang et al. 2011).The relative motion takes place on a belt of extensional andleft-lateral transtensional faulting north of the SCR, between itand the Siberian part of the Eurasian SCR. The belt of faultingcomprises the Hangay and Khubsugul extensional systems, theTunka fault system, the Baikal rift system, and the Stanovoyfault zone (Figure 2). Judging from the motions that Liu et al.used to compute Quaternary rates of fault slips, the MongoliaSCR appears to be moving eastward with respect to Siberia atmuch less than 1 mm/yr, and possibly as little as 0.1 mm/yr.Tomography shows that S-wave velocities at 50 km depthare similar across the Mongolia SCR, and the same is true for100 km depth (Feng and An 2010). From this it appears thatcrustal and lithospheric thicknesses vary little across the SCR.Seismicity is sparse and geodetically measured velocity andstrain are small (Broadbent and Allan Cartography 1994; Liuet al. 2007; Feng and An 2010) (see also the earthquake catalogsat http://earthquake.usgs.gov/earthquakes; last accessedJuly 21, 2011).The compilation map of Ren et al. (2002) shows normalfaults that bound Paleogene and older basins throughout theSCR, but no Neogene faults or basins. Active systems of northerlystriking normal faults and easterly striking left-lateralstrike-slip faults in western Mongolia do not appear to extendinto the SCR, except in its northwestern corner at the northeast-strikingnormal faults of the Hangay extensional system(Figure 2) (McCalpin and Khromovskikh 1995; Walker et al.2007; Yin 2010). Walker (2009) did not find active faults ineastern Mongolia and adjacent China, which include most ofthe Mongolia SCR.Figure 2 shows that Neogene volcanic rocks are less numerousper unit area within the Mongolia SCR than in more tectonicallyactive regions, such as the Indochina SCR and Chinaeast of the South China block and the Tanlu fault (Ren et al.2002; Liu et al. 2007; Yin 2010). The Neogene volcanic rocksin the Mongolia SCR are largely alkaline although older volcanicrocks range more widely in compositions (Whitford-Stark1987; Basu et al. 1991). Barry et al. (2003) cited computationsby McKenzie and Bickle (1988), which imply that generationof significant amounts of alkaline melts would require muchmore Neogene horizontal extension than appears to haveoccurred in most of the Mongolia SCR. Consequently, sincethe definition of the Mongolia SCR in 1994, new informationdoes not demonstrate extension younger than Paleogene exceptin the northwestern corner of the Mongolia SCR. The rest ofthe Mongolia SCR meets the criteria of Table 1 and retains itsclassification as an SCR.2011 China SCRNorth China BlockThe North China craton is the Chinese part of the Sino-Korean craton, with the remainder being the northern part ofthe Korean peninsula (for example, Zhang et al. 1984, Zhaoet al. 2009, Yang et al. 2010). I follow Kwon et al. (2009) incalling both cratons “blocks” because, as explained later, theyunderwent Mesozoic and Cenozoic metamorphism, extension,intrusion, and volcanism so that they are no longer cratoniccrust (Figure 2; note that the craton boundary is northwestof the boundary between North and South Korea). The Tanlufault splits the North China block into two parts. The largerpart of the block lies entirely west of the 1994 and 2011 ChinaSCRs, whereas the smaller part is within both versions of theSCR. The smaller part is of more interest here, but most of theinformation on the North China block comes from the activecrust of the larger part. Therefore, I will discuss the NorthChina block as a whole. The block is a triangular region innorthern China (Figure 2) that is made of early Precambriancrust (Zhang et al. 1984; Zhao et al. 2001; Kwon et al. 2009).The North China block is moving eastward with respect to afixed Siberia and the Mongolia SCR (Yin 2010). Geologic dataincluding slip rates of individual faults indicate eastward movementwith respect to Siberia of 1–2 mm/yr in the eastern halfof the North China block and 2–4 mm/yr in the western half(Liu et al. 2007). Geodetic data indicate rates consistent withthose of Liu et al. (2007) (Wang et al. 2011).The eastern part of the block is seismically active, whereasthe western part is less so (Liu et al. 2007). For example, the2008 version of the “Centennial” earthquake catalog ofEngdahl and Villasenor (2002) lists 17 earthquakes of magnitude6.0 or larger in the eastern part of the block but onlyone in the western part. Results of P- and S-wave tomographyshow a low-velocity zone that extends to 300–400 km depthbeneath the eastern part of the block (Zhao et al. 2009). S-wavetomography, deep seismic-reflection profiles, and receiver-functionimaging show that the crust thins eastward from approximately45 km in the western part of the block to about 30 kmin the eastern part (Li et al. 2006; Zheng et al. 2006; Chen etal. 2009; Feng and An 2010). Zheng et al. (2006) concludedthat most of the thinning took place in the lower crust and ina transitional zone between the crust and mantle. The thinningresulted from extension that began with widespread EarlyCretaceous eruption and intrusion of alkaline basaltic and graniticrocks (Ren et al. 2002; Wu et al. 2005; Zhu et al. 2010).The entire North China block underwent extension by normaland transtensional faulting of early Neogene age, whereas theeastern part of the block and the northern part of the Koreanpeninsula also underwent late Neogene alkaline igneous activity(Liu et al. 2001; Ren et al. 2002; Zheng et al. 2006; Yu etal. 2008; Yang et al. 2010; Yin 2010). Zhao et al. (2009) inter-Seismological Research Letters Volume 82, Number 6 November/December 2011 975

preted the low-velocity zone in the mantle beneath the easternpart of the block in terms of warm mantle material that mighthave caused the rifting and alkaline igneous activity. Extensionis demonstrated in the upper crust by the mapped extensionalfaults, in the lower crust by the geophysical evidence, anddeeper than 70 km (McKenzie and Bickle 1988) in the mantleby the compositions of the igneous rocks. The Neogene extensiondemonstrates that the North China block and the rest ofthe Sino-Korean block are no longer a craton (Yang et al. 2008,2010). The extension also requires reclassifying this part of the1994 China SCR as active crust (Table 1).Eastern China Continental MarginThe eastern China continental margin comprises all of theYellow Sea, the East China and South China seas within200–600 km of the mainland Chinese coast, and the landeast of the North China and South China blocks (Figure 2;Zhou et al. 1995; Yin 2010). Bathymetric, geologic, and geophysicaldata show that the continental crust of the marginextends seaward approximately to the edge of the continentalshelf (GEBCO World Map Editorial Board 2006; Wang et al.2006; Yin 2010). Fault-slip data show that the margin is movingsouth-southwestward past the North and South Chinablocks (Figure 2; Yin 2010); geodetic data give the rate as 0.7mm/yr in the north and 1.8 mm/yr in the south (Wang et al.2011). The continental margin is more seismically active thanthe South and North China blocks, especially west of Taiwan(Liu et al. 2007; Wang et al. 2011).The compilations of Yin (2010), Ren et al. (2002), andSengor and Natal’in (2001) show the continental marginas having undergone Paleogene and older extension. S-wavetomography indicates thick sediments and thin crust and lithosphereunder the margin (Feng and An 2010). China east ofthe North and South China blocks contains numerous basinsbounded by normal faults; most of the basins are of Paleogeneage (Ren et al. 2002). In the same part of China and in offshorebasins on the continental shelf, abundant basaltic volcanicrocks of late Neogene ages are exposed from Hainan Islandnortheastward to the southern Tanlu fault (Figure 2; Ren et al.2002; Yin 2010). Many of the basaltic rocks are alkaline (Ho etal. 2003). In the South China Sea, early Neogene thermal subsidenceof several kilometers was followed by middle Neogenenormal faulting (Zhou et al. 1995). The alkaline basalts andnormal faulting indicate Neogene extension of the continentalmargin from Hainan Island to the southern Tanlu fault.Yin (2010) restricted the definition of the margin northeastof North Korea to offshore continental crust that has beenthinned by back-arc extension. However, three onshore areasbetween the coast and the Tanlu fault must also be consideredbecause they are in the 1994 China SCR, or adjacent to it andalong its tectonic grain (Figure 2):1. The northern Korean peninsula and the northern YellowSea are part of the 1994 China SCR. The northern peninsulaunderwent Neogene extension as described earlier.Thus, ACR of the Sino-Korean block and the easternChina continental margin surround the northern YellowSea on three sides (Figure 2). The Neogene extensionimplies that the northern Korean peninsula, and probablythe adjacent part of the Yellow Sea, should be reclassifiedas active crust (Table 1).2. Northeast of the North China block and east of the Tanlufault, northeastern China and adjacent Russia are notin the 1994 China SCR. This area contains numerousNeogene basalts, many of them alkaline (Liu et al. 2001;Ren et al. 2002). The alkaline basalts indicate Neogeneextension, and Sengor and Natal’in (2001) show latestCretaceous–early Neogene grabens in the same area as thebasalts. The evidence for Neogene extension implies thatthe area was correctly classified as active crust in 1994.3. The southern part of the Korean peninsula is part of the1994 China SCR. Yin (2010) reported that the peninsulaappears to have had little Paleogene or Neogene extension.The southern part of the peninsula is outside the Sino-Korean block. Summaries of Korean geology and tectonicsdo not show any evidence of foreland deformation ororogenic activity of Cretaceous or younger age (Table 1)(Exxon Production Research Company 1985; Chang1997; Kwon et al. 2009). However, the southeastern partof the peninsula exposes a few extensional or transtensionalfaults that bound Neogene basins, as well as sparseNeogene alkaline volcanic rocks (Exxon ProductionResearch Company 1985; Ren et al. 2002; Yang et al.2010; Yin 2010). Two small islands approximately 170 kmeast of the mainland and 120 km south of it also exposesparse Neogene alkaline volcanic rocks (Reedman andKim 1997). Kanter (1994) classified the southern island asbeing within the same northeast-trending belt of Mesozoiccontinental crust as the southern part of the Korean peninsulaand most of the Yellow Sea (Broadbent and AllanCartography 1994). The Neogene alkaline rocks exposedon the southern island, together with the Neogene alkalinerocks exposed along trend to the southwest in theeastern China continental margin (Figure 2), suggest thatsimilar Neogene alkaline rocks may be hidden beneath theYellow Sea. Thus, the evidence for Neogene extension issparse and whether the southern part of the Korean peninsulaand the Yellow Sea are SCR crust or ACR crustis arguable. Overall, I judge that neither area meets therequirements for classification as an SCR (Table 1).Regional geologic relations are consistent with the classificationof the southern part of the Korean peninsula as probable ACRcrust. As pointed out earlier in this section, all other areas eastof the Tanlu fault, including the northern part of the Koreanpeninsula, have undergone Neogene extension. As a result,500–600 km of the active crust of the North China block separatesthe southern Korean peninsula from the nearest SCR,which is the Mongolia SCR. This isolation allows the possibilitythat the southern Korean peninsula might be a small fragmentof comparatively intact ACR crust that is being carriedeast-southeastward by Neogene extension of the surroundingmore active continental crust. The possibility is consistent with976 Seismological Research Letters Volume 82, Number 6 November/December 2011

