GUIDE WAVE ANALYSIS AND FORECASTING - WMO

WORLD METEOROLOGICAL ORGANIZATIONGUIDETOWAVE ANALYSISAND FORECASTING1998(second edition)WMO-No. 702Secretariat of the World Meteorological Organization – Geneva – Switzerland1998

© 1998, World Meteorological OrganizationISBN 92-63-12702-6NOTEThe designations employed and the presentation of material in this publication do not imply theexpression of any opinion whatsoever on the part of the Secretariat of the World MeteorologicalOrganization concerning the legal status of any country, territory, city or area, or of its authorities orconcerning the delimitation of its fontiers or boundaries.

CONTENTSPageFOREWORD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .ACKNOWLEDGEMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .VVIVIIChapter 1 – AN INTRODUCTION TO OCEAN WAVES1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 The simple linear wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.3 Wave fields on the ocean . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6Chapter 2 – OCEAN SURFACE WINDS2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152.2 Sources of marine data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162.3 Large-scale meteorological factors affecting ocean surface winds . . . . . . . . . . . . . . . . . 212.4 A marine boundary-layer parameterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272.5 Statistical methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32Chapter 3 – WAVE GENERATION AND DECAY3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353.2 Wind-wave growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353.3 Wave propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363.4 Dissipation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393.5 Non-linear interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403.6 General notes on application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41Chapter 4 – WAVE FORECASTING BY MANUAL METHODS4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434.2 Some empirical working procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454.3 Computation of wind waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454.4 Computation of swell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474.5 Manual computation of shallow-water effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52Chapter 5 – INTRODUCTION TO NUMERICAL WAVE MODELLING5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 575.2 Basic concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 575.3 The wave energy-balance equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 585.4 Elements of wave modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 585.5 Model classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62Chapter 6 – OPERATIONAL WAVE MODELS6.1 Introductory remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 676.2 Wave charts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 696.3 Coded wave products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 696.4 Verification of wave models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 706.5 Wave model hindcasts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 756.6 New developments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

IVCONTENTSPageChapter 7 – WAVES IN SHALLOW WATER7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 817.2 Shoaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 817.3 Refraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 827.4 Diffraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 847.5 Wave growth in shallow waters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 857.6 Bottom friction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 867.7 Wave breaking in the surf zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 877.8 Currents, set-up and set-down . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88Chapter 8 – WAVE DATA, OBSERVED, MEASURED AND HINDCAST8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 898.2 Differences between visual and instrumental data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 898.3 Visual observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 898.4 Instruments for wave measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 918.5 Remote sensing from large distances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 938.6 Wave data in numerical models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 988.7 Analysis of wave records . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 988.8 Sources of wave data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100Chapter 9 – WAVE CLIMATE STATISTICS9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1019.2 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1019.3 Presentation of data and wave climate statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1039.4 Estimating return values of wave height . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1059.5 Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1099.6 Wave climatologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109ANNEXESI ABBREVIATIONS AND KEY TO SYMBOLS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119IIFM 65-IX WAVEOB – REPORT OF SPECTRAL WAVE INFORMATION FROMA SEA STATION OR FROM A REMOTE PLATFORM (AIRCRAFT ORSATELLITE) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123III PROBABILITY DISTRIBUTIONS FOR WAVE HEIGHTS . . . . . . . . . . . . . . . . . . . . . 133IV THE PNJ (PIERSON-NEUMANN-JAMES, 1955) WAVE GROWTH CURVES. . . . . . 139REFERENCES AND SELECTED BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143INDEX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

FOREWORDThe national Meteorological Services of a large number of maritime countries have, for many years now, been engagedin the provision of ocean wave forecast and hindcast services in support of the requirements of users in the whole rangeof maritime activities (shipping, fisheries, offshore mining, commerce, coastal engineering, construction, recreation, andso on). In recognition of this, and of the relative lack of easily accessible guidance material on wave forecasting methodologysuitable for use by national Meteorological Services in developing countries, the WMO Guide to wave analysis andforecasting was prepared by a group of experts and published in 1988 as publication WMO-No. 702. This formal WMOGuide updated and replaced the earlier, very popular, WMO Handbook on the same subject, first published in 1976.In further recognition, both of the requirements of national Meteorological Services for the provision of oceanwave-related services and also of the rapid developments which were occurring in wave measurement, analysis andforecast techniques, the WMO Commission for Marine Meteorology (CMM) established in 1984 a WMO waveprogramme. The various elements of this programme are implemented, reviewed and updated by the CMM Subgroupon Wave Modelling and Forecasting. One of these elements involves the continuous review and revision, asnecessary, of the Guide to wave analysis and forecasting. To this end, the sub-group established, in 1991, an ad hocgroup of experts, under the chairmanship of Dr A. K. Laing (New Zealand), to undertake a complete revision andupdating of the Guide, in the light of new developments and especially of feedback from users of the first edition.The international team of experts, directed by Dr Laing, individually prepared substantially revised versions ofthe different chapters of the Guide. These individual contributions were subsequently coordinated, assembled andedited by Dr Laing into a draft, which was then submitted to a wide network of wave experts for review and comment.Reviewers’ comments were incorporated to the extent possible and a final editing of a new, second edition of theGuide was made by Dr Laing.No publication such as this can ever be perfect, particularly in such a continuously developing field of scienceand technology, and further additions and modifications will undoubtedly be required in the future. Nevertheless, it isfirmly believed that this second edition of the Guide to wave analysis and forecasting will continue to prove a veryvaluable publication in support of the marine services provided by WMO’s maritime Members. It is also believed thatit continues to meet very well its two-fold objectives: to provide introductory but self-sufficient guidance material foruse in the provision of basic wave forecast services, while at the same time acting as a source text and a guide tofurther reading on the subject.Detailed acknowledgements to authors are given with each section and chapter as appropriate, but I shouldlike here, on behalf of the World Meteorological Organization, to express my sincere appreciation to all theexperts (authors, reviewers and particularly Dr Laing) who have contributed so much to this important and valuablepublication.(G. O. P. Obasi)Secretary-General

ACKNOWLEDGEMENTSThe revision of this Guide to wave analysis and forecasting has been very much a team effort, involving anumber of experts from several countries in various aspects of ocean waves. Overall direction of the project hasbeen undertaken by Dr Andrew Laing (New Zealand). The chapter editors/authors have had overall responsibilityfor the revision of each chapter and, in cases where substantial assistance has been given, co-authors areacknowledged.Much of the material is derived from the first edition of the Guide and the present editor would like toonce again acknowledge the efforts put into that publication by the editor E. Bouws, and the chapter editorsE. Bouws, L. Draper, A. K. Laing, D. J. T. Carter, L. Eide and J .A. Battjes.The present edition has been produced with specific attributions due to:• Overall direction and introduction:A. K. Laing, National Institute of Water and Atmospheric Research (NIWA), Wellington, New Zealand• Chapter 1:A. K. Laing, NIWA, Wellington, New Zealand• Chapter 2:W. Gemmill, National Meteorological Center (NMC), Washington D.C., USA• Chapter 3:A. K. Magnusson, Norwegian Meteorological Institute (DNMI), Bergen, NorwayM. Reistad, DNMI, Bergen, Norway• Chapter 4:L. Burroughs, National Meteorological Center (NMC), Washington D.C., USA• Chapter 5:M. Reistad, DNMI, Bergen, NorwayA. K. Magnusson, DNMI, Bergen, Norway• Chapter 6:M. Khandekar, Atmospheric Environment Service (AES), Toronto, Canada• Chapter 7:L. Holthuijsen, Delft Technical University, Delft, the Netherlands• Chapter 8:J. A. Ewing, United KingdomD. J. T. Carter, Satellite Observing Systems, Godalming, Surrey, United Kingdom• Chapter 9:D. J. T. Carter, Satellite Observing Systems, Godalming, Surrey, United KingdomV. Swail, Atmospheric Environment Service (AES), Toronto, CanadaIllustrations have been acknowledged in the captions. Otherwise they have been taken from the previousedition of the Guide or specially prepared for this edition.Contact with contributors can be made through the Ocean Affairs Division of the WMO Secretariat.The Editor would like to thank the Secretariat staff who assisted in the preparation of this publication.

INTRODUCTIONOverviewThe subject of this Guide is ocean waves, and specificallythose generated by the wind. Such waves affect ourcoasts and activities of all kinds which we pursue at thecoast, near the coast and out to sea. At any given timethe waves are a result of the recent history of winds overoften quite broad expanses of ocean. Indeed, knowledgeof the winds allows us to diagnose the wave conditions.Just as the winds vary considerably in time and space, sodo the waves. Hence, because there is some short-termpredictability in the winds, there is an associatedpredictability for wind waves, enabling operational forecaststo be made. On the other hand, longer termestimates of wave conditions must be based on climatologicalinformation for which measurements,observations or “synthetic” data are needed. Estimates oflikely future conditions or extremes may be required andthere are often limited data to work with, making thecorrect choice of technique critical.The objective of this Guide is to provide basicinformation and techniques in the analysis and forecastingof ocean waves. The Guide is not intended to be acomprehensive theoretical treatment of waves, nor doesit contain details of present research activity, rather itfocuses on providing a general overview with moredetail on aspects considered useful in the practice ofwave analysis and forecasting. The latest research intowave processes and modelling is given a thorough treatmentin the recently published book, Dynamics andmodelling of ocean waves (Komen et al., 1994), and anextensive treatment of problems related to wave data andtheir use is provided in Waves in ocean engineering:measurement, analysis, interpretation (Tucker, 1991).For problems specifically related to coastal engineering,a comprehensive reference is the Shore protectionmanual (CERC 1973, 1984).The primary users of this publication are seen asbeing professionals and technicians involved in operationswhich are affected by ocean waves, that is a fairlywide community of marine operators and those providingspecialized services to them. Marine weatherforecasters form a key group, but the Guide is equallyintended for the potential users of wave data analyses,forecasts and climatological products.This edition of the WMO Guide to wave analysisand forecasting replaces the 1988 edition of the samename, which in turn replaced the 1976 WMO Handbookon wave analysis and forecasting. In all editions it hasbeen recognized that the interpretation of wave data andproducts demands a good understanding of the processesby which these products are derived. An overview of theelementary theory is therefore necessary in this Guideand, wherever a technique or source of data is introduced,sufficient background information has beenincluded to make this publication as self-contained aspossible.Whilst many of the basics have remained consistentbetween editions, there are developments which arecreating new opportunities in wave information services.In recent years, an increasing number of wave productsincorporate data from satellites or are synthesized fromsimulations using numerical wave models. Indeed, manyforecasting centres are operationally using numericalmodels. Effective use of these products is only possibleif the forecasters and other users have sufficient knowledgeabout the physical background of wave modellingand satellite observations.Wave modelling was given prominence in the 1988Guide, an emphasis which has been retained in thisedition. The concerted international effort of the 1980sto develop physically realistic wave models has culminatedin a generation of wave models which have beenthoroughly researched, tested and are now becomingoperational tools. Wave modelling has “come of age”with this development. This does not mean that the problemis solved, far from it. For example, dissipation playsa critical role in the energy balance, but is poorly knownand is generally formulated to satisfy closure of theenergy-balance equation. Further, for computationalreasons, operational models must all use parameterizationsof the wave-wave interactions which control thedistribution of energy within the wave spectrum. Theproblems of wave evolution in shallow water and interactionwith surface currents also require continuingefforts.The present edition also includes a cataloguedescribing present day operational wave models and it isworth noting that regular updates of such models aregiven in WMO Marine Meteorology and RelatedOceanographic Activities Report No. 12 and its supplements(see WMO, 1994(a)). Further, wave modellinghas now been used extensively in synthesizing wave datafor climatological purposes. Hence, in the chapter onwave statistics specific treatment of wave climatologiesis presented, including the use of wave models inhindcasting.Another exciting development since the last editionhas been the increase in wave and wind data availablefrom satellites. Thus, new material on this source of data

VIIIGUIDE TO WAVE ANALYSIS AND FORECASTINGhas been included particularly in the chapters on windfields, wave data, climate and statistics. These data arealso now being used in conjunction with wave models toprovide improved initializations and to validate hindcastdata. Reference is given to some of the work beingcarried out.Notwithstanding such advances, it is recognizedthat many wave products are based on visual observationsand manual analyses and forecasts are still widelyused. Hence, material on manual methods has beenretained in the present Guide. These manual methods arelinked with the numerical by demonstrating their jointphysical basis.TerminologyAnalysis of waves is used in this Guide to refer to abroad range of procedures. The conventional meteorologicalcontext of an analysis involves data assimilationand this has had limited application in deriving waveproducts since wave data have been too scarce. That isuntil recently, when the large volume of data from satelliteshas brought to the fore the problem of assimilatingwave data into numerical models.In this Guide, wave analysis incorporates the proceduresfor estimating, calculating or diagnosing waveconditions. There is a direct relationship between windand waves which, in principle, is a matter of calculationby manual or automatic means. Sometimes this relationshipis quite simple and can be represented by a tableshowing the wave height for a given wind speed (andpossibly wind direction). In many cases, however, amore complicated approach is required, depending on(a) the amount of detail that is needed, i.e. informationon wave periods, wave steepness, etc.; (b) the environmentalconditions of the forecasting area, including thegeometry of coastlines, bathymetry and currents; and (c)the nature of the wind which can sometimes be quitevariable, i.e. with changes taking place before a stationarywave condition has been reached.Forecasting has a meaning here which is slightlydifferent from that which is common in meteorology.When issuing a wave forecast, one can forecast the propagationof wave energy, but the evolution (growth) of thewave energy is dependent on the wind and so a majorpart of the procedure is actually referring to the forecastof the winds that cause the waves. The wave growth is infact diagnosed from the forecast wind.Hindcasting, on the other hand, refers to the diagnosisof wave information based on historical wind data.A computation based on present wind data is commonlyreferred to as wave analysis. The term nowcast isincreasingly used in meteorology in a similar context.A number of conventions have been adopted in theformulations used in the Guide. In particular, vectorquantities are written in bold italic type (e.g. a) to distinguishthem from scalar quantities (e.g. a). A glossary ofvariables, symbols and acronyms is included as Annex I.Climatological issuesWhatever the main objective of the wave analysis orforecast, an appreciation of wave climatology is essential.The aim of some readers may be to derive the waveclimate and estimate extremes, and for others to produceoperational wave forecasts. For the latter it must benoted that verification of an estimate of wave conditionsis often not possible due to the lack of wave observations.Hence, as with all computations, sensible resultsare more likely if the local wave climate is known,particularly the ranges and likelihood of various waveparameters (for example, height and period) in the sea orocean area of interest. Furthermore, experience shouldgive a feeling for the probable values which might occurunder the given wind conditions. It is of great importance,therefore, that any investigation, including trainingfor wave forecasting, starts with a detailed study of thegeography and climatology of the area of interest, inorder to appreciate the limitations of wind fetch forcertain wind directions, the existence of strong oceancurrents, the typical configuration of wind fields in theweather patterns prevailing over the area and theclimatological probability of wind speeds and directions.It is useful to know the range of wave heights andperiods which may occur at sea in general: individualwaves higher than 20 m are very rare — the highestwave reported and checked being 33 m (North Pacific,1933) — which implies that characteristic or significantwave heights will rarely exceed 10 m; characteristicwave periods usually vary between four and 15 secondsand are seldom greater than 20 seconds. Furthermore,one should be aware that waves can travel long distancesand still retain appreciable height and energy; forinstance, waves generated in mid-latitude storms in theNorth Atlantic Ocean have been observed as swell in theSouth Atlantic, and certain atolls in the equatorialPacific have been damaged by swell waves which musthave travelled several thousands of kilometres.If regular wave forecasting is required for a fixedposition or area (e.g. in support of coastal or offshoreengineering or other marine operations such as shiploading) it is preferable to arrange for regular wavemeasurements at suitable points. This provides data forverification of wave forecasts or validation of models forhindcasting. In a few instances it may even be possibleto develop a sufficiently homogeneous set of measureddata for determining the wave climate from statisticalanalysis. In many applications, the only way to obtain asatisfactory data set is to hindcast the waves for a sufficientlylong time period, using wind fields derived fromhistorical weather charts or archived air-pressure datafrom atmospheric models. However, the current availabilityof upwards of 10 years of satellite altimeter dataglobally allows wave height climatology to be accuratelydescribed, at least in areas of the world’s oceans notaffected by tropical storms, down to the spatial resolutionof the satellite data.

XGUIDE TO WAVE ANALYSIS AND FORECASTINGconsiderable attention. Annex III, which is related to thischapter, contains a list of various distribution functionsand their properties. It has been noted that knowledge ofthe wave climate is important in operational forecastingas well as in the estimation of extremes. Hence, thesecond part of Chapter 9 is devoted to the derivation ofwave climatologies, from hindcasts as well as wave data.In respect to hindcasts we still face the problem ofgenerating reliable historical wind fields, particularlyfrom the time before the introduction of fine-meshmodels in the early 1970s. The chapter includes tablesidentifying sources of major climatologies, both frommeasured and observed data and from hindcast wavedata.It is acknowledged that material has been includedthat is dated and “perishable”. It is inevitable thatinformation on the status of data acquisition, modelsbeing used, products being delivered, studies in progressand other activities, which was correct at the time ofwriting, is going to change. However, inclusion of theseitems is necessary in a publication of this type.This publication is intended to provide guidance insolving day-to-day problems. In some cases, however,the problems concerned go beyond the scope of thisGuide. If the reader wishes to pursue particular topicsthen sufficient references are included to provide a gatewayto the open literature. The editor of the relevantchapter may also be contacted; the names and affiliationsare listed in the “Acknowledgments”. Further, the Guideis not intended to be a self-contained training course. Inparticular, specialist training courses are recommendedfor marine forecasters who are required to make waveforecasts. For example, the interpretation of the guidancewhich the output from a particular wave model providesfor a particular national forecast service goes wellbeyond the scope of this Guide.

CHAPTER 1AN INTRODUCTION TO OCEAN WAVESA. K. Laing: editor1.1 IntroductionOcean surface waves are the result of forces acting onthe ocean. The predominant natural forces are pressureor stress from the atmosphere (especially through thewinds), earthquakes, gravity of the Earth and celestialbodies (the Moon and Sun), the Coriolis force (due tothe Earth’s rotation) and surface tension. The characteristicsof the waves depend on the controlling forces.Tidal waves are generated by the response to gravity ofthe Moon and Sun and are rather large-scale waves.Capillary waves, at the other end of the scale, are dominatedby surface tension in the water. Where the Earth’sgravity and the buoyancy of the water are the majordetermining factors we have the so-called gravity waves.Waves may be characterized by their period. This isthe time taken by successive wave crests to pass a fixedpoint. The type and scale of forces acting to create thewave are usually reflected in the period. Figure 1.1 illustratessuch a classification of waves.On large scales, the ordinary tides are ever presentbut predictable. Less predictable are tsunamis (generatedby earthquakes or land movements), which can be catastrophic,and storm surges. The latter are associated withthe movement of synoptic or meso-scale atmosphericfeatures and may cause coastal flooding.Wind-generated gravity waves are almost alwayspresent at sea. These waves are generated by windssomewhere on the ocean, be it locally or thousands ofkilometres away. They affect a wide range of activitiessuch as shipping, fishing, recreation, coastal and offshoreindustry, coastal management (defences) and pollutioncontrol. They are also very important in the climateprocesses as they play a large role in exchanges of heat,energy, gases and particles between the oceans andatmosphere. It is these waves which will be our subjectin this Guide.To analyse and predict such waves we need to havea model for them, that is we need to have a theory forhow they behave. If we observe the ocean surface wenote that the waves often form a rather complex pattern.To begin we will seek a simple starting model, which isconsistent with the known dynamics of the oceansurface, and from this we will derive a more completepicture of the wind waves we observe.The model of the ocean which we use to developthis picture is based on a few quite simple assumptions:• The incompressibility of the water. This means thatthe density is constant and hence we can derive acontinuity equation for the fluid, expressing theconservation of fluid within a small cell of water(called a water particle);• The inviscid nature of the water. This means thatthe only forces acting on a water particle are gravityand pressure (which acts perpendicular to thesurface of the water particle). Friction is ignored;• The fluid flow is irrotational. This means that theindividual particles do not rotate. They may movearound each other, but there is no twisting action.This allows us to relate the motions of neighbouringparticles by defining a scalar quantity, called thevelocity potential, for the fluid. The fluid velocity isdetermined from spatial variations of this quantity.From these assumptions some equations may be written todescribe the motion of the fluid. This Guide will not presentthe derivation which can be found in most textbookson waves or fluids (see for example Crapper, 1984).1.2 The simple linear waveThe simplest wave motion may be represented by a sinusoidal,long-crested, progressive wave. The sinusoidaldescriptor means that the wave repeats itself and has thesmooth form of the sine curve as shown in Figure 1.2. Thelong-crested descriptor says that the wave is a series ofRelative energyCapillarywaves0.1 sGravitycapillarywavesOrdinarygravitywavesInfra-gravitywaves,wave groupsLong-periodwavesSeiches,storm surges,tsunamisTrans-tidalwavesOrdinarytidalwavesSunandMoon1 s 30 s 5 min 12 h 24 hWave periodFigure 1.1 —Classification ofocean waves bywave period(derived from Munk,1951)

2GUIDE TO WAVE ANALYSIS AND FORECASTINGηCrestZero levela H = 2aaTroughFigure 1.2 — A simple sinusoidal wavelong and parallel wave crests which are all equal in height,and equidistant from each other. The progressive nature isseen in their moving at a constant speed in a directionperpendicular to the crests and without change of form.1.2.1 Basic definitions*The wavelength, λ, is the horizontal distance (in metres)between two successive crests.The period, T, is the time interval (in seconds)between the passage of successive crests passed a fixedpoint.The frequency, f, is the number of crests which passa fixed point in 1 second. It is usually measured innumbers per second (Hertz) and is the same as 1/T.The amplitude, a, is the magnitude of the maximumdisplacement from mean sea-level. This is usually indicatedin metres (or feet).The wave height, H, is the difference in surfaceelevation between the wave crest and the previous wavetrough. For a simple sinusoidal wave H = 2a.The rate of propagation, c, is the speed at which thewave profile travels, i.e. the speed at which the crest andtrough of the wave advance. It is commonly referred toas wave speed or phase speed.The steepness of a wave is the ratio of the height tothe length (H/λ).1.2.2 Basic relationshipsFor all types of truly periodic progressive waves one canwrite:λ = cT , (1.1)i.e. the wavelength of a periodic wave is equal to theproduct of the wave speed (or phase speed) and theperiod of the wave. This formula is easy to understand.Let, at a given moment, the first of two successive crestsarrive at a fixed observational point, then one period later(i.e. T seconds later) the second crest will arrive at thesame point. In the meantime, the first crest has covered adistance c times T.The wave profile has the form of a sinusoidal wave:η(x,t) = a sin (kx – ωt) . (1.2)In Equation 1.2, k = 2π/λ is the wavenumber andω =2π/T, the angular frequency. The wavenumber is a_______* See also key to symbols in Annex I.Tλx,tcyclic measure of the number of crests per unit distanceand the angular frequency the number of radians persecond. One wave cycle is a complete revolution whichis 2π radians.Equation 1.2 contains both time (t) and space (x)coordinates. It represents the view as may be seen froman aircraft, describing both the change in time and thevariations from one point to another. It is the simplestsolution to the equations of motion for gravity wavemotion on a fluid, i.e. linear surface waves.The wave speed c in Equation 1.1 can be written asλ/T or, now that we have defined ω and k, as ω/k. Thevariation of wave speed with wavelength is calleddispersion and the functional relationship is called thedispersion relation. The relation follows from the equationsof motion and, for deep water, can be expressed interms of frequency and wavelength or, as it is usuallywritten, between ω and k:ω 2 = gk , (1.3)where g is gravitational acceleration, so that the wavespeed is:λ ω gc = = = .T k kIf we consider a snapshot at time t = 0, the horizontalaxis is then x and the wave profile is “frozen” as:η(x) = a sin (kx) .However, the same profile is obtained when the wavemotion is measured by means of a wave recorder placedat the position x = 0. The profile then recorded isη(t) = a sin (– ωt) . (1.4)Equation 1.4 describes the motion of, for instance, amoored float bobbing up and down as a wave passes by.The important parameters when wave forecastingor carrying out measurements for stationary objects,such as offshore installations, are therefore wave height,wave period (or wave frequency) and wave direction. Anobserver required to give a visual estimate will not beable to fix any zero level as in Figure 1.2 and cannottherefore measure the amplitude of the wave. Instead,the vertical distance between the crest and the precedingtrough, i.e. the wave height, is reported.In reality, the simple sinusoidal waves describedabove are never found at sea; only swell, passing throughan area with no wind, may come close. The reason forstarting with a description of simple waves is that theyrepresent the basic solutions of the physical equationswhich govern waves on the sea surface and, as we shallsee later, they are the “building blocks” of the real wavefields occurring at sea. In fact, the concept of simplesinusoidal waves is frequently used as an aid to understandingand describing waves on the sea surface. Inspite of this simplified description, the definitions andformulae derived from it are extensively used in practiceand have proved their worth.