the small horizontal relative motions and dilational strainsbetween the southern part of the peninsula and adjacent ACRsto the west and southwest (Liu et al. 2007; Wang et al. 2011).The 1994 and 2011 versions of the China SCR extend asfar offshore as 700 km southeast of Hainan Island (Figures 1,2; Broadbent and Allan Cartography 1994). Bathymetric datashow that only the near-shore half of this part of the SCR is onthe continental shelf, at water depths of 0–200 m (GEBCOWorld Map Editorial Board 2006). The offshore half is beyondthe shelf edge, at depths of 200–5,000 m. As shown on themap of Broadbent and Allan Cartography (1994), the offshorehalf was interpreted as SCR crust that had been highlyextended during the Paleogene Period. However, interpretationsof more recent marine seismic-reflection profiles betweenHainan Island and southern Taiwan show thinned continentalcrust in the landward part of the surveyed area and unusuallythick oceanic crust in the seaward part (Wang et al. 2006). Theprofile interpretations imply that the crust that was consideredto be highly extended SCR crust should now be reclassified asthick oceanic crust.In summary, information published since 1994 shows thatalmost the entire eastern China continental margin underwentNeogene extension. The only potential exception is the southernpart of the Korean peninsula.South China BlockThe South China block is part of the 1994 China SCR. Theblock consists of two smaller blocks that joined during latePrecambrian time to form crust that is now tectonically stable(Qiu et al. 2000; Li et al. 2002; Zheng et al. 2006; Liu et al.2007). S-wave tomography shows that the South China blockhas thinner sedimentary cover and higher upper mantle velocitiesthan the active North China block (Feng and An 2010).Feng and An (2010) inferred that the different upper-mantlevelocities indicate thicker lithosphere in the South China blockthan in the North China block. Seismicity and geodetic strainrates are approximately as low as those of the Mongolia SCR.These comparisons allowed Liu et al. (2007) to suggest that theSouth China block is being extruded southeastward, away fromthe India-Eurasia collision zone, as a single intact entity. Shen etal. (2005) and Liu et al. (2007) used the geodetic data to computethe rate of southeastward motion as 7–8 mm/yr and 4–6mm/yr, respectively, with respect to a stable Siberia, and Wanget al. (2011) used the data to calculate a left-lateral slip rate of0.9 mm/yr between the South China and North China blocks.Active continental crust bounds the South China blockon all sides. On the north, the Qinling fold belt of Paleogeneor Neogene age separates the South China and North Chinablocks (Figure 2; Terman 1974; Ren et al. 2002). The foldbelt itself is active continental crust because it is the productof orogenic activity younger than Early Cretaceous (Table 1).The southernmost component of the fold belt is the left-lateralQinling fault system of Neogene age (Yin 2010). Thus, theQinling fault system marks the northern boundary of the SouthChina block. On the east, the South China block adjoins theactive continental crust of the eastern China continental margin.On the south, the Neogene Ailao Shan–Red River shearzone separates the South China block from the IndochinaSCR. The Ailao Shan–Red River shear zone is tens of kilometerswide, many hundreds of kilometers long, and constitutesactive continental crust, as described in a later section on theIndochina SCR. The northern edge of the shear zone marksthe south boundary of the South China block. On the west,the left-lateral Xiangshuihe-Xiaojiang fault system forms partof the block’s boundary. The fault system formed during themiddle Neogene Period and remains active today. The rest ofthe western boundary of the South China block is along theeast-dipping Sichuan Basin fault and the west-dipping thrustfaults that crop out at the eastern front of the Longmen Shanrange (Figure 2; Yin 2010; Burchfiel et al. 2008).The Longmen Shan thrust fault comprises several strandsthat crop out along different sections of the northeast-trendingLongmen Shan range front (Hubbard et al. 2010). Mostof the exposed strands of the fault crop out within the rangeand along its southeastern front. The 400-km-wide SichuanBasin is southeast of the range front and is part of the 1994China SCR (Kanter 1994). Jia et al. (2006) and Burchfiel et al.(2008) explained that the Sichuan Basin is the foreland basinproduced by orogenic movements farther west, including thoseon the Longmen Shan thrust fault. The orogenic movementscontinue, as shown by the occurrence of the M 7.9 Wenchuanearthquake in 2008 (Figure 2) and by a similar earthquake thatpaleoseismic and historical evidence dates at approximately 2ka (Liu et al. 2010).From the range front southeastward into the SichuanBasin is a belt of northeast-trending anticlines and northeaststriking,northwest-dipping reverse faults (Burchfiel et al.1995). Interpretations of well and seismic-reflection data andof geologic mapping imply that the anticlines and reverse faultsare underlain by a nearly horizontal strand of the LongmenShan fault, which propagated southeastward into the basin(Jia et al. 2006; Hubbard et al. 2010). Southeastward movementon the buried fault strand buckled the overlying stratato form the anticlines and reverse faults (Burchfiel et al. 1995,2008). Elsewhere, similar combinations of anticlines, reversefaults, and one or more underlying thrust faults deformedsedimentary strata of foreland basins and moved the strataoutward away from growing mountain ranges, as in the southern,central, and northern Appalachian mountains, the southernCanadian Rocky Mountains, the Idaho-Montana thrustbelt, and the Ouachita mountains of Oklahoma and Arkansas(for example, Rich 1934; Boyer and Elliott 1982; Perry et al.1984; Rodgers 1970; Arbenz 1988; Hatcher et al. 1990). In theSichuan Basin, deformation of the basin deposits continuedas late as Paleogene and Neogene time (Burchfiel et al. 1995;Kirby et al. 2002; Jia et al. 2006). Thus, the northwestern partof the Sichuan Basin is a deformed orogenic foreland that hasbeen active more recently than the Early Cretaceous. This conclusionresults in the reclassification of the northwestern partof the basin as active crust (Table 1).Geologic maps and cross-sections show that several of thereverse faults dip southeastward, particularly at or near theSeismological Research Letters Volume 82, Number 6 November/December 2011 977

southeast edge of the foreland deformation (Burchfiel et al.1995; Kirby et al. 2002; Jia et al. 2006; Burchfiel et al. 2008;Hubbard et al. 2010). Such backward-dipping reverse faultsat or near the fronts of fold-and-thrust belts are recognizedelsewhere (for example, Rodgers 1950; Malik et al. 2010).Accordingly, a line drawn along the southeasternmost outcropsof southeast-dipping thrust faults may approximate theeastern limit of deformation in the Sichuan Basin. Figure 1of Yin (2010) shows such a line as a southeast-dipping reversefault labeled the Sichuan Basin fault. I use the Sichuan Basinfault to approximate the western boundary of the South Chinablock (Figure 2).Figure 2 shows an isolated right-lateral strike-slip faultwithin the northeastern South China block. Yin (2010) showsthe fault as having Neogene movement but provides no otherinformation on the fault. The North America SCR containsfour similarly isolated active faults: the normal Cheraw faultin eastern Colorado, the strike-slip Meers fault in southernOklahoma, the reverse Reelfoot fault in the New Madrid seismiczone of southeastern Missouri, and the reverse Ungavafault in northern Quebec (Russ 1979; Madole 1988; Croneand Luza 1990; Adams et al. 1991; Crone, Machette, Bradleyet al. 1997). The active faults inside the North America SCRimply that the fault inside the South China block need not disqualifythat part of the block from being SCR crust. NeitherYin (2010) nor Ren et al. (2002) show any Neogene igneousrocks or additional faults in the South China block. To summarize,the South China block retains its classification as SCRcrust except for the northwestern Sichuan Basin.Indochina SCRSince middle Paleogene time the Indochina SCR has beenextruded southeastward in response to subduction of the Indianplate beneath the Eurasian plate (Figure 2; Leloup et al. 1995;Yin 2010). The SCR is moving rapidly southeastward relativeto a fixed Siberia (Liu et al. 2007; Simons et al. 2007). Wang etal. (2011) calculated rates of left-lateral strike slip of 18 and 10mm/yr along two sections of the Xiangshuihe-Xiaojiang faultsystem (Figure 2). The northeastern boundary of the extrudingmass is the northwest-striking Ailao Shan–Red River shearzone. The western boundary of the extruding mass is a complexfault network that includes the Sangaing fault and other northandnorthwest-striking, mostly right-lateral faults (Leloup etal. 1995; Ren et al. 2002; Yin 2010) and diversely oriented,offshore Paleogene and Neogene grabens (Sengor and Natal’in2001). Since early Neogene time most of the interior of theIndochina SCR was fragmented by right-lateral transtensionalfaulting and underwent eruption of alkaline basalts (Ranginet al. 1995; Ren et al. 2002; Ho et al. 2003). S-wave tomographyand teleseismic receiver function analyses indicate thinnedcrust and lithosphere under the Indochina SCR (Feng andAn 2010; Bai et al. 2010). Seismicity is negligible in the SCR(Broadbent and Allan Cartography 1994; Tarr et al. 2010).The Ailao Shan–Red River shear zone is 20–200 km wideand 800–1,200 km long onshore, with another 500–1,400km of length suggested under the South China Sea (Terman1974; Exxon Production Research Company 1985; Leloup etal. 1995). The shear zone underwent roughly 500–700 km ofleft-lateral strike slip during 40–15 Ma (Leloup et al. 1995).The zone reversed its slip sense at approximately 5 Ma and hasaccumulated 20–50 km of right-lateral slip since that time. Theshear zone contains a core of high-temperature metamorphicand igneous rocks that were uplifted from mid-crustal depths.Flanking the core on its northeast and southwest sides are beltsof lower-temperature metamorphic rocks and, farther out fromthe core, sedimentary basins whose strata were folded and cutby reverse faults that dip inward toward the core. Presumablymost of the characteristics of the shear zone and related structuresformed during the left-lateral majority of the zone’s evolution.For this reason Figure 2 shows the dominant left-lateralslip sense.One of the groups of structures related to the shear zone isthe Lanping-Simao fold belt (Leloup et al. 1995), which overlapsthe north tip of the Indochina SCR (Figure 2). The foldbelt comprises elongated folds and related reverse faults, whichtogether deformed Late Cretaceous–early Paleogene shales ofone of the sedimentary basins that flank the core of the AilaoShan–Red River shear zone. The folds and reverse faults accommodatedapproximately 40 km of horizontal transport to thewest-northwest. This structural style is the same as those of thenorthwestern Sichuan Basin and the North American analogslisted in the earlier description of the Sichuan Basin. The directionof horizontal transport is kinematically consistent withthe orientation and movement sense of late Paleogene and earlyNeogene left-lateral movement on the Ailao Shan–Red Rivershear zone. Leloup et al. (1995) described field relations thatled them to conclude that the fold belt formed at the same timeas the shear zone and formed in the same stress field. Thesecharacteristics of the Lanping-Simao fold belt and its similaritiesto the Sichuan Basin and the North American analogs ofthe Sichuan Basin motivate me to classify the Lanping-Simaofold belt as a deformed foreland that is younger than the EarlyCretaceous Epoch. Thus, new information that demonstratesNeogene rifting and post-Cretaceous foreland deformationrequires reclassification of the Indochina SCR as active crust(Table 1).DISCUSSIONIn this section I draw on the preceding descriptions to summarizewhether the Mongolia SCR, North China block, northernand southern parts of the Korean peninsula, Yellow Sea,eastern China continental margin, South China block, andIndochina SCR satisfy the criteria of Kanter (1994) as summarizedin Table 1. Figure 3 summarizes the following discussion.The paucity of alkaline basaltic rocks, the absence of reportedextensional faulting, and the laboratory experiments in whichheating of mantle rock under mantle pressures produced alkalicmelts first, when considered together indicate that theMongolia SCR may have undergone incipient Neogene extensionat upper-mantle melting depths exceeding 70 km. Theextension does not appear to have propagated upward to rift978 Seismological Research Letters Volume 82, Number 6 November/December 2011