AN INTRODUCTION TO OCEAN WAVES 31.2.3 Orbital motion of water particlesIt is quite evident that water particles move up and downas waves travel through water. By carefully watchingsmall floating objects, it can be seen that the water alsomoves backwards and forwards; it moves forward on thecrest of a wave and backward in the trough. If the wateris not too shallow relative to the wavelength, thedisplacements are approximately as large in the horizontalas in the vertical plane. In fact, during one cycle of asimple wave (i.e. a wave period) the particles describe acircle in a vertical plane. The vertical plane is the crosssectionwhich we have drawn in Figure 1.2. In shallowwater the motion is an ellipse. Figure 1.3 illustrates thisparticle motion for a simple sinusoidal wave in deepwater.Consider the speed at which a water particlecompletes its path. The circumference of the circle isequal to πH. This circumference is covered by a particlewithin a time equal to one period T. The speed of thewater is therefore πH/T. This is also the greatest forwardspeed reached in the crests. The speed of individualwater particles should not be confused with the speed atwhich the wave profile propagates (wave speed). Thepropagation rate of the wave profile is usually fargreater, as it is given by λ/T, and the wavelength λ isgenerally much greater than πH.Figure 1.3 has been slightly simplified to show theprogression of wave crests and troughs as the result ofwater particle motion. In reality, depending on the wave1234steepness, a water particle does not return exactly to thestarting point of its path; it ends up at a slightly advancedposition in the direction in which the waves are travelling(Figure 1.4). In other words, the return movement inthe wave trough is slightly less than the forward movementin the wave crest, so that a small net forward shiftremains. This difference increases in steep waves (seeSection 1.2.7).1.2.4 Energy in wavesWe have noted that waves are associated with motion inthe water. Therefore as a wave disturbs the water there iskinetic energy present which is associated with the wave,and which moves along with the wave. Waves alsodisplace particles in the vertical and so affect the potentialenergy of the water column. This energy also movesalong with the wave. It is an interesting feature of thewaves that the total energy is equally divided betweenkinetic energy and potential energy. This is called theequipartition of energy.It is also important to note that the energy does notmove at the same speed as the wave, the phase speed. Itmoves with the speed of groups of waves rather thanindividual waves. The concept of a group velocity willbe discussed in Section 1.3.2 but it is worth noting herethat, in deep water, group velocity is half the phasespeed.The total energy of a simple linear wave can beshown to be ρ w ga 2 /2 which is the same as ρ w gH 2 /8,where ρ w is the density of water. This is the total of theFigure 1.3 —Progression of a wavemotion. Thirteen snapshotseach with an interval of1/12th period (derived fromGröen and Dorrestein, 1976)5678Figure 1.4 —Path shift of a water particleduring two wave periods910111213

4GUIDE TO WAVE ANALYSIS AND FORECASTINGFigure 1.5 — Paths of the water particles at various depths ina wave on deep water. Each circle is one-ninthof a wavelength below the one immediatelyabove itpotential and kinetic energies of all particles in the watercolumn for one wavelength.1.2.5 Influence of water depthAs a wave propagates, the water is disturbed so thatboth the surface and the deeper water under a wave arein motion. The water particles also describe verticalcircles, which become progressively smaller withincreasing depth (Figure 1.5). In fact the decrease isexponential.Below a depth corresponding to half a wavelength,the displacements of the water particles in deep water areless than 4 per cent of those at the surface. The result isthat, as long as the actual depth of the water is greaterthan the value corresponding to λ/2, the influence of thebottom on the movement of water particles can beconsidered negligible. Thus, the water is called deepwith respect to a given surface wave when its depth is atleast half the wavelength.In practice it is common to take the transition fromdeep to transitional depth water at h = λ/4. In deep water,the displacements at this depth are about 20 per cent ofthose at the surface. However, so long as the water isdeeper than λ/4, the surface wave is not appreciablydeformed and its speed is very close to the speed ondeep water. The following terms are used to characterizethe ratio between depth (h) and wavelength (λ):• Deep water h > λ/4;• Transitional depth λ/25 < h < λ/4;• Shallow water h < λ/25.Note that wave dissipation due to interactions with thebottom (friction, percolation, sediment motion) is not yettaken into account here.When waves propagate into shallow water, forexample when approaching a coast, nearly all thecharacteristics of the waves change as they begin to“feel” the bottom. Only the period remains constant. Thewave speed decreases with decreasing depth. From therelation λ = cT we see that this means that the wavelengthalso decreases.From linearized theory of wave motion, an expressionrelating wave speed, c, to wavenumber, k = 2π/λ,and water depth, h, can be derived:c2 g= tanh kh ,(1.5)kwhere g is the acceleration of gravity, and tanh x denotesthe hyperbolic tangent:x – xe – etanh =x – x.e + eThe dispersion relation for finite-depth water ismuch like Equation 1.5. In terms of the angularfrequency and wavenumber, we have then the generalizedform for Equation 1.3:ω 2 = gk tanh kh .(1.3a)In deep water (h > λ/4), tanh kh approaches unityand c is greatest. Equation 1.5 then reduces to2 g gc = = λ (1.6)k 2πor, when using λ = cT (Equation 1.1)T = 2πλ(1.7)gandλ = gT 2(1.8)2πandgT g gc = = =(1.9)2π 2πfω .Expressed in units of metres and seconds (m/s), theterm g/2π is about equal to 1.56 m/s 2 . In this case, onecan write λ = 1.56 T 2 m and c = 1.56 T m/s. When, onthe other hand, c is given in knots, λ in feet and T inseconds, these formulae become c = 3.03 T knots andλ = 5.12 T 2 feet.When the relative water depth becomes shallow(h < λ/25), Equation 1.6 can be simplified toc= gh.(1.10)This relation is of importance when dealing with longperiod,long-wavelength waves, often referred to as longwaves. When such waves travel in shallow water, thewave speed depends only on water depth. This relationcan be used, for example, for tsunamis for which theentire ocean can be considered as shallow.If a wave is travelling in water with transitionaldepths (λ/25 < h < λ/4), approximate formulae can be usedfor the wave speed and wavelength in shallow water:c = c 0 tanh k 0 h,(1.11)λ = λ 0 tanh kh 0 , (1.12)with c 0 and λ 0 the deep-water wave speed and wavelengthaccording to Equations 1.6 and 1.8, and k 0 thedeep-water wavenumber 2π/λ 0 .

AN INTRODUCTION TO OCEAN WAVES 5xxxBeach linexxxIsobathsxxxxxxWave crestsxxxBreakersxxxFigure 1.6 — Refraction along a straight beach with parallelbottom contoursBeach lineIsobathsBeach lineIsobathsBayIsobathsOrthogonalsOrthogonalsHeadlandFigure 1.7(a) —Refraction by asubmarine ridgeFigure 1.7(b) —Refraction by asubmarine canyonFigure 1.8 — Refraction along an irregular shorelineBayOrthogonalsA further feature of changing depth is changingwave height. As a wave approaches the shore its heightincreases. This is a result of the changes in groupvelocity. The energy propagating towards the coast mustbe conserved, at least until friction becomes appreciable,so that if the group velocity decreases and wavelengthdecreases, the energy in each wavelength must increase.From the expression for energy in Section 1.2.4 we seethat this means that the height of the wave must increase.1.2.6 Refraction and diffractionAs waves begin to feel the bottom, a phenomenon calledrefraction may occur. When waves enter water of transitionaldepth, if they are not travelling perpendicular tothe depth contours, the part of the wave in deeper watermoves more rapidly than the part in shallower water,according to Equation 1.11, causing the crest to turnparallel to the bottom contours. Some examples ofrefraction patterns are shown in Figures 1.6, 1.7 and 1.8.Generally, any change in the wave speed, forinstance due to gradients of surface currents, may leadto refraction, irrespective of the water depth. In Section4.5.1 a few examples are given to illustrate refractionunder simplified conditions. A more complete descriptionof methods for the analysis of refraction anddiffraction can be found in Sections 7.3 and 7.4, and inCERC (1984).Finally, the phenomenon of wave diffraction shouldbe mentioned. It most commonly occurs in the lee ofobstructions such as breakwaters. The obstruction causesenergy to be transformed along a wave crest. This transferof energy means that waves can affect the water inthe lee of the structure, although their heights are muchreduced. An example is illustrated in the photograph inFigure 1.9.1.2.7 Breaking wavesIn Section 1.2.3, it was noted that the speed of the waterparticles is slightly greater in the upper segment of theorbit than in the lower part. This effect is greatly magnifiedin very steep waves, so much so that the maximumforward speed may become not πH/T but 7H/T. Should7H become equal to the wavelength λ (i.e. H/λ = 1/7),the forward speed of the water in the crest would then beequal to the rate of propagation which is λ/T. There canbe no greater forward speed of the water because thewater would then plunge forward out of the wave: inother words, the wave would break.According to a theory of Stokes, waves cannotattain a height of more than one-seventh of the wavelengthwithout breaking. In reality, the steepness ofwaves is seldom greater than one-tenth. However, atvalues of that magnitude, the profile of the wave haslong ceased to be a simple undulating line and looksmore like a trochoid (Figure 1.10). According to Stokes’theory, at the limiting steepness of one-seventh, theforward and backward slopes of the wave meet in thecrest under an angle of 120° (Figure 1.11).

6GUIDE TO WAVE ANALYSIS AND FORECASTINGFigure 1.9 — Wave diffraction at Channel Islands harbour breakwater (California) (from CERC, 1977)Mean levelFigure 1.10 —Trochoidal wave profile. Here thecrests project farther above the meanlevel than the troughs sink under it120 °71Figure 1.11 —Ultimate form which water wavescan attain according to Stokes’theoryWhen waves propagate into shallow water theircharacteristics change as they begin to feel the bottom,as we have already noted in Section 1.2.5. The waveperiod remains constant, but the speed decreases as doesthe wavelength. When the water depth becomes less thanhalf the wavelength, there is an initial slight decrease inwave height*. The original height is regained when theratio h/λ is about 0.06 and thereafter the height increasesrapidly, as does the wave steepness, until breaking pointis reached:h b = 1.28 H b (1.13)_________* Tracking a wave into shallow water, the wavelengthdecreases and the wave slows down but, initially, itsenergy does not. The energy then spreads over relativelymore waves and the height reduces This is only temporary.The wave energy soon also slows and the heightbegins to increase.in which h b is termed the breaking depth and H b thebreaker wave height.1.3 Wave fields on the ocean1.3.1 A composition of simple wavesActual sea waves do not look as simple as the profileshown in Figure 1.2. With their irregular shapes, theyappear as a confused and constantly changing watersurface, since waves are continually being overtaken andcrossed by others. As a result, waves at sea are oftenshort-crested. This is particularly true for waves growingunder the influence of the wind (wind sea).A more regular pattern of long-crested and nearlysinusoidal waves can be observed when the waves areno longer under the influence of their generating winds.Such waves are called swell and they can travelhundreds and thousands of kilometres after having leftthe area in which they were generated. Swell fromdistant generating areas often mixes with wind wavesgenerated locally.

AN INTRODUCTION TO OCEAN WAVES 710–110IFigure 1.12 —The upper profile is equalto the sum (superposition)of two simple waves, I andII, shown in the lower partof the figure. The horizontaldimensions are greatlyshortened with respect tothe vertical ones–1IIThe simple waves, which have been described inSection 1.2, can be shown to combine to compose theobserved patterns. To put it differently, any observedwave pattern on the ocean can be shown to comprise anumber of simple waves, which differ from each other inheight, wavelength and direction.As a first step, let us consider waves with long,parallel crests but which differ in height, for example,the profile as shown in the top curve of Figure 1.12.Although this curve looks fairly regular, it is certainlyno longer the profile of a simple sinusoidal wave,because the height is not everywhere the same, nor arethe horizontal distances between crests. This profile,however, can be represented as the sum of two simplewave profiles of slightly different wavelength (see I andII in Figure 1.12). In adding the vertical deviations of Iand II at corresponding points of the horizontal axis, weobtain the vertical deviations of the sum of wave I andwave II, represented by the top wave profile in Figure1.12.Thus, the top profile can be broken down, ordecomposed, into two simple waves of different wavelength.The reason why the crests are of varying heightin the sum of I and II is that at one place waves I and IIare “in phase” and their heights therefore add up,whereas the resulting height is reduced at those placeswhere the waves are out of phase.Taking this idea one step further, we can see how anirregular pattern of wind waves can be thought of as asuperposition of an infinite number of sinusoidal waves,propagating independently of each other. This is illustratedin Figure 1.13, which shows a great number ofsinusoidal waves piled up on top of each other. Think,for example, of a sheet of corrugated iron as representinga set of simple sinusoidal waves on the surface of theocean and caught at an instant in time. Below this, thereis another set of simple sinusoidal waves travelling in aslightly different direction from the one on top. Belowthat again is a third one and a fourth one, etc. — all withdifferent directions and wavelengths. Each set is aclassic example of simple sinusoidal waves.It can be shown that, as the number of differentsinusoidal waves in the sum is made larger and largerand the heights are made smaller and smaller, and theperiods and directions are packed closer and closertogether (but never the same and always over a considerablerange of values), the result is a sea surface just likethe one actually observed. Even small irregularities fromthe sinusoidal shape can be represented by superpositionsof simple waves.Figure 1.13 — The sea surface obtained from the sum ofmany sinusoidal waves (derived from Pierson,Neumann and James, 1955)

81.3.2 Wave groups and group velocityWe have seen how waves on the ocean are combinationsof simple waves. In an irregular sea the number of differingwavelengths may be quite large. Even in regularswell, there are many different wavelengths present butthey tend to be grouped together. In Figure 1.12 we seehow simple waves with close wavelengths combine toform groups of waves. This phenomenon is common.Anyone who has carefully observed the waves of the seawill have noticed that in nature also the larger wavestend to come in groups.Although various crests in a group are neverequidistant, one may speak of an average distance andthus of an average wavelength. Despite the fact thatindividual crests or wave tops advance at a speed correspondingto their wavelength, the group, as a coherentunit, advances at its own velocity — the group velocity.For deep water this has magnitude (group speed):cc(1.14)= 2GUIDE TO WAVE ANALYSIS AND FORECASTINGg .A more general expression also valid in finite depthwater is:c 2khcg = ( 1 + ) (1.15)kh .2 sinh 2The general form for the group speed can be shownto be:dωc g =d k .Derivations may be found in most fluid dynamics texts(e.g. Crapper, 1984).We can also show that the group velocity is thevelocity at which wave energy moves. If we consider theenergy flow (flux) due to a wave train, the kinetic energyis associated with the movement of water particles innearly closed orbits and is not significantly propagated.The potential energy however is associated with the netdisplacement of water particles and this moves alongwith the wave at the phase speed. Hence, in deep water,the effect is as if half of the energy moves at the phasespeed, which is the same as the overall energy moving athalf the phase speed. The integrity of the wave is maintainedby a continuous balancing act between kinetic andpotential energy. As a wave moves into previously undisturbedwater potential energy at the front of the wavetrain is converted into kinetic energy resulting in a lossof amplitude. This leads to waves dying out as theyoutrun their energy. At the rear of the wave train kineticenergy is left behind and is converted into potentialenergy with the result that new waves grow there.One classical example of a wave group is the bandof ripples which expands outwards from the disturbancecreated when a stone is cast into a still pond. If you fixyour attention on a particular wave crest then you willnotice that your wave creeps towards the outside of theband of ripples and disappears. Stating this slightlydifferently, if we move along with waves at the phasevelocity we will stay with a wave crest, but the wavesahead of us gradually disappear. Since the band ofripples is made up of waves with components from anarrow range of wavelengths the wavelength of our wavewill also increase a little (and there will be fewer wavesimmediately around us). However, if we travel at thegroup velocity, the waves ahead of us may lengthen andthose behind us shorten but the total number of wavesnear us will be conserved.Thus, the wave groups can be considered as carriersof the wave energy (see also Section 1.3.7), and thegroup velocity is also the velocity with which the waveenergy is propagated. This is an important considerationin wave modelling.1.3.3 Statistical description of wave recordsThe rather confusing pattern seen in Figure 1.13 can alsobe viewed in terms of Equation 1.4 as the motion of thewater surface at a fixed point. A typical wave record forthis displacement is shown in Figure 1.14, in which thevertical scale is expressed in metres and the horizontalscale in seconds. Wave crests are indicated with dashesand all zero-downcrossings are circled. The wave periodT is the time distance between two consecutive downcrossings(or upcrossings*), whereas the wave height His the vertical distance from a trough to the next crest asit appears on the wave record. Another and morecommonly used kind of wave height is the zero-crossingwave height H z , being the vertical distance between thehighest and the lowest value of the wave record betweentwo zero-downcrossings (or upcrossings). When thewave record contains a great variety of wave periods, thenumber of crests becomes greater than the number ofzero-downcrossings. In that case, there will be somedifference between the crest-to-trough wave height andH z . In this chapter, however, this difference will beneglected and H z will be used implicitly. A simple andcommonly used method for analysing wave records byhand is the Tucker-Draper method which gives goodapproximate results (see Section 8.7.2).A measured wave record never repeats itselfexactly, due to the random appearance of the sea surface.But if the sea state is “stationary”, the statistical propertiesof the distribution of periods and heights will besimilar from one record to another. The most appropriateparameters to describe the sea state from a measuredwave record are therefore statistical. The following arefrequently used:_________* There is no clear convention on the use of either zeroupcrossingsor downcrossings for determining the waveheight and period of zero-crossing waves. Generally, if therecord in sufficiently long, no measurable differences willbe found among mean values.

AN INTRODUCTION TO OCEAN WAVES 9Figure 1.14 —Sample of awave record(dashes showwave crests; zerodown-crossingsare circled)H max10 s 1 mH – = Average wave height;H max = Maximum wave height occurring in aT – z =record;Average zero-crossing wave period; the timeobtained by dividing the record length byH – 1/n =the number of downcrossings (or upcrossings)in the record;The average height of the 1/n highest waves(i.e. if all wave heights measured from therecord are arranged in descending order, theone-nth part, containing the highest waves,should be taken and H – 1/n is then computed asthe average height of this part);T – H 1/n= The average period of the 1/n highest waves.A commonly used value for n is 3:H – 1/3 = Significant wave height (its value roughlyapproximates to visually observed waveT – H 1/3=height);Significant wave period (approximatelyequal to the wave period associated with thespectral maximum, see Section 1.3.8).1.3.4 Duration of wave recordsThe optimal duration of wave records is determined byseveral factors. First of all, for a correct description of thesea state, conditions should be statistically stationaryduring the sampling period. In fact, this will never beachieved completely as wave fields are usually evolving(i.e. growing or decaying). On the other hand, to reducestatistical scatter, the wave record should contain at least200 zero-downcrossing waves. Hence, the optimal timeover which waves are usually measured is 15–35 minutes,as this reasonably accommodates both conditions.So far, we have introduced the manual analysis ofanalogue “stripchart” records. Most analyses areperformed by computer for which digital records areused, i.e. the vertical displacement of the ocean surface(or the position of the pen at the chart recorder) is givenwith a sampling rate of 1–10 times per second (1–10Hz). For example, a record of 20 minutes duration with asampling rate of 4 Hz contains 4 800 values.When wave records are processed automatically,the analysis is always preceded by checks on the qualityof the recorded data points to remove outliers and errorsdue to faulty operation of sensors, in data recordingequipment, or in data transmission.1.3.5 Notes on usage of statistical parametersIn this Guide, the term sea state is used as a wave conditionwhich is described by a number of statisticalparameters. It is common to use the significant waveheight, H – 1/3, and the average zero-crossing period, T – z, orsome other characteristic period, to define the sea state.The corresponding maximum wave height can bededuced as shown in Section 1.3.6.The use of the average zero-crossing period, T – z, hasits drawbacks. The distribution of individual zerodowncrossingperiods of a record is usually very wide andis also somewhat sensitive to noise, in contrast with thedistribution of periods of, say, the highest one-third ofwaves. Moreover, the average period of the highest wavesof a record is usually a good approximation of the periodassociated with the peak of the wave spectrum (seeSection 1.3.8). It has been found that average wave periodsof the 1/n highest waves with n > 3 are not essentiallydifferent from T – H 1/n, but exhibit a larger scatter.In this Guide, as elsewhere, various definitions ofwave steepness are used. The general form is ξ = H/λwhich becomes, using Equation 1.8:2πHξ =gT , 2where H represents a wave height (e.g. H – 1/3, H m0 , H rms ,√m 0 ) and T the wave period (e.g. T – z, T – H 1/3, T p , T m02 ).Some of these parameters are introduced in Section1.3.8.1.3.6 Distribution of wave heightsThe elevation of the sea surface is denoted η(x,t). Thisformulation expresses the variations of sea surface inspace and time for both simple waves (see Equation 1.2)and more complicated sea states. If the range of

10GUIDE TO WAVE ANALYSIS AND FORECASTINGwavelengths in a given sea state is not too broad, it hasbeen shown (Longuet-Higgins, 1952) that the elevation,η, has a statistical distribution which is Gaussian (i.e.normal).For a normally distributed parameter, such as η, themaximum values are known to be distributed with aRayleigh distribution. For a sea state these maximumvalues are directly related to the wave heights. Hence,the distribution of (zero-downcrossing) wave heights canbe represented by the Rayleigh distribution. This featurehas been shown theoretically and verified empirically. IfF(H) denotes the probability of heights not exceeding agiven wave height H 1 in a sea state characterized by aknown value of H – 1/3, F(H) is given by:F(H 1 ) = 1 – exp [– 2 (H 1 /H – 1/3) 2 ] . (1.16)The probability Q(H 1 ) of heights exceeding H 1 is thenQ(H 1 ) = 1 – F(H 1 ) . (1.17)Example:Given a sea state for which H – 1/3 = 5 m,what is the probability of observing waves higherthan 6 m?Since F(H 1 ) = 1 – exp [– 2 (6/5) 2 ] = 0.94,the probability of heights exceeding 6 m isQ(H 1 ) = 1 – 0.94 = 0.06.If H – 1/3 is computed from a wave record of finitelength, the record length or the number of waves usedfor the computation should be taken into account. If, ona record containing N waves, n (n ≤ N) waves exceed agiven height H 1 , the probability of heights exceeding H 1is:nQH ( 1) = .(1.18)NInserting the relationships from Equations 1.16 and1.17 into Equation 1.18 leads to:H = H ( 05 . ln N .n )1/3 1– 05 .(1.19)Equation 1.19 provides a quick method for thedetermination of H – 1/3 from a wave record. On the otherhand, if H – 1/3 is known, the distribution of a wave recordcan be compared with the Rayleigh distribution byusing:H 1 = H 1/ 3 05 . ln N . (1.19a)nFor the prediction of the maximum wave heightH max for a sequence of N waves with known H – 1/3, it iscommon to take the mode of the distribution of maximumvalues:H (1.20)max = H 1/ 3 05 . lnN.Alternatively, if the 50-percentile of the distribution ofmaximum values is taken, we get a more conservativeestimate of H max because of the asymmetry of the distribution,i.e. about 5 per cent greater than according toEquation 1.20:H max = H / 05 . ln 145 . N.(1.21)The prediction of H max must be based on a realisticduration, e.g. six hours, apart from the usual confidencelimits of the H – 1/3 forecast. This implies N = 2 000–5 000(in 6 hours there are about 2 700 waves if the peakperiod is 8 seconds). Using Equation 1.20 we get*:H max ≅ 2.0 H – 1/3 ≅ 1.9 H m0 .1.3.7 The wave spectrumWe have noted in Section 1.3.1 that a sea surface with arandom appearance may be regarded as the sum of manysimple wave trains. A way of formalizing this concept isto introduce the wave spectrum. A wave record may bedecomposed by means of harmonic (or Fourier) analysisinto a large number of sinusoidal waves of differentfrequencies, directions, amplitudes and phases. Eachfrequency and direction describes a wave component,and each component has an associated amplitude andphase.The harmonic (Fourier) analysis thus provides anapproximation to the irregular but quasi-periodic form ofa wave record as the sum of sinusoidal curves. For asurface elevation varying in time in a single direction:n∑η() t = η + a sin (jω t+φ )01 3j=1in which:η(t) = recorded elevation of the water surface attime t;η 0 = mean elevation (as shown for instance inFigure 1.14);ω 0 = angular wave frequency of the longest wavefitted to the record;j = number of wave component;a j = amplitude of the jth component;φ j = phase angle of the jth component;n = total number of components.The phase angle allows for the fact that the componentsare not all in phase, i.e. their maxima generally occur atdifferent times. The high frequency components tend tobecome insignificant and hence there is a reasonablelimit to n.Each wave component travels at its own speed(which depends on the wave frequency — or period —as expressed in Equation 1.10). Hence, the spectrum ofwave components is continuously changing across thesea surface as the low frequency (large period or longwavelength) components travel faster than the highfrequency components.The expected values of the squares of the amplitudesa j are the contribution to the variance of the surfaceelevation (η) from each of the wave components (i.e. the_________* See Section 1.3.8 for definition of H m0 and its relationwith H – 1/3.j0j

AN INTRODUCTION TO OCEAN WAVES 11variance is E [Σ j a j 2 ]). The resulting function is known asthe wave-variance spectrum S(f)*. Typical spectra ofwave systems have a form as shown in Figure 1.15where the squared amplitudes for each component areplotted against their corresponding frequencies. Thefigure shows the spectrum from a measured wave record,along with a sample of the data from which it was calculated† . On the horizontal axis, the wave components arerepresented by their frequencies (i.e. 0.1 Hertz (Hz)corresponds to a period of ten seconds).In actual practice, wave spectra can be computed bydifferent methods. The most commonly used algorithm isthe fast-Fourier transform (FFT), developed by Cooleyand Tukey (1965). A much slower method, now supersededby the FFT, is the auto-correlation approachaccording to the Wiener-Kinchine theorem, introducedfor practical use by Blackman and Tukey (1959) (see alsoBendat and Piersol, 1971). Experience has shown thatthe difference between spectra computed by any twomethods does not exceed confidence limits of each ofthem.Since the wave energy E equals ρ w gH 2 /8 orρ w ga 2 /2 (H = 2a), wave spectra in the earlier literaturewere expressed in terms of E and called wave-energyspectra. However, it has become common practice todrop the term ρ w g and to plot a 2 /2 or, simply, a 2 alongthe vertical axis. The wave-energy spectrum isthus usually regarded synonymously with the “variancespectrum”.Wave spectra are usually given as a continuouscurve connecting the discrete points found from theFourier analysis and systems typically have a generalform like that shown in Figure 1.16. The curve may notalways be so regular. Irregular seas give rise to broadspectra which may show several peaks. These may beclearly separated from each other or merged into a verybroad curve with several humps. Swell will generallygive a very narrow spectrum concentrating the energy ina narrow range of frequencies (or wavelengths) around apeak value. Such a narrow spectrum is associated withthe relatively “clean” appearance of the waves. Recallfrom Section 1.3.2 (and Figure 1.12) that this was oftena condition where wave groups were clearly visible.It is important to note that most measurements donot provide information about the wave direction andtherefore we can only calculate an “energy” distributionover wave frequencies, E(f). On the vertical axis, ameasure for the wave energy is plotted in units of m 2 /Hz.This unit is usual for “frequency spectra”. We have seenEnergy density (m 2 /Hz)201020 0 0 021 metre1 minute2 2 220 0.1 0.2 0.3Frequency (Hz)Figure 1.15 — Example of a wave spectrum with the correspondingwave record (12 November 1973,21 UTC, 53°25'N, 4°13'E, water depth 25 m,wave height 4.0 m, wave period 6.5 s, Westwind 38 kn (19.6 m/s) (derived from KNMI)earlier that, although the spectrum may be continuous intheory, in practice the variances (or energies) arecomputed for discrete frequencies. Even when a highspeedcomputer is used, it is necessary to regard thefrequency domain (or the frequency-directional domain)as a set of distinct or discrete values. The value of a 2 at,Figure 1.16 — Typical wave-variance spectrum for a singlesystem of wind waves. By transformation ofthe vertical axis into units of ρ w gS(f), a waveenergyE(f) spectrum is obtained22_________* The variance of a wave record is obtained by averaging thesquares of the deviations of the water surface elevation, η,from its mean η 0 . In Section 1.3.8, this variance is relatedto the area, m 0 , under the spectral curve.†This example shows a case with pure wind sea. However,the spectrum may often have a more complicated appearance,with one or more peaks due to swell.Variance (density)Frequency

12GUIDE TO WAVE ANALYSIS AND FORECASTINGfor instance, a frequency of 0.16 Hz is considered to be amean value in an interval which could be 0.155 to0.165 Hz. This value, divided by the width of the interval,is a measure for the energy density and expressed inunits of m 2 /Hz (again omitting the factor ρ w g). In factthe wave spectrum is often referred to as the energydensityspectrum.Thus, this method of analysing wave measurementsyields a distribution of the energy of the variouswave components, E(f,θ). It was noted in Section 1.3.2that wave energy travels at the group velocity c g , andfrom Equation 1.15 we see that this is a function ofboth frequency and direction (or the wavenumbervector) and possibly water depth. The energy in eachspectral component therefore propagates at the associatedgroup velocity. Hence it is possible to deduce howwave energy in the local wave field disperses across theocean.It is important to note that a wave record and thespectrum derived from it are only samples of the seastate (see Section 1.3.4). As with all statistical estimates,we must be interested in how good our estimate is,and how well it is likely to indicate the true state. Thereis a reasonably complete statistical theory to describethis. We will not give details in this Guide but refer theinterested reader to a text such as Jenkins and Watts(1968). Suffice to say that the validity of a spectralestimate depends to a large extent on the length of therecord, which in turn depends on the consistency of thesea state or statistical stationarity (i.e. not too rapidlyevolving). The spectral estimates can be shown to havethe statistical distribution called a χ 2 distribution forwhich the expected spread of estimates is measured bya number called the “degrees of freedom”. The largerthe degrees of freedom, the better the estimate is likelyto be.1.3.8 Wave parameters derived from thespectrumA wave spectrum is the distribution of wave energy (orvariance of the sea surface) over frequency (or wavelengthor frequency and direction, etc.). Thus, as astatistical distribution, many of the parameters derivedfrom the spectrum parallel similar parameters from anystatistical distribution. Hence, the form of a wave spectrumis usually expressed in terms of the moments* ofthe distribution (spectrum). The nth-order moment, m n ,of the spectrum is defined by:nm n = ∫ ∞f E( f ) df (1.22)0(sometimes ω = 2πf is preferred to f). In this formula,E(f) denotes the variance density at frequency, f, as inFigure 1.16, so that E(f) df represents the variance a 2 i /2contained in the ith interval between f and f + df. In practice,the integration in Equation 1.22 is approximated bya finite sum, with f i = i df:Namnfi n2i= ∑ . (1.22a)i = 0 2From the definition of m n it follows that themoment of zero-order, m 0 , represents the area under thespectral curve. In finite form this is:which is the total variance of the wave record obtainedby the sum of the variances of the individual spectralcomponents. The area under the spectral curve thereforehas a physical meaning which is used in practical applicationsfor the definition of wave-height parametersderived from the spectrum. Recalling that for a simplewave (Section 1.2.4) the wave energy (per unit area), E,was related to the wave height by:then, if one replaces the actual sea state by a single sinusoidalwave having the same energy, its equivalent heightwould be given by:8EHrms= ,ρ gthe so-called root-mean-square wave height. E nowrepresents the total energy (per unit area) of the seastate.We would like a parameter derived from the spectrumand corresponding as closely as possible to thesignificant wave height H – 1/3 (as derived directly fromthe wave record) and, equally, the characteristic waveheight H c (as observed visually). It has been shown thatH rms should be multiplied by the factor √2 in order toarrive at the required value. Thus, the spectral waveheight parameter commonly used can be calculatedfrom the measured area, m 0 , under the spectral curve asfollows:Hm0Na a= ∑i = ,2 2i = 0E = 1 ρ w gH ,88E= 2 = 4m .m0 0ρ w gNote that we sometimes refer to the total varianceof the sea state (m 0 ) as the total energy, but we must bemindful here that the total energy E is really ρ w gm 0 . Intheory, the correspondence between H m0 and H – 1/3 is validonly for very narrow spectra which do not occur often innature. However, the difference is relatively small in2w22_________* The first moment of a distribution of N observations X 1 , X 2 , …, X n is defined as the average of the deviations x 1 , x 2 , …, x nfrom the given value X 0 . The second moment is the average of the squares of the deviations about X 0 ; the third moment is theaverage of the cubes of the deviations, and so forth. When X 0 is the mean of all observations, the first moment is obviouslyzero, the second moment is then known as the “variance” of X and its square root is termed the “standard deviation”.

AN INTRODUCTION TO OCEAN WAVES 13most cases, with H m0 = 1.05 H – 1/3 on average. Thesignificant wave height is also frequently denoted by H s .In that case, it must be indicated which quantity (4√m 0or H – 1/3) is being used.The derivation of parameters for wave period is amore complicated matter, owing to the great variety ofspectral shapes related to various combinations of seaand swell. There is some similarity with the problemabout defining a wave period from statistical analysis(see Section 1.3.5). Spectral wave frequency and waveperiod parameters commonly used are:f p = Wave frequency corresponding to the peakof the spectrum (modal or peak frequency);T p = Wave period corresponding to f p : i.e.T p = f –1 p ;T m01 = Wave period corresponding to the meanfrequency of the spectrum:Tm01m0= ;m1(1.24)T m02 = Wave period theoretically equivalent withmean zero-downcrossing period T – z:Tm02=m0;m2(1.25)T m–10 = Energy wave period, so-called for its role incomputing wave power, J, in wave energystudies:where J, the wave power in kW/m of wave front, iscomputed as J = 0.49 H 2 m0 T m–10 .Note that the wave period T m02 is sensitive to thehigh frequency cut-off in the integration (Equation 1.22)which is used in practice. Therefore this cut-off shouldbe noted when presenting T m02 and, in particular, whencomparing different data sets. For buoy data, the cut-offfrequency is typically 0.5 Hz as most buoys do notaccurately measure the wave spectrum above thisfrequency. Fitting a high frequency tail before computingthe spectral moments can be a useful conveniencewhen high frequency information is not available (forexample in a wave model hindcast).Goda (1978) has shown that, for a variety of cases,average wave periods of the higher waves in a record,e.g. T – H 1/3(see Section 1.3.5), remain within a range of0.87 T p to 0.98 T p .Finally, the width of the spectral peak can be usedas a measure of the irregularity of the sea state. Thespectral width parameter ε is defined by:ε =Tm– 10m=mm0m4– mm m0 4– 1;0Parameter ε varies between 0 (very narrow spectrum;regular waves) and 1 (very broad spectrum; many differentwave periods present; irregular wave pattern).22.The use of ε is not recommended, however,because of its sensitivity to noise in the wave record dueto the higher order moments, in particular m 4 . Rye(1977) has shown that the peakedness parameter, Q p , byGoda (1970) is a good alternative:2 ∞ 2Q p = f S( f ) df .2 ∫(1.26)m00Q p = 1 corresponds with ε = 1, while Q p becomes verylarge for very narrow spectra. Under natural conditions,Q p usually remains within the interval 1.5–5.1.3.9 Model forms for wave spectraThe concept of a wave spectrum is commonly used formodelling the sea state. Models of the spectrum enablethe spectrum to be expressed as some functional form,usually in terms of frequency, E(f), frequency and direction,E(f,θ), or alternatively in terms of the wavenumber,E(k). Since the wavenumber and frequency are relatedby the dispersion relation (see Equations 1.3 and 1.3a),the frequency and wavenumber forms can be transformedfrom one to the other.Models of the spectrum are used to obtain an estimateof the entire wave spectrum from known values of alimited number of parameters such as the significantwave height and wave period. These may be obtained byhindcast calculations, by direct measurement or visualobservation. To give an idea of the various factors whichneed to be taken into account, a few models are givenbelow as examples. In the first three models no bottomeffects have been taken into account. The TMA spectrum(see p. 14) is proposed as a general form for amodel spectrum in depth-limited waters. In all cases E isused to represent the variance density spectrum.The Phillips’ spectrum describes the shape particularlyof the high frequency part of the spectrum, abovethe spectral peak. It recognizes that the logarithm of thespectrum is generally close to a straight line, with a slopethat is about –5. Hence, the general form:E( f ) = 0.005 g2f 5 ,iff ≥ g u= 0 elsewhere.(1.27)The Pierson-Moskowitz spectrum (Pierson andMoskowitz, 1964) is often used as a model spectrum fora fully developed sea, an idealized equilibrium statereached when duration and fetch are unlimited. Thisspectrum is based on a subset of 420 selected wavemeasurements recorded with the shipborne waverecorder — developed by Tucker (1956) — on boardBritish ocean weather ships during the five-year period1955–1960. In its original form, this model spectrum is:E( f ) =α g24(2π)5fe⎛ g ⎞– 0.74⎜⎟⎝ 2πuf ⎠4(1.28)

14GUIDE TO WAVE ANALYSIS AND FORECASTINGin which E(f) is the variance density (in m 2 /s), f the wavefrequency (Hz), u the wind speed (m/s) at 19.5 m abovethe sea surface, g the acceleration due to gravity (m/s 2 )and α a dimensionless quantity, α = 0.0081.It can be shown that the peak frequency of thePierson-Moskowitz spectrum is:f p = 0.877g(1.29)2πu .Equation 1.28, together with Equations 1.22 and1.23, allows us to calculate m 0 as a function of windspeed. Hence H m0 (the significant wave height) for afully grown sea is:H m0 = 0.0246 u 2 , (1.30)with H m0 in metres and u in m/s, with the wind speednow related to 10 m height*. This agrees well with limitingvalues of wave-growth curves in Chapter 4 (e.g.Figure 4.1). Equations 1.29 and 1.30 are valid for fullydeveloped sea only, as is their combination:H m0 = 0.04 f p –2 . (1.31)The JONSWAP spectrum is often used to describewaves in a growing phase. Observations made during theJoint North Sea Wave Project (JONSWAP) (Hasselmannet al., 1973) gave a description of wave spectra growingin fetch-limited conditions, i.e. where wave growthunder a steady offshore wind was limited by the distancefrom the shore. The basic form of the spectrum is interms of the peak frequency rather than the wind speed,i.e as in Equation 1.28 but after the substitution forg/(2πu) using Equation 1.29:E( f ) =α g24(2π)5f(1.32)The function γ is the peak enhancement factor, whichmodifies the interval around the spectral peak making itmuch sharper than in the Pierson-Moskowitz spectrum.Otherwise, the shape is similar. The general form of theJONSWAP spectrum is illustrated in Figure 1.17.Using JONSWAP results, Hasselmann et al. (1976)proposed a relation between wave variance and peakfrequency for a wide range of growth stages. Transformingtheir results into terms of H m0 and f p we get:H m0 = 0.0414 f –2 p (f p u) 1/3 , (1.33)again with H m0 in m, f p in Hz and u in m/s at 10 m abovemean water level.e4⎛ f ⎞–1.25⎜⎝ f⎟ p ⎠γ ( f ).3E(f)/E(f p )10JONSWAP peakenhancementPierson-Moskowitzspectral shape1 2Figure 1.17 — General form of a JONSWAP spectrum as afunction of f/f pThis equation is connected with developing wavesand so is not exactly comparable with Equation 1.30 forfully developed waves. The peak frequency can beobtained by reversing Equation 1.33:f p = 0.148 H m0–0.6u 0.2 . (1.34)Equation 1.34 can be applied for estimating theapproximate spectrum and characteristic wave periodswhen wave height and wind speed are known. This iscommon practice when predicting waves using growthcurves relating wave height to wind speed and fetch orduration.The TMA (Texel-Marsen-Arsloe) spectrum, proposedas a model in depth-limited waters, takes the form:E(f) = E JONSWAP (f) Φ(f,h) , (1.35)where Φ is a function of frequency, f, and depth, h (seeBouws et al., 1985).Finally, it should be noted that the spectra shownhere are all of the typeE(f) = E(f, parameters),with no account taken of the directional distribution ofthe sea state. Further information on directionality canbe found in Section 3.3.f/f p_________* The usual reference height for wind speeds is 10 m. Wind speed at 19.5 m is reduced to 10 m height by applying a correctionfactor; in this case the wind speed has been divided by 1.075 (see Section 2.4.1).

CHAPTER 2OCEAN SURFACE WINDSW. Gemmill: editor2.1 IntroductionIn order to be able to make good wave forecasts, it isessential that accurate surface wind analyses and forecastsare available. A detailed analysis of the wind fieldand of its development from the past into the future isrequired.The effects of erroneous wind speeds are cumulativein time, and the effect on forecast wave heightsmay be appreciable. For example, the wave-growthdiagram in Figure 3.1 shows that a wind of 15 m/s (29knots) is capable of raising a sea of 4 m height after12 h and of 5 m height after 24 h. If the wind speed wasassumed to be 17.5 m/s (34 knots), the forecast wouldindicate wave heights of 4.9 m and 6.3 m, respectively.An initial error of 16 per cent in the wind speed wouldthus give rise to errors of 25–30 per cent in the forecastwave heights.If, in the above example, the “forecast” was in factan analysis of the current situation and the analysis wasused as a basis for forecasts of future wave conditions,the error would tend to become still greater and errors inwave height of 2 m or even more would be the result.Wave-height differences of this magnitude are generallyof great importance to marine operations and, for thisreason, the wave forecaster must be aware of the sourcesof errors in his forecast and he should also be able tomake an assessment of their magnitude.Ocean surface wind fields can be determinedfrom:• Manual analysis of the marine meteorologicalobservations with extrapolation in time;• Automated numerical objective analysis andprediction models; or• A combination of both — the “man-machine mix”.The physical processes that determine the oceansurface wind field are contained in the basic equations ofmotion. But, these equations are mathematically verycomplex and can only be solved by using numericalmethods on high-speed, large-scale computers. Thisapproach is commonly referred to as weather predictionby numerical methods, pioneered originally by L. F.Richardson (1922). A hierarchy of numerical weatherprediction models are currently used by various countries.For a description of the basic principles seeHaltiner and Williams (1980), for example. For operationalrequirements it would, of course, be preferable touse the automated procedures of a numerical modelsince they provide the ocean surface winds in a quickand efficient manner.When numerical weather prediction models areemployed, the ocean surface wind field can be obtainedfrom the lowest layer of the analysis and forecast modelsystem. However, additional diagnostic calculations maybe necessary to adjust the model winds to obtain the windat a pre-assigned height over the ocean surface (10 or20 m). The closer the lowest level of the model is to theocean, the smaller is the potential for introducing errorswhen extrapolating from that level to the near surface.If one does not have access to a large computersystem with a real-time observational database, it willbe necessary to perform a manual analysis of marinevariables to construct the wind fields with the aid ofsome simple dynamical relationships. These can bederived from the more complex and complete set ofequations of motion. Although time consuming,manual analyses can produce more detailed wind fieldsthan models. For hindcast studies, where accuracy isparticularly important and more time may be available,the combined use of models and manual interventionprovides perhaps the best wind fields (see alsoSection 9.6.2). Statistical techniques can also be used toinfer surface winds from selected model outputs. Thisapproach is particularly useful for site specific areaswhere local effects can be important in determining thewind field.The basic force that drives ocean waves is thesurface stress imparted by the wind. However, thatparameter is not measured directly. It is estimated byknowing the wind at a specified height (10 or 20 m) andapplying an appropriate value for drag coefficient. Thedrag coefficient is dependent on height and stability (airminus sea temperature difference) of the atmosphereabove the ocean surface. In unstable conditions the sea iswarmer than the overlying air and there is more turbulentmixing in the lower atmosphere. This increases the stresson the surface, so for the same wind speed at a givenheight, waves will become higher under unstable conditionsthan under stable conditions. Hence, characterizingfriction and the stability in the lower boundary layer isan important step in deriving winds or surface stressessuitable for use in wave estimation.In the following sections, a brief discussion on thesources of marine data is first presented (2.2). This isfollowed by a discussion on the methods to obtain theocean surface wind field using simple dynamical relations(between pressure gradient, Coriolis and frictionalforces, 2.3), with particular attention to characterizingthe marine boundary layer (2.4), and statisticalapproaches (2.5).

16GUIDE TO WAVE ANALYSIS AND FORECASTING2.1.1 Wind and pressure analyses — generalconsiderationsThe wind field is usually analysed by indirect means,starting with a surface pressure analysis in mid- andhigher latitudes, or a streamline analysis in low latitudes.In routine analyses of weather charts, the analyst usesthe latest available analysis from which to obtain a firstguess. An isobaric pattern is drawn using pressure observations,and hopefully some satellite imagery. Theobservations of wind speed are used as a check on thepressure gradient, and those of wind direction on theorientation of the isobars. Both pressure and wind directionare used for streamline analysis (see the end of thissection for more discussion on wind analysis in thetropics). At best the analyses are coarse because ofsparseness of the surface data.It takes time to carefully examine and fit the pressureand wind data in the final analysis. Sanders (1990)has shown that producing a good subjective analysis ofpressure, even when sufficient surface pressure data areavailable, is a time consuming activity and requires carefulediting for quality control. Then, to prepare the windfield analysis, an additional step is required. Wind observationsare used to specify the wind speed and directionat observation locations. However, over large areas ofthe ocean where no (or very few) observations are available,one must resort to determining wind speed anddirection from parameters which can be obtained fromweather charts (i.e. pressure gradients, curvature ofisobars, and air-sea temperature differences). The correspondingwind speeds and directions are analysed in sucha way that a consistent pattern emerges which shows thecharacteristic features of wind fields associated withweather disturbances over the ocean.Thus, determining the wind field requires an analysisof a weather chart in two consecutive steps. First, astandard analysis of the isobaric pattern is performed todetermine the main locations of weather systems. Thenfollows a close examination of the exact position of theisobars, with a view to adjusting their position at placeswhere that is needed to arrive at a logical and consistentwind-isotach field. In such reviews, one soon finds that,even in areas where the network of ship observations iscomparatively dense, local adjustments to the isobarspacing can be made which may easily result ingradient-wind correction of about 2 m/s. Without observationsthe potential for error is much higher. Hence, themain errors in determining the wind field result from alack of pressure and wind observations over the oceans.Today, operational meteorologists generally usenumerical forecast guidance as the starting point toproduce ocean wind charts. Figure 2.1(a) shows asample 48-hour forecast extracted from the NCEP(National Center for Environment Prediction, USA)global forecast model of 10 m surface winds and sealevelpressures for the north-west Atlantic Ocean. Theforecaster studies the movement and development ofweather systems through the forecast period (out to 72hours at 12-hour increments), as well as the consistencyof the forecasts from run cycle to run cycle. Many forecastcentres also maintain statistics on their own forecastmodel performance (i.e. the accuracy of movement andintensity of weather systems) so that this informationcan be used to improve the forecasts. Meteorologists caneither accept the guidance as is, or use the additionalinformation to modify the forecasts. Figure 2.1(b) showsthe final interpretation for the ocean surface weatherforecast chart using the information in Figure 2.1(a). Ofcourse, the work is all done within restrictive deadlines.The forecast chart is “broadcast” in “real time” to themarine community through various distribution systemsincluding marine radio facsimile and private companies.In the tropics it is not possible to determine thewind field directly from the pressure analyses. This isbecause the geostrophic relationship is weaker in thelower latitudes and breaks down completely at theEquator. Also, errors in pressure measurements canbecome significant compared to the pressure gradientswhich are to be analysed. Direct analysis of the wind inthe form of streamlines and isotachs gives a usefuldepiction of the low-level wind field.The procedure for streamline analysis is similar topressure analysis in that all weather systems haveconsistency over time and they need to be located andtracked from chart to chart. Each system should befollowed from genesis, to maturity and decay, and itsmovement tracked. A knowledge of conceptual modelsof weather systems is required to carry this out so thatstreamlines and isotachs can be correctly analysed wherethere are few observations. It is particularly importantfor the analyst to monitor the evolution of the anticyclonesin the subtropical ridge as sudden increases inintensity can result in surges of wind penetrating wellinto the tropics, with resultant increases in wave andswell height. A useful reference for analysis in thetropics is the Forecasters guide to tropical meteorology(AWS, 1995).2.2 Sources of marine dataThere are three sources of surface observations that aregenerally used to make analyses of pressure and windsover the ocean. These are observations taken routinely atsix-hour synoptic intervals by ships, ocean buoys (fixedand drifting) and land (coastal) weather stations. Theyare disseminated worldwide via the Global TelecommunicationSystem (GTS) to be available in realtime for use in operational centres. In addition, remotemeasurements of winds using active and passivemicrowave sensors on board satellites are now availableand used by the operational meteorological world.2.2.1 Ship weather reportsToday, ship weather reports provide the standard sourceof marine data. They are prepared by deck officers aspart of their routine duties. Wind speed and direction are

OCEAN SURFACE WINDS 17Figure 2.1(a) — Numerical model forecast from the NCEP global forecast model. The output shows isobars of surface pressure(hPa) and wind barbs for 10 m winds (knots (kn), 1.94 kn = 1 m/s) for the 48-hour forecast valid at 00 UTC on21 December 1995estimated either indirectly, by the observer using seastate and feel of the wind, or directly by anemometer ifthe vessel is so equipped.Estimated wind observations are subject to a widevariety of errors. Such reports are often made by firstdetermining the wind speed in terms of the Beaufortscale, where each scale number represents a range ofpossible wind speeds. Then, a single speed is chosen forreporting purposes. The scale is based on the appearanceof the sea state. However, a substantial time lag mayoccur for the sea to reach a state that truly reflects theprevailing wind conditions. In addition, it is obvious thatnight-time wind reports based on visual sea state will besubject to great error.

18GUIDE TO WAVE ANALYSIS AND FORECASTINGFigure 2.1(b) — Surface weather 48-hour forecast based on information in Figure 2.1(a). Valid at 00 UTC, 21 December 1995(from National Meteorological Center, Marine Forecast Branch, US Department of Commerce/NOAA/NWS)

OCEAN SURFACE WINDS 19The descriptions of wind effects on the sea surfacewhich are used by mariners for observing the variousintensities of the wind have not been chosen arbitrarily.The criteria are the result of long experience and representindividually distinguishable steps on a specificscale. Originally, these steps were characterized bydescriptive terms, such as light breeze and moderatebreeze, or gale and strong gale. When, in the lastcentury, weather reporting became an operational practice,the steps were given numbers, and the wind scalewas named after Admiral Beaufort who introduced thiscoding system in 1814 in the British navy. TheBeaufort scale was recommended for international usein 1874.Since it is possible for an experienced observer todistinguish one step from another, a wind observationshould be accurate to within a half of one scale interval— a fact which has indeed been verified empirically.However, wind effects on the sea surface are sometimesmodified by other phenomena, of which the observermay be unaware. The plankton content of the sea waterhas an influence on foam forming, air stability affectsthe steepness of waves to some extent, while strongcurrents may also change the form of waves and hencethe general appearance of the sea state. As a result, thestandard deviation of a wind observation is, on the average,somewhat greater than a half of one scale interval: itamounts to 0.58 I (I being the width of the scale interval)for each of the steps 1–10 of the Beaufort scale. Thismeans that the standard deviation of an individualwind-speed observation varies from 0.76 m/s at step 1(wind speed 2.0 m/s) to 1.34 m/s at step 5 (wind speed10.2 m/s) and 2.6 m/s at step 10 (wind speed 24.2 m/s)(Verploegh, 1967).It should be stressed that wave height is not a criterionused in wind observations at sea. Any mariner isall too well aware that wave height depends on otherfactors as well — such as the duration and fetch of thewind. The influence of fetch, for instance, is clearlynoticeable when the ship is in the lee of land, or whenthere are abrupt wind changes in open seas.Many studies have been made to determine windspeeds equivalent to the steps of the Beaufort scale(WMO, 1970) and such studies are continuing at present(Lindau, 1994). A scale recommended for scientific useis given in Table 2.1. It shows the equivalent wind speedfor each Beaufort number and equivalent scale intervalsin metres/second (m/s) and in knots.Since this scientific scale has not been introducedfor the operational reporting of wind speeds from ships,ships’ observers use an older conversion table which wasintroduced internationally in 1948. This table (usingknots) is given in Table 2.2 (WMO, 1990). For thepurpose of wave forecasting, therefore, a wind report(expressed in knots) should be converted back into theoriginal Beaufort scale number and then converted intothe correct wind speed of the scientific scale. Thisslightly complicates the work, but many years ofBeaufortTABLE 2.1Conversion scales for Beaufort wind forceDescriptiveScale recommended for usein wave forecastingnumber term Equivalent Intervalswind speed(m/s) m/s kn0 Calm 0.8 1 or 0 0–21 Light air 2.0 2 3–52 Light breeze 3.6 3–4 6–83 Gentle breeze 5.6 5–6 9–124 Moderate breeze 7.8 7–9 13–165 Fresh breeze 10.2 9–11 17–216 Strong breeze 12.6 12–14 22–267 Near gale 15.1 14–16 27–318 Gale 17.8 17–19 32–379 Strong gale 20.8 19–22 38–4310 Storm 24.2 23–26 44–5011 Violent storm 28.0 26–30 51–5712 Hurricane - 31 and 58 andabove abovewave-forecasting practices have shown that importantsystematic errors are avoided in this way.The wind direction is much easier to determinefrom the orientation of the crests of wind waves. Thestandard deviation of an individual observation of winddirection amounts to 10° (Verploegh, 1967) and appearsto be independent of wind speed. The direction ofconstant trade or monsoon winds can be determined withgreater accuracy.The mix of wind observations between visualestimates and anemometers varies considerably fromone ocean area to another. Anemometer winds areprevalent in the Pacific whereas most observations in theAtlantic are visual estimates. Surprisingly, a study byTABLE 2.2Scales in use in international reportsBeaufort Descriptive Range of values reportednumber term by observers (kn)0 Calm

20GUIDE TO WAVE ANALYSIS AND FORECASTINGFigure 2.2(a) (right) —Usual way of computingthe true wind (3) from adiagram plot of head wind(1) and measured (apparent)wind speed anddirection (2)Figure 2.2(b) (far right) —The computed true windmay be erroneous both asregards direction andspeedHeadwind(1)(1)(2)Apparent wind (2)True wind (3)Directionalerrors ofcomputedtrue wind(3)Errorellipseof measured(apparent)wind vectorWilkerson and Earle (1990) shows that the quality ofwind reports from ships with anemometers is not muchbetter than those without. Pierson (1990) also reportedon the characteristics of ship wind observations madewith and without anemometers and concluded that shipreports, with or without anemometers, were inferior tobuoy measurement. Errors from anemometer measurementson board ships are generally introduced by poorinstrument exposure, improper reading of wind speedand direction indicators, vessel motion and maintenanceproblems.Winds measured by an anemometer on board shipwill be the combined effect of the wind over the oceanand the ship’s motion. Therefore, in order to obtain abest estimate of the prevailing (true) wind over the sea,the measured wind must be corrected to exclude theship’s movement. This is done by means of a vectordiagram as illustrated in Figure 2.2(a).Whilst an anemometer may give an accuraterepresentation of the air flow at the location of theanemometer, the main problem with measuring wind onboard ship is that a fixed anemometer cannot always beproperly exposed to all wind directions. The vessel’ssuperstructure may interfere with the flow of air, andconsequently the measurement may not be representativeof the true air flow over the ocean surface. It shouldbe obvious that errors in the apparent wind, and/or theincorrect application of the vector diagram will producepoor quality wind reports, as illustrated in Figure 2.2(b).On larger ships, anemometers are usually installedat great heights: heights of 40 m above the surface of thesea are by no means uncommon. The wind speednormally increases with height, but the rate of speedincrease depends on stability of the air. However, routineobservations are not corrected for height, and that is yetanother source of error when comparing wind data frommany platforms. For a more detailed review of problemsof wind measurements at sea, see Dobson (1982) andalso Taylor et al. (1994).2.2.2 Fixed buoy reportsSince 1967, moored buoys equipped with meteorologicalinstruments have provided surface atmosphericand oceanographic data. Buoys can be expected toprovide better quality data than those reported by shipsfor several reasons:• The sensor’s location on the buoy is carefullyconsidered to avoid exposure problems;• Sampling and averaging periods for the measurementsare determined after accounting for buoymotion;• Duplicate sensors are used for redundancy and eachis calibrated before deployment;• Monitoring in near-real time allows detection ofinstrument errors. (This is the practice in the USAfor all data gathered by buoys deployed by theNational Data Buoy Center.)Buoys are presently providing measurements whichare within the original accuracy specification.The National Data Buoy Center’s specified accuracyrequires that winds, averaged for 8.5 minutes, shouldhave a root-mean-square error for speed of less than1.0 m/s (or 10 per cent) and for direction of less than10°. This has been verified through field calibrationstudies (Gilhousen, 1987).2.2.3 Land (coastal) stationsLand (coastal) reporting stations provide data of variablequality and applicability. In some cases they mayprovide reliable data (NOAA, 1990). However, usingthese reports requires knowledge of the exposure, localtopography, proximity to the coast and type of station:whether it is a buoy, light tower, or coastguard station.Consideration should be given also to the time of day

OCEAN SURFACE WINDS 21(for the possible influence of land-sea breezes) and levelof station maintenance.2.2.4 Satellite dataDuring the past decade or so, there have been a numberof oceanographic satellites on which active and passivemicrowave sensors, such as scatterometers and radiometers(1978 on SEASAT) and radar altimeters(1985–89 on GEOSAT), have demonstrated the capabilityof measuring ocean surface winds. Since 1987,the Special Sensor Microwave/Imager (SSM/I) (USDefense Meteorological Satellite Program spacecraft)has been providing surface wind speed data over theglobal oceans in real time. These microwave measurements,from polar-orbiting satellites in 102-minuteorbits around the Earth, provide several orders ofmagnitude greater surface wind observations comparedto conventional ship and buoy data. Most recently theEuropean Space Agency also has started to providevector wind data from scatterometers and wind speedand wave data from radar altimeters on board theERS-1 satellite and its successor ERS-2 launched inApril 1995.Research efforts now are concentrating on how bestto use these remotely sensed ocean surface wind data inimproving the initial ocean surface wind analyses andsubsequent forecasts. Table 2.3 is a summary of thecurrent operational wind data available from satellites*.An example of an ocean surface wind data plotusing SSM/I wind speed data is shown in Figure 2.3.This plot represents data centred at 00 UTC with a± 3-hour window. The wind directions are assignedusing the lowest level winds (about 45 m) from the USNational Meteorological Center global operational analysissystem. Winds are missing either where there isprecipitation or where wind speeds are in excess of50 kn (~ 25 m/s) (the satellite sensor’s upper limit).It should be noted that the quality of the windmeasurements obtained from satellite-borne sensorsdepends on the accuracy of the algorithms used to derivewind-related parameters (speed and, if applicable, direction)from the sensor measurements (brightnesstemperatures from passive microwave sensors, and radarbackscatter cross-section and antenna parameters fromactive microwave sensors) and various corrections thatneed to be applied for atmospheric water vapour andliquid water contamination. Furthermore, the sensorresponse may drift in time and careful quality-controlprocedures should be used to monitor the retrievals._________* The altimeter footprint depends on the pulse duration,satellite height, the sea state and the method for retrievingthe backscatter value. The footprint is about 10 km.Measurements are reported as one second averages (so thearea sampled is effectively about 17 km x 10 km) centredat about 7 km intervals along the sub-satellite track.TABLE 2.3Satellite derived windsInstrument Mode Swath Footprint MeasurementAltimeter Active Nadir ≈10 km Surface windmicrowavespeedScatterometer Active 500 km 50 km Surface windmicrowavespeed anddirectionSpecial Sensor Passive 1 500 km 25 km Surface windMicrowave/ microwave speedImager2.2.5 Common height adjustmentIn order to perform wind and pressure analyses, the datamust be adjusted to a standard height. However, there isno fixed height at which the observations are made.Anemometer heights on fixed buoys range between 3 mand 14 m, on ships they range from 15 m to over 40 m,and on platforms or at coastal stations the heights mayrange up to 200 m or more above the sea.The theory of Monin-Obukhov (1954) is wellknown for calculating the shape of the wind profilethrough the lowest part of the atmosphere — known asthe constant flux layer. The calculation requires the windspeed, at a known height within the layer, and thestability, which is derived from the air-sea temperaturedifference. Methods for adjusting winds to a standardheight are reviewed and wind adjustment ratios arepresented in table form in a WMO document byShearman and Zelenko (1989). In practice, when anocean surface wind field is being analysed from observations,the wind field is assumed to be prescribed at 10 m.For further discussion of wind profiles in the loweratmospheric boundary layer see Section 2.4.2.3 Large-scale meteorological factorsaffecting ocean surface windsIn common practice a forecaster may have to work fromsurface weather maps, because prognostic or diagnosticmodels for producing ocean surface wind fields may notbe available. The forecaster can subjectively produce anocean surface wind chart based on a knowledge of a fewprinciples of large-scale atmospheric motions and someknowledge of boundary-layer theory.The simplest approach to obtain ocean surfacewinds would be to:• Calculate the geostrophic wind speed;• Correct it for curvature to derive the gradient windspeed;• Simulate the effect of friction by reducing this windto approximately 75 per cent and rotating the winddirection by approximately 15° (anti-clockwise inthe northern hemisphere and clockwise in thesouthern hemisphere). In relation to the geostrophicwind that is towards the region of lower pressure.