80°100°120°140°EXPLANATIONBorder of stable continentalregion (SCR) (Kanter, 1994)80°Mongolia SCR140°Border of stable continentalregion (SCR) (this paper)40°40°Epicenter in activecontinental crustSouth China SCREpicenter in stablecontinental crust20°20°0 1,000Kilometers100°120°▲▲Figure 3. Summary of the changes from the 1994 stable continental regions (SCRs) to the SCRs of this paper (see Discussion text).the upper-crustal seismogenic zone. Therefore, nearly all of theMongolia SCR retains its SCR classification, at least in theupper crust and for the purpose of reassessing the 1994 ChinaSCR to improve estimates of CEUS Mmax.Similarly, the South China block should retain its classificationas SCR crust (Figure 3). Its lithosphere is thick. Lowseismicity and strain rates led Liu et al. (2007) to the interpretationthat the block is being extruded southeastward as anintact block. I did not find any published evidence of Neogeneextensional faulting, Neogene alkaline igneous rocks, or forelanddeformation or orogeny younger than Early Cretaceouswithin the South China block. I suggest that the block becalled the South China SCR.In contrast, the North China block, eastern China continentalmargin, and Indochina SCR fail to meet the requirementsfor classification Wheeler, Figure as 3 SCR crust (Figure 3). All threeareas contain widespread Neogene extensional faulting andabundant Neogene alkaline basaltic rocks. Thus, all threeareas underwent Neogene rifting. The Korean peninsula andthe Yellow Sea contain lesser amounts of Neogene extensionalfaulting and alkaline rocks. In addition, the northern tip of theIndochina SCR overlaps the Lanping-Simao fold belt, in whichfolding and reverse faulting deformed foreland-basin strataafter Early Cretaceous time. A better classification for all fiveareas is active continental crust.In the introduction I explained that moderate to large historicalearthquakes in the 1994 China SCR provide valuableconstraints on the value of CEUS Mmax. The realization thatmost of the 1994 China SCR is active crust reduces the numberof Southeast Asian earthquakes that are available to constrainCEUS Mmax. Figure 2 shows epicenters of 43 earthquakesof magnitude 6.0 or larger on any magnitude scale (Wheeler,in preparation). Some epicenters coincide so that the figureappears to show only 39 of them. Thirty-seven of the earthquakes,or 86 percent of the 43, are now recognized as havingoccurred in ACRs, leaving only six SCR earthquakes. The 37active-crust earthquakes had estimated moment magnitudesranging up to approximately 8. Two of the six stable-crustearthquakes occurred in the eastern part of the Mongolia SCRand four were in or on the border of the South China SCR(Figure 3). Table 2 lists the six SCR earthquakes, and doublecircles identify their epicenters in Figure 2.Moderate earthquakes can provide lower bounds onestimates of Mmax, but larger earthquakes can providetighter constraints on Mmax. Specifically, the tightest constraintson CEUS Mmax are sizes of earthquakes with aboutM 6.7 and larger (for example, see Figure 3 of Petersen et al.2008). As explained earlier, many of Earth’s larger historicalearthquakes occurred in China. As explained in this paper,recently published information indicates that nearly all ofthose larger earthquakes occurred in ACR crust instead ofSCR crust. These reclassified earthquakes are lost to estimationof CEUS Mmax. Fortunately, the development of paleoseismologyin recent decades could counterbalance this loss.Paleoseismic data allow estimation of the sizes, locations, andages of moderate to large prehistoric earthquakes (Tuttle 2001;McCalpin 2009). Wheeler (2008) estimated that earthquakesof approximately M 6.5 and larger are the most likely to yieldpaleoseismic estimates of their magnitudes, locations, and ages.Therefore, paleoseismic magnitude estimates of large prehistoricSCR earthquakes may partly counterbalance the loss oflarge Chinese earthquakes. Determination of the impact ofSeismological Research Letters Volume 82, Number 6 November/December 2011 979

TABLE 2Moderate to Large Earthquakes in the Mongolia and South China SCRs *SCR †Date(YearMoDay)Origin Time(HrMnSecs) Source 1 ‡ Latitude Longitude Source 2 § M Source 3 ||SCH 16310814 unknown J94 29.300 111.700 J94 5.8 J94, J96SCH 19170124 004812.00 ISS 31.000 114.000 ISS 6.5 C08, J96MO 19220814 114104.70 C08 52.069 130.539 C08 6.6 C08, J96SCH 19360401 unknown J94 22.500 109.400 J94 6.8 J94, J96MO 19410505 151827.00 J94 46.500 126.900 ISS 6.0 J94, J96SCH 20080525 082149.99 PDE 32.560 105.420 PDE 6.1 GCMT* Stable Continental Regions (see text).† SCR in which earthquake occurred. SCH, South China SCR; MO, Mongolia SCR.‡ Source catalog for date and origin time. C08, 2008 version of “Centennial” catalog of Engdahl and Villasenor (2002); ISS,International Seismological Service (available at http://www.isc.ac.uk/); J94, Johnston et al. (1994); PDE, PreliminaryDetermination of Epicenters, U. S. Geological Survey (available at http://earthquake.usgs.gov/earthquakes/eqarchives/epic/).§ Source catalog for latitude and longitude of epicenter.|| Source catalog for moment magnitude M. GCMT, Global Centroid Moment Tensor (available at http://www.globalcmt.org/CMTsearch.html). First entry is source catalog of original size estimate (M S or maximum intensity). Second entry indicatesthat size estimate was converted to M with the look-up tables of Johnston (1996a,b).this counterbalance on CEUS Mmax is beyond the scope ofthis paper and will be examined in a future report.CONCLUSIONS1. New information that was not available in the early 1990sshows that most of the 1994 China SCR is active continentalcrust. Only the South China and Mongolia partsof the 1994 SCR meet the criteria for retaining their classificationsas SCRs.2. This finding reduces the number of Southeast Asian moderateto large earthquakes that are available to constrainCEUS Mmax from 43 to six.ACKNOWLEDGMENTSA. C. Johnston and his colleagues produced the global SCRcatalog. The catalog and its extensive documentation of the geologicand tectonic contexts of each earthquake form the mostvaluable source of information with which to constrain CEUSMmax. Many of the ideas in this paper stemmed from discussionswith J. Adams, J. Ake, A. C. Johnston, C. S. Mueller, R. L.Wesson, and the other participants in the 2008 Mmax workshop.Suggestions by R. Gold, C. S. Mueller, M. D. Petersen,and two anonymous reviewers improved the manuscript.REFERENCESAdams, J., R. J. Wetmiller, H. S. Hasegawa, and J. Drysdale (1991). Thefirst surface faulting from a historical intraplate earthquake inNorth America. Nature 352, 617–619.Arbenz, J. K., comp. (1988). Ouachita system: Cross sections. In TheAppalachian-Ouachita Orogen in the United States, ed. R. D.Hatcher Jr., W. A. Thomas, and G. W. Viele, plate 11. Volume F-2of The Geology of North America. Boulder, CO: Geological Societyof America.Bai, L., X. Tian, and J. Ritsema (2010). Crustal structure beneath theIndochina peninsula from teleseismic receiver functions. GeophysicalResearch Letters 37 (5), L24308, doi:10.1029/2010GL044874.Bailey, D. K. (1974). Continental rifting and alkaline magmatism. In TheAlkaline Rocks, ed. H. Sorensen, 148–159. New York: John Wileyand Sons.Bakun, W. H., and M. G. Hopper (2004). Magnitudes and locations ofthe 1811–1812 New Madrid, Missouri, and the 1886 Charleston,South Carolina, earthquakes. Bulletin of the Seismological Societyof America 94, 64–75.Barry, T. L., A. D. Saunders, P. D. Kempton, B. F. Windley, M. S. Pringle,D. Dorjnamjaa, and S. Saandar (2003). Petrogenesis of Cenozoicbasalts from Mongolia: Evidence for the role of asthenospheric versusmetasomatized lithospheric mantle sources. Journal of Petrology44, 55–91.Basu, A. R., W. Junwen, H. Wankang, X. Guanghong, and M. Tatsumoto(1991). Major element, REE, and Pb, Nd and Sr isotopic geochemistryof Cenozoic volcanic rocks of eastern China: Implications fortheir origin from suboceanic-type mantle reservoirs. Earth andPlanetary Science Letters 105, 149–169.Boyer, S. E., and D. Elliott (1982). Thrust systems. American Associationof Petroleum Geologists Bulletin 66, 1,196–1,230.Broadbent, T., and Allan Cartography (1994). Maps of the stable continentalregions. In The Earthquakes of Continental Regions,vol. 5, ed. A. C. Johnston, K. J. Coppersmith, L. R. Kanter, andC. A. Cornell, appendix G. Report for Electric Power ResearchInstitute (EPRI), Palo Alto, CA. EPRI TR-102261, 15 plates, scales1:12,500,000 and 1:20,000,000.Burchfiel, B. C., L. H. Royden, R. D. van der Hilst, B. H. Hager, Z.Chen, R. W. King, C. Li, J. Lu, H. Yao, and E. Kirby (2008). Ageological and geophysical context for the Wenchuan earthquakeof 12 May 2008, Sichuan, People’s Republic of China. GSA Today18; doi:10.1130/GSATG18A.1.Burchfiel, B. C., C. Zhiliang, L. Yuping, and L. H. Royden (1995).Tectonics of the Longmen Shan and adjacent regions, CentralChina. International Geology Review 37, 661–735.980 Seismological Research Letters Volume 82, Number 6 November/December 2011

Chang, K. H. (1997). Korean peninsula. In Encyclopedia of European andAsian Regional Geology, ed. E. M. Moores and R. W. Fairbridge,465–473. London: Chapman and Hall.Chen, L., C. Cheng, and Z. Wei (2009). Seismic evidence for significantlateral variations in lithospheric thickness beneath the central andwestern North China craton. Earth and Planetary Science Letters286, 171–183.Chinnery, M. A. (1979). Investigations of the Seismological Input to theSafety Design of Nuclear Power Reactors in New England. NuclearRegulatory Commission Report NUREG/CR–0563, 72 pp.Coppersmith, K. J. (1994). Introduction to The Earthquakes ofContinental Regions, ed. A. C. Johnston, K. J. Coppersmith, L.R. Kanter, and C. A. Cornell, vol. 1, 1:1–1:10. Report for ElectricPower Research Institute (EPRI), Palo Alto, CA.Coppersmith, K. J., A. C. Johnston, A. G. Metzger, and W. J. Arabasz(1987). Methods for Assessing Maximum Earthquakes in the Centraland Eastern United States. EPRI RP-2556–12, report for ElectricPower Research Institute (EPRI), Palo Alto, CA, 312 pp.Cox, R. T., J. Gordon, S. Forman, T. Brezina, M. Negrau, A. Hill, C.Gardner, and S. Machin (2010). Paleoseismic sand blows in northLouisiana and south Arkansas. Seismological Research Letters 81,1,032–1,047.Crone, A. J., and K. V. Luza (1990). Style and timing of Holocene surfacefaulting on the Meers fault, southwestern Oklahoma. Bulletinof the Geological Society of America 102, 1–17.Crone, A. J., M. N. Machette, and J. R. Bowman (1997). Episodicnature of earthquake activity in stable continental regions revealedby palaeoseismicity studies of Australian and North AmericanQuaternary faults. Australian Journal of Earth Science 44, 203–214.Crone, A. J., M. N. Machette, L. A. Bradley, and S. A. Mahan (1997).Late Quaternary Surface Faulting on the Cheraw Fault, SoutheasternColorado. USGS Miscellaneous Investigations Map I–2521 (1sheet), various scales, 7 pp.Ebel, J. E. (1996). The seventeenth-century seismicity of northeasternNorth America. Seismological Research Letters 67, 51–68.Ebel, J. E. (2011). A new analysis of the magnitude of the February 1663earthquake at Charlevoix, Quebec. Bulletin of the SeismologicalSociety of America 101, 1,024–1,038.Engdahl, E. R., and A. Villasenor (2002). Global seismicity, 1900–1999. In International Handbook of Earthquake and EngineeringSeismology, Part A, ed. W. H. K. Lee, H. Kanamori, P. C. Jennings,and C. Kisslinger, 665–690. Amsterdam: Academic Press.Exxon Production Research Company (1985). Tectonic Map of the World.Twenty-one sheets, scale 1:10,000,000 at the equator. Tulsa, OK:American Association of Petroleum Geologists Foundation.Feng, M., and M. An (2010). Lithospheric structure of the Chinesemainland determined from joint inversion of regional and teleseismicRayleigh-wave group velocities. Journal of Geophysical Research115, B06317; doi:10.1029/2008JB005787.GEBCO World Map Editorial Board (2006). World Ocean Bathymetry,General Bathymetric Chart of the Oceans (GEBCO). One sheet,scale 1:35,000,000 at the equator; available at http://www.gebco.net/. Last accessed July 21, 2011.Gradstein, F., J. Ogg, and A. Smith, eds. (2004). A Geologic Time Scale2004. Cambridge: Cambridge University Press, 589 pp., 1 foldedplate.Hatcher, R. D. Jr., P. H. Osberg, A. A. Drake Jr., P. Robinson, and W.A. Thomas (1990). Tectonic map of the U.S. Appalachians. InThe Appalachian-Ouachita Orogen in the United States, ed. R. D.Hatcher Jr., W. A. Thomas, and G. W. Viele, plate 1. Volume F-2of The Geology of North America. Boulder, CO: Geological Societyof America.Ho, K.-S., J.-C. Chen, C.-H. Lo, and H.-L. Zhao (2003). 40Ar–39Ar datingand geochemical characteristics of late Cenozoic basaltic rocksfrom the Zhejiang-Fujian region, SE China: Eruption ages, magmaevolution and petrogenesis. Chemical Geology 197, 287–318.Hough, S. E., J. G. Armbruster, L. Seeber, and J. F. Hough (2000). Onthe modified Mercalli intensities and magnitudes of the 1811–1812New Madrid earthquakes. Journal of Geophysical Research 105,23,839–23,864.Hubbard, J., J. H. Shaw, and Y. Klinger (2010). Structural setting ofthe 2008 Mw 7.9 Wenchuan, China, earthquake. Bulletin of theSeismological Society of America 100, 2,713–2,735.Jaques, A. L., and D. H. Green (1980). Anhydrous melting of peridotite at0–15 kb pressure and the genesis of tholeiitic basalts. Contributionsto Mineralogy and Petrology 73, 287–310.Jia, D., G. Wei, Z. Chen, B. Li, Q. Zeng, and G. Yang (2006). LongmenShan fold-thrust belt and its relation to the western Sichuan Basinin central China: New insights from hydrocarbon exploration.American Association of Petroleum Geologists Bulletin 90, 1,425–1,447.Johnston, A. C. (1996a). Seismic moment assessment of earthquakes instable continental regions: I. Instrumental seismicity. GeophysicalJournal International 124, 381–414.Johnston, A. C. (1996b). Seismic moment assessment of earthquakesin stable continental regions: II. Historical seismicity. GeophysicalJournal International 125, 639–678.Johnston, A. C. (1996c). Seismic moment assessment of earthquakes instable continental regions: III. New Madrid 1811–1812, Charleston1886 and Lisbon 1755. Geophysical Journal International 126,314–344.Johnston, A. C., K. J. Coppersmith, L. R. Kanter, and C. A. Cornell(1994). The Earthquakes of Stable Continental Regions. Five vols.Report for Electric Power Research Institute (EPRI), Palo Alto,CA, EPRI TR-102261.Kanter, L. R. (1994). Tectonic interpretation of stable continental crust.Chapter 2 iin The Earthquakes of Stable Continental Regions, ed. A.C. Johnston, K. J. Coppersmith, L. R. Kanter, and C. A. Cornell,vol. 1, 1–98. Report for Electric Power Research Institute (EPRI),Palo Alto, CA, EPRI TR 102261.Keller, G. R., and J. D. Hoover (1988). A comparison of the Newark andRio Grande rift systems and associated magmatism. In Triassic-Jurassic Rifting, ed. W. Manspeizer, 911–932. New York: Elsevier.Kirby, E., P. W. Reiners, M. A. Krol, K. X. Whipple, K. V. Hodges, K.A. Farley, W. Tang, and Z. Chen (2002). Late Cenozoic evolutionof the eastern margin of the Tibetan Plateau: Inferences from40Ar/39Ar and (U-TH)/He thermochronology. Tectonics 21,1,001; doi:10.1029/2000TC0012146.Kwon, S., K. Sajeev, G. Mitra, Y. Park, S. W. Kim, and I.-C. Ryu (2009).Evidence for Permo-Triassic collision in Far East Asia: The Koreancollisional orogen. Earth and Planetary Science Letters 279, 340–349.Leloup, P. H., R. Lacassin, P. Tapponnier, U. Scharer, Z. Dalai, L.Xiaohan, Z. Liangshang, J. Shaocheng, and P. T. Trinh (1995). TheAilao Shan–Red River shear zone (Yunnan, China), Tertiary transformboundary of Indochina. Tectonophysics 251, 3–84.Li, S., W. D. Mooney, and J. Fan (2006). Crustal structure of mainlandChina from deep seismic sounding data. Tectonophysics 420, 239–252.Li, Z.-X., X.-H. Li, H. Zhou, and P. D. Kinny (2002). Grenvillian continentalcollision in south China: New SHRIMP U-Pb zirconresults and implications for the configuration of Rodinia. Geology30, 163–166.Liu, J., J. Han, and W. S. Fyfe (2001). Cenozoic episodic volcanism andcontinental rifting in northeast China and possible link to Japan Seadevelopment as revealed from K-Ar geochronology. Tectonophysics339, 385–401.Liu, J. F., J. Chen, J. H. Yin, Y. C. Lu, A. Murray, L. C. Chen, J. Thompson,and H. L. Yang (2010). OSL and AMS 14 C dating of the penultimateearthquake at the Leigu trench along the Beichuan fault, LongmenShan, in the northeast margin of the Tibetan Plateau. Bulletin of theSeismological Society of America 100, 2,681–2,688.Seismological Research Letters Volume 82, Number 6 November/December 2011 981

Liu, M., Y. Yang, S. Zhengkang, S. Wang, W. Min, and W. Yongge(2007). Active tectonics and intracontinental earthquakes inChina; the kinematics and geodynamics. In Continental IntraplateEarthquakes: Science, Hazard, and Policy Issues, ed. S. Stein and S.Mazzotti, 299–318. Geological Society of America Special Paper425. Boulder, CO: Geological Society of America.Madole, R. F. (1988). Stratigraphic evidence of Holocene faulting in themid-continent: The Meers fault, southwestern Oklahoma. Bulletinof the Geological Society of America 100, 392–401.Malik, J. N., A. A. Shah, A. K. Sahoo, B. Puhan, C. Banerjee, D. P.Shinde, N. Juyal, A. K. Singhvi, and S. K. Rath (2010). Activefault, fault growth and segment linkage along the Janauri anticline(frontal foreland fault), NW Himalaya, India. Tectonophysics 483,327–343.McCalpin, J. P., ed. (2009). Paleoseismology. Amsterdam: AcademicPress, 649 pp.McCalpin, J. P., and V. S. Khromovskikh (1995). Holocene paleoseismicityof the Tunka fault, Baikal rift, Russia. Tectonics 14, 594–605.McKenzie, D., and M. J. Bickle (1988). The volume and composition ofmelt generated by extension of the lithosphere. Journal of Petrology29, 625–679.McNulty, W. E., and S. F. Obermeier (1999). Liquefaction evidence for atleast two strong Holocene paleo-earthquakes in central and southwesternIllinois, USA. Environmental and Engineering Geoscience5, 133–146.Mueller, C. S. (2010). The influence of maximum magnitude on seismichazardestimates in the central and eastern United States. Bulletinof the Seismological Society of America 100, 699–711.Neumann, E. R., and I. B. Ramberg (1978). Paleorifts: Concludingremarks. In Tectonics and Geophysics of Continental Rifts, ed.I. B. Ramberg and E. R. Neumann, 409–424. Dordrecht, theNetherlands: D. Reidel Publishing Co.Obermeier, S. F. (1998). Liquefaction evidence for strong earthquakesof Holocene and latest Pleistocene ages in the states of Indiana andIllinois, USA. Engineering Geology 50, 227–254.Office of Nuclear Regulatory Research (2007). A Performance-basedApproach to Define the Site-specific Earthquake Ground Motion.Nuclear Regulatory Commission Guide 1.208, Nuclear RegulatoryCommission, 53 pp.; http://www.nrc.gov/reading-rm/doc-collections/reg-guides/power-reactors/rg/01-208/01-208.pdf.Lastaccessed July 21, 2011.Olafsson, M., and D. H. Eggler (1983). Phase relations of amphibole,amphibole-carbonate, and phlogopite-carbonate peridotite:Petrologic constraints on the asthenosphere. Earth and PlanetaryScience Letters 64, 305–315.Peizhen, Z., B. C. Burchfiel, P. Molnar, Z. Weiqi, J. Dechen, D. Qidong,W. Yipeng, L. Royden, and F. Song (1991). Amount and style oflate Cenozoic deformation in the Liupan Shan area, NingxiaAutonomous Region, China. Tectonics 10, 1,111–1,129.Perry, W. J., D. H. Roeder, and D. R. Lageson, comps. (1984). NorthAmerican Thrust-faulted Terranes: Selected Papers Reprinted fromthe AAPG Bulletin and other Geological Journals. Tulsa, OK:American Association of Petroleum Geologists, 466 pp.Petersen, M. D., A. D. Frankel, S. C. Harmsen, C. S. Mueller, K. M.Haller, R. L. Wheeler, R. L. Wesson, Y. Zeng, O. S. Boyd, D. M.Perkins, N. Luco, E. H. Field, C. J. Wills, and K. S. Rukstales(2008). Documentation for the 2008 Update of the United StatesNational Seismic Hazard Maps. USGS Open-File Report 2008–1128, 119 pp.; http://earthquake.usgs.gov/hazards/products/conterminous/2008/.Last accessed July 21, 2011.Qiu, Y. M., S. Gao, N. J. McNaughton, D. I. Groves, and W. Ling (2000).First evidence of >3.2 Ga continental crust in the Yangtze craton ofsouth China and its implications for Archean crustal evolution andPhanerozoic tectonics. Geology 28, 11–14.Rangin, C., P. Huchon, X. Le Pichon, H. Bellon, C. Lepvrier, D. Roques,N. D. Hoe, and P. V. Quynh (1995). Cenozoic deformation of centraland south Vietnam. Tectonophysics 251, 179–196.Reedman, A. J., and D. H. Kim (1997). Korea—South Korea. InEncyclopedia of European and Asian Regional Geology, ed. E. M.Moores and R. W. Fairbridge, 473–483. London: Chapman andHall.Ren, J., K. Tamaki, S. Li, and Z. Junxia (2002). Late Mesozoic andCenozoic rifting and its dynamic setting in eastern China and adjacentareas. Tectonophysics 344, 175–205.Rich, J. L. (1934). Mechanics of low-angle overthrust faulting as illustratedby Cumberland Thrust Block, Virginia, Kentucky, andTennessee. Bulletin of the American Association of PetroleumGeologists 18, 1,584–1,596.Risk Engineering Inc., Geomatrix Consultants Inc., Woodward-ClydeConsultants, and Cygna Corporation (1986). Seismic HazardMethodology for the Central and Eastern United States. NP-4726,11 vols. Report for Seismicity Owners Group and Electric PowerResearch Institute, Palo Alto, CA.Rodgers, J. (1950). Mechanics of Appalachian folding as illustratedby Sequatchie anticline, Tennessee and Alabama. AmericanAssociation of Petroleum Geologists Bulletin 34, 672–681.Rodgers, J. (1970). The Tectonics of the Appalachians. New York: Wiley,271 pp.Russ, D. P. (1979). Late Holocene faulting and earthquake recurrencein the Reelfoot Lake area, northwestern Tennessee. Bulletin of theGeological Society of America 90, 1,013–1,018.Sengor, A. M. C., and B. A. Natal’in (2001). Rifts of the world. InMantle Plumes: Their Identification through Time, ed. R. E. Ernstand K. L. Buchan, 389–482. Geological Society of America SpecialPaper 352. Boulder, CO: Geological Society of America.Shen, Z.-K., J. Lu, M. Wang, and R. Burgmann (2005). Contemporarycrustal deformation around the southeast borderland of theTibetan Plateau. Journal of Geophysical Research 110, B11409,doi:10.1029/2004JB003421.Simons, W. J. F., A. Soquet, C. Vigny, B. A. C. Ambrosius, S. H. Abu, C.Promthong, C. Subarya, D. A. Sarsito, S. Matheussen, P. Morgan,and W. Spakman (2007). A decade of GPS in southeast Asia—resolving Sundaland motion and boundaries. Journal of GeophysicalResearch 112, B06420, doi:10.1029/2005JB003868.Takahashi, E., and I. Kushiro (1983). Melting of a dry peridotite at highpressures and basalt magma genesis. American Mineralogist 68,859–879.Tarr, A. C., A. Villasenor, K. P. Furlong, S. Rhea, and H. M. Benz (2010).Seismicity of the Earth 1900–2007. USGS Scientific InvestigationsMap I-3064 (1 sheet), scale 1:25,000,000.Terman, M. J. (1974). Tectonic map of China and Mongolia. Two sheets,scale 1:5,000,000. Boulder, CO: Geological Society of America.Tuttle, M. P. (2001). The use of liquefaction featurees in paleoseismology—lessonslearned in the New Madrid seismic zone, centralUnited States. Journal of Seismology 5, 361–380.Tuttle, M. P., H. Al-Shukri, and H. Mahdi (2006). Very large earthquakescentered southwest of the New Madrid seismic zone 5,000–7,000 years ago. Seismological Research Letters 77, 755–770.Walker, R. T. (2009). The active tectonics of NE Asia. Abstracts andPrograms of the Geological Society of America 41, 658 (abstract).Walker, R. T., E. Nissen, E. Molor, and A. Bayasgalan (2007).Reinterpretation of the active faulting in central Mongolia. Geology35, 759–762.Wang, H., M. Liu, J. Cao, X. Shen, and G. Zhang (2011). Slip ratesand seismic moment deficits on major active faults in mainlandChina. Journal of Geophysical Research 116, B02405;doi:10.1029/2010JB007821.Wang, T. K., M.-K. Chen, C.-S. Lee, and K. Xia (2006). Seismic imagingof the transitional crust across the northeastern margin of theSouth China Sea. Tectonophysics 412, 237–254.Wheeler, R. L. (1997). Boundary separating the seismically activeReelfoot rift from the sparsely seismic Rough Creek graben,Kentucky and Illinois. Seismological Research Letters 68, 586–598.982 Seismological Research Letters Volume 82, Number 6 November/December 2011