22GUIDE TO WAVE ANALYSIS AND FORECASTINGFigure 2.3 — Example of a combined Aviation Model (global spectral model) — Special Sensor Microwave/Imager (SSM/I)ocean surface wind analysis, valid at 00 UTC on 6 April 1992 for the area of the Grand Banks, North Atlantic(from NWS)As a quick approximation of ocean surface winds thisapproach may be satisfactory. However, there are severalimportant factors which should be considered whenparticular meteorological situations are identified.Some important meteorological relationships thatgovern the speed and direction of ocean surface winds are:(1) Surface-pressure gradient — geostrophic wind;(2) Curvature of isobars — gradient wind;(3) Vertical wind shear of the geostrophic wind —thermal wind;(4) Rapidly changing pressure gradient in time —isallobaric wind;(5) Rapidly changing pressure gradient downstream —difluence and confluence;(6) Friction — Ekman and Prandtl layers;(7) Stability of air over the sea — air-sea temperaturedifference.These relationships (discussed in the followingsections) may be considered independently and thencombined to give an estimate of the wind field. It should

OCEAN SURFACE WINDS 23be noted that this approach is an over-simplificationbecause each relationship yields a wind componentwhich is derived from rather specific assumptions butthen combined under very general conditions in order toapproximate the wind field.2.3.1 Geostrophic windThe primary driving force for atmospheric motions is thepressure gradient force. One of the most importantbalances in large-scale atmospheric motions is thatbetween the Coriolis force and the pressure gradientforce, with the resulting balanced motion called thegeostrophic wind. This balance is generally valid:• For large-scale flows;• In the free atmosphere above the friction layer;• Under steady state conditions; and• With straight isobars.The geostrophic flow is parallel to the isobars and isexpressed by the following relationship:1 ⎛ δpδp⎞( ug, vg) = ⎜ – , ⎟ , (2.1)fρa⎝ δyδx⎠where p is the atmospheric pressure, f is the Coriolisparameter (f = 2 Ω sin Θ), ρ a is the air density, Ω is theangular speed of the Earth’s rotation and Θ is the latitude.u g and v g are the geostrophic winds in the x(positive towards east), and y (positive towards north)directions.Equation 2.1 shows that the wind blows in such amanner that, looking downwind, the high pressure is tothe right and the low pressure is to the left in the northernhemisphere (f > 0) and vice versa in the southernhemisphere (f < 0). It should be observed that, for agiven pressure gradient, the geostrophic wind willincrease with decreasing latitude and, in fact, goes toinfinity at the Equator. The geostrophic wind relation isassumed to be not valid in low latitudes, betweenapproximately 20°N and 20°S. Also, over the ocean, airdensity decreases about 10 per cent from cold high pressuresystems to warm low pressure systems. Thegeostrophic wind speed can be estimated from a pressureanalysis using Table 2.4.2.3.2 Gradient windGenerally atmospheric flow patterns are not straight, butmove along curved trajectories. This implies an additionalacceleration along the radius of curvature, i.e. theaddition of centripetal force to balance the flow. Thisbalanced motion is known as the gradient wind. Thegradient wind (Gr) equation is given by:Gr = fr2⎡⎢⎣– 1 + 1 + 4Gfr⎤⎥ ,⎦(2.2)where f is the Coriolis force, G is the geostrophic windspeed (which is used as proxy for the pressure gradientas seen from Equation 2.1) and r is the radius of curvatureof an isobar at the point of interest.For a low pressure system (cyclone), the circulationis in a counter-clockwise direction in the northern hemisphereand in a clockwise direction in the southernhemisphere. The sense of direction around a high pressuresystem (anticyclone) is opposite to that of a cyclonein each of the hemispheres.Around a low pressure centre, the Coriolis andcentrifugal forces act together to balance the pressuregradient force, compared to geostrophic flow where onlythe Coriolis force balances the pressure gradient.Consequently, the speed of the gradient wind around acyclone is less than that of a geostrophic wind correspondingto the same pressure gradient. The balance offorces shows that for a high pressure system the Coriolisforce is balanced by pressure gradient and centrifugalforces acting together. Hence, the gradient flow around ahigh is larger (in magnitude) than the geostrophic flowcorresponding to the same pressure gradient.There is an upper limit to the anticyclonic gradientwind which is obtained when the pressure gradient termreaches:21 δprf= – .(2.3)δr4ρ aWhen the pressure gradient reaches this value, the quantityunder the square root in Equation 2.2 becomes zero,resulting in a maximum (in magnitude) gradient windspeed of:Gr = fr(2.4)2 .Using the geostrophic relation with Equation 2.3and combining with Equation 2.4, the upper limit for thegradient wind speed for anticyclonic flow is twice thegeostrophic wind speed:Gr ≤ 2 G . (2.5)There is no such corresponding lower limit to thecyclonic gradient wind speed in relation to the pressuregradient.The gradient wind speed can be estimated usingTable 2.5 by measuring the radius of curvature of anisobar from a weather map and using the geostrophicwind speed determined from Table 2.4.The balance diagrams for the simple frictionlessflows reviewed above are illustrated in Figure 2.4.2.3.3 Thermal windSo far, the sea-level pressure field has been considered tobe constant from the ocean surface to the top of theboundary layer. However, if there are non-zero horizontaltemperature gradients at the surface, it can beshown through the thermal wind equation that the pressuregradient will change with height. This results in achange of the geostrophic wind with height so that thegeostrophic wind at the top of the boundary layer willdiffer from that at the surface. The vertical shear of thegeostrophic wind is given by:

24GUIDE TO WAVE ANALYSIS AND FORECASTINGTABLE 2.4Geostrophic wind tableSpeed is given (in knots) as a function of latitude and distance (in degrees of latitude)for 4 hPa change in pressure; pressure is 1015.0 hPa; temperature is 285 K; density is 0.001241 gm/cm 3Dist. Latitude (°)(°lat.) 20 25 30 35 40 45 50 55 60 65 701.0 113 92 77 68 60 55 51 47 45 43 411.1 103 83 70 61 55 50 46 43 41 39 371.2 94 76 65 56 50 46 42 39 37 36 341.3 87 70 60 52 46 42 39 36 34 33 321.4 81 65 55 48 43 39 36 34 32 31 291.5 75 61 52 45 40 37 34 32 30 28 271.6 71 57 48 42 38 34 32 30 28 27 261.7 67 54 46 40 35 32 30 28 26 25 241.8 63 51 43 38 33 30 28 26 25 24 231.9 60 51 43 38 33 30 28 26 25 24 232.0 57 46 39 34 30 27 25 24 22 21 212.1 54 44 37 32 29 26 24 23 21 20 202.2 51 42 35 31 27 25 23 21 20 19 192.3 49 40 34 29 26 24 22 21 19 19 182.4 47 38 32 28 25 23 21 20 19 18 172.5 45 37 31 27 24 22 20 19 18 17 162.6 44 35 30 26 23 21 19 18 17 16 162.7 42 34 29 25 22 20 19 18 17 16 152.8 40 33 28 24 22 20 18 17 16 15 152.9 39 32 27 23 21 19 17 16 15 15 143.0 38 31 26 23 20 18 17 16 15 14 143.1 37 30 25 22 19 18 16 15 14 14 133.2 35 29 24 21 19 17 16 15 14 13 133.3 34 28 23 20 18 17 15 14 14 13 123.4 33 27 23 20 18 16 15 14 13 13 123.5 32 26 22 19 17 16 14 14 13 12 123.6 31 25 22 19 17 15 14 13 12 12 113.7 31 25 21 18 16 15 14 13 12 12 113.8 30 24 20 18 16 14 13 12 12 11 113.9 29 23 20 17 15 14 13 12 11 11 114.0 28 23 20 17 15 14 13 12 11 11 104.2 27 22 18 16 14 13 12 11 11 10 104.4 26 21 18 15 14 12 11 11 10 10 94.6 25 20 17 15 13 12 11 10 10 9 94.8 24 19 16 14 13 11 11 10 9 9 95.0 23 18 15 14 12 11 10 9 9 9 85.2 22 18 15 13 12 11 10 9 9 8 85.4 21 17 14 13 11 10 9 9 8 8 85.6 20 16 14 12 11 10 9 8 8 8 75.8 20 16 13 12 10 9 9 8 8 7 76.0 19 15 13 11 10 9 8 8 7 7 76.2 18 15 12 11 10 9 8 8 7 7 76.4 18 14 12 11 9 9 8 7 7 7 66.6 17 14 12 10 9 8 8 7 7 6 66.8 17 13 11 10 9 8 7 7 7 6 67.0 16 13 11 10 9 8 7 7 6 6 68.0 14 11 10 8 8 7 6 6 6 5 59.0 13 10 9 8 7 6 6 5 5 5 510.0 11 9 8 7 6 5 5 5 4 4 4

OCEAN SURFACE WINDS 25TABLE 2.5Gradient wind tableWind speeds (in knots) are shown at latitude 40°N for given geostrophic wind speed (from Table 2.4) and radius of curvature (° lat.).(Note: for any other latitude, φ, the winds should be scaled by the ratio f φ /f 40 , where f is the value of the Coriolis parameter.)Radius ofcurvature(° lat.)Geostrophic wind (kn)5 10 15 20 25 30 35 40 45 50 55 60 65 70 75Anticyclonic flow25 5 10 15 21 26 32 38 44 50 56 63 70 77 84 9224 5 10 15 21 26 32 38 44 50 57 63 70 77 85 9323 5 10 16 21 27 32 38 44 51 57 64 71 78 86 9422 5 10 16 21 27 32 38 44 51 57 64 72 79 87 9621 5 10 16 21 27 32 39 45 51 58 65 72 80 89 9720 5 10 16 21 27 33 39 45 52 58 66 73 81 90 10019 5 10 16 21 27 33 39 45 52 59 67 75 83 92 10218 5 10 16 21 27 33 39 46 53 60 68 76 85 95 10617 5 10 16 21 27 33 40 46 53 61 69 78 87 98 11116 5 10 16 21 27 33 40 47 54 62 70 80 90 103 11915 5 10 16 22 28 34 40 47 55 63 72 83 95 110 13814 5 10 16 22 28 34 41 48 56 65 75 87 102 129 013 5 10 16 22 28 35 42 49 58 67 79 93 120 0 012 5 10 16 22 28 35 42 51 60 71 85 111 0 0 011 5 10 16 22 29 36 44 52 63 76 102 0 0 0 010 5 11 16 23 29 37 45 55 68 92 0 0 0 0 09 5 11 17 23 30 38 47 60 83 0 0 0 0 0 08 5 11 17 23 31 40 51 74 0 0 0 0 0 0 07 5 11 17 24 32 43 65 0 0 0 0 0 0 0 06 5 11 18 25 35 55 0 0 0 0 0 0 0 0 05 5 11 18 28 46 0 0 0 0 0 0 0 0 0 04 5 12 20 37 0 0 0 0 0 0 0 0 0 0 03 6 13 28 0 0 0 0 0 0 0 0 0 0 0 02 6 18 0 0 0 0 0 0 0 0 0 0 0 0 01 9 0 0 0 0 0 0 0 0 0 0 0 0 0 0Cyclonic flow25 5 10 15 19 24 28 33 37 42 46 50 54 58 62 6624 5 10 15 19 24 28 33 37 41 46 50 54 58 62 6623 5 10 15 19 24 28 33 37 41 46 50 54 58 62 6622 5 10 15 19 24 28 33 37 41 45 49 54 58 61 6521 5 10 15 19 24 28 33 37 41 45 49 53 57 61 6520 5 10 14 19 24 28 32 37 41 45 49 53 57 61 6519 5 10 14 19 24 28 32 37 41 45 49 53 57 60 6418 5 10 14 19 23 28 32 36 40 45 49 52 56 60 6417 5 10 14 19 23 28 32 36 40 44 48 52 56 60 6316 5 10 14 19 23 28 32 36 40 44 48 52 55 59 6315 5 10 14 19 23 27 32 36 40 44 48 51 55 59 6214 5 10 14 19 23 27 31 36 39 43 47 51 54 58 6213 5 10 14 19 23 27 31 35 39 43 47 50 54 57 6112 5 10 14 19 23 27 31 35 39 43 46 50 53 57 6011 5 10 14 18 23 27 31 35 38 42 46 49 53 56 5910 5 10 14 18 22 27 30 34 38 41 45 48 52 55 589 5 10 14 18 22 26 30 34 37 41 44 48 51 54 578 5 9 14 18 22 26 30 33 37 40 43 47 50 53 567 5 9 14 18 22 25 29 33 36 39 42 45 48 51 546 5 9 13 17 21 25 28 32 35 38 41 44 47 50 525 5 9 13 17 21 24 27 31 34 37 40 42 45 48 504 5 9 13 17 20 23 26 29 32 35 38 40 43 45 473 5 9 12 16 19 22 25 27 30 33 35 37 39 41 442 4 8 12 15 17 20 22 25 27 29 31 33 35 37 381 4 7 10 12 15 16 18 20 22 23 25 26 27 29 30

26GUIDE TO WAVE ANALYSIS AND FORECASTINGGr∇pLowCCnf∇pGCCnf∇pGrCHighThe isallobaric wind component is given by:⎡ ⎛ p⎞⎛ p⎞⎤⎢δ⎜δ ⎟ δ⎜δ ⎟1 ⎝ δt⎠ ⎝ δt⎠ ⎥( ui, vi) = –2⎢ , ⎥ . (2.7)ρaf ⎢ δxδy⎥⎣⎢⎦⎥The modification of the geostrophic wind fieldaround a moving low pressure system is illustrated inFigure 2.5.Figure 2.4 — Balance of forces for basic types of frictionlessflow (northern hemisphere) (Gr = gradientwind, G = geostrophic wind, ∇p = pressuregradient force; C = Coriolis force andCnf = centrifugal force)⎛ δugδvg⎞g ⎛ δTδT⎞⎜ , ⎟ = ⎜ – , ⎟ . (2.6)⎝ δzδy⎠ T ⎝ δyδx⎠In the southern hemisphere the left hand side of theequation needs to be multiplied by –1.It is clear from Equation 2.6 that the geostrophicwind increases with height if higher pressure coincideswith higher temperatures (as in the case of mid-latitudewesterlies) and decreases with height if higher pressurecoincides with lower temperatures. Furthermore, if thegeostrophic wind at any level is blowing towards warmertemperatures (cold advection), the wind turns to the left(backing) as the height increases and the reverse happens(veering) if the geostrophic wind blows towards lowertemperatures (warm advection).The vector difference in the geostrophic wind attwo different levels is called the “thermal wind”. It canbe shown geometrically that the thermal wind vectorrepresents a flow such that high temperatures are locatedto the right and low temperatures to the left. The thermalwind, through linear vertical wind shear, can be incorporateddirectly into the solution of the Ekman layer, andthus incorporated in the diagnostic models which will bedescribed in more detail below.2.3.4 Isallobaric windIn the above discussions, the wind systems have beenconsidered to be evolving slowly in time. However, whena pressure system is deepening (or weakening) rapidly, ormoving rapidly, so that the local geostrophic wind ischanging rapidly, an additional wind componentbecomes important. This is obtained through the isallobaricwind relation. An isallobar is a line of equalpressure tendency (time rate of change of pressure). Thestrength of the isallobaric wind is proportional to the isallobaricgradient, and its direction is perpendicular to thegradient — away from centres of rises in pressure andtoward centres of falls in pressure. Normally, this componentis less than 5 kn (2.5 m/s), but can become greaterthan 10 kn (5 m/s) during periods of rapid or explosivecyclogenesis.2.3.5 Difluence of wind fieldsDifluence (confluence) of isobars also creates flows thatmake the winds deviate from a geostrophic balance.When a difluence of isobars occurs (isobars spreadapart), the pressure gradient becomes weaker than itsupstream value, so that as an air parcel moves downstream,the pressure gradient is unbalanced by theCoriolis force associated with the flow speed. This thenresults in the flow being deflected towards high pressurein an effort to restore the balance of forces through anincrease in the pressure gradient force. In the case ofconverging isobars, the pressure gradient increases fromit upstream value. Hence, the pressure gradient forcebecomes larger than the Coriolis force and the flow turnstowards the low pressure in a effort to decrease the pressuregradient force. In either case, it is clear that anon-geostrophic cross-isobaric flow develops of amagnitude U dG/ds (Haltiner and Martin, 1957), whereG is the geostrophic speed, U is the non-geostrophiccomponent normal to the geostrophic flow that hasdeveloped in response to the confluence or difluence ofthe isobars, and s is in the direction of the geostrophicflow.+2IGGI+4 LowGGFigure 2.5 — Examples of the isallobaric wind field. Solidline = isobar, dashed line = isallobar, G = (u g ,v g ) = geostrophic wind, and I =(u i , v i ) = isallobaricwindIIGGGIII-4GI-2

OCEAN SURFACE WINDS 27Pp 0GPp u1Dfp 1Figure 2.6 — Examples of difluent/confluent wind fields (p 0and p 1 are isobars. D f and C f are the difluentand confluent wind components, respectively,and G and u are the geostrophic and surfacewinds)p 0In reality, since surface friction turns the flowtowards low pressure, confluence will increase theinflow angle over the effect of friction alone, and difluencewill decrease the inflow angle, with the result thatflow towards high pressure will seldom exist. Figure 2.6illustrates the modification to geostrophic wind due toeffects by difluence and confluence.2.3.6 Surface friction effectsThe effect of friction is to reduce the speed of the free airflow. As a result of the balance of forces this will turnthe direction of flow toward lower pressure, which is tothe left in the northern hemisphere and to the right in thesouthern hemisphere. As one approaches the surface ofthe Earth the wind speed tends to zero, and the inflowangle tends to reach a maximum. A simple balancebetween the pressure gradient, Coriolis and frictionalforces (Figure 2.7) describes this effect through thewell-known Ekman spiral. The lower part of the atmospherein which this effect is prominent is called theEkman layer. In nature the predicted 45° angle of turningat the surface is too large, and predicted speeds near thesurface are too low.These effects of friction on the geostrophic flow inthe Ekman layer are further influenced by the thermalwind discussed above. Studies have shown the importanceof the thermal wind in explaining deviations fromthe typical Ekman spiral (Mendenhall, 1967). A resultof the thermal wind influence is larger surfacecross-isobaric (inflow) flow angles during cold advection,and smaller inflow angles during warm advection,as described above. An illustration of the effect of thermalwind on the wind in the Ekman layer is shown for alow pressure system in Figure 2.8.In order to represent the effects of friction in amore realistic manner several approaches have beendeveloped that relate the free atmospheric wind to astress at the ocean surface. These often use the conceptof a two-regime boundary layer; a constant flux layerat the surface, and the Ekman layer (spiral) above.Observations fit much better with this representation ofC fuGthe planetary boundary layer, with a turning angle of10–15° predicted over the ocean for a neutrally stableatmosphere in contact with the ocean.2.3.7 Stability effectsStability within the boundary layer is important in determiningthe wind speed near the ocean surface. Overmuch of the oceans, the surface air temperature is inequilibrium with the sea-surface temperature so thatnear-neutral stability dominates. Under these conditionsthe structure of the wind profile in the constant flux layeris dominated by friction and can be described by a logarithmicprofile.For unstable cases (air temperature colder than thewater temperature) convection becomes active and thehigher wind speeds aloft are brought to the surfacequickly, reducing dissipation by friction, and increasingthe stress on the ocean surface. A stable atmosphere(warm air over cold water) acts to increase the frictionforces in the boundary layer, resulting in lighter windsand a weaker wind stress. Table 2.6 shows the effects ofstability on wind profiles above the ocean.Stability effects are most important for ocean areasnear large land masses, which act as source areas for airmasses with very different physical properties fromoceanic air masses. An expanded discussion of thetheory of friction and stability in the boundary layer ispresented in the next section.2.4 A marine boundary-layerparameterizationThe boundary layer of the atmosphere is that regionwhich extends from the surface to the free atmosphere(at approximately 1 km height). At the surface frictionalforces dominate. In the free atmosphere frictional forcesbecome less important and, to a first order of approximation,the atmospheric flow is close to being ingeostrophic balance.Because all flows in nature are turbulent, the frictionalforces we are concerned with are those arisingfrom turbulent fluctuations — the so-called Reynoldsstresses. A fundamental difficulty of the turbulencetheory is to relate these turbulent stresses to the propertiesof the mean flow. Through extensive research someinsights into these relationships have evolved, at least in∇puGFCFigure 2.7 —Balances of friction,pressure gradient andCoriolis forces (northernhemisphere)(u = surface wind;G = geostrophic wind,∇p = pressure gradientforce; C = Coriolis forceand F = friction)

28GUIDE TO WAVE ANALYSIS AND FORECASTINGPOINT B — The effect of a cold low inincreasing the speed of the surface windPOINT C — The effect of warm airadvection in decreasing cross-isobarflow for the surface windG(SURFACE)zzEKMANBCEKMANPLUSTHERMALEKMANG (SURFACE)zzEKMANPLUSTHERMALAPOINT A — The effect of cold air advectionin increasing cross-isobar flow for thesurface windKEYGSURFACEGGRADIENT WINDLEVELTHERMAL WINDG (SURFACE )zzEKMAN PLUSTHERMALEKMANACTUAL WIND AT zACTUAL WIND AT zWITHOUTTHERMAL EFFECTFigure 2.8 — The thermal wind. Sketch of surface winds, surface isobars (solid contours) and a constant pressure surface (dashed contours) for an extra-tropical cyclone (after Pierson, 1979)