Wheeler, R. L. (2008). Paleoseismic targets, seismic hazard, and urbanareas in the central and eastern United States. Bulletin of theSeismological Society of America 98, 1,572–1,580.Wheeler, R. L. (2009a). Methods of Mmax Estimation East of the RockyMountains. USGS Open–File Report 2009–1018, 43 pp.; http://pubs.usgs.gov/of/2009/1018. Last accessed July 21, 2011.Wheeler, R. L. (2009b). Sizes of the largest possible earthquakes inthe Central and Eastern United States: Summary of a workshop,September 8—9, 2008, Golden, Colorado, U.S. Geological SurveyOpen–File Report 2009–1263, 304 pp. (available at http://pubs.usgs.gov/of/2009/1263/, accessed July 21, 2011).Whitford-Stark, J. L. (1987). A Survey of Cenozoic Volcanism onMainland Asia. Geological Society of America Special Paper 213.Boulder, CO: Geological Society of America, 74 pp.Wilson, M. (1989). Igneous Petrogenesis. London: Chapman and Hall,466 pp.Wu, F.-Y., J.-Q. Lin, S. A. Wilde, X. O. Zhang, and J.-H. Yang (2005).Nature and significance of the Early Cretaceous giant igneousevent in eastern China. Earth and Planetary Science Letters 233,103–119.Yang, J.-H., S. O’Reilly, R. J. Walker, W. Griffin, F.-Y. Wu, M. Zhang,and N. Pearson (2010). Diachronous decratonization of the Sino-Korean craton: Geochemistry of mantle xenoliths from NorthKorea. Geology 38, 799–802.Yang, J.-H., F.-Y. Wu, and S. A. Wilde (2008). Mesozoic decratonizationof the North China block. Geology 36, 467–470.Yin, A. (2010). Cenozoic tectonic evolution of Asia: A preliminary synthesis.Tectonophysics 488, 293–325.Yu, Z., S. Wu, D. Zou, D. Feng, and H. Zhao (2008). Seismic profilesacross the middle Tan-Lu fault zone in Laizhou Bay, Bohai Sea,eastern China. Journal of Asian Earth Science 33, 383–394.Zhang, Z. M., J. G. Liou, and R. G. Coleman (1984). An outline ofthe plate tectonics of China. Bulletin of the Geological Society ofAmerica 95, 295–312.Zhao, G., S. A. Wilde, and P. A. Cawood (1998). Thermal evolution ofArchean basement rocks from the eastern part of the North Chinacraton and its bearing on geologic setting. International GeologyReview 40, 706–721.Zhao, G., S. A. Wilde, P. A. Cawood, and M. Sun (2001). Archean blocksand their boundaries in the North China craton: Lithological, geochemical,structural and P-T path constraints and tectonic evolution.Precambrian Research 107, 45–73.Zhao, L., R. M. Allen, T. Zheng, and S.-H. Hung (2009). Reactivationof an Archean craton: Constraints from P- and S-wave tomographyin North China. Geophysical Research Letters 36, L17306;doi:10.1029/2009GL039781.Zheng, J., W. L. Griffin, S. Y. O’Reilly, M. Zhang, N. Pearson, and Y.Pan (2006). Widespread Archean basement beneath the Yangtzecraton. Geology 34, 417–420.Zheng, T., L. Chen, L. Zhao, W. Xu, and R. Zhu (2006). Crust-mantlestructure difference across the gravity gradient zone in NorthChina craton: Seismic image of the thinned continental crust.Physics of Earth and Planetary Interiors 159, 43–58.Zhou, D., K. Ru, and H.-Z. Chen (1995). Kinematics of Cenozoic extensionon the South China Sea continental margin and its implicationsfor the tectonic evolution of the region. Tectonophysics 251,161–177.Zhu, G., M. Niu, C. Xie, and Y. Wang (2010). Sinistral to normal faultingalong the Tan-Lu fault zone: Evidence for geodynamic switchingof the east China continental margin. Journal of Geology 118,277–293.U.S. Geological SurveyMS 966, Box 25046Lakewood, Colorado 80225 U.S.A.wheeler@usgs.govSeismological Research Letters Volume 82, Number 6 November/December 2011 983

Moment Magnitude (M W ) Conversion Relationsfor Use in Hazard Assessment in EasternCanadaAllison L. BentAllison L. BentCanadian Hazards Information Service, Geological Survey of CanadaABSTRACTTo be unbiased and uniform across a wide geographical area,seismic hazard assessments based primarily on earthquakerecurrence rates require that the same magnitude scale be usedfor all earthquakes evaluated. Increasingly, moment magnitude,M W , is seen as the magnitude of preference. Momentmagnitude, however, was not routinely calculated in the pastfor earthquakes in Canada, necessitating the conversion fromother magnitude types in common use. This step is complicatedby the fact that several magnitude scales are routinely reportedfor Canadian earthquakes with the choice being influencedprimarily by geography and to a lesser extent by the size of theearthquake. This paper focuses on eastern Canada, where m Nis the most commonly used magnitude scale. Conversions toM W are established and evaluated. The simple conversion ofapplying a constant is sufficient. However, the conversion istime dependent with the constant changing from 0.41 to 0.53in the mid-1990s.INTRODUCTIONMagnitude recurrence rates are an important factor in seismichazard assessment in Canada and elsewhere. The (Canadian)National Earthquake Database (NEDB 2010), hereafterreferred to as the NEDB, routinely reports several earthquakemagnitude scales for Canadian earthquakes, with m N and M Lbeing the most commonly used for eastern Canada. 1 Whenevaluating magnitude recurrence curves for use in seismic hazardassessment there exists the possibility that a mixed dataset will lead to non-uniform or even erroneous results. Thus,it becomes imperative to use the same magnitude scale for all1. For hazard purposes a conversion equation only for offshore eventswith M L is required. For these, m N is not appropriate because theirS wavetrain lacks Lg energy or the Lg energy is clearly attenuated.There are at least two other types of M L used in eastern Canada: pre-1980 onshore earthquakes for which magnitudes were computedbefore m N was defined (m N s for most of these back to about 1940have subsequently been determined from amplitude data but someevents remain as M L s) and small earthquakes up to the present forwhich there is no amplitude data at a station beyond 50 km (these arenot important for seismic hazard).earthquakes in the data set. Moment magnitude, or M W , hasbecome the preferred magnitude scale as it can be related tothe physical properties of the earthquake rupture and does notsaturate at high magnitudes. However, this magnitude has notbeen routinely calculated in the past in Canada and using it forhazard assessment requires that reliable M W s be determinedfor all earthquakes used in the hazard calculations. This paperfocuses on eastern Canada but similar studies have been undertakento derive M W conversions for western Canada (Ristau etal. 2003, 2005).Moment magnitude has been determined for many ofthe largest earthquakes in eastern Canada and for some of themoderate ones. In a recent study Bent (2009) evaluated data forthe 150 largest earthquakes that met the completeness criteriafor use in hazard assessment in eastern and northern Canadaand determined M W s for each of them. While instrumentallydetermined M W s were given preference, conversions fromother magnitude types or felt information were sometimesnecessary given the long time period covered. Furthermore, itis now almost always possible to determine M W for eastern andnorthern Canadian earthquakes of magnitude 5.0 or greaterand also possible for many of the magnitude 4 to 5 earthquakes.Therefore, in developing a conversion scale, the emphasis is onthe smaller (less than magnitude 5.0) earthquakes.The approach used in this study is to assemble a databaseof eastern and northern Canadian earthquakes for which aninstrumentally determined m N and M W are available, establisha conversion relation, and evaluate it with respect to several previouslypublished relations based on smaller data sets. The M Wdata set for offshore regions where M L is the primary magnitudewas insufficient to establish a reliable conversion relation.It may be necessary to employ a two-step conversion from M Lto m b and then m b to M W . This conversion relation is underinvestigation by the author.A variety of conversions of varying complexity were testedbut the author verified that the more complex conversions didnot result in a statistically significant improvement over thesimple application of a constant, and in some cases did notprovide a better absolute fit in terms of mean residual. Thesenew simple conversions were, however, found to be statisticallysignificant improvements over previously published relations.984 Seismological Research Letters Volume 82, Number 6 November/December 2011 doi: 10.1785/gssrl.82.6.984

A significant outcome of this study is that the m N – M Wconversion was found to be time dependent with the constantchanging by 0.1 around 1995, which corresponds to a timewhen the Canadian National Seismograph Network (CNSN)underwent several changes and upgrades and the procedureused for routine earthquake locations and magnitude calculationsin eastern Canada was also updated. Further studies areunderway to better understand the exact nature of the reason(s)for this change.m N to M WA data set of earthquakes for which both m N and M W were independentlydetermined from instrumental data was establishedby searching the NEDB, published literature, and relevant Websites. Except in a few cases where it was unavailable, the m Nvalue is that found in the NEDB. Note that based on the recommendationsof Wetmiller and Drysdale (1982), 1) the Nuttli(1973) formula for distances greater than 4° (444 km) is usedfor all distances, and 2) the magnitudes are calculated over awider range of frequencies than that for which the scale was originallydefined. The Nuttli magnitude scale was derived for usewith World-wide Standard Seismograph Network (WWSSN)instruments. Amplitudes used for magnitude determination inthe NEDB are read from the raw waveforms and corrected forinstrumental magnification at the period at which they are read.That is, there is no transfer to a WWSSN instrument response.However, the instrument responses for the earlier versions ofthe CNSN were very similar to those of the WWSSN instruments.With the most recent upgrades to the network, that is nolonger the case, as the “flat” frequency range has been extendedfor both the short-period and broadband instruments.The M W s used in this study come from a wide variety ofsources (noted in Table 1) but the author verified that theywere instrumentally determined using well established methods,primarily moment tensor inversion, forward waveformmodeling, and spectral analysis. The final data set consists of154 earthquakes (Table 1, Figure 1). Note that m N 2.5 is theminimum magnitude used in this study as it is generally thelowest magnitude used in hazard calculations for reasons ofcompleteness.A precursory evaluation of the data showed that for themost part there was a narrow range of M W s for any given m N(Figure 2). There were, however, a few events for which the relationappeared anomalous. On closer inspection it was foundthat all of these events occurred in the Byam Martin Channelat the northwest extreme of the area of interest. All ByamMartin Channel events (six total) were excluded from furtheranalysis. Therefore, the resulting m N – M W conversion relationmay not be applicable to that region.The simplest conversion relation would be the straightforwardapplication of a constant, which was determined by subtractingM W from m N and calculating the mean value. The bestfit for the complete data set is M W = m N – 0.43 with a standarddeviation of 0.18. Subdividing the data set into magnitude binsdid not result in a significantly different constant for any bin.In the course of updating the hazard calculations foreastern Canada (J. Adams, personal communication 2006) itwas noticed that there was a slight change in the magnituderecurrence curves for the lower magnitudes when the data setwas updated by adding events that occurred since 1990 (theadded earthquakes were mostly lower magnitudes). EvaluatingM W – m N as a function of time suggested that there was achange in the relation around 1995. This date corresponds wellto a previously identified date (Bent, unpublished data) whena number of factors that could influence magnitude changed.These factors include• a decrease in the average period at which magnitude wasmeasured;• an increase in the number of stations used to calculatemagnitude;• an increase in the average station-epicenter distance (probablyrelated to station increase); and• an improvement in the precision to which amplitudes andperiods were calculated.All of these are likely linked to improvements and changes madeto the CNSN as well as a change in the software used to calculatelocations and magnitudes that occurred in the mid-1990s.In light of the above, the calculations were redone subdividingthe data into pre-1995 and post-1995 bins with the1995 events included in the post-1995 bin. The resulting constantswere 0.41 ± 0.18 for the pre-1995 events (116 events) and0.53 ± 0.19 for the post-1995 group (32 events). The calculationswere redone using dividing dates ranging from 1993 to1998 but the slight differences in the constants were not statisticallysignificant and the exact date at which the changeoccurred could not be further refined. To help equalize the sizeof the data sets the calculations were redone using only thoseevents of magnitude 4.0 or greater. The same time dependencewas seen. Given that the network changes occurred over aperiod of several years, there may not be an exact date at whichthe relation changed. All further calculations use 1995 as thedividing date. Subdividing each group further by magnitudedid not affect the results, and all further calculations use data atall magnitudes unless stated otherwise. The data set was sent toanother researcher (R. Youngs, written communication 2010),who confirmed that there was a time dependence in the difference;1997 was his preferred date as it gave the lowest standarddeviation but again the date is not precisely constrained.Least squares straight line fits were also determined for thedata, with the results as follows:M W = 0.99m N – 0.36 ± 0.16 (pre-1995)M W = 0.93m N – 0.22 ± 0.19 (post-1995)The F-test was used to compare the fit of the straight line andconstant for each group of data. For the pre-1995 group theconstant with a mean residual (converted M W vs. true M W ) of–0.004 results in a slightly better fit than the straight line witha mean residual of 0.007. The p value from the F-test is 0.601,implying that the difference between the two is not statisticallySeismological Research Letters Volume 82, Number 6 November/December 2011 985