OCEAN SURFACE WINDS 29TABLE 2.6The effects of stability on wind profilesWind speed profiles are shown (in m/s) for a range of stabilities (characterized by the air-sea temperature difference).Profiles are given both for a fixed wind speed at 50 m and for a fixed surface friction velocity (0.36 m/s)HeightAir-sea temperature difference (°C)(m) –10 –8 –6 –4 –3 –2 –1 0 1 2 3 4 6 8 10Fixed wind speed at 50 m50 10.0 10.0 10.0 10.0 10.0 10.0 10.0 10.0 10.0 10.0 10.0 10.0 10.0 10.0 10.045 10.0 10.0 10.0 10.0 10.0 10.0 9.9 9.9 9.8 9.8 9.7 9.7 9.7 9.7 9.740 9.9 9.9 9.9 9.9 9.9 9.9 9.9 9.8 9.6 9.5 9.4 9.4 9.4 9.4 9.435 9.9 9.9 9.9 9.9 9.8 9.8 9.8 9.7 9.4 9.2 9.1 9.1 9.1 9.1 9.130 9.8 9.8 9.8 9.8 9.8 9.8 9.7 9.6 9.2 8.9 8.8 8.7 8.7 8.7 8.728 9.8 9.8 9.8 9.8 9.7 9.7 9.7 9.5 9.1 8.8 8.6 8.6 8.5 8.5 8.626 9.8 9.8 9.7 9.7 9.7 9.7 9.6 9.4 9.0 8.6 8.5 8.4 8.3 8.4 8.424 9.7 9.7 9.7 9.7 9.7 9.6 9.6 9.4 8.9 8.5 8.3 8.2 8.2 8.2 8.222 9.7 9.7 9.7 9.6 9.6 9.6 9.5 9.3 8.7 8.4 8.1 8.0 8.0 8.0 8.020 9.7 9.6 9.6 9.6 9.6 9.5 9.5 9.2 8.6 8.2 8.0 7.8 7.8 7.8 7.818 9.6 9.6 9.6 9.5 9.5 9.5 9.4 9.1 8.5 8.0 7.8 7.6 7.5 7.5 7.616 9.6 9.5 9.5 9.5 9.4 9.4 9.3 9.0 8.4 7.9 7.6 7.4 7.3 7.3 7.314 9.5 9.5 9.5 9.4 9.4 9.3 9.2 8.9 8.2 7.7 7.4 7.2 7.0 7.0 7.012 9.4 9.4 9.4 9.3 9.3 9.2 9.1 8.8 8.1 7.5 7.1 6.9 6.7 6.7 6.710 9.3 9.3 9.3 9.2 9.2 9.1 9.0 8.6 7.9 7.3 6.9 6.6 6.4 6.4 6.49 9.3 9.2 9.2 9.1 9.1 9.0 8.9 8.5 7.8 7.2 6.7 6.5 6.2 6.2 6.28 9.2 9.2 9.1 9.1 9.0 8.9 8.8 8.4 7.7 7.0 6.6 6.3 6.1 6.0 6.07 9.1 9.1 9.1 9.0 8.9 8.8 8.7 8.3 7.6 6.9 6.4 6.2 5.9 5.8 5.76 9.0 9.0 9.0 8.9 8.8 8.7 8.6 8.2 7.4 6.7 6.3 6.0 5.7 5.5 5.55 8.9 8.9 8.8 8.7 8.7 8.6 8.5 8.0 7.3 6.6 6.1 5.8 5.4 5.3 5.24 8.8 8.7 8.7 8.6 8.5 8.4 8.3 7.8 7.1 6.4 5.9 5.5 5.2 5.0 4.93 8.6 8.5 8.5 8.4 8.3 8.2 8.0 7.6 6.9 6.2 5.6 5.3 4.9 4.7 4.52 8.3 8.2 8.1 8.0 8.0 7.9 7.7 7.2 6.6 5.9 5.3 5.0 4.5 4.3 4.11 7.7 7.6 7.5 7.4 7.4 7.3 7.1 6.7 6.1 5.4 4.9 4.5 4.0 3.7 3.6Friction velocity fixed (0.36 m/s)50 8.2 8.3 8.5 8.7 8.8 9.0 9.3 10.0 11.1 12.0 12.8 13.5 14.5 15.2 16.145 8.1 8.3 8.4 8.6 8.8 9.0 9.2 9.9 10.9 11.8 12.5 13.1 14.1 14.9 15.640 8.1 8.3 8.4 8.6 8.7 8.9 9.2 9.8 10.7 11.5 12.1 12.7 13.7 14.5 15.135 8.1 8.2 8.4 8.6 8.7 8.9 9.1 9.7 10.4 11.2 11.8 12.3 13.2 13.9 14.630 8.0 8.2 8.3 8.5 8.6 8.8 9.0 9.6 10.2 10.8 11.4 11.9 12.7 13.4 13.928 8.0 8.2 8.3 8.5 8.6 8.8 9.0 9.5 10.1 10.7 11.2 11.7 12.5 13.1 13.726 8.0 8.1 8.3 8.5 8.6 8.7 9.0 9.4 10.0 10.6 11.1 11.5 12.2 12.9 13.424 8.0 8.1 8.2 8.4 8.6 8.7 8.9 9.4 9.9 10.4 10.9 11.3 12.0 12.6 13.122 8.0 8.1 8.2 8.4 8.5 8.7 8.9 9.3 9.7 10.2 10.7 11.1 11.7 12.3 12.820 7.9 8.1 8.2 8.4 8.5 8.6 8.8 9.2 9.6 10.1 10.5 10.8 11.5 12.0 12.518 7.9 8.0 8.2 8.3 8.4 8.6 8.7 9.1 9.5 9.9 10.3 10.6 11.2 11.7 12.116 7.9 8.0 8.1 8.3 8.4 8.5 8.7 9.0 9.3 9.7 10.1 10.4 10.9 11.4 11.814 7.8 8.0 8.1 8.2 8.3 8.4 8.6 8.9 9.2 9.5 9.8 10.1 10.6 11.0 11.412 7.8 7.9 8.0 8.2 8.2 8.4 8.5 8.8 9.0 9.3 9.6 9.8 10.2 10.6 11.010 7.7 7.9 7.9 8.1 8.2 8.3 8.4 8.6 8.8 9.1 9.3 9.5 9.9 10.2 10.59 7.7 7.8 7.9 8.0 8.1 8.2 8.3 8.5 8.7 8.9 9.1 9.3 9.7 10.0 10.28 7.6 7.8 7.8 8.0 8.0 8.1 8.2 8.4 8.6 8.8 9.0 9.1 9.4 9.7 10.07 7.6 7.7 7.8 7.9 8.0 8.0 8.1 8.3 8.5 8.6 8.8 8.9 9.2 9.5 9.76 7.5 7.7 7.7 7.8 7.9 8.0 8.0 8.2 8.3 8.5 8.6 8.7 9.0 9.2 9.45 7.4 7.6 7.6 7.7 7.8 7.8 7.9 8.0 8.1 8.3 8.4 8.5 8.7 8.9 9.14 7.3 7.5 7.5 7.6 7.6 7.7 7.8 7.8 7.9 8.0 8.1 8.2 8.4 8.5 8.73 7.2 7.3 7.3 7.4 7.4 7.5 7.5 7.6 7.7 7.8 7.8 7.9 8.0 8.1 8.32 7.1 7.1 7.1 7.1 7.2 7.2 7.2 7.2 7.3 7.4 7.4 7.5 7.6 7.6 7.71 6.6 6.6 6.6 6.7 6.7 6.7 6.7 6.7 6.8 6.8 6.8 6.8 6.9 6.9 7.0

30GUIDE TO WAVE ANALYSIS AND FORECASTINGa general sense, and can be applied to the atmosphere.These concepts have been developed to formulate diagnosticmodels that allow us to deduce surface turbulentflow fields from the free atmospheric flow.The marine boundary layer itself can be separatedinto two regimes: the constant flux (or constant stress)layer (from the surface to about 50 m) and the Ekmanlayer (from about 50 m up to the free atmosphere≈ 1 km). In the surface layer it can be assumed that thefrictional forces contributing to turbulence are constantwith height, and the effects of the Coriolis and pressuregradient forces, as well as the horizontal gradient ofturbulent fluxes, are negligible. The wind direction isconsequently constant with height. Using the mixinglength theory developed by Prandtl, it can be shownthat the flow in the constant flux layer (or Prandtl layer)depends only on the surface roughness length.2.4.1 Constant flux layerUnder neutral conditions, Prandtl’s solution shows thatthe horizontal flow over the ocean surface follows thewell known “log” (logarithmic) profile in the verticaldirection,u = u *(2.8)κ ln ⎛ z ⎞⎜ ⎟⎝ z 0 ⎠,where κ is the von Kármán constant, z 0 is the constant ofintegration, known as the roughness length, and u * is thefriction velocity, which has magnitude:τu * = ,(2.9)ρ aτ is the magnitude of the surface stress, and ρ a thedensity of air. u * can be thought of as a proxy for thesurface stress.It is common also to express stress (τ) through thebulk transfer relation:τ = ρ a C d u 2 , (2.10)where C d is the drag coefficient. In general, C d and u areboth functions of height. Determining C d has been theobjective of many field research programmes over theyears.One of the problems of specifying the wind in theturbulent layer near the ocean is the formulation of z 0and its relationship to u * . Using dimensional argument,Charnock (1955) related the roughness length of the seasurface to the friction velocity of the wind as follows:z 0 = αu * 2,(2.11)gwhere α is the Charnock constant, with a value as determinedby Wu (1980) of 0.0185 , and g is the accelerationby gravity.If the boundary layer is in a state of neutral stratification,the drag coefficient, which is a function ofheight, can be expressed as:orCd( z)=C d⎛ ⎡ z ⎤⎞/ ⎜ln ⎢ ⎥⎝ ⎣ z⎟ ,⎦⎠κ 2 0⎛ u( z) = *⎞⎜ ⎟⎝ uz⎠( )(2.12)(2.13)However, the boundary layer over the ocean is notnecessarily neutral. A stability dependence was originallyderived by Monin-Obukhov (1954) from profilesimilarity theory. This is used to modify the simple logrelation provided above:u = u ⎡*(2.14)κ ln ⎛ z ⎞⎜ ⎟⎝ z 0 ⎠– ψ ⎛ z ⎞ ⎤⎢⎝ L ⎠⎥⎣⎢⎦⎥⎛ ⎡ ⎛ z ⎞C z⎛ z ⎞⎤⎞d( )= ⎜κ / ⎢ln ⎜ ⎟ – ψ(2.15)⎝ z ⎠ ⎝ L⎠⎥⎟.⎝ ⎣⎢0 ⎦⎥⎠The function ψ has been derived for both stable andunstable conditions. L is the Monin-Obukhov mixinglength. For neutral conditions ψ (z/L) = 0. Businger et al.(1971) proposed functional relationships between thenon-dimensional wind shear and z/L which can be usedto determine the stability function ψ (z/L) in Equation2.14.2.4.2 The use of the drag coefficientIn the above discussion, we developed the concept of thedrag coefficient, as defined in Equation 2.10. There havebeen many studies to determine C d under varying windspeeds and stabilities and a summary can be found inRoll (1965). More recently, Wu (1980, 1982) has shownthrough empirical studies that the drag coefficient at agiven height depends linearly on the wind speed, andthat the following formulation holds for a wide range ofwinds under near neutrally stable conditions:10 3 C 10 = (0.8 + 0.65 U 10 ) (2.16)where: C 10 = Drag coefficient at the 10 m level; andU 10 = Wind speed (m/s) at the 10 m level.This simple linear empirical relationship, however,needs to be modified when temperature stratification ispresent. Stratification affects turbulent momentum transport,thereby causing the wind profile to deviate from thelogarithmic form.Schwab (1978) determined C d over water for awide range of wind speeds and atmospheric stabilities.Figure 2.9 shows the results of his calculation. A crucialissue to be addressed at this juncture concerns the effectof changing stability and wind stress on the predictionof wave growth. It can be inferred from Figure 2.9 that,for a given 10 m level wind speed, unstable conditionsresult in higher drag coefficients (or surface stress) andhence larger wave growth than stable conditions. Liu etal. (1979) have developed a set of equations whichcompute the surface variables u * , z 0 and the boundarylayerstability length (L), so that the wind profile of2.22

OCEAN SURFACE WINDS 3110 m wind speed (m/s)20151052.01.81.6C d × 10 31.41.21.00.80.60.40.20-20 -15 -10 -5 0 5 10 15 20Air-sea temperature difference (°C)Figure 2.9 — Drag coefficient as a function of stability (airseatemperature difference) and 10 m windspeed (derived from Schwab, 1978)the constant flux layer including stability can bedetermined.For wave forecast models, it appears appropriate toexpress the input wind in terms of u * . This is calculatedtaking stability into account. The wind is then expressedat some nominal height by applying the neutral logarithmicprofile (Equation 2.8, given u * with stability effectalready taken into account and ψ = 0). This wind is thencalled the “equivalent neutral” wind at that height.However, if winds are given at heights above theconstant flux layer (above 50 m), i.e. well into theEkman spiral, the above techniques should not be used.Instead one needs to deal with the more complicatedtwo-regime (constant flux and Ekman) boundary layer.Several approaches have been developed using the tworegimeboundary-layer model, whereby the surfacestress and its direction are determined using the freeatmospheric parameters.2.4.3 Analytic modelsOne analytic approach is to solve the equations for boththe constant flux and Ekman layers with a matching ofthe wind speed, wind direction and stress across theinterface between the layers. This enables the surfacevariables to be related to variables in the free atmosphere.This type of model was initially developed byBlackadar (1965) over land, and modified for use overthe ocean by Cardone (1969, 1978), Brown and Liu(1982) and Krishna (1981).Overland and Gemmill (1977) compared theCardone model to observations from fixed buoys in theNew York Bight and found mean absolute speed errorsof about 3 m/s. Brown and Liu (1982) compared theirmodel with data taken during two large-scale experiments,JASIN and GOASEX. They indicate that theoverall accuracy of their model compared to observationswas 2 m/s and 20°.2.4.4 Rossby number similarity theoryThe other common approach is based on Rossby numbersimilarity theory. The governing equations are the sameas before, but the matching is by using two similarityapproximations: one for the flow of the Ekman layer,and another for the constant flux layer (Stull, 1988).Clark and Hess (1974) have produced a modified versionof the Rossby number similarity theory which gives relationsfor u * and the turning angle at the ocean surface interms of the geostrophic wind, ocean current, surfaceroughness and similarity functions A(µ,m) and B(µ,m),where m is related to the thermal wind and µ is related tostability. A(µ,m) and B(µ,m) have been evaluated usingthe measurements taken during the Wangara Experiment(Clark, et al., 1971). Possible inconsistencies in the windfields produced by the two approaches described abovemay be traced to at least three sources:(a) The non-dimensional wind shear, and hence thestability function ψ (analytic approach) and thefunctions A and B (Rossby number similarity) areempirically based;(b) The specification of the depth of the constant fluxlayer by the analytic approach is not well known;(c) Significant error is introduced by the large-scalemodel which provides the wind and temperaturefields.Figure 2.10 summarizes the properties of thetwo-regime boundary layer.2.4.5 Prognostic boundary-layer modelsPrognostic boundary-layer models, with generally highhorizontal and vertical resolution, are potentially themost accurate. However, the very high spatial resolutionrequired makes it difficult to run such models in anoperational setting because of the large computationalresources required.A typical prognostic boundary-layer model willdescribe the space and time evolution of wind, temperatureand moisture fields from prescribed initialdistribution of these fields over the domain of interest.At the upper boundary, typically around 2 km, boundaryconditions are prescribed from an overlying large-scalemodel. If the model covers a limited area, lateral boundariesmust also be prescribed. An example of such anoperationally run model is the US National WeatherService atmospheric boundary-layer model (Long et al.,1978). Unfortunately, the model has only limited use asa forecast tool for ocean surface winds because itsdomain covers only a small part of the coastal area of theUSA.Another approach is to incorporate sophisticatedboundary-layer physics in large-scale atmosphericmodels. This permits the forecasting of more

32GUIDE TO WAVE ANALYSIS AND FORECASTINGFree atmosphere(above 1 km)u = GEkman layerK = const.Constant flux layerMatched layer(below 50 m)Ocean surfaceρK = du/dz, du/dz, and u continuousK = κu * z, κ = 0.4, τ = const.(z = z 0 ), u * , u = 0Figure 2.10 — Schematic of the two-regime boundary layer, K is the mixing coefficientrepresentative ocean surface winds as input to oceanforecast models. Examples may be found in theECMWF model (Tiedtke, et al., 1979; Simmons et al.,1988) and the US National Meteorological Center spectralforecast model (Sela, 1982; Kanamitsu et al., 1991).It should be emphasized that, even with these improvements,a large-scale model cannot be considered a trueboundary-layer model because of its incomplete formulationof boundary-layer physics and horizontal andvertical resolution. In order to obtain wind forecasts inthe boundary layer, one generally takes the wind forecastfrom the lowest level of these large-scale models andapplies the diagnostic boundary-layer considerationsdiscussed earlier. In this approach, there is a tacitassumption that the winds in the boundary layer are ininstantaneous equilibrium with those from the modelfirst level.A new generation of regional models, with highhorizontal resolution and several additional verticallevels in the boundary layer, are currently under developmentwith greater emphasis on boundary-layer physics.One such model is the so-called ETA model (Mesingeret al., 1988), which has a horizontal resolution of 20 kmand 40 layers in the vertical. The model is centred overthe North American continental domain of the USA. Italso covers large ocean areas adjacent to the continent.When this model becomes operational, the winds in theboundary layer are directly given by the forecast modelitself at the ocean surface (20 m) and there is no need toperform the above diagnostic calculations separately. Aspecial advantage of this model is that it is easilyportable to any other region of the world to producemeteorological forecasts, provided the lateral boundaryconditions are supplied by some means.2.5 Statistical methodsStatistical methods provide another approach to generatesurface wind fields from large-scale forecast models.These techniques do well in situations where forecastmodels do not or cannot depict the surface winds withadequate accuracy.In general, the statistical approach to the generationof wind analyses and forecasts calls for the derivation ofa mathematical relationship between a dependent variable(the predictand) and selected independent variables(the predictors). The predictors may either be madeavailable to forecasters for manual use or may be entereddirectly to computers for automated use.2.5.1 ExtrapolationIn the simplest approach, observed surface winds overthe land can be compared with concurrent winds overthe water and a relationship derived, usually in the formof a simple ratio. Forecasts of wind over water can thenbe made from the projected values of wind over land.Examples of this approach can be found in Richards etal. (1966), Overland and Gemmill (1977), Phillips andIrbe (1978) and Burroughs (1984). An advantage of thisapproach is that it can easily be applied manually.Clearly, the technique would be most effectivewhen used in situations only requiring estimations ofwinds relatively near the coast. It may also be useful onenclosed bodies of water, such as the Great Lakes, wherethe surrounding wind field is sufficiently specified. Forproducing wind forecast fields that extend from coastalareas to far distances on the high seas, however, this typeof approach has only limited usefulness.2.5.2 Perfect prognosisA more sophisticated approach is to derive a relationshipbetween analysed atmospheric parameters and the windsat the ocean surface. The predictors are obtained directlyfrom gridded analysed fields and include parameterswhich are directly available, such as winds, geopotentialheight or pressure, etc. at various model levels. Computedparameters such as gradients, vorticity, divergence,etc. are also included. Values of the predictors anywhereon the grid may be considered in addition to valuescollocated with the predictand.This approach to statistical forecasting is called“perfect prognosis” because it is assumed that, when thescheme is put into operation, the predictors supplied

OCEAN SURFACE WINDS 33from the numerical models as input will approach aperfect simulation of the actual atmosphere.The predictands are the wind direction and speedobtained from either fixed locations, such as mooredbuoys and ocean weather ships, or from voluntaryobserving ships.A number of different statistical techniques fordeveloping relationships between predictors and predictandsare available. The most prominent among them iscalled step-wise screening multiple linear regression andresults in equations of the following form:Y = C 0 + C 1 X 1 + … + C n X n , (2.17)where: Y = Predictand (dependent variable);C n = Constants;X n = Predictors (independent variables).An example of the use of this approach over marineareas can be found in Thomasell and Welsh (1963).When tested against independent data, their systemachieved RMS errors of 3.7 m/s and 35° in speed anddirection, respectively.An important consideration in using this type oftechnique is that the wind-specification equations areindependent of the model supplying the predictors.Because of this the accuracy will improve as bettermodels are developed.2.5.3 Model output statisticsThis approach (Glahn and Lowry, 1972) is similar to theperfect prognosis technique, i.e. a relationship is derivedbetween observed ocean surface winds and independentpredictors. However, predictors are not derived from ananalysis of the actual observations but from a particularnumerical forecast model. This has a number of importantimplications:• Biases associated with the particular numericalmodel used to supply the predictors are suppressed;TABLE 2.7Data requirements for various statistical approachesApproachDevelopment data Operational dataDependent Independent IndependentExtrapolation Observations Observations Manual or modelforecast outputPerfect Observations Analyses Model forecastprognosisoutputModel output Observations Model forecast Model forecaststatistics output output• The number of potential predictors available isgreatly enlarged since predictors may include thosefrom different forecast times; but• As the model producing the predictors is changedand improved, the wind-specification equationsmust also be re-derived using the latest version ofthe numerical model as a predictor source.An example of this technique is given by Feit andPore (1978) for use on the Great Lakes.Table 2.7 summarizes the data requirement for eachof the approaches described above.The major difference between the approaches liesin the type of independent data used in the developmentstage. In the extrapolation approach observations areused as potential predictors, in the perfect prognosisapproach analyses are used and in the model outputstatistics approach model forecasts are used. Anotherdifference to note is that during actual operations ofthe extrapolation approach the predictors may be manuallyapplied, while in the operations of the perfectprognosis and model output statistics approach thepredictors are derived in an automated fashion frommodel forecasts.

CHAPTER 3WAVE GENERATION AND DECAYA. K. Magnusson with M. Reistad: editors3.1 IntroductionThis chapter gives an overview of the processes involvedin wave generation and decay. An indication of howthese processes are formulated is given.Wave forecasting is the process of estimating howwaves evolve as changing wind fields act on the surfaceof the ocean. To understand this we need to identify theprocesses affecting the energy of the waves. In simpleterms, wave energy at a given location is changedthrough advection (rate of energy propagated into andaway from the location), the wave energy gains from theexternal environment and wave energy losses due todissipation. In wave modelling, the usual approach is torepresent these influences as a wave energy conservationequation, as presented in Chapter 5 (Equation 5.1), andthen to solve it.The sources of wave energy (gains and losses) areidentified as three major processes, the external gains(S in ), the dissipative loss (S ds ) and the shifting of energywithin the spectrum due to weakly non-linear wavewaveinteractions (S nl ). A description of these terms aswell as of propagation are presented in this chapter.3.2 Wind-wave growthThe only input of energy to the sea surface over the timescaleswe are considering comes from the wind. Transferof energy to the wave field is achieved through the surfacestress applied by the wind and this varies roughly as thesquare of the wind speed. Thus, as already noted inSection 2.1, an error in wind specification can lead to alarge error in the wave energy and subsequently in parameterssuch as significant wave height.After the onset of a wind over a calm ocean thereare thought to be two main stages in the growth of windwaves. First, the small pressure fluctuations associatedwith turbulence in the airflow above the water are sufficientto induce small perturbations on the sea surfaceand to support a subsequent linear growth as thewavelets move in resonance with the pressure fluctuations.This mechanism is called the Phillips’ resonance(see Phillips, 1957). Formulations can be found inBarnett (1968) and Ewing (1971). However, this mechanismis only significant early in the growth of waves ona calm sea.Most of the development commences when thewavelets have grown to a sufficient size to start affectingthe flow of air above them. The wind now pushes anddrags the waves with a vigour which depends on the sizeof the waves themselves.This growth is usually explained by what is called ashear flow instability: the airflow sucking at the crestsand pushing on the troughs (or just forward of them). Auseful theory has been presented by Miles (1957). Therate of this growth is exponential as it depends on theexisting state of the sea. This is usually described interms of the components of the wave energy-densityspectrum (see Section 1.3.7).From the formulation of Miles (1960):orkE f , f g P f 2π ftµθk, e4π ρ µ– 12( ) = ( )( )( ) =Sin f,θwhere E(f,θ) is a frequency-direction component, k is thewavenumber, P(k,f) is the spectrum of wave-inducedturbulence, µ is a coupling coefficient to be defined, g isgravitational acceleration and ρ w is the water density.It has been noted that the rates of growth predicted byMiles are much smaller than observed growth ratesfrom laboratory and field studies. Based on a field experiment,Snyder and Cox (1966) proposed a simple form:µ = ρ aρ wwhere c and θ are the phase speed and the direction,respectively, of the component being generated, ψ and uare the direction and speed of the wind and ρ a is the airdensity.Measurements made in the Bight of Abaco in theBahamas in 1974 enabled Snyder et al. (1981) topropose a revision which can be expressed by:S in ( f ,θ) = E( f,θ)⎡max 0,K 1 2πf ρ a ⎛ U 5ρ w⎝ c cos( θ – ψ )–1 ⎞ ⎤⎢⎣⎠ ⎥.⎦The height at which the wind speed was specified in theoriginal work was 5 m. Since application of this to othersituations can be affected by the structure of the lowerpart of the atmospheric boundary layer (see Section 2.4),it may be better to express the wind input in terms of thefriction velocity u * with magnitude:u * =w( ) = ( )δE f,θδt⎡u⎣⎢ c cos( θ – ψ )–1 ⎤⎦⎥ ,τρ a= u C d2 πµ f E f, θ ,where τ is the magnitude of the wind stress and C d is thedrag coefficient.

36GUIDE TO WAVE ANALYSIS AND FORECASTINGThe drag coefficient, which relates u * to u, varieswith u. Komen et al. (1984) have used an approximateform to write the wind input term as:( ) = E( f,θ) (3.1)S in f ,θ⎡ ρmax 0,K a1 2πf⎛ uK *2ρ w⎝ c cos( θ – ψ )–1 ⎞ ⎤⎢⎣⎠ ⎥.⎦The constants K 1 (≈ 0.25) and K 2 (≈ 28) allow for someflexibility in specifying this term.Later research has shown that the aerodynamic dragon the sea surface depends on the state of the waves.Janssen (1991) and Jenkins (1992), using quasi-lineartheories, have found that the drag coefficient depends onthe wave age parameter defined as c p /u * , where c p is thephase velocity of the peak frequency of the actual wavespectrum. These theories imply that the aerodynamicdrag and the growth rate of the waves are higher foryoung wind sea than for old wind sea. This result is ingood agreement with experimental data by Donelan(1982) and Maat et al. (1991).There are also many empirical formulae for wavegrowth which have been derived from large data sets.These formulae make no attempt to separate the physicalprocesses involved. They represent net wave growthfrom known properties of the wind field (wind speed anddirection, fetch and duration). In representing such relations,it simplifies comparisons if the variables are allmade dimensionless:Peak frequency f p * = uf p /gFetch X* = gX/u 2Duration t* = gt/uHeight H* = gH/u 2Energy E* = Eg 2 /u 4 etc.For example, from the JONSWAP 1973 data (seeHasselmann et al., 1973, 1976; and Section 1.3.9), thepeak frequency and total energy in the spectrum arerelated to the fetch (X) and wind speed at 10 m (U 10 ) bythe following:f * p = 3.5 X *–0.33 , E * = 1.6 x 10 –7 X * (3.2)Operationally useful graphical presentations of suchempirical relations have existed since the mid-1940s andthe Sverdrup and Munk (1947) and Pierson-Neumann-James (1955) curves have been widely used (seeAnnex IV). Such relationships may also need to considerthe depth (see Section 7.5); since wave growth is affectedby the depth, with additional dissipative processes in play,the deep water curves will over-estimate the wave growthin shallow water.A more recent set of curves are those developed byGröen and Dorrestein (1976). These comprise a varietyof formats for calculating wave height and period, giventhe wind speed, fetch length and wind duration, and theeffects of refraction and shoaling. In Figure 3.1, thebasic non-dimensional graphs are displayed, showingcharacteristic height and period versus fetch and duration.It should be noted that these graphs have beenconstructed to fit visually assessed wave heights andperiods (see Section 8.3) and are thus called “characteristic”height (H c ) and period (T c ), as distinct fromsignificant height (H – 1/3) and mean period (T – z). The set ofcurves and the applications of them are given inSection 4.1.There is some uncertainty about the relationshipbetween the visually and instrumentally derived quantities,but it appears that some bias (H c and T c both beingslightly higher than H – 1/3 and T – z, respectively) may needto be kept in mind when using this type of graph. On theother hand, the systematic errors are generally smallcompared to random errors in individual observations.So far, we have considered wind waves which arein the process of growing. When the wind stops or whenthe waves propagate out of the generating area, they areoften called “swell”. Swell looks different from anordinary running sea. The backs of the waves aresmoother and the crests are long. Whereas wind wavesgrow under the influence of the wind, swell wavesdecrease in its absence. Since propagation is the mostimportant characteristic of swell waves, they arediscussed more fully in the next section.3.3 Wave propagationA disturbance on the water will travel away from thepoint at which it was generated. For a wave train withperiod T (frequency f = 1/T) and wavelength λ, the wavespeed (phase speed) is c = λ/T. This can be written alsoas ω/k, where ω is the angular frequency (2πf) and k isthe wavenumber 2ω/λ (the number of crests per unitdistance). Wave energy travels at the group velocity,which is not in general the same. For the dispersivewaves in deep water, the group velocity (c g ) is only halfof the phase velocity. As given in Section 1.3.2, this canbe worked out from the dispersion relation by usingc g =dω/dk. In very shallow water, the waves are nondispersiveas the bottom dominates the fluid’s responseto a perturbation and the group velocity equals the phasevelocity.In general, for water of finite depth, h, the dispersionrelation is:ω 2 = gk tanh khand the group speed is:khcg = 1 ω ⎛⎜k ⎝+ 2 ⎞1 ⎟2 sinh 2kh⎠(see also Section 1.3.2). For large h, this reduces to ωk/2and, for very small h, to ω/k [= √(gh)].In wave modelling, we are interested in how alocal average of the energy moves. This is not as simpleas moving the energy at one point in a straight line (ormore correctly a great circle path) across the ocean. Forany one location the energy is actually spread over a