TABLE 1Earthquakes for m N – M W Conversionyyyymmdd Location m N*M W Source †19391019 Charlevoix QC 5.6 5.3 B09, J9419660101 Attica NY 4.7 4.3 H7819670613 Attica NY 4.5 4.1 H7819711002 W of Coral Harbor NU 5.1 4.7 B09, J9419720121 Baffin Bay 5.1 4.6 B09, J9419721227 Byam Martin Channel ‡ 5.4 6.3 B09, J9419721121 Byam Martin Channel ‡ 5.7 6.0 B09, J9419721119 Byam Martin Channel ‡ 5.6 5.9 B09, J9419721228 Byam Martin Channel ‡ 5.1 5.9 B09, J9419790627 NW of Spence Bay 5.0 5.0 B09, J9419790819 Charlevoix QC 5.0 4.8 B09, J9419800311 St-Basile QC 3.7 3.4 Bo9419800403 Lwr St Lawrence 4.0 3.6 Bo9419810616 Charlevoix QC 3.6 3.2 Bo9419810704 ON-QC-NY border 3.8 3.2 Bo9419810713 Lwr St. Lawrence 3.8 3.3 Bo9419810918 Ste-Adele QC 3.6 3.0 Bo9419810930 Lac-du-Cerf QC 3.4 3.0 Bo9419811028 Lwr St. Lawrence 3.9 3.5 Bo9419820109a Miramichi NB 5.8 § 5.6 B09, J9419820109b Miramichi NB 5.0 § 4.9 B09, J9419820109c Miramichi NB 3.8 3.4 Bo9419820109d Miramichi NB 3.7 3.3 Bo9419820111 Miramichi NB 5.5 5.0 B09, J9419820113 Miramichi NB 4.0 3.2 Bo9419820115 Miramichi NB 3.8 3.4 Bo9419820117 Miramichi NB 3.6 3.3 Bo9419820119 Gaza NH 4.5 4.5 Bu8719820316 Miramichi NB 3.5 3.3 Bo9419820402 Miramichi NB 4.3 3.7 Bo9419820411 Miramichi NB 4.0 3.5 Bo9419820418 Miramichi NB 4.1 3.5 Bo9419820506 Miramichi NB 4.0 3.5 Bo9419820623 SE of Val d’Or QC 3.5 2.9 Bo9419820713 St-Jovite QC 3.8 3.3 Bo9419820728 Miramichi NB 3.7 3.3 Bo9419820806 Western QC 3.7 3.1 Bo9419820813 N of North Bay ON 4.3 3.6 Bo9419820903 Western QC 3.7 3.2 Bo9419821026 Miramichi NB 3.5 3.1 Bo9419821204 Charlevoix QC 3.9 3.4 Bo9419830117 Lwr St. Lawrence 4.1 3.6 Bo9419830513a Miramichi NB 3.5 3.1 Bo9419830513b Miramichi NB 3.9 3.7 Bo9419830516 Charlevoix NB 3.8 3.4 Bo9419830529 Lac-Megantic QC 4.1 3.7 Bo9419830812 St-Stephen NB 3.5 3.0 Bo94TABLE 1 (continued)Earthquakes for m N – M W Conversionyyyymmdd Location m N*M W Source †19831007a Goodnow NY 5.3 4.8 NS8919831007b Goodnow NY 3.6 3.2 Bo9419831011 Kanata, ON 4.1 3.6 Bo9419831117 Miramichi NB 3.7 3.3 Bo9419831228 Western QC 3.5 2.9 Bo9419840224 Miramichi NB 3.7 3.3 Bo9419840411 Lwr St. Lawrence 3.8 3.4 Bo9419840923 Southern NB 3.6 3.3 Bo9419841130 Miramichi NB 3.8 3.4 Bo9419850303 Charlevoix QC 3.1 2.8 Bo9419850412 Maine 3.5 3.0 Bo9419851005 Miramichi NB 4.0 3.5 Bo9419851019 Pennsylvania 4.1 3.6 Bo9419860111 Charlevoix QC 4.0 3.4 Bo9419860131 Painesville OH 5.0 4.6 N8819860806 Western QC 3.5 3.2 Bo9419860818 Charlevoix QC 3.0 2.7 Bo9419860919 Charlevoix QC 4.2 3.6 Bo9419861109 Lwr St. Lawrence 4.2 3.7 Bo9419870318 Charlevoix QC 3.3 2.8 Bo9419870713 Ashtabula OH 4.1 3.6 Bo9419870806 Charlevoix QC 3.4 2.9 Bo9419870926 New York 3.8 3.3 Bo9419871023 Kilmar QC 3.7 3.2 Bo9419871111a Western QC 3.5 3.0 Bo9419871111b Western QC 3.2 3.0 Bo9419880102 Charlevoix QC 3.6 3.1 Bo9419880124 Charlevoix QC 3.1 2.7 Bo9419880128 Lwr St. Lawrence 3.8 3.6 Bo9419880310 Western QC 3.7 3.3 Bo9419880313 Charlevoix QC 3.1 2.8 Bo9419880424 Southern NB 3.6 3.3 Bo9419880509 Miramichi NB 3.5 3.1 Bo9419880515 S of Ottawa ON 3.3 3.0 Bo9419880809 Cornwall ON 3.4 2.8 Bo9419880826 Miramichi NB 3.8 3.5 Bo9419881020 Northern NH 3.9 3.5 Bo9419881123 Saguenay QC 4.8 4.3 H9619881125 Saguenay QC 6.5 5.9 B09, N8919881126 Sageunay QC 4.1 3.4 H9619881126 Saguenay QC 2.9 2.5 H9619881126 Saguenay QC 2.6 2.5 H9619881211 Saguenay QC 2.8 2.5 H9619890119 Saguenay QC 3.6 3.3 Bo9419890131 Charlevoix QC 3.1 2.9 Bo9419890210 Lwr St. Lawrence 4.3 3.8 Bo9419890309 Charlevoix QC 4.3 3.8 Bo94986 Seismological Research Letters Volume 82, Number 6 November/December 2011

TABLE 1 (continued)Earthquakes for m N – M W Conversionyyyymmdd Location m N*M W Source †19890311 Charlevoix QC 4.4 3.6 SA0119890316 E coast of Ungava QC 5.7 5.0 B09, BH9219890810 Blackville NB 3.5 3.3 Bo9419891013 Charlevoix NB 3.2 3.0 Bo9419891104 Western QC 3.4 2.9 Bo9419891112 Charlevoix QC 3.4 2.9 SA0119891116 Western QC 4.0 3.6 Bo9419891122 Charlevoix QC 3.4 3.0 Bo9419891225 Ungava QC 6.1 6.2 B09, B9419900303 Charlevoix QC 3.6 3.5 Bo9419900313 Charlevoix QC 3.2 2.9 Bo9419900421 Charlevoix QC 3.1 2.9 Bo9419900423 Charlevoix QC 3.0 2.7 Bo9419901007 Mont-Laurier QC 3.9 3.6 Bo9419901019 Mont-Laurier QC 5.0 4.5 B09, L9419901021 Charlevoix QC 3.3 3.0 Bo9419901218 Charlevoix QC 3.3 3.1 Bo9419901220 Mont Laurier QC 2.7 2.4 H9619910101 Mont Laurier QC 2.9 2.5 H9619910131 Mont-Laurier QC 3.2 2.7 H9619910801 Mont-Laurier QC 2.6 2.0 H9619911127 Mont-Laurier QC 2.5 2.2 H9619911208 Charlevoix QC 4.3 3.7 SA0119930807 Charlevoix QC 3.1 2.5 SA0119931116 Napierville QC 4.1 3.9 D0319940116 Reading PA 4.0 3.9 D0319940116 Reading PA 4.6 4.6 D0319950616 Lisbon NH 3.8 3.7 D0319960314 Ste-Agathe QC 4.4 4.2 BCMT19960607 Charlevoix QC 3.1 2.6 SA0119960821 Berlin NH 3.6 3.4 D0319960924 Charlevoix QC 3.1 2.4 SA0119970524 Christieville QC 4.2 3.6 D0319971028 Charlevoix QC 4.7 4.3 D0319971106 Quebec City 5.1 4.9 B09, BCMT19971206 Wager Bay NU 5.7 5.0 B09, BCMT19980730 La Conception QC 4.4 3.7 D0319980925 OH-PA border 5.4 4.5 B09, USGS19990316 Lwr St. Lawrence 5.1 4.5 B09, L0420000101 Kipawa QC 5.2 4.7 B09, B0220000420 Saranac L. NY 4.0 3.6 D0320010126 Ashtabula OH 4.4 3.9 D03TABLE 1 (continued)Earthquakes for m N – M W Conversionyyyymmdd Location m N*M W Source †20010814 Byam Martin Channel ‡ 5.6 5.2 B09, GCMT20010814 Byam Martin Channel ‡ 4.9 5.2 GCMT20020420 Au Sable Forks NY 5.5 5.1 B09, GCMT20020605 Radisson QC 4.5 3.7 BCMT20030613 Charlevoix QC 4.2 3.8 BCMT20040628 Illinois 4.7 4.2 H1020040804 Lake Ontario 3.8 3.1 K0620040826 Wager Bay NU 5.0 4.3 B09, BCMT20050306 Charlevoix QC 5.4 4.7 B09, BCMT20051020 Thornbury QC 4.3 3.6 H1020060109 Huntingdon QC 4.2 3.6 BCMT20060225 Thurso QC 4.5 3.9 BCMT20061129 Sudbury ON 4.1 3.7 A0820061207 Cochrane ON 4.2 3.7 BCMT20060407 Charlevoix QC 4.1 3.8 H1020061003 Maine 4.3 3.9 H1020081115 Charlevoix QC 4.2 3.6 H1020090321 Ivujivik QC 4.4 3.8 H1020090721 Lwr St. Lawrence 4.3 3.6 BCMT* m N from NEDB (2010) unless otherwise stated.† Sources for M WA08: Atkinson et al. (2008)B02: Bent et al. (2002)B09: Bent (2009)B94: Bent (1994)BCMT: Bent unpublished CMTBH92: Bent and Hasegawa (1992)Bo94: Boatwright (1994)Bu87: Burger et al. (1987)D03: Du et al. (2003)GCMT: Global CMT Project (2010)H10: Herrmann (2010)H78: Herrmann (1978)H96: Haddon (1996)J94: Johnston et al. (1994) and references thereinK06: Kim et al. (2006)L04: Lamontagne et al. (2004)L94: Lamontagne et al. (1994)N88: Nicholson et al. (1988)N89: North et al. (1989)NS89: Nábélek and Suaréz (1989)SA01: Sonley and Atkinson (2001)USGS: United States Geological Survey (2010)‡ Event not used in analysis (see comments in text).§ m N from Chael (1987).Seismological Research Letters Volume 82, Number 6 November/December 2011 987