WAVE GENERATION AND DECAY 37H c * T c *0,24-2H c * versus X*10 -15H c * versus t*210 -25106,28T c * versus X*2T c * versus t*10 -32 5 2 5 2 5 2 5 2 510 10 2 10 3 10 4 10 5t* and X*1Figure 3.1 — Basic diagram for manual wave forecasting. The curves are drawn for non-dimensional parameters(H c * = gH c /u 2 ; T c * = gT c /u) (derived from Gröen and Dorrestein, 1976)range of directions. Also, waves at different frequencieswill propagate at different speeds. Thus, from a pointsource, each component of the spectrum E(f,θ) can bepropagated in the direction θ with a speed c g (f,θ).3.3.1 Angular spreadingIt is often assumed that the directional part of the windwavedistribution has a cos 2 (θ − ψ) form where ψ is thepredominant direction of the waves and θ is the directionof the spectral component concerned. Most of theenergy is propagated in the mean direction of the sea.At deviating angles, less energy is transportedand, for all practical purposes, the energy propagatedat right angles to the mean direction is negligible. Thereis considerable evidence indicating that the spreadis dependent on the lengths of waves. Several suchformulations for the directional distribution basedon observations have been given. Using the formcos 2s (θ − ψ)/2, Mitsuyasu et al. (1975) andHasselmann et al. (1980) give functional forms for swhich are dependent on the ratio of frequency to peakfrequency. These indicate the narrowest spread at thepeak of the spectrum, with widening at both lower andhigher frequencies. However, there is still uncertaintywith respect to the real functional form of the s parameterin simple sea states (e.g. wind-sea generation andlong swells). This is due to the fact that different directionalwave instruments located very close to oneanother give widely different results (see for example,Allender et al., 1989).Waves which have left the generating area (i.e.swell) are reduced by the angular spreading. Numericalmodels automatically take care of this by splitting thespectrum into components and propagating each of thecomponents independently. Manual methods requiremore action by the operator. Angular spreading anddispersion factors must be applied.As illustrated in Figure 3.2, a point P will receivewave energy from points all along the front of the fetch.It is possible to compute the sum of all the contributions.For a cosine-squared distribution of energy at the fetchfront, results are shown in Figure 3.3. At any fixedfrequency, the curved lines in this diagram represent thepercentages of the wave energy from the fetch frontwhich reach there. These are the angular spreadingfactors. The spatial coordinates are expressed in terms ofthe width AB of the fetch area. For example, at adistance of 2.5 AB along the predominant swelldirection, the wave energy has decreased to about 25 percent of the energy per unit area which was present at thefetch front AB. The reduction in wave height due to

38GUIDE TO WAVE ANALYSIS AND FORECASTINGFetchABFigure 3.2 — Possible directions for swell originating at astorm front, AB, and incident on a point Pangular spreading is by the square root of this factor.These heights are the maximum heights which the swellmay attain.3.3.2 DispersionA further reduction needs to be applied to account fordispersion. It has been explained how long waves andtheir energy travel faster than short waves and theirenergy. The wave field leaving a generating area has amixture of frequencies. At a large distance from thegenerating fetch, the waves with low frequencies (longwaves) will arrive first, followed by waves of increasingfrequency. If the spectrum at the front edge of the generatingfetch is that represented in Figure 3.4, then, givenPthe distance from this edge and the time elapsed, it iseasy to work out the speed of the slowest wavewhich could possibly reach your point of observation.Hence, we know the maximum frequency we wouldobserve. Similarly, the length of fetch and the time atwhich generation ceases may limit the low frequencies:all the fastest waves may have passed through the monitoringpoint. The swell spectrum is therefore limited to anarrow band of frequencies (indicated by the shadedareas of the spectra in Figure 3.4). The shaded portion ofthe spectrum is the maximum that can be expected at thepoint of observation. The ratio of this area to the totalarea under the spectrum is called the wave-energydispersion factor. A consequence of this dispersion effectis that by analysing the way in which swell arrives at anobserving point, and noting the changes in the frequenciesin the swell spectrum, it is possible to get an idea ofthe point of origin of the waves.Thus, given the spectrum of wave energy exiting agenerating area and a point at which you wish to calculatethe swell, you can calculate the angular spreading factorand the dispersion factor to get an estimate of the swell.Dispersion and spreading can be considered themain causes of a gradual decrease of swell waves. Someenergy is also lost through internal friction and air resistance.This acts on all components of the wave spectrum,but is strongest on the shorter waves (higher frequencies),contributing to the lengthening of swell at increasingdistances from the source. This dissipation is often2 X101.5 XX20250.5 XA95 90 80 70 60 50 40300Fetch0.5 XBX1.5 X2 X0X2 X3 X4 X5 XFigure 3.3 — Angular spreading factors (as percentages) for swell energy

WAVE GENERATION AND DECAY 39small enough that swell can survive over large distances.Some interesting observations of the propagation ofswell were made by Snodgrass et al. (1966) when theytracked waves right across the Pacific Ocean from theSouthern Ocean south of Australia and New Zealand tothe Aleutian Islands off Alaska.The spectra of swell waves need not be narrow. Forpoints close to large generating areas, faster waves fromfurther back in the generating area may catch upwith slower waves from near the front. The result is arelatively broad spectrum. The dimensions of the wavegeneratingarea and the distance from it are thereforealso important in the type of swell spectrum which willbe observed.In Chapter 4, more detail will be given on how todetermine the spectrum of wind waves and swell and onhow to apply the above ideas.Other considerations in propagating waves are thewater depth and currents. It is not too difficult to adaptthe advection equation to take account of shoaling andrefraction. This is considered in Sections 7.2 and 7.3.For operational modelling purposes, currents have oftenbeen ignored. The influence of currents on wavesdepends on local features of the current field and wavepropagation relative to the current direction. Althoughcurrent modulation of mean parameters may be negligiblein deep oceans, and even in shelf seas for ordinarypurposes, the modulation of spectral density in the highfrequencyrange may be significant (see, for instance,Tolman, 1990). Waves in these frequencies may even beblocked or break when they propagate against a strongcurrent, as in an estuary. For general treatment of thissubject see Komen et al. (1994). The gain in wave forecastquality by taking account of currents in aforecasting model is not generally considered worthwhile compared to other factors that increase computingtime (e.g. higher grid resolution). Good quality currentfields are also rarely available in larger ocean basins.3.4 DissipationWave energy can be dissipated by three different processes:whitecapping, wave-bottom interaction and surfbreaking. Surf breaking only occurs in extremely shallowwater where depth and wave heights are of thesame order of magnitude (e.g. Battjes and Janssen,1978). For shelf seas this mechanism is not relevant. Anumber of mechanisms may be involved in the dissipationof wave energy due to wave-bottom interactions. Areview of these mechanisms is given by Shemdin et al.(1978), which includes bottom friction, percolation(water flow in the sand and the sea-bed) and bottommotion (movement of the sea-bed material itself). InSections 7.6 and 7.7, dissipation in shallow water willbe discussed more fully.The primary mechanism of wave-energy dissipationin deep and open oceans is whitecapping. As wavesgrow, their steepness increases until a critical point whenthey break (see Section 1.2.7). This process is highlynon-linear. It limits wave growth, with energy being lostinto underlying currents. This dissipation depends on theexisting energy in the waves and on the wave steepness,and can be written:( ) = ( ) ( )S f E f ds , θ – ψ E f, θ ,f(3.3)where ψ(E) is a property of the integrated spectrum, E.ψ may be formulated as a function of a wave steepnessparameter (ξ = Ef –4 /g 2 , where f – is the mean frequency).Forms for ψ have been suggested by Hasselmann (1974)and Komen et al. (1984).There are also the processes of micro-scale breakingand parasitic capillary action through which waveenergy is lost. However, there is still much to learn aboutdissipation and usually no attempt is made to distinguishthe dissipative processes. The formulation of S ds stillrequires research.Manual wave calculations do not need to payspecific attention to dissipative processes. Generally,the dissipation of wind waves is included implicitly inthe overall growth curves used. Swell does suffer alittle from dissipative processes, but this is minor. It isobserved that swell can travel over large distances.Swell is mostly reduced by dispersion and angularspreading.3.5 Non-linear interactionsIn our introduction we noted that simple sinusoidalwaves, or wave components, were linear waves. This isan approximation. The governing equations admit more2Figure 3.4 —The effect of dispersion on waves leavinga fetch. The spectrum at the fetchfront is shown in both (a) and (b). Thefirst components to arrive at a point fardownstream constitute the shaded partof the spectrum in (a). Higher frequenciesarrive later, by which time some ofthe fastest waves (lowest frequencies)have passed by as illustrated in (b)EFrequency(a)Frequency(b)

40GUIDE TO WAVE ANALYSIS AND FORECASTING0.2MEAN JONSWAPSPECTRUMFigure 3.5 —Growth from non-linear interactions(S nl ) as a function offrequency for the meanJONSWAP spectrum (E)24Energy E(m 2 /Hz)0.1S nlE168S x 10 6 (m 2 )0.00S nl-80.0 0.2 0.4 0.6 0.8Frequency Hzdetailed analysis. Whilst the theory is restricted by arequirement that the waves do not become too steep, theweakly non-linear interactions have been shown to bevery important in the evolution of the wave spectrum.These weakly non-linear, resonant, wave-waveinteractions transfer energy between waves of differentfrequencies, redistributing the energy within the spectrumin such a way that preserves some characteristics ofthe spectral shape, i.e. a self-similar shape. The processis conservative, being internal to the wave spectrum andnot resulting in any change to the overall energy contentin the wave field.The resonance which allows this transfer betweenwaves can be expressed by imposing the conditions thatthe frequencies of the interacting waves must sum tozero and likewise the wavenumbers. This first occurs atthe third order of a perturbation analysis in wave energy(as shown by Hasselmann, 1962) and the integrals whichexpress these energy transfers are complicated cubicintegrals:Snl ( f,θ)=f∫ ∫ ∫ dk1dk2dk3δ ( k1+ k2– k3–k)δ ( f1+ f2–f3−f)(3.4)nn n + n – nn n n[ 1 2( 3 ) 3 ( 1+2)]( )K k , k , k , k .1 2 3In this integral, the delta functions, δ, enforce the resonanceconditions, the (f i , k i ) for i = 1, 2, 3 are thefrequency and wavenumber pairs for the interactingwave components (related as in Equation 1.3), then i = E(f i ,θ i )/f i , are the wave action densities, and theKernel function, K, gives the magnitude of the energytransfer to the component k, (or (f,θ)) from each combinationof interacting wave components.This interactive process is believed to be responsiblefor the downshift in peak frequency as a wind seadevelops. Figure 3.5 illustrates the non-linear transferfunction S nl (f) calculated for a wind sea with an energydistribution, E(f i ,θ i ), given by the mean JONSWAP spectrum(see Section 1.3.9). The positive growth just belowthe peak frequency leads to this downshift.Another feature of the non-linear wave-wave interactionsis the “overshoot” phenomenon. Near the peak,Energy, ELinear growthExponential growthOvershootFetch, XSaturationFigure 3.6 — Development of wave energy at a singlefrequency along an increasing fetch, illustratingthe various growth stages

WAVE GENERATION AND DECAY 41Figure 3.7 —Structure of spectral energy growth.The upper curves show the componentsS nl , S in , and S ds . The lowercurves are the frequency spectrumE(f), and the total growth curve SEnergy density (E) Rate of growth (S)E (f)fSfgrowth at a given frequency is dominated by the nonlinearwave-wave interaction. As a wind sea develops (oras we move out along a fetch) the peak frequencydecreases. A given frequency, f e , will first be well belowa peak frequency, resulting in a small amount of growthfrom the wind forcing, some non-linear interactions, anda little dissipation. As the peak becomes lower andapproaches f e , the energy at f e comes under the influenceof a large input from non-linear interactions. This can beseen in Figures 3.5 and 3.7 in the large positive region ofS or S nl just below the peak. As the peak falls below f ethis input reverses, and an equilibrium is reached(known as the saturation state). Figure 3.6 illustrates thedevelopment along a fetch of the energy density at sucha given frequency f e .Although the non-linear theory can be expressed,as in Equation 3.4, the evaluation is a problem. Theintegral in Equation 3.4 requires a great deal ofcomputer time, and it is not practical to include it inthis form in operational wave models. Some wavemodels use the similarity of spectral shape, which is amanifestation of this process, to derive an algorithm sothat the integral calculation can be bypassed. Havingestablished the total energy in the wind-sea spectrum,these models will force it into a pre-defined spectralshape. Alternatively, it is now possible to use integrationtechniques and simplifications which allow areasonable approximation to the integral to be evaluated(see the discrete interaction approximation (DIA)of Hasselmann and Hasselmann, 1981, Hasselmannand Hasselmann, 1985, and Hasselmann et. al., 1985;or the two-scale approximation (TSA) of Resio et al.,1992). These efficient computations of the non-lineartransfer integral made it possible to develop thirdgeneration wave models which compute the non-linearsource term explicitly without a prescribed shape forthe wind-sea spectrum.Resonant weakly non-linear wave-wave interactionsare only one facet of the non-linear problem.When the slopes of the waves become steeper, and thenon-linearities stronger, modellers are forced to resort toweaker theories and empirical forms to representprocesses such as wave breaking. These aspects havebeen mentioned in Section 3.4.3.6 General notes on applicationThe overall source term is S = S in + S ds + S nl . Ignoringthe directional characteristics (i.e. looking only at thefrequency dependence), we can construct a diagram forS such as Figure 3.7. This gives us an idea of the relativeimportance of the various processes at different frequencies.For example, we can see that the non-linear transferis the dominant growth agent at frequencies near thespectral peak. Also, for the mid-frequency range (fromthe peak to about twice the peak frequency) the growthis dominated by the direct input from the atmosphere.The non-linear term relocates this energy mostly to thelower frequency range. The dissipation term, so far as isknown, operates primarily on the mid- and highfrequencyranges.The development of a frequency spectrum along afetch is illustrated in Figure 3.8 with a set of spectrameasured during the JONSWAP experiment. The downshiftin peak frequency and the overshoot effect at eachfrequency are evident.

42GUIDE TO WAVE ANALYSIS AND FORECASTINGEnergy, E (m 2 /Hz)0.7 Figure 3.8 —Growth of a frequency spectrum along afetch. Spectra 1–5 were measured at0.65distances 9.5, 20, 37, 52, and 80 km,respectively, offshore. The wind is7 m/s (after Hasselmann et al., 1973)0.50.40.3430.220.1100.1 0.2 0.3 0.4 0.5 0.6 0.7Frequency (Hz)Application of the source terms is not alwaysstraightforward. The theoretical-empirical mix in mostwave models allows for some “tuning” of the models.This also depends on the grid, the boundary configuration,the type of input winds, the time-step, the influenceof depth, computer power available, etc. Manualmethods, on the other hand, conform to what may beregarded as fairly universal rules and can usually beapplied anywhere without modification.For manual calculations, the distinction betweenwind sea and swell is quite real and an essential part ofthe computation process. For the hybrid parametricmodels (see Chapter 5) it is necessary, although theproblem of interfacing the wind-wave and swellregimes arises. For numerical models which use spectralcomponents for all the calculations (i.e. discretespectral models) the definition of swell is quite arbitrary.From the spectrum alone there is no hard and fastrule for determining the energy which was generatedlocally and that which was propagated into the area.This can pose problems when attempting to interpretmodel estimates in terms of features traditionallyunderstood by the “consumer”. Many users are familiarwith, and often expect, information in terms of “windsea” and “swell”. One possible algorithm is to calculatethe Pierson-Moskowitz spectrum (Section 1.3.9) forfully developed waves at the given local wind speedand, specifying a form for directional spreading(Section 3.3.1), allocate energy in excess of this toa swell spectrum. More sophisticated algorithmswill assess a realistic cut-off frequency for the localgeneration conditions, which may be considerablyhigher than the the peak frequency from the Pierson-Moskowitz spectrum and hence also designate as swellexcess low-frequency wave energy at directions nearthe wind.

CHAPTER 4WAVE FORECASTING BY MANUAL METHODSL. Burroughs: editor4.1 IntroductionThere are many empirical formulae for wave growthwhich have been devised from large visually observeddata sets. There are also formulae, more recentlyderived, which are based on wave measurements. Theseformulae make no attempt to separate the physicalprocesses involved. They represent the net wave growthfrom known properties of the wind field (wind speed anddirection, fetch and duration).There are some inherent differences betweenvisually and instrumentally observed wave heightsand periods which affect wave prediction. In general,the eye concentrates on the nearer, steeper waves and sothe wave height observed visually approximatesthe significant wave height (H – 1/3), while thevisually observed wave periods tend to be shorter thaninstrumentally observed periods. There are severalformulae which have been used to convert visual datato H – 1/3 more accurately. For almost all practicalmeteorological purposes, it is unlikely to be worth thetransformation. Operationally useful graphical presentationsof such empirical relations have existed since themid-1940s.The curves developed by Sverdrup and Munk(1947) and Pierson, Neumann and James (1955) (PNJ)are widely used. These two methods are similar in thatthe basic equations were deduced by analysing a greatnumber of visual observations by graphical methodsusing known parameters of wave characteristics.However, they differ fundamentally in the way in whichthe wave field is specified. The former method describesa wave field by a single wave height and wave period(i.e. H – 1/3 and T – H 1/3) while the latter describes a wave fieldin terms of the wave spectrum. The most obviousadvantage of the PNJ method is that it allows for amore complete description of the sea surface. Its majordisadvantage is the time necessary to make thecomputations.A more recent set of curves has been developed byGröen and Dorrestein (1976) (GD). These curvescomprise a variety of formats for calculating waveheight and period given the wind speed, fetch length,wind duration, and the effects of refraction and shoaling.These curves differ little from those found in PNJexcept that the wave height and period are called thecharacteristic wave height (H c ) and period (T c ) ratherthan H – 1/3 and T – – H1/3 , and mks units are used rather thanfeet and knots. Both PNJ and GD are derived fromvisually assessed data. The only difference between“characteristic” and “significant” parameters (i.e. H cand H – 1/3, and T c and T – H 1/3) is that H c and T c are biasedslightly high when compared to H – 1/3 and T – H 1/3, whichare assessed from instruments. However, the differencesare insignificant for all practical purposes, and H cand T c are used throughout this chapter to be consistentwith the GD curves.Figure 4.1 shows the GD curves for deep water.This figure is of the form introduced in Figure 3.1(Section 3.2) and will be used in wave calculations inthis chapter. The PNJ curves are presented for comparisonin Annex IV. For example, in Figure 4.1, thickdark lines represent the growth of waves along increasingfetch, which is shown by thin oblique lines. Eachthick line corresponds to a constant wind speed. Thecharacteristic wave height, H c , is obtained from thehorizontal lines and the characteristic period T c fromthe dotted lines. The vertical lines indicate duration atwhich that stage of development will be reached. If theduration is limited, the waves will not develop along thethick dark lines beyond that point irrespective of thefetch length.It should be noted that the curves are nearly horizontalon the right-side of the diagram. This implies thatfor a given wind speed, the waves stop growing when theduration or fetch is long enough.Of the formulae which have been devised frommeasured wave data, the most noteable are from theJONSWAP experiment which was introduced inChapter 1, Section 1.3.9 (see also Figure 1.17 andEquation 4.1).In this chapter several manual forecasting examplesare presented. Each example is designed to show how tomake a forecast for a given set of circumstances and/orrequirements. Some empirical working procedures arebriefly mentioned in Section 4.2. These procedures haveproved their value in actual practice and are alluded to inSections 4.3 and 4.4. In Section 4.3 examples highlightingthe various aspects of computing wind waves areexplained. Examples of swell computations are given inSection 4.4. In Sections 4.3 and 4.4 all the examples arerelated to deep-water conditions. In Chapter 7 shallowwatereffects on waves will be discussed, and in Section4.5 a few examples of manual applications related toshallow (finite-depth) water conditions are presented.Table 4.1 provides a summary of each example given,and indicates the sub-section of Chapter 4 in which itcan be found.The explanations given in this chapter and inChapter 2 provide, in principle, all the material

44GUIDE TO WAVE ANALYSIS AND FORECASTING302014108654321.410.80.70.60.50.40.30.6Wave height: H c in mWind speed uin m/s301251.52017.51512.5102346100 n. mi. = 185 km; 100 km = 54 n. mi.1 kn = 0.51 m/s; 1 m/s = 1.94 kn10Fetch X in km15203040602100200150300 400 600Wave period T c in s310004150020003000 4000568791017161514131211Wind speed u.in m/s30252017.515.012.510.07.550.27.55Wind duration in hours0.10.1 0.5 1 2 2 6 9 12 18 24 36 48 72 96Figure 4.1 — Manual wave forecasting diagram (from Gröen and Dorrestein, 1976)TABLE 4.1Specific examples of manual forecasting presented in Chapter 4and associated sub-section numberDescriptionSub-sectionNo.1 Determining sea-state characteristics for a given wind speed and fetch 4.3.12 Determining sea state for an increasing wind speed 4.3.23 Extrapolating an existing wave field with further development from a constant wind 4.3.34 Extrapolating an existing wave field with further development from an increasing wind 4.3.45 Computing when swell will arrive, what periods it will have and how these periods will change for 36 hafter arrival begins 4.4.16 Same as 5, except for a long fetch in the generation area 4.4.27 Computing swell characteristics at Casablanca for a storm of nearby origin 4.4.38 Estimating the swell heights for the cases described in 5, 6 and 7 4.4.49 Determining wave number and shoaling factor for two wave periods and several representative depths 4.5.1.110 Finding the shallow water refraction factor and angle, for a given deep water wave angle with the bottom,shallow depth, and wave period 4.5.1.211 Finding the refraction factor by Dorrestein’s method 4.5.1.312 Finding the shallow water height and period for a given wind speed, shallow water depth, and fetch 4.5.2

WAVE FORECASTING BY MANUAL METHODS 45necessary to forecast waves for a particular location andalso to perform a spatial wave analysis (i.e. analyse awave chart) by manual methods. A great deal of experienceis necessary to analyse an entire chart withinreasonable time limits, mainly because one has to workwith constantly changing wind conditions.Generally, one starts from known wave and windconditions, say 12 h earlier, and then computes, usingthe present analysed wind chart, the correspondingwave chart. In cases of sudden wind changes, the intermediatewind chart from 6 h earlier may also beneeded. The forecast wind in the generation area andthe forecast movement of the generation area are alsonecessary to produce the best wave forecast over thenext 24 to 36 h.4.2 Some empirical working proceduresThree empirical procedures are briefly discussed in thissection. They are concerned with freshening of the windat constant direction, changing wind direction, andslackening of the wind. These procedures are usefulwhen time is short for preparing a forecast.4.2.1 Freshening of the wind at constantdirectionFreshening wind at constant direction is a frequentoccurrence and the procedure described in Section 4.3.4should be used. For quick calculations: subtract onequarterof the amount the wind has increased from thenew wind speed, and work with the value thus obtained.Example: the wind speed has increased from 10 to 20 kn(1.94 kn = 1 m/s) over the last 12 h; to compute thecharacteristic wave height, use a wind speed of 17.5 knover a duration of 12 h. When sharp increases of windspeed occur, it is advisable to perform the calculation intwo stages.4.2.2 Changing wind directionIf the direction changes 30° or less, wave heights andperiods are computed as if no change had occurred; thewave direction is assumed to be aligned with the newdirection. At greater changes, the existing waves aretreated as swell, and the newly generated waves arecomputed with the new wind direction.4.2.3 Slackening of the windWhen the wind speed drops below the value needed tomaintain the height of existing waves, the waves turn intoTABLE 4.2Characteristic parameters of wind waves–H c T c T p TH1/3 fRange ofp Hperiods max(m) (s) (s) (s) (s –1 ) (s) (m)5 9 10.3 9.3 0.097 5 –15 10swell and should be treated as such. As a first approximation,swell may be reduced in height by 25 per cent per12 h in the direction of propagation. For instance, swellwaves 4 m high will decrease to 3 m in 12 h.4.3 Computation of wind waves4.3.1 Determining sea-state characteristics for agiven wind speed and fetchProblem:Determine the characteristics of the sea state for a windspeed of 15 m/s (about 30 kn), with a fetch of 600 km(about 325 nautical miles (n.mi.)) and after a duration of36 h.Solution:According to the diagram given in Figure 4.1, the fetchis the limiting factor. For a fetch of 600 km, the characteristicwave height (H c ) is 5 m and the characteristicperiod (T c ) is 9 s.Other characteristics can be obtained as well. Fromthe JONSWAP spectrum (see Section 1.3.9, Figure 1.17,and Equation 4.1), we can find the peak frequency andperiod, and from them the range of important waveperiods and the period of the highest one-third of thewaves (significant wave period, T – H 1/3). From T – H 1/3wecan derive the highest wave in a 6-h period.The peak frequency f p is given by:f p = 0.148 H –0.6 m0 u 0.2 (4.1)where H m0 is the model wave height in metres, and u iswind speed in metres per second. The model period isT p =1/f p . Assume H m0 ≈ H c = 5 m, and u = 15 m/s, thenf p = 0.097 s –1 , and T p = 10.3 s.To determine T – , we use Goda’s (1978) results inH1/3an approximate form:T – H1/3 ≈ 0.9 T p . (4.2)For our problem, T – ≈ 0.9 x 10.3 = 9.3 s.H1/3The range of important wave periods can be determinedfrom Figure 1.17, where f (= 1/T) varies fromabout 0.7 f p to about 2.0 f p . This translates to a range of5 to 15 s. The maximum energy in the spectrum will benear the period of 10 s.For a record of wave heights with about 2 000 wavemeasurements, we can use the following approximationto find the maximum wave height:H max ≈ 2.0 H – 1/3 ≈ 10 m.Table 4.2 presents all these results.4.3.2 Determining sea state for an increasingwind speedProblem:An aeroplane has had to ditch at sea 200 km from shore.The closest ship is positioned 600 km from shore. Thewind speed over the last 24 h has been steady at 17 m/s.