▲▲Figure 1. Map showing locations of events used to determine the m N – M W conversion relation.significant. Given that the constant is simpler to apply andprovides a marginally better fit, there is clearly no advantage tousing a more complicated relation. For the post-1995 group themean residual is −0.005 for both the straight line and the constantand the p value is 0.922, meaning that there is no advantageto using one over the other. Thus, the constant is retainedas the preferred value.There are several existing M W – m N conversion relationsin the published literature. Table 2 shows a comparison of theresults from this study with existing conversion relations. Itis not surprising that the conversion relations from this studygive the lowest residuals, since they were derived from the dataset used in the comparison while the others were based on varioussubsets of it. Additionally, some relations were stated asbeing valid only for a specific range of magnitudes and thereforemight not be expected to provide a good fit for every eventin this data set. The principal question, however, is whether thelower residuals of this study represent a statistically significantimprovement over the earlier conversion relations. Applyingthe F-test to all of the relations included in Table 2 gives a pvalue of 0.000 for both the pre- and post-1995 groups, implyingthat the differences are statistically significant at a 99.9%or higher level. The worst fitting relations in terms of residualswere removed and the comparison was redone for thosemodels where the absolute value of the mean residual was lessthan 0.1—the two models from this study as well as Atkinson(1993) and Sonley and Atkinson (2005). For the pre-1995 datathe differences are again significant at the 99.9% level. For thepost-1995 data the results are equivocal. The p value of 0.508indicates that it is about equally likely that the differences arereal or due to random chance.The conversion relations presented here are sufficientlywell determined and tested that they may be used for convertinglarge data sets for use in hazard assessments. However, morework needs to be done to better understand the time dependence.Some preliminary tests have been performed to evaluatethe likelihood that specific changes made during the 1990s ledto the apparent change in magnitude. By rounding the amplitudedata of the recent events to the precision of the earlierevents and recalculating the magnitudes, the added precisioncould be ruled out. The other factors (number and distributionof stations, frequency, processing software) could not be ruledout easily and more detailed examination of them is required todetermine which is/are the cause(s) and why.CONCLUSIONSRelations for converting from m N to M W for eastern Canadahave been established. The conversion from m N requires simplythe subtraction of a constant although there is a time-dependenceto that value. Prior to 1995 the value is 0.41 and forearthquakes occurring since 1995 it is 0.53. The cut-off date isnot precise, but differences caused by changing the cut-off dateby up to two years in either direction were not found to be sta-988 Seismological Research Letters Volume 82, Number 6 November/December 2011

▲▲Figure 2. Earthquakes used in the m N – M W conversion. Gray symbols represent the pre-1995 events, black symbols the post-1995ones, and white symbols the Byam Martin Channel events not used in the analysis. The corresponding diagonal lines show the best fitstraight line in a least squares sense to each group. Most dots represent individual earthquakes but some represent two to six earthquakesthat lie on the same point.TABLE 2Comparison of M W – m N Conversion RelationsPre-1995Post-1995RelationMean Residual S. D. Mean Residual S.D.Constant (this study) –0.004 0.16 –0.005 0.19Line (this study) 0.007 0.16 –0.005 0.19Atkinson (1993) –0.061 0.16 0.046 0.19Boore and Atkinson (1987) 0.149 0.22 0.124 0.21Johnston et al. (1994) 0.405 0.30 0.301 0.26Hasegawa-a(1983) * –0.192 0.31 –0.284 0.27Hasegawa-b (1983) * –0.391 0.24 –0.140 0.27Nuttli (1983) 0.216 0.29 0.147 0.25Sonley and Atkinson (2005) –0.088 0.17 0.048 0.20* a refers to relation for m N < 4.2 and b for m N 4.2–6.6.tistically significant. This date corresponds to known changesin the CNSN and operating system. Further investigation isunderway to better understand the underlying reason for thechange in the magnitude relation.It should be cautioned that these relations were establishedusing earthquakes of m N 2.5 or greater and may not be appropriatefor smaller magnitudes. While it is also possible thatthey may not be appropriate for the largest earthquakes, thisfactor should not be a major problem as it should always be possibleto determine M W directly for these events. Comparisonswith instrumentally determined moment magnitudes showthat these conversions are very reliable on average when a largedata set is considered but will not always give the correct M Wfor any individual earthquake. While it would be expected thatthe conversions should hold true for eastern North America ingeneral, it must be emphasized that they are based on magni-Seismological Research Letters Volume 82, Number 6 November/December 2011 989

tudes calculated by the GSC and should be used with cautionfor magnitudes calculated by other agencies unless it can beverified that the same formulae and procedures were used. Inparticular, until the underlying causes of the m N – M W timedependenceare better understood, it should not be assumedthat this time dependence applies to all data bases for easternNorth America.ACKNOWLEDGMENTSI thank Bob Youngs for providing a second opinion on the timedependence of m N . John Adams, John Cassidy, and MartinChapman provided constructive reviews. Janet Drysdale verifiedthe pre-1995 magnitude calculation procedure. NaturalResources Canada contribution number 20110070.REFERENCESAtkinson, G. M. (1993). Earthquake source spectra in eastern NorthAmerica. Bulletin of the Seismological Society of America 83, 1,778–1,798.Atkinson, G. M., S.-L. I. Kaka, D. Eaton, A. Bent, V. Peci, and S. Halchuk(2008). A very close look at a moderate earthquake near Sudbury,Ontario. Seismological Research Letters 79, 119–131.Bent, A. L. (1994). The 1989 (M S 6.3) Ungava, Quebec earthquake: Acomplex intraplate event. Bulletin of the Seismological Society ofAmerica 84, 1,075–1,088.Bent, A. L. (2009). A Moment Magnitude Catalog for the 150 LargestEastern Canadian Earthquakes. Geological Survey of CanadaOpen-File Report 6080, 23 pp.Bent, A. L., and H. S. Hasegawa (1992). Earthquakes along the northwesternboundary of the Labrador Sea. Seismological ResearchLetters 63, 587–602.Bent, A. L., M. Lamontagne, J. Adams, C. R. D. Woodgold, S. Halchuk,J. Drysdale, R. J. Wetmiller, S. Ma, and J.-B. Dastous (2002). TheKipawa, Quebec “Millennium” earthquake. Seismological ResearchLetters 73, 285–297.Boatwright, J. (1994). Regional propagation characteristics and sourceparameters of earthquakes in eastern North America. Bulletin ofthe Seismological Society of America 84, 1–15.Boore, D. M., and G. M. Atkinson (1987). Stochastic prediction ofground motion and spectral response parameters at hard-rock sitesin eastern North America. Bulletin of the Seismological Society ofAmerica 77, 440–467.Burger, R. W., P. G. Somerville, J. S. Barker, R. B. Herrmann, and D. V.Helmberger (1987). The effect of crustal structure on strong groundmotion attenuation relations in eastern North America. Bulletin ofthe Seismological Society of America 77, 420–439.Chael, E. P. (1987). Spectral scaling of earthquakes in the Miramichiregion of New Brunswick. Bulletin of the Seismological Society ofAmerica 77, 347–365.Du, W.-X., W.-Y. Kim, and L. R. Sykes (2003). Earthquake sourceparameters and state of stress for the northeastern United Statesand southeastern Canada from analysis of regional seismograms.Bulletin of the Seismological Society of America 93, 1,633–1,648.Global CMT Project (2010). Online database, http://www.globalcmt.org.Haddon, R. A. W. (1996). Use of empirical Green’s functions, spectralratios, and kinematic source models for simulating strong groundmotion. Bulletin of the Seismological Society of America 86, 597–615.Hasegawa, H. S. (1983). Lg spectra of local earthquakes recorded bythe Eastern Canada Telemetered Network and spectral scaling.Bulletin of the Seismological Society of America 73, 1,041–1,061.Herrmann, R. B. (1978). A seismological study of two Attica, New Yorkearthquakes. Bulletin of the Seismological Society of America 68,641–651.Herrmann, R. (2010). Online database, http://www.eas.slu.edu/eqc/eqc_mt/MECH.NA/.Johnston, A. C., K. J. Coppersmith, L. R. Kanter, and C. A. Cornell(1994). The Earthquakes of Stable Continental Regions. Assessmentof Large Earthquake Potential, TR-102261-V1. Five-volume proprietaryreport prepared for Electric Power Research Institute, PaloAlto, CA.Kim, W.-Y., S. Dineva, S. Ma, and D. Eaton (2006). The 4 August 2004,Lake Ontario, earthquake. Seismological Research Letters 77, 65–73.Lamontagne, M., A. L. Bent, C. R. D. Woodgold, S. Ma, and V. Peci(2004). The 16 March 1999 m N 5.1 Côte-Nord earthquake: Thelargest earthquake ever recorded in the Lower St. Lawrence seismiczone, Canada. Seismological Research Letters 75, 299–316Lamontagne, M., H. S. Hasegawa, D. A. Forsyth, G. G. R. Buchbinder,and M. Cajka (1994). The Mont-Laurier, Quebec, earthquake of 19October 1990 and its seismotectonic environment. Bulletin of theSeismological Society of America 84, 1,506–1,522.Nábélek, J., and G. Suaréz (1989). The 1983 Goodnow earthquake in thecentral Adirondacks, New York: Rupture of a simple, circular crack.Bulletin of the Seismological Society of America 79, 1,762–1,778.National Earthquake Database (2010). Digital database, http://www.seismo.nrcan.gc.ca, Geological Survey of Canada, Ottawa, Ontario.Nicholson, C., E. Roeloffs, and R. L. Wesson (1988). The northeasternOhio earthquake of 31 January 1986: Was it induced? Bulletin ofthe Seismological Society of America 78, 188–217.North, R. G., R. J. Wetmiller, J. Adams, F. M. Anglin, H. S. Hasegawa,M. Lamontagne, R. DuBerger, L. Seeber, and J. Armbruster(1989). Preliminary results from the November 25, 1988 Saguenay(Quebec) earthquake. Seismological Research Letters 60, 89–93.Nuttli, O. W. (1973). Seismic wave attenuation and magnitude relationsfor eastern North America. Journal of Geophysical Research 78,876–885.Nuttli, O. W. (1983). Average seismic source-parameter relations formid-plate earthquakes. Bulletin of the Seismological Society ofAmerica 73, 519–535.Ristau, J., G. Rogers, and J. Cassidy (2003). Moment magnitude calibrationfor earthquakes off Canada’s west coast. Bulletin of theSeismological Society of America 93, 2,296–2,300.Ristau, J., G. C. Rogers, and J. F. Cassidy (2005). Moment magnitudelocalmagnitude calibration for earthquakes in western Canada.Bulletin of the Seismological Society of America 95, 1,994–2,000;doi:10.1785/0120050028.Sonley, E., and G. M. Atkinson (2001). Apparent source spectra forearthquakes in the Charlevoix seismic zone: A comparison ofdirect and empirical Green’s function methods. Bulletin of theSeismological Society of America 91, 1,729–1,740.Sonley, E., and G. M. Atkinson (2005). Empirical relationship betweenmoment magnitude and Nuttli magnitude for small-magnitudeearthquakes in southeastern Canada. Seismological Research Letters76, 752–755.United States Geological Survey (2010). Online database, http://neic.usgs.gov.Wetmiller, R. J., and J. A. Drysdale (1982). Local magnitude of easternCanadian earthquakes by an extended m b (Lg) scale. EarthquakeNotes 53 (3), 40.Canadian Hazards Information ServiceGeological Survey of Canada7 Observatory CrescentOttawa, Ontario K1A 0Y3 Canadabent@seismo.nrcan.gc.ca990 Seismological Research Letters Volume 82, Number 6 November/December 2011

Meeting CalendarM E E T I N GC A L E N D A R20114 November. Consortium of Organizations for StrongMotion Observation Systems (COSMOS) Annual Meetingand Technical Session, “Recent Major Earthquakes andtheir Influence on Strong Ground Motion Determinationsand Design,” Emeryville, California.www.cosmos-eq.org.6–7 December. Geotechnical Short Course, Virginia Tech,Blacksburg, Virginia.www.cpe.vt.edu/gee/201211–13 January. Magmatic Rifting and Active VolcanismConference, Addis Ababa, Ethiopia.http://www.see.leeds.ac.uk/afar/new-afar/conference/conference.html22–25 January. 7th Gulf Seismic Forum (GSF 2012),Jeddah, Saudi Arabia.http://7gsf.info/.28 February−2 March. 10th International Workshop onSeismic Microzoning and Risk Reduction, Tsukuba,Japan.http://www.jaee.gr.jp/event/10IWSMRR/index.html3–4 March. International Symposium One Year after the2011 Eastern Japan Earthquake, Kenchiku-kaikan Hall,Tokyo.kawashima.k.ae@m.titech.ac.jp10−14 April. Earthquake Engineering Research InstituteAnnual Meeting/Federal Emergency ManagementAgency National Earthquake Conference, Peabody Hotel,Memphis, Tennessee.www.eeri.org17−19 April. 2012 SSA Annual Meeting, San Diego,Californiawww.seismosoc.org and page 966 for details.23−27 April. National Earthquake Conference, Memphis,Tennesseewww.earthquakeconference.org13–15 June. Incorporated Research Institutions forSeismology (IRIS) Workshop, Boise, Idaho.www.iris.edu26−29 June. 45th Rock Mechanics / GeomechanicsSymposium, San Francisco, Californiawww.armasymposium.org5−10 August. 4th International Geological Congress,Brisbane, Australia.www.34igc.org24−28 September. 15th World Conference on EarthquakeEngineering (15WCEE), Lisbon, Portugalwww.15wcee.org4–7 November. Geological Society of America AnnualMeeting, Charlotte, North Carolina.www.geosociety.org/meetings/2012/Please send notices of meetings you would like to appear in the“Meeting Calendar” three months before the expected date ofpublication. Send announcements to SRL Editor Jonathan M.Lees in care of the SRL managing editor at srl@seismosoc.org.doi: 10.1785/gssrl.82.6.991Seismological Research Letters Volume 82, Number 6 November/December 2011 991

992 Seismological Research Letters Volume 82, Number 6 November/December 2011

201 Plaza Professional Bldg. • El Cerrito, CA 94530Phone: +1-510-525-5474 • Fax: +1-510-525-7204 • http://www.seismosoc.orgMEMBERSHIP APPLICATIONTitle: ☐ Dr. ☐ Mr. ☐ Ms. ☐ Mrs. ☐ Miss ☐ Fr.Name:__________________________________________________________________Company/Institution:___________________________________________________Address:_ _____________________________________________________________________________________________________________________________________Phone: ____________ Fax: ____________ E‐mail: __________________________SSA publishes two journals: The Bulletin of the Seismological Society of America (BSSA) and Seismological Research Letters (SRL.) Bothof these journals are available in electronic form to SSA members. In addition, some membership options include the print version ofone or both journals. All members are entitled to meeting and print journal discounts.SSA Membership for 2012All memberships include online access to both BSSA and SRL.☐ Membership with print BSSA and print SRL: $140 in North America, $150 other countries.☐ Membership with just print SRL: $110 in all countries☐ Membership with only electronic journals: $100 all countries☐ Student membership (all electronic, must have current student id): $25☐ Student membership (print SRL, must have current student id): $35☐ Developing country (all electronic, must reside in country not listed as high-income by the World Bank): $45Additional for expedited shipping of BSSA outside the United States: $75Additional for priority shipping of BSSA within the United States: $35Additional for expedited shipping of SRL outside the United States: $35Additional for priority shipping of SRL within the United States: $30Eastern Section Membership $5SubtotalFor applications dated after 1 July, applicants may opt for an 18-month membership (e.g., 1 July 2012 through 31December 2013). Calculate the additional cost by multiplying the subtotal by 0.5.Total enclosed________________________________________________________Payment Options:☐ Check drawn on U.S. bank in U.S. dollars payable to SSA☐ International Postal Money Order☐ Credit Card: ☐ Visa ☐ Mastercard ☐ American ExpressAccount Number _________________________________ Expiration Date _____________________Signature_ _________________________________________________________________________Check to receive further information on:☐ EERI Associate membership ☐ Back issues of BSSA ☐ Corporate membership☐ Institutional subscriptions☐ Other: ____________________________Seismological Research Letters Volume 82, Number 6 November/December 2011 993

Permission to Reproduce Material from SRLThe Seismological Society of America (SSA) no longer requires that permission be obtainedfrom SSA or the author(s) for the use of tables, figures, or short extracts of papers publishedin SSA journals, provided that the source be appropriately and accurately cited. The author’spermission must be obtained for modifying the material in any way beyond simple redrawing,however.SSA recommends that the credit line read “authors, article title, journal title, volume number,page number(s), year, © Seismological Society of America.”SSA does require that the article be published by SSA before any material may be reproducedtherefrom.BACK ISSUES AVAILABLEBack issues of volumes 66 to date of SRL are still available. Limited quantities of issues of volumesbefore 66 are also available. Issues from volumes 66 to 72 are $10 each, except for 68:1 ($25 each)and 68:4 ($20 each). Issues from volumes 73 and 74 are $15 each, except for 73:5 and 74:5, whichare $25 each. Issues from volume 75 forward are $20 each. Available issues from volumes before66 are $15 each, except for issue 58:4, which is $25. Orders should be sent to the SeismologicalSociety of America, 201 Plaza Professional Building, El Cerrito, California 94530. For more information,visit http://www.seismosoc.org/publications/orders/srl-back-issue.php.RECOMMENDATION FOR MEMBERSHIPSEISMOLOGICAL SOCIETY OF AMERICAI suggest the following for membership in the Society.Name (please print or type)_ _________________________________________________Address_ ________________________________________________________________Name___________________________________________________________________Address_ ________________________________________________________________You (may) (may not) use my name in your invitation.Signed_ _________________________________________________________________Please print name__________________________________________________________Address_ ________________________________________________________________Return this form to:The Seismological Society of America201 Plaza Professional Bldg. • El Cerrito, CA 94530Phone +1-510-525-5474; Fax +1-510-525-7204http://www.seismosoc.org994 Seismological Research Letters Volume 82, Number 6 November/December 2011

Seismological Research Letters (SRL) is published bimonthly bythe Seismological Society of America. SRL is a unique journal—the first to serve as a general forum for informal communicationamong seismologists, as well as between seismologists and thosenonspecialists interested in seismology and related disciplines.The contents of SRL include contributed articles on topicsof broad seismological and earthquake engineering interest,opinion pieces on current seismological topics, news and notesabout seismology and seismologists in the U.S. and internationally,special earthquake reports, the Electronic Seismologistcolumn, strong-motion network reports, equipment news, andletters to the editor. A special section in SRL for the EasternSection of the Seismological Society of America focuses onSUBSCRIBE* TOSeismological Research Letterseastern North American seismology. Abstracts of papers to bepresented at the Annual Meeting of the Seismological Societyof America and the annual meeting of the SSA Eastern Sectionalso appear in SRL.The subscription rate for North American institutions is$150*; for countries outside North America the rate is $160*.Subscription rates include normal shipping; for priority maildelivery within the United States please include $30, outsidethe United States please include an additional payment of $35.Note: Subscribers to both SRL and our sister publicationBulletin of the Seismological Society of America (BSSA) are eligiblefor a combined discounted rate if both are ordered together.See below for combined rate information.☐ We wish to receive SRL for 2012 (vol. 83)Subscription within North America$150* per year _____Subscription outside North America$160* per year _____Additional for priority mail outside the United States$35 per year _____Additional for priority mail inside the United States$30 per year _____☐ We wish to receive a combined print subscriptions to both BSSA and SRL for 2012Subscription within North America$690* per year _____Subscription outside North America$710* per year _____Additional for priority mail outside the United States$110 per year _____Additional for priority mail inside the United States$65 per year _____☐ We wish to receive an electronic subscription to BSSA and print subscription SRL for 2012Subscription within North America$665* per year _____Subscription outside North America$675* per year _____Additional for priority mail outside the United States$35 per year _____Additional for priority mail inside the United States$30 per year _____☐ We wish to receive only the electronic subscription to BSSA for 2012Electronic BSSA only subscription$540* per year _____Total_____Payment Options:☐ Check drawn on U.S. bank in U.S. dollars payable to SSA☐ International Postal Money Order☐ Credit Card: ☐ Visa ☐ Mastercard ☐ American ExpressAccount Number _________________________________ Expiration Date _____________________Signature_ _________________________________________________________________________Name_____________________________________________________________________________Address_ __________________________________________________________________________Phone/Fax_________________________________________________________________________For institutions and other nonmembers.The Seismological Society of America201 Plaza Professional Bldg. • El Cerrito, CA 94530Phone +1-510-525-5474; Fax +1-510-525-7204http://www.seismosoc.orgSeismological Research Letters Volume 82, Number 6 November/December 2011 995

Last Page996 Seismological Research Letters Volume 82, Number 6 November/December 2011

Christa von Hillebrandt-Andrade, PresidentWilliam U. (Woody) Savage, Vice-PresidentSSA OFFICERSKeith L. Knudsen, SecretaryMitchell M. Withers, TreasurerRick Aster (2014)Laurie Baise (2014)Eric Calais (2014)Michael Campillo (2014)Steven M. Day (2012)Robert Graves (2012)SSA BOARD OF DIRECTORSChrista von Hillebrandt-Andrade (2013)Klaus-G. Hinzen (2013)Thomas H. Jordan (2013)Lisa Grant Ludwig (2013)William U. (Woody) Savage (2012)David Wald (2012)Janet Drysdale, PresidentMartin Chapman, Vice-PresidentChris Cramer, SecretaryEASTERN SECTION OF SSA BOARD OF DIRECTORSCharles Scharnberger, TreasurerDavid Eaton, Fifth MemberAndy Michael (Chair)Roland BurgmannSteven DayDavid WaldSSA PUBLICATIONS COMMITTEEEx Officio:Christa von Hillebrandt-AndradeDiane I. DoserJonathan M. LeesKim OlsenBechtel CorporationP.O. Box 193965San Francisco, California 94119Degenkolb EngineersNatalia Schoeck350 Sansome Street Suite 900San Francisco, California 94104ExxonMobil Upstream ResearchCompanyTechnical Information CenterURC-URC-SW104PO Box 2189Houston, Texas 77252-2189Fugro William Lettis &Associates, Inc.Robin K. McGuire4155 Darley Avenue, Suite ABoulder, Colorado 30305CORPORATE MEMBERS, SEISMOLOGICAL SOCIETY OF AMERICAGeotech InstrumentsAttn.: Lani Oncescu10755 Sanden Dr.Dallas, Texas 75238-1366Guralp Systems LimitedHorst Rademacher12 Southwood Dr.Orinda, California 94563Kinemetrics Inc.Ogie Kuraica222 Vista AvenuePasadena, California 91107Lighthouse R & D Enterprises, Inc.16945 Northchase DriveSuite 100Houston, Texas 77060Nanometrics, Inc.Neil Spriggs250 Herzberg RoadKanata, OntarioCanada K2K 2A1Pacific Gas & Electric Co.Lloyd S. Cluff, Geosciences Dept.245 Market Street, Room 403San Francisco, California 94105Puerto Rico Strong MotionProgramP.O. Box 9041Mayagüez, Puerto Rico 00681Refraction Technology1600 Tenth Street, Suite APlano, Texas 75074URS CorporationIvan G. WongSeismic Hazards Group1333 Broadway, Suite 800Oakland, California 94612

Seismological Society of America201 Plaza Professional BuildingEl Cerrito, CA 94530