46GUIDE TO WAVE ANALYSIS AND FORECASTING600 km200 kmAeroplaneShipX 1WindDuring the preceding 24 h, it had gradually increasedfrom 13 to 17 m/s. The wind direction remained constantfor the entire period and at an angle of 30° from land tosea. Figure 4.2 illustrates the situation.Forecast the sea conditions for:• The point at which the aeroplane ditched in order todetermine whether a search and rescue seaplanecould land and rescue the pilot, or whether thenearest vessel should be dispatched — whichwould take more time; and• The position of the ship.Solution:The effective fetch for the landing place of the aeroplaneis:200 200X 1 =400sin 30° = 05 .= kmand for the position of the ship is600 600X 2 =1200sin 30° = 05 .= km.This is an example of more complicated winddurationconditions. During the first full 24 h period, thewind increased steadily from 13 to 17 m/s. RecallingSection 4.2.1, we divide the 24 h period into two 12 hperiods with the wind increasing from 13 and 15 m/s inthe first half and from 15 to 17 m/s in the second. Onequarterof the increase in each period is 0.5 m/s.Subtracting this from 15 m/s and 17 m/s, gives 14.5 m/sand 16.5 m/s, respectively.TABLE 4.3Values of wave parameters at the locations of the ditchedaeroplane and the shipX 2CoastFigure 4.2 — Illustration of the situation in Problem 4.3.2Use of the diagram in Figure 4.1 shows that, for awind speed of 14.5 m/s, wind waves reach a height of3.7 m after 12 h. The same height would be attainedafter 8 h, at a speed of 16.5 m/s. Therefore, instead ofworking with a wind speed of 14.5 m/s over the firstinterval, we work with a wind speed of 16.5 m/s, but anequivalent duration of 8 h. The equivalent duration forthe two periods amounts to 8 h + 12 h = 20 h. Thediagram shows that, for a duration of 20 h and windspeed of 16.5 m/s, the wave height is 5.3 m.At the beginning of the second full 24 h period,wave heights are 5.3 m, and the wind speed is 17 m/sand remains constant throughout the period. In order towork with a wind speed of 17 m/s, we determine theequivalent duration necessary to raise waves to a heightof 5.3 m. It is 16 h. Thus, the equivalent duration for a17 m/s wind speed is 16 h + 24 h or 40 h. For theseconditions, H c = 6.5 m and T c = 10 s.For these conditions to occur, a minimum fetch of1 050 km is required. At the position of the nearest ship,there is no fetch limitation, but at the position of theditched aeroplane there is. From Figure 4.1 we see thatfor a fetch of 400 km and a wind speed of 17 m/s,H c = 5.6 m and T c = 9 s.Table 4.3 shows the values of other wave parameterswhich can be computed from u and H c for bothlocations.4.3.3 Extrapolation of an existing wave fieldwith further development from a constantwindProblem:Figure 4.3 shows a wave field at time t 0 . Forecast thecharacteristics of the sea state at point B at the timet 0 + 12 h, with a constant west wind of 17.5 m/s blowingat and to the west of point B.Solution:The starting point for the calculation cannot be atpoint B because the waves there move away from it. Wemust look for a starting point upwind at a point wherewaves have travelled from time t 0 to arrive at B at timet 0 + 12 h.To estimate how far upwind A should be from B,we pick a point where the wave height is 4 m. FromFigure 4.1 we see that waves with H c = 4 m and u = 17.5m/s have T c = 7 s. After 12 h, T c increases to about 9 s,2 m3 m4 mA BH c T c u T p– TH 1/3 H max(m) (s) (m/s) (s) (s) (m)Aeroplane 5.6 9 17 10.8 9.7 11.2Ship 6.5 10 17 11.8 10.6 13Figure 4.3 — Wave field at time t 0 indicated by lines of equalwave height

WAVE FORECASTING BY MANUAL METHODS 47TABLE 4.4Additional wave information for Problem 4.3.3TABLE 4.5Values of parameters computed from H c for Problem 4.3.4H c T c u T p– TH 1/3 H max(m) (s) (m/s) (s) (s) (m)Point B 6 9.3 17.5 11.2 10 12H c T c u T p– TH 1/3 H max(m) (s) (m/s) (s) (s) (m)Point B 9.2 10.6 27.5 13.2 11.9 18.4and the mean period is about 8 s. The wave propagationrate is given by c = gT c /2π and the group velocity isgiven by c g = c/2 in deep water. Since g/2π = 1.56 m/s 2(or 3.03 kn/s) then the group velocity c g ≈ 0.78 T c m/s or≈ 1.5 T c kn. Waves with T c = 8 s have c g ≈ 12 kn and, in12 h, cover a distance of 144 n.mi., or 2.4° of latitude.Dividing 2.4 by T c = 8 gives 0.3. This example showsthat the travel distance over 12 h, expressed in degrees oflatitude, is 0.3 T c . This is an easy formula to use to determinehow far upwind to put A.H c equals 4.2 m at time t 0 at point A. To producewaves of that height, a wind of 17.5 m/s needs an equivalentduration of 8 h. H c at point B at t 0 + 12 h can bedetermined by taking a total duration 8 h + 12 h = 20 h.From Figure 4.1, H c = 6 m and T c = 9.3 s. Table 4.4gives additional information.4.3.4 Extrapolation of an existing wave fieldwith further development from anincreasing windProblem:The situation at t 0 is the same as in Figure 4.3, but nowthe wind increases from 17.5 m/s at t 0 to 27.5 m/s att 0 + 12 h over the area which includes the distance AB.Forecast the sea state at Point B.Solution:Because the increase in wind speed is so great, divide thetime period into two 6 h periods with winds increasingfrom 17.5 m/s at t 0 to 22.5 m/s, and 22.5 m/s to 27.5 m/s,respectively. Recalling Section 4.2.1, the correspondingspeeds used to compute the waves are 21 m/s (21.25)and 26 m/s (26.25).At t 0 , H c equals 4.2 m. Waves grow to this heightwith 21 m/s winds after 5 h. Over the first period thenthe equivalent duration with u = 21 m/s is 5 h + 6 h =11 h. At the end of the interval, H c = 6.5 m.Waves would reach the 6.5 m height after an equivalentduration of 5.5 h with a wind of 26 m/s. Over thesecond 6 h period then, we may work with an equivalentduration of 11.5 h and a wind of 26 m/s. Thus,H c = 9.2 m, and T c = 10.6 s (see Table 4.5).Actually these waves would have passed point B,since the average wave period during the 12 h period inthis case was nearly 9 s. Their travel distance would beslightly greater than 2.6° of latitude. This small adjustmentto the distance AB would not have influenced thecalculations. In routine practice the travel distances arerounded to the nearest half degree of latitude.4.4 Computation of swellFor most practical applications two different types ofsituation need to be distinguished:(a) Swell arriving at the point of observation from astorm at a great distance, i.e. 600 n.mi. or more. Inthis case, the dimensions of the wave-generatingarea of the storm (e.g., tropical cyclone) can beneglected for most swell forecasting purposes, i.e.the storm is regarded as a point source. The importanteffect to be considered is wave dispersion;(b) Swell arriving at the point of observation from anearby storm. Swell fans out from the points alonga storm edge. Because of the proximity of the stormedge, swell may reach the point of observationfrom a range of points on the storm front. Therefore,apart from wave dispersion, the effect ofangular spreading should also be considered.In swell computations, we are interested in thepropagation of wave energy. Therefore, the group velocityof individual wave components, as approximated byrepresentative sinusoidal waves, should be considered.Since large distances are often involved, it is moreconvenient to measure distances in units of nauticalmiles (n.mi.) and group velocities in knots (kn). Thewave period T is measured in seconds as usual. We thenhave (as in Section 1.3.2):c gTcg = = 1= 13. 03T= 1. 515T(kn).2 2 2π2(4.3)4.4.1 Distant stormsIn the case of a distant storm (Figure 4.4), the questionsto be answered in swell forecasting are:(a) When will the first swell arrive at the point ofobservation from a given direction?(b) What is the range of wave periods at any giventime?(c) Which wavelengths are involved? and, possibly(d) What would the height of the swell be?The known data to start with are the distance,R p (n.mi.), from the storm edge to the point of observation,P, the duration, D p , of wave generation in thedirection of P, and the maximum wave period in thestorm area.Because they travel faster, the wave componentswith the maximum period are the first to arrive at P.Their travel time is:

48GUIDE TO WAVE ANALYSIS AND FORECASTINGStormwith period T 1 has started out D p hours later and has thustravelled over a time of (t – D p ) hours. We have, for theslow component:RpRpT2= = 0. 660 ( s)(4.5)1.515 t tStormand, for the fast component:orTR c t– D 1. 515 T 1 t–D= ( ) = ( )p g p pR1 = ( ) =p1. 515 t–DpRp0.660t–Dp() s(4.6)R pA possible new swellarriving later(R p is measured in n.mi.; t and D p in hours.)T 1 and T 2 are the limits of all wave periods whichcan possibly be present at P at a given time of observation.Some periods within this range may actually notbe present in the observed wave spectrum; the componentscould be dissipated during their long traveloutside the storm area. It can easily be shown from theabove equations that the range of possible wavefrequencies is given by:Dpf1 – f2 = 1. 515 .R(4.7)p21PRThis means that the band width (range) of frequencies ofwave components, which exist at a given point of observation,is a constant for that point and depends on theduration of wave generation, D p. The range of frequenciesbecomes smaller at greater distances from the storm.This result, obtained from a schematic model, is indeedobserved. Thus, as the result of wave dispersion, swellattains a more regular appearance at greater traveldistances.Example of swell from a distant stormFigure 4.4 — Swell from a distant storm. Waves travelling inthe direction of P have been generated over atime period D pRpRpRpt = = = 0. 660 (h).c 1.515TTg(4.4)These components continue to arrive over a period of D phours, then they disappear. D p is determined fromweather charts by examining how long a given fetchremained in a given area. In the meantime, slower wavecomponents have arrived and each component isassumed, in the present example, to last for D p hours.There is a wave component which is so slow that it startsto arrive at the moment when the fastest wave componentpresent is about to disappear (compare boxes 1and 2 in Figure 4.4). The first wave of the slow component(box 2) with period T 2 has travelled over a timeof t hours; the last wave of the fast component (box 1)Problem:Waves were generated in the direction R for a period of18 h. The highest wave period generated in the stormwas 15 s. Forecast swell for point A at 600 n.mi. and forpoint B at 1 000 n.mi. from the area of generation.Compute when the first waves arrive and which periodscould possibly be present during the subsequent 36 h.Solution:At point A, R p = 600 n.mi.; D p = 18 h; T max = 15 s.From Equation 4.4 the first waves arrive att = 0.660 x 600/15 = 26.4 h after the beginning of thestorm. These waves continue for 18 h after they firstarrive.The range of periods (T 1 – T 2 ) at point A arecomputed for the 36 h subsequent to the arrival timeof the first wave at 6 h intervals beginning with 30 hafter the storm as shown in Table 4.6. The rangeof wavelengths can also be given using the relationλ = 1.56 T 2 (m). The wave components with a period of15 s disappear after t = 44.4 h.

WAVE FORECASTING BY MANUAL METHODS 49TABLE 4.6Ranges of swell periods and wavelengths at point A forarrival time after storm beginningArrival time (h) Periods (s) Wavelengths (m)30 15.0–13.2 351–27236 15.0–11.0 351–18942 15.0–9.4 351–13848 13.2–8.2 272–10554 11.0–7.3 189–8360 9.4–6.6 138–6866 8.2–6.0 105–56SOStorm edgeR pFigure 4.5 — Swell from a quasi-stationary, distant storm, inwhich the waves travel over a distance R p , andthe generation area has a long fetchPRFor point B, R p = 1 000 n.mi.; D p = 18 h; T max = 15 s.The first waves arrive 44 h after the beginning ofthe storm. Table 4.7 shows the same information asTable 4.6, but now for point B, starting at 48 h andending at 84 h. The wave components with a period of15 s disappear after t = 62 h.Rather long forecast times have been given for bothpoints to demonstrate the gradual change of waveperiods. In practice, the shorter waves may not benoticeable after two to three days’ travel time, and alsoafter displacement of the storm in the case of a tropicalcyclone. A cyclone does not, however, generate swell inall directions; this depends on the structure of the windfields in the cyclone.4.4.2 Distant storm with long fetchForecasting the waves from a distant storm with long fetchis a more complicated case, since the distance travelled byindividual wave components inside the wave-generatingarea will generally not be the same for the various components.The longer and larger waves will generally befound in the downwind part of the storm area.For all practical purposes, one may choose an appropriatemean value for the distance S (see Figure 4.5) andapply a corrected duration D ṕ by adding to D p the timeneeded by wave components to cover the distance S:SSDp′ = Dp+ = Dp+c 1.515T1giTABLE 4.7Same as Table 4.6 except at point B(4.8)in which c gi is the group speed of the component considered.It can be shown that in this case the range of wavefrequencies of the swell does not remain constant for agiven point P, but increases slightly as the larger componentsdisappear, and the spectrum consists ofprogressively smaller components.Example of swell from distant storm with long fetchProblem:Referring to Figure 4.5, waves were generated in thedirection R. The “mean” generation fetch is 180 n.mi.for waves with periods between 12 and 15 s; R p =600 n.mi.; D p = 18 h. Find the wave conditions at P.Solution:The corrected duration for waves with T = 15 s, isD p = 18 h + (0.660 x 180/15) = 18 h + 8 h = 26 h. Thiscomponent arrives at P, 26.4 h after the storm as before,but disappears 26 h later. Likewise the corrected durationfor generation of waves with T = 12 s is D p = 27.9 h. Thetravel time for this component to arrive at point P ist = 0.660 x (600/12) = 33 h. The last waves with T = 12 spass point P at t = 33 h + 27.9 h = 60.9 h.The ranges of periods and wavelengths arereflected in Table 4.8.Comparison of the examples in Sections 4.4.1 and4.4.2 shows that, in the latter, the wave spectrum remainsTABLE 4.8Ranges of swell periods and wavelengths at point P for arrivaltimes after the beginning the stormArrival time (h) Periods (s) Wavelengths (m)48 15.0–13.8 351–29754 15.0–12.2 351–23260 15.0–11.0 351–18466 13.8–10.0 297–15672 12.2–9.2 232–13278 11.0–8.5 189–11384 10.0–7.9 156–97Arrival time (h) Periods (s) Wavelengths (m)30 15.0–13.2 351–27236 15.0–11.0 351–18942 15.0–9.4 351–13848 15.0–8.2 351–10554 14.1–7.3 310–8360 12.0–6.6 225–6866 10.4–6.0 169–56

50GUIDE TO WAVE ANALYSIS AND FORECASTINGbroader for a longer period. Because of the wide rangeof energetic periods, the swell will also have a less regularappearance.4.4.3 Swell arriving at a point of observationfrom a nearby storm.As pointed out in the introduction to Section 4.4, swell,fanning out from different points of a nearby storm edge(less than 600 n.mi.), may reach the observation point.When forecasting swell from a nearby storm, therefore,the effect of angular spreading should be considered inaddition to wave dispersion.For swell estimates:• The sea state in the fetch area which has influenceon the forecast point must be computed;• The distance from the leading edge of the fetch areato the observation point measured;• The period of the spectral peak, and the range ofwave periods about the peak found;• The arrival time of the swell at the forecast pointdetermined;• The range of periods present at different timescalculated; and• The angular spreading factor and the wave dispersionfactor at each forecast time determined.The angular spreading can be calculated by usingthe width of the fetch area and the distance from thefetch area to the forecast point in Figure 3.3 (see alsoSection 3.3). This factor is a percentage of the energy,so, when applied to a wave height the square root mustbe taken.From the JONSWAP results, Hasselman et al.(1976) proposed a relation between sea-surface variance(wave energy) and peak frequency for a wide range ofgrowth stages. Transforming their results into terms ofH m0 and f p givesH m0 = 0.414 f p –2 (f p u) 1/3 . (4.9)Equation 4.9 together with theJONSWAP spectrum (Figure 1.17)and PNJ can be used to find thewave dispersion factor at eachforecast time at the forecast point.Figures 3.4(a) and (b) illustratehow the wave spectrum dispersesover time. This is illustrated in thefollowing example.Problem:Figure 4.6 shows the storm thatproduced waves which we want toforecast at Casablanca. A review ofprevious weather charts showedthat, in the past 24 h, a cold fronthad been moving eastward. It travelledslowly, but with sufficientspeed to prevent any waves fromLmoving out ahead of the fetch. At chart time, the front wasslowing down, and a secondary low started to develop. Theforecast indicated that, as the secondary low intensified,the wind in the fetch area would change to become a crosswindfrom the south. It was also expected that the frontwould continue its movement, but the westerly winds inthe rear would decrease in strength. At chart time the welldevelopedsea that existed in the fetch area would nolonger be sustained by the wind.The fetch area (hatched area in Figure 4.6) was480 n.mi. long and 300 n.mi. wide at chart time. Thewinds in the area were WNW. The distance toCasablanca from the leading edge of the fetch (R c ) is600 n.mi. Determine all of the swell characteristics atCasablanca and their direction.Solution:During the past 24 h, the average wind speed wasu = 15 m/s in the fetch area; for that wind H c = 4.8 m,and T c = 8.6 s. From Equation 4.1, T p = 10.1 s, and therange of important wave periods is from 14.4 s down to5.0 s (2 f p to 0.7 f p ).The first waves with a period of 14.4 s will arrive atthe coast at0.660 x 600t = = 27.5h14.4after chart time. Also, the waves with a period of 14.4 scease to arrive at the coast in( ) =0.660 x 480 + 600t =14.449.5h.Since the time to get from the rear of the fetch areafor the wave components with T = 14.4 s is less than24 h, Equations 4.5 and 4.6 are used to determine therange of frequencies at each forecast time at Casablanca.These ranges are shown in Table 4.9.The angular spreading factor is determined fromthe ratio of R c to the fetch width, i.e. 600/300 = 2. ForFigure 4.6 — Weather situation over the North Atlantic at t = 0LCasablanca

WAVE FORECASTING BY MANUAL METHODS 51TABLE 4.9Forecast of swell periods, wavelengths and heights at Casablanca for various times after the arrival of the longest period waves.H c and angular spreading and dispersion multipliers are also shown(The bracketed values in the last two columns are based on the “uniform distribution” approximation.)Arrival Periods Wave- H c Angular Dispersion Arrivingtime lengths spreading multiplier wave heights(h) (s) (m) (m) multiplier (m)30 14.4 – 13.2 323 – 272 4.8 0.55 0.35 (0.23) 0.9 (0.6)36 14.4 – 11.0 323 – 189 4.8 0.55 0.63 (0.41) 1.7 (1.1)42 14.4 – 9.4 323 – 138 4.8 0.55 0.75 (0.54) 2.0 (1.4)48 14.4 – 8.2 323 – 105 4.8 0.55 0.90 (0.64) 2.4 (1.7)54 13.2 – 7.3 272 – 83 4.8 0.55 0.84 (0.70) 2.2 (1.8)60 11.0 – 6.6 189 – 68 4.8 0.55 0.73 (0.69) 1.9 (1.8)66 9.4 – 6.0 138 – 56 4.8 0.55 0.62 (0.68) 1.6 (1.8)72 8.2 – 5.5 105 – 47 4.8 0.55 0.40 (0.68) 1.1 (1.8)this situation, the angular spreading factor is 30 per cent(from Figure 3.3). This means that the height of swellobserved at the shore of Casablanca should be less than√0.30 x 4.8 or 2.6 m.The wave dispersion factor at any given forecast timeis the most difficult part of the forecast. A model spectrummust be chosen, and the mathematical integration ofthe frequency spectrum, S(f), should be carried out forthe range of frequencies at each forecast time. In thisexample, Equation 4.1 is used in conjunction with theJONSWAP spectrum (Figure 1.17) to determine the rangeof important frequencies corresponding to f p = 1/T p =0.099, i.e. between 0.7 f p and 2 f p (1/14.4 = 0.069 Hz and1/5.04 = 0.199 Hz), and PNJ is used to determine the fractionof energy in given ranges of frequencies (dispersionfactor) arriving at any given time at Casablanca.To determine the energy values, use the distortedco-cumulative spectra for wind speeds 10–44 kn as afunction of duration from PNJ (shown in Annex IV).The wind speed is 15 m/s (30 kn). Determine where theupper and lower frequencies intersect the 30 kn curve,and find the E values that intersect with those points. Forf = 0.069 Hz, E = 51. For f = 0.199, E = 2.5. The differenceis 48.5. To calculate the dispersion factor at 36 h,find the E values for the periods arriving at that time anddivide by the total energy, i.e. (51 – 31.5)/48.5 = 0.402.Note that the PNJ curves are used only to find theproportion of energy in the given frequency ranges; forthe significant wave height it is preferable to use the GDcurves (Figure 4.1).The wave height is determined by multiplying thesquare root of the dispersion factor (i.e. the dispersionmultiplier) times the square root of the angular spreadingfactor (angular spreading multiplier) times H c . For 36hours this is √0.402 x √0.3 x 4.8 = 1.7 m.At t = 48 h a much larger part of the wave spectrumis still present with periods shifted to lower values(wave frequencies to higher values). The dispersionfactor is (51 – 12)/48.5 = 0.804, and the wave heightis given by √0.804 x √0.3 x 4.8 = 2.4 m. The variouswave heights computed in this way are given inTable 4.9.A short cut: A rough approximation to the dispersionfactor may be obtained by simply assuming that thefraction of the energy we are seeking is dependent onlyon the proportion of the frequency range we are considering,i.e. the ratio of the frequency range to the totalsignificant frequency range. For this example, the significantfrequency range is (0.199 – 0.069) = 0.13 Hz. Forthe first row in Table 4.9 the frequency range is (0.076 –0.069) = 0.007 Hz, which leads to a dispersion multiplierof √(0.007/0.13) = 0.23. The appropriate estimates havebeen entered in Table 4.9 in brackets.It must be emphasised that this is a rough approximation.It assumes that the energy distribution isuniform across the frequencies, which we know to bewrong (see Figure 1.17). As the worked example inTable 4.9 shows, the errors may be acceptable if a rapidcalculation is required, although they are significant. Thepeak of the storm will inevitably be underestimated andthe decay will also characteristically be too gradual.Using these observations some subjective correctionscan be developed with experience.This example has shown the common features ofswell development as a function of time: a generalincrease in wave height at first and, over a long periodof time, a more or less constant height as the spectrumreaches its greatest width. Since factors such as internalfriction and air resistance have not been taken intoaccount, it is likely that the swell would have died outafter about 60 h.4.4.4 Further examplesProblem:Assuming that swell originates from a small fetch areaof hurricane winds in a tropical cyclone, wind waveshave a characteristic height of 12 m over a width of

52GUIDE TO WAVE ANALYSIS AND FORECASTING120 n.mi., estimate the swell heights in the casesdescribed in Sections 4.4.1 and 4.4.2 with the experiencegained from Section 4.4.3.Solution 4.4.1:In Section 4.4.1, the distance to point A is 600 n.mi., orfive times the fetch width. From Figure 3.3, the angularspreading factor is about 12 per cent. Thus, the characteristicheight of the swell arriving at A should be lessthan √0.12 x 12 = 4.2 m.We can determine the dispersion factor from thesame PNJ graph as before (see Annex IV), but we haveto determine an effective wind speed since only thewave height is known. The effective wind speed isdetermined by following the line for a 15 s period onthe PNJ graph to where it intersects the given waveheight. We can determine the effective wind speed fromthe intersection point, in this case 21.5 m/s (43 kn).Then we determine the E values, and compute thedispersion factors as in Section 4.3.3. Since the swellspectrum is broadest at 42 h, the wave heights are highestat that time. The dispersion factor is about 0.8giving a multiplier of about 0.9 which leads to a characteristicwave height of about 3.7 m when angularspreading is included in the computation.The height is very small at first, when only longwaves arrive at A. The wave heights are highest inthe period between 40 h and 50 h (see Table 4.6), sincethe swell spectrum has its greatest width during thatperiod.The distance to point B equals about eight timesthe width of the fetch area, which leads to an angularspreading factor of about 6 per cent; therefore the swellheights at point B should be no more than 2.9 m.Considering that the wave dispersion must haveprogressed further from point A to point B and that thesmaller components may have been dissipated on theirlong journey as a result of internal friction and airresistance, we can compute, from PNJ, wave dispersionfactors for each arrival time. The widest part of thespectrum, with the highest heights, passes point Babout 60 h after generation. The dispersion factor forthis time is about 0.6, so the multiplier is about0.8; therefore, the characteristic swell height will beabout 2.3 m.Solution 4.4.2:In Section 4.4.2, the swell spectrum at point A is morecomplex. There is a range of generation wave periodsfrom 12 to 15 s. There is also a limited fetch (180n.mi.). The highest characteristic waves possible differfor each of these wave limits.To determine these wave period limits, we use thedistorted co-cumulative spectra for wind speeds10–44 kn as a function of fetch from PNJ. To ascertainthe maximum characteristic wave height for the 15 swaves, we trace the 15 s period line to where it intersectsthe 180 n.mi. fetch line and read the wave height(4.5 m = 14.8 ft). We also find an effective wind speedfor these waves (13 m/s = 26 kn). Likewise we discoverthese values for the 12 s waves (characteristic waveheight = 4.8 m = 15.8 ft, and effective wind speed = 14m/s = 28 kn). The E values (energy proportions) aredetermined in the same way as before.By comparing the 15 s and 12 s wave heights afterthe dispersion factors have been taken into account foreach forecast hour, we can show that the waves arrivingat 60 h have the greatest characteristic heights. Recall,the angular spreading factor at point A is 12 per cent;this means the highest characteristics possible forthese conditions have a height of about 1.7 m. Thedispersion factor at 60 h is about 0.82 with the multiplierbeing about 0.9; thus, the waves arriving at pointA at 60 h can have a characteristic height of no morethan 1.5 m.(b)4.5 Manual computation of shallow-watereffectsSeveral kinds of shallow-water effects (shoaling, refraction,diffraction, reflection, and bottom effects) aredescribed elsewhere in this Guide. In this section, a fewpractical methods are described which have been takenfrom CERC (1977) and Gröen and Dorrestein (1976). Inabsolute terms, a useful rule of thumb is to disregard theeffects of depth greater than about 40 m unless the wavesare very long, i.e. if a large portion of the wave energy isin waves with periods greater than 10 s. Distinction ismade between:(a) Swell originating from deep water entering a shallowarea with variable depth; andWind waves with limited wave growth in shallowwater with constant depth.More complicated cases with combinations of (a)and (b) will generally require the use of numericalmodels.Section 4.5.1 deals with shoaling and refraction ofswell whose steepness is sufficiently small to avoid wavebreaking after shoaling and focusing due to refraction. InSection 4.5.2 a diagram is presented for estimating waveheights and periods in water with constant depth.4.5.1 Shoaling and refraction of swell in acoastal zoneIn this section, wave decay due to dissipation by bottomfriction and wave breaking is neglected. Shoaling andrefraction generally occur simultaneously; however, theywill be considered separately.4.5.1.1 Variation in wave height due to shoalingTo obtain the shoaling factor, K s , which represents thechange of wave height (H) due to decreasing depth(without refraction), we need to consider the basic rulethat energy flux must be conserved. Since energy isrelated to the square of the wave height (Sections 1.2.4and 1.3.8) and wave energy travels at the group velocity,

WAVE FORECASTING BY MANUAL METHODS 53the energy flux is c g H 2 . This is a constant. Hence theshoaling factor depends on the ratio between the deepwatergroup velocity c g0 and the group velocity c g atdepth h, and is given by:KsH cg= = =H c01 cc0 0g 2 g(4.10)where c 0 is the phase velocity in deep water (√g/k 0 ),k 0 is the wavenumber in deep water and H 0 is the waveheight in deep water. Since the group velocity c g is givenby:gkhcg = khk= 1 ⎛⎜⎝+ ⎞β tanh , β⎟ ,2 1 2sinh 2 kh⎠the shoaling factor can also be expressed as:,Ks = 2 – 1/2β tanh kh .(4.11)The wavenumber k = 2π/λ at depth h can beapproximated as follows:k = k 0 (tanh kh) –1 ,k = k 0 [tanh (k 0 h tanh kh) –1 ] –1 ≈ k 0 (tanh k 0 h) –1 . (4.12)Figure 4.7(a) shows various curves involved in thetransformation of properties of a wave propagating fromdeep to shallow water. The shoaling factor, K s, is alsoshown in this figure. But for the convenience of theuser interested in obtaining only the shoaling factor K sassociated with a deep water wave of a given wavenumberk 0 (= 2π/λ 0 ), the curve of K s vs h/λ 0 is presentedin Figure 4.7(b) in an expanded scale.10K s12β0.1c/c 0 . λ/λ 0c g /c 0h/λ0.010.0001 0.001 0.01 0.1 1h/λ 0Figure 4.7(a) — Illustration of various functions of h/λ 0 (derived from CERC, 1984)

54GUIDE TO WAVE ANALYSIS AND FORECASTING1.8Figure 4.7(b) —A graph of shoaling factorK s versus h/λ 0 (derived fromCERC, 1984)1.61.4Ks1.21.00.80.01 0.1 1h/λoExample:For a deep water wavelength λ 0 = 156 m and T 0 = 10 s,then k 0 = 0.04 m –1 , and for T 0 = 15 s, k 0 = 0.018 m –1 .Table 4.10 shows the shoaling factor K s for anumber of shallow depths. It also shows that the heightof shoaling waves is reduced at first, but finally increasesup to the point of breaking which also depends on theinitial wave height in deep water (see also Figures 4.7(a)and (b)).4.5.1.2 Variation of wave height due to refractionIn the previous example, no refraction was taken intoaccount, which implies propagation of waves perpendicularto parallel bottom depth contours. In naturalconditions this will rarely occur. So the angle of incidencewith respect to the bottom depth contours usuallydiffers from 90°, which is equivalent to α, the anglebetween a wave crest and the local isobath, being differentfrom 0°. This leads to variation of the width betweenwave rays. Using Snell’s law,H=HThe refraction factor is:0K r =cgcgcos α.cos α0 0cos α0cos α(4.13)with α 0 the angle between a wave crest and a localisobath in deep water. Figure 4.8, taken from CERC(1984), is based on Equation 4.13. For a given depth andwave period the shallow-water angle of incidence (solidlines) and the refraction factor (broken lines) can be readoff easily for a given deep-water angle of incidence α 0 . Itis valid for straight parallel depth contours only.Problem:Given an angle α 0 = 40° between the wave crests in deepwater and the depth contours of the sloping bottom, findα and the refraction at h = 8 m for T = 10 s.TABLE 4.10Wavenumber k and shoaling factor K s for two waveperiods, at several depths, h, in metres, usingEquations 4.11 and 4.12h T = 10 s, k 0 = 0.04 m –1 T = 15 s, k 0 = 0.018 m –1(m) k (m –1 ) K s k (m –1 ) K s100 0.040 1.00 0.018 0.9450 0.041 0.95 0.021 0.9225 0.046 0.91 0.028 0.9815 0.055 0.94 0.035 1.0610 0.065 1.00 0.043 1.155 0.090 1.12 0.060 1.332 0.142 1.36 0.095 1.64

WAVE FORECASTING BY MANUAL METHODS 5580°70°60°0.500.600.7070°60°50°40°α0, degrees50°40°30°20°10°010-45°Lines of equal K r0.800.850.9010°0.952 5 2 5 2 515°0.97Lines of equal α (angle betweenwave crest and contour)10-3 10-2 0.1h/(gT 2 )Figure 4.8 — Changes in wave direction and height due to refraction on slopes with straight parallel bottomcontours. α 0 is the deep water angle of incidence, measured between the wave crest and the localisobath. The continuous curves are lines of equal incidence for various combinations of period anddepth. To estimate the refraction as a wave moves into shallower water, starting from a given α 0follow a horizontal line from right (deep water) to left. The broken lines are lines of equal refractionfactor, K r (derived from CERC, 1977)20°0.9930°5°Solution:h/(gT 2 ) = 8/(9.8 x 100) = 0.0082, and from Figure 4.8the refraction factor is 0.905, and α = 20°.4.5.1.3 Dorrestein’s methodDue to the fact that in reality the bottom depth contoursrarely are straight, we generally find sequences ofconvergence and divergence (see also Section 1.2.6 andSection 7.3). Dorrestein (1960) devised a method formanually determining refraction where bottom contoursare not straight. This method requires the construction ofa few wave rays from a given point P in shallow water todeep water, including all wave directions that must betaken into account according to a given directionaldistribution in deep water.We assume that, in deep water, the angular distributionof wave energy is approximately a uniformdistribution in the azimuthal range α 1´ and α 2´. Theseangles correspond, respectively, to the angles of incidenceα 1 and α 2 at point P. Rays must at least beconstructed for waves at these outer limits of the distribution.It may be sufficient to assume straight isobathsand use Figure 4.8 to calculate these angles. Then,according to Dorrestein, the refraction factor is:(4.14)with c 0 and c the phase velocities in deep water and atpoint P, respectively.Problem:As in the example in Section 4.5.1.2, h = 8 m at point P.T = 10 s, so h/(gT 2 ) = 0.0082. With the help ofFigure 4.8, α = 20° for α 0 = 40°. Find K r by Dorrestein’smethod.Solution:Snell’s law gives:cccKr = 0c⋅ α1 – α2α′ – α′1 2sin sin °= = 188sin sin 20° = . .α0 α 0 40

56GUIDE TO WAVE ANALYSIS AND FORECASTING42T ___ c√hIn general the results for K r are sensitive to thechoice of the spread between α 1 and α 2 (or α 1´ and α 2´).4.5.2 Wind waves in shallow water10.80.60.40.20.10.080.060.040.020.01Deep waterDeep waterX/h = 3 000X/h = 3 000H c ___√h1 2 4 6 8 10 20 40u/√hFigure 4.9 — Diagram to estimate characteristic wave heightsand period of wind waves in shallow water ofconstant depth. Solid lines indicate limitingvalues for sufficiently long fetches. Brokenlines indicate conditions with relatively shortfetches, specified by X/h = 3 000 (from Gröenand Dorrestein, 1976)In general it is prudent to use a small window∆α = α 1 – α 2 centered around α (= 20° in this case) inapplying the ideal linearly sloping bottom results toobtain the answers to a real case of bottom topography.If we take α 1 = 21° and α 2 = 19°, we find fromFigure 4.8, α 1 ≈ 42° and α 2 ≈ 37.5°. Then, Equation 4.14gives K r = 0.904.In shallow water the bottom topography and compositionhas a dissipating effect on waves. Here we present asimple manual method for prediction of the characteristicwave height, H c , and the associated wave period inshallow water with constant depth. To that end, we useFigure 4.9, taken from Gröen and Dorrestein (1976),based on the same deep-water growth curve, but withadditional terms for taking the limitation of wave growthby the bottom into account. These are similar to theshallow-water growth curves of CERC (1984), but withsomewhat different coefficients.As in the previous sections of this chapter, weassume that H c is approximately equal to the significantwave height H – 1/3 or H m0 , and T c is approximately equalto the significant wave period T – H 1/3or T p (see Chapter1). For the sake of compactness, only ratios areshown: H c /h and T c /√h, both as a function of u/√h. udenotes wind speed at standard level (usually 10 mabove water) in m/s. Wave height, H c , and depth, h, areexpressed in m, and wave period, T c , in s. The solidcurves show H c /h and T c /√h for sufficiently long fetches,e.g. a X/h ratio much greater than 3 000. Then a sort ofbalance between wind input and bottom dissipation canbe assumed. The effect of limited fetch is shown forX/h = 3 000.Problem:Find H c and T c for u = 20 m/s, h = 10 m and X = 200 kmand 30 km.Solution:u/√h = 20/√10 = 6.3. For X = 200 km, X/h = 200 000/10= 20 000. From Figure 4.9, we use the solid curves, andH c /h = 0.23, while T c /√h = 2.1, or H c = 0.23 m x 10 =2.3 m is the maximum characteristic height for the givenu and h, and T c = 2.1√10 = 6.6 s.For X = 30 km, X/h = 3 000. Using the curves forX/h = 3 000 in Figure 4.9, H c /h = 0.2, T c /√h = 1.5, orH c = 0.2 x 10 = 2 m, and T c = 1.5√10 = 4.7 s.

CHAPTER 5INTRODUCTION TO NUMERICAL WAVE MODELLINGM. Reistad with A.K. Magnusson: editors5.1 IntroductionNational Meteorological Services in maritime countrieshave experienced a rapidly growing need for wave forecastsand for wave climatology. In particular the offshoreoil industry needs wave data for many purposes: designsea states, fatigue analysis, operational planning andmarine operations. Furthermore, consulting companiesoperating in the maritime sector have an increasing needfor wave information in their projects.To meet this growing requirement for wave information,wave conditions must be estimated over largetracts of ocean at regular intervals, often many times aday. The volume of data and calculations makes computersindispensable. Furthermore, measured wave dataare often sparse and not available when and where theyare desired. Using wind information and by applicationof the basic physical principles that are described inChapter 3, numerical models have been developed tomake the required estimates of wave conditions.In wave modelling, we attempt to organize ourtheoretical and observational knowledge about wavesinto a form which can be of practical use to the forecaster,engineer, mariner, or the general public. The mostimportant input to the wave models is the wind at the seasurface and the accuracy of the wave model output isstrongly dependent on the quality of the input windfields. Chapter 2 is devoted to the specification of marinewinds.In the WMO Handbook on wave analysis and forecasting(1976), one particular model was described indetail to exemplify the structure and methodology ofnumerical wave models. Since then new classes ofmodels have appeared and were described in the WMOGuide to wave analysis and forecasting (1988). Ratherthan giving details of one or a few particular models, thischapter will give general descriptions of the three modelclasses that were defined in the SWAMP project(SWAMP Group, l985). A short description of the “thirdgeneration” WAM model developed by an internationalgroup of wave modellers is added.The basic theory of wave physics was introduced inChapter 3. In this chapter, Section 5.2 gives a brief introductionto the basic concepts of wave modelling. Section5.3 discusses the wave energy-balance equation.Section 5.4 contains a brief description of some elementsof wave modelling. Section 5.5 defines and discusses themost important aspects of the model classes. The practicalapplications and operational aspects of the numerical wavemodels are discussed in Chapter 6.5.2 Basic conceptsThe mathematical description of surface waves has alarge random element which requires a statisticaldescription. The statistical parameters representing thewave field characterize conditions over a certain timeperiod and spatial extent. Formally, over these scales, weneed to assume stationarity (steadiness in time) andspatial homogeneity of the process describing the seasurface. Obviously, no such conditions will hold over thelarger scales that characterize wave growth and decay.To model changing waves effectively, these scales (timestepor grid length) must be small enough to resolve thewave evolution, but it must be recognized that in time orspace there are always going to be smaller scale eventswhich have to be overlooked.The most used descriptor of the wave field is theenergy-density spectrum in both frequency and directionE(f,θ), where f is the frequency, and θ the direction ofpropagation (see Section 1.3.7). This representation isparticularly useful because we already know how tointerpret what we know about wave physics in terms ofthe spectral components, E(f,θ). Each component can beregarded as a sinusoidal wave of which we have areasonably well-understood theory. From this spectrum,we can deduce most of the parameters expected of anoperational wave model, namely: the significant waveheight, the frequency spectrum, the peak frequency andsecondary frequency maxima, the directional spectrum,the primary wave direction, any secondary wave directions,the zero-crossing period, etc. (see Chapter 1).Not all models use this representation. Simplermodels may be built around direct estimation of thesignificant wave height, or on the frequency spectrum,with directional characteristics often diagnosed directlyfrom the wind.There is a reasonable conception of the physicalprocesses which are thought to control wave fields. Tobe of general use in wave modelling, these processes aredescribed by the response of wave ensembles, i.e. theyare translated into terms of useful statistical quantitiessuch as the wave spectrum. Not all the processes are yetfully understood and empirical results are used to varyingdegrees within wave models. Such representationsallow a certain amount of “tuning” of wave models, i.e.model performance can be adjusted by altering empiricalconstants.Although models for different purposes may differslightly, the general format is the same (see Figure 5.1for a schematic representation).

58GUIDE TO WAVE ANALYSIS AND FORECASTINGTime tTime t + δtInitial stateE(f, θ; x, t)Wave growth∂E (f, θ; x, t)∂tModified stateE(f, θ; x, t + δt)=Wave dataPropagation-∇ • (c gE)RefractionShoaling∂-∂θ(∇ θ • c gE)HINDCASTnumerical/manualNon-linearinteractionsDissipation+S nl+S dsFORECASTWindAtmosphericforcing+S inAtmosphericANALYSISAtmosphericFORECASTFigure 5.1 — The elements of wave modelling5.3 The wave energy-balance equationThe concepts described in Chapter 3 are represented inwave models in a variety of ways. The most generalformulation for computer models based on the elementsin Figure 5.1 involves the spectral energy-balance equationwhich describes the development of the surfacegravity wave field in time and space:∂E∂t+∇• ( c g E) = S = S in + S nl + S ds(5.1)where:E = E(f,θ,x,t) is the two-dimensional wave spectrum(surface variance spectrum) dependingon frequency, f, and direction of propagation,θ;c g = c g (f,θ) is the deep-water group velocity;Sis the net source function, consisting ofthree terms:S in :S nl :energy input by the wind;non-linear energy transfer by wavewaveinteractions; andS ds : dissipation.This form of the equation is valid for deep water with norefraction and no significant currents.5.4 Elements of wave modellingThe essence of wave modelling is to solve the energybalanceequation written down in Equation 5.1. This firstrequires the definition of starting values for the waveenergy, or initial conditions which in turn requiresdefinition of the source terms on the right hand side ofEquation 5.1 and a method for solving changes as timeprogresses.5.4.1 Initial conditionsIt is rare that we have a flat sea to work from, or that wehave measurements which completely characterize thesea state at any one time.For computer models, the usual course of action isto start from a flat sea and “spin up” the model with thewinds from a period of several days prior to the periodof interest. We then have a hindcast derived for theinitial time. For operational models this has to be done

INTRODUCTION TO NUMERICAL WAVE MODELLING 59only once, since it is usual to store this hindcast andprogressively update it as part of each model run.In some ocean areas in the northern hemisphere,there is sufficient density of observations to make adirect field analysis of the parameters observed (such assignificant wave height or period). Most of the availabledata are those provided visually from ships; these dataare of variable quality (see Chapter 8). Such fields arenot wholly satisfactory for the initialization of computermodels. Most computer models use spectral representationsof the wave field, and it is difficult to reconstruct afull spectral distribution from a height, period and direction.However, the possibility of good quality wave datafrom satellite-borne sensors with good ocean coverage,has prompted attempts to find methods of assimilatingthese data into wave models. The first tested methodshave been those using significant wave heights measuredby radar altimeters on the GEOSAT and ERS-1 satellites.It has been shown that a positive impact on thewave model results can be achieved by assimilatingthese data into the wave models (e.g. see Lionello et al.,1992, and Breivik and Reistad, 1992). Methods to assimilatespectral information, e.g. wave spectra derived fromSAR (Synthetic Aperture Radar) images, are also underdevelopment. More details on wave data assimilationprojects are given in Section 8.6.5.4.2 WindPerhaps the most important element in wave modellingis the motion of the atmosphere above the sea surface.The only input of energy to the sea surface over thetime-scales we are considering comes from the wind.Transfer of energy to the wave field is achieved throughthe surface stress applied by the wind, which variesroughly as the square of the wind speed. Thus, an errorin wind specification can lead to a large error in the waveenergy and subsequently in parameters such as significantwave height.The atmosphere has a complex interaction with thewave field, with mean and gust wind speeds, windprofile, atmospheric stability, influence of the wavesthemselves on the atmospheric boundary layer, etc., allneeding consideration. In Chapter 2, the specification ofthe winds required for wave modelling is discussed.For a computer model, a wind history or prognosisis given by supplying the wind field at a series of timestepsfrom an atmospheric model. This takes care of theproblem of wind duration. Similarly, considerations offetch are taken care of both by the wind field specificationand by the boundary configuration used in thepropagation scheme. The forecaster using manualmethods must make his own assessment of fetch andduration.5.4.3 Input and dissipationThe atmospheric boundary layer is not completely independentof the wave field. In fact, the input to the wavefield is dominated by a feedback mechanism whichdepends on the energy in the wave field. The rate atwhich energy is fed into the wave field is designated byS in .This wind input term, S in , is generally accepted ashaving the form:S in = A(f,θ) + B(f,θ) E(f,θ) (5.2)A(f,θ) is the resonant interaction between waves andturbulent pressure patterns in the air suggested byPhillips (1957), whereas the second term on the righthandside represents the feedback between growingwaves and induced turbulent pressure patterns assuggested by Miles (1957). In most applications, theMiles-type term rapidly exceeds the Phillips-type term.According to Snyder et al. (1981), the Miles termhas the form:UBf, θ ⎡ ρ ⎛max ,g f⎞( )= ⎜ cos ( θ– ψ)– ⎟ πρ ⎝⎠f ⎤a 5⎢0 K1 K21 2 ⎥⎣ w⎦(5.3)where ρ a and ρ w are the densities of air and water,respectively; K 1 and K 2 are constants; ψ is the directionof the wind; and U 5 is the wind speed at 5 m (see alsoSection 3.2).Equation 5.3 may be redefined in terms of the frictionspeed u * = √(τ/ρ a ), where τ is the magnitude of thewind shear stress. From a physical point of view, scalingwave growth to u * would be preferable to scaling withwind speed U z at level z. Komen et al. (1984) haveapproximated such a scaling, as illustrated in Equation3.1, however, lack of wind stress data has precludedrigorous attempts. U z and u * do not appear to be linearlyrelated and the drag coefficient, C d , used to determine τ(τ = ρ a C d U z 2 ), appears to be an increasing function of U z(e.g. Wu, l982; Large and Pond, 1981). The scaling is animportant part of wave modelling but far from resolved.Note that C d also depends on z (from U z ) (see forexample Equation 2.14). Recent advances with a quasilineartheory which includes the effects of growingwaves on the mean air flow have enabled further refinementof the formulation (Janssen, 1991; Jenkins, 1992;Komen et al., 1994).Note also that in the case of a fully developed sea,as given by the Pierson-Moskowitz spectrum E PM (seeEquation 1.28), it is generally accepted that the dimensionlessenergy, ε,ε = g2 ∫ E PM fu *4( ) df= g2 E totalu *4is a universal constant. However, if ε is scaled in termsof U 10 this saturation limit will vary significantly withwind speed since C d is a function of U 10 . NeitherEquation 5.3 nor the dependence of C d on U z are welldocumented in strong wind.The term S ds describes the rate at which energy islost from the wave field. In deep water, this is mainlythrough wave breaking (whitecapping). In shallow water

60GUIDE TO WAVE ANALYSIS AND FORECASTINGit may also be dissipated through interaction with thesea-bed (bottom friction). More details are given inSections 3.4 and 7.6.5.4.4 Non-linear interactionsGenerally speaking, any strong non-linearities in thewave field and its evolution are accounted for in thedissipation terms. Input and dissipation terms can beregarded as complementary to those linear and weaklynon-linear aspects of the wave field which we are able todescribe dynamically. Into this category fall the propagationof surface waves and the redistribution of energywithin the wave spectrum due to weak, non-linear interactionsbetween wave components, which is designatedas a source term, S nl . The non-linear interactions arediscussed in Section 3.5.The effect of the term, S nl , is briefly as follows: inthe dominant region of the spectrum near the peak, thewind input is greater than the dissipation. The excessenergy is transferred by the non-linear interactionstowards higher and lower frequencies. At the higherfrequencies the energy is dissipated, whereas the transferto lower frequencies leads to growth of new wavecomponents on the forward (left) side of the spectrum.This results in migration of the spectral peak towardslower frequencies. The non-linear wave-wave interactionspreserve the spectral shape and can be calculatedexactly.The source term S nl can be handled exactly but therequirement on computing power is great. In third generationmodels, the non-linear interactions between wavecomponents are indeed computed explicitly by use ofspecial integration techniques and with the aid of simplificationsintroduced by Hasselmann and Hasselmann(1985) and Hasselmann et al. (1985). Even with thesesimplifications powerful computers are required toproduce real-time wave forecasts. Therefore manysecond generation models are still in operational use. Insecond generation numerical wave models, the nonlinearinteraction is parameterized or treated in asimplified way. This may give rise to significant differencesbetween models. A simplified illustration of thethree source terms in relation to the wave spectrum isshown in Figure 3.7.5.4.5 PropagationWave energy propagates not at the velocity of the wavesor wave crests (which is the phase velocity: the speed atwhich the phase is constant) but at the group velocity(see Section 1.3.2). In wave modelling we are dealingwith descriptors such as the energy density and so it isthe group velocity which is important.The propagative effects of water waves are quantifiedby noting that the local rate of change of energy isequal to the net rate of flow of energy to or from thatlocality, i.e. the divergence of energy-density flux. Thepractical problem encountered in computer modelling isto find a numerical scheme for calculating this. Inmanual models, propagation is only considered outsidethe generation area and attention is focused on thedispersion and spreading of waves as they propagate.Propagation affects the growth of waves throughthe balance between energy leaving a locality and thatentering it. In a numerical model it is the propagation ofwave energy which enables fetch-limited growth to bemodelled. Energy levels over land are zero and so downwindof a coast there is no upstream input of waveenergy. Hence energy input from the atmosphere is propagatedaway, keeping total energy levels near the coastlow.Discrete-grid methodsThe energy balance, Equation 5.1, is often solvednumerically using finite difference schemes on a discretegrid as exemplified in Figure 5.2. ∆x i (i = 1, 2) is the gridspacing in the two horizontal directions. Equation 5.1may take such a form as:E( x, t+∆t) = E( x,t)⎡( E) – ( E)⎤2 cgc⎢i gx ii xi– ∆xi⎥– ∆t∑ ⎢⎥ (5.4)i=1∆x⎣⎢i⎦⎥∆tSx,t+ ( )Deep or constant water depthVarying shallow water depthj+2∆yj+1jj-1i-1(x, y)θ 0 θ(x, y)θ0 θ 0(x 0 , y 0 ) (x 0 , y 0 )i i+1 i+2∆xFigure 5.2 —Typical grid for numerical wavemodels (x stands for x 1 and y forx 2 ). In grid-type models the energyin (f,θ) bins is propagated betweenpoints according to an equation likeEquation 5.4. In ray models theenergy is followed alongcharacteristic lines

INTRODUCTION TO NUMERICAL WAVE MODELLING 61Frequency0.0500.0670.0830.1000.167Period201512106Distance travelledin km363027222185181510926 sLength ofdisturbancein km55623233327816710 s12 s15 s20 s15°yStormFront1111 kmx0°16 110112 115120–15°Figure 5.3 — Propagation of discrete spectral components from a storm front. This illustrates the effects of angular spreadingand dispersion on the wave energy. The “garden sprinkler” effect resulting from the discretization of the spectrumcan be seenwhere ∆t is the time-step; E and S are functions ofwavenumber (k), or frequency and direction (f,θ).Using the spectral representation E = E(f,θ), wehave energy density as an array of frequency directionbins (f,θ). The above approach collects a continuum ofwave components travelling at slightly different groupvelocities into a single frequency bin, i.e. it uses asingle frequency and direction to characterize eachcomponent. Due to the dispersive character of oceanwaves, the area of a bin containing components within(∆f, ∆θ) should increase with time as the waves propagateaway from the origin; the wave energy in this binwill spread out over an arc of width ∆θ and stretch outdepending on the range of group velocities. In the finitedifference approach, all components propagate at themean group velocity of the bin, so that eventually thecomponents separate as they propagate across themodel’s ocean. This is called the “sprinkler” effect as itresembles the pattern of droplets from a garden sprinkler.It should be stressed that this is only an artifice ofthe method of modelling. All discrete-grid models inthe inventory suffer from the sprinkler effect (seeFigure 5.3), although usually the smoothing effect ofcontinual generation diminishes the potential ill effects,or it is smoothed over as a result of numerical error(numerical diffusion).There are many finite difference schemes in use:from first-order schemes, which use only adjacent gridpoints to work out the energy gradient, to fourth-orderschemes, which use five consecutive points. The choiceof time-step, ∆t, depends on the grid spacing, ∆x, as fornumerical stability the distance moved in a time-stepmust be less than one grid space. Typically models usefrom 20–200 km grid spacings and time-steps of severalminutes to several hours.The discrete-grid models calculate the complete(f,θ) spectrum at all sea points of the grid at eachtime-step.Ray tracing methodsAn alternative method is to solve the energy balance(Equation 5.1) along characteristics or rays. The timeintegration is still performed by finite-differencing butthe spatial integration is not needed and the sprinklereffect is avoided. However, the number of output pointsis usually reduced for reasons of computational costs.For ocean waves there is a dispersion relation, relatingthe wave frequency to the wavenumber (seeEquations 1.3 and 1.3a), of the form:f x, t σ k x, t , Ψ x, t . (5.5)[ ]( ) = ( ) ( )

62GUIDE TO WAVE ANALYSIS AND FORECASTINGHere, σ is used to denote the particular frequency associatedwith wavenumber, k, and the property of themedium, Ψ, which in this context will be the bottomdepth and/or currents. Characteristic curves are thenobtained by integration ofdxδσ= c(5.6)g =dtδkand in an ocean with steady currents these curves needonly be obtained once. Examples including refractiondue to bottom topography are shown in Figures 5.2 and7.1. For more details on ray theory see LeBlond andMysak (1978).Thus, starting from the required point of interest,rays or characteristics are calculated to the boundary ofthe area considered necessary to obtain reliable waveenergy at the selected point. Since we are consideringthe history of a particular wave frequency, our referenceframe moves with the component and all we needconsider is the source function along the rays, i.e.(5.7)Rays are calculated according to the required directionalresolution at the point of interest; along each ray,Equation 5.7 may be solved either for each frequencyseparately or for the total energy. In the former approachS nl is not considered at all. In the latter case interactionsin frequency domain are included but the directions areuncoupled.The ray approach has been extensively usedin models where wind sea and swell are treated separately.In such cases, swell is propagated alongthe rays subjected only to frictional damping andgeometric spread. Interactions with the wind sea maytake place where the peak frequency of the Pierson-Moskowitz spectrum (≈ 0.13g/U 10 ) is less than the swellfrequency.5.4.6 Directional relaxation and wind-sea/swellinteractionMany of the differences between numerical wave modelsresult from the way in which they cater for the weakly nonlinearwave-wave interactions (S nl ). The differences areparticularly noticeable in the case of non-homogeneousand/or non-stationary wind fields. When the wind directionchanges, existing wind sea becomes partly swell and a newwind sea develops. The time evolution of these componentsresults in a relaxation of the wave field towards a newsteady state that eventually approaches a fully developedsea in the new wind direction.Three mechanisms contribute to the directionalrelaxation:(a) Energy input by the wind to the new wind sea;(b)(c)δEδt= S.Attenuation of the swell; andWeak non-linear interactions, resulting in energytransfer from swell to wind sea.The way these mechanisms are modelled may yieldsignificant deviations between models. The third mechanismappears to be dominant in this respect.5.4.7 DepthWater depth can considerably affect the properties ofwaves and how we model them. We know that wavesfeel the sea-bed and are changed significantly by it atdepths less than about one-quarter of the deep-waterwavelength (see also Section 1.2.5). In a sea with a widespectrum, the longer waves may be influenced by thedepth without much effect on the short waves.One major effect of depth is on the propagationcharacteristics. Waves are slowed down and, if the seabedis not flat, may be refracted. Also, the non-linearinteractions tend to be enhanced and, of course, there aremore dissipative processes involved through interactionwith the sea-bed. The framework we are describing forwave modelling is broad enough in its concept to be ableto cope with depth-related effects without drastic alterationsto the form of the model outlined in Figure 5.1. InChapter 7 the effect of shallow water will be discussedin more detail.5.4.8 Effects of boundaries, coastlines andislandsWith the exception of global models, most existing wavemodels have an open ocean boundary. Wave energy maythen enter the modelled area. The best solution is toobtain boundary data from a model operating over alarger area, e.g. a global model. If there is no knowledgeof the wave energy entering the model area, a possibleboundary condition is to let the energy be zero at theboundaries at all times. Another solution may be tospecify zero flux of energy through the boundary. Ineither case it will be difficult to get a true representationof distantly generated swell. The area should thereforebe sufficiently large to catch all significant swell thataffects the region of interest.In operational models, with grid resolution in therange 25–400 km, it is difficult to get a true representationof coastlines and islands. A coarse resolution willstrongly affect the shadowing effects of islands andcapes. To obtain a faithful representation of the sea statenear such a feature we need to take special precautions.One solution may be to use a finer grid for certain areas,a so-called “nested” model, where results from thecoarse grid are used as boundary input to the fine grid. Itmay also be necessary to increase the directional resolutionso as to model limited depth and shadowing effectsbetter. Another way may be to evaluate the effects of thetopographic feature for affected wave directions at acertain number of grid points and tabulate these as“fudge factors” in the model.5.5 Model classesWave models compute the wave spectrum by numericalintegration of Equation 5.1 over a geographical region.

INTRODUCTION TO NUMERICAL WAVE MODELLING 63The models may differ in several respects, e.g.: the representationof the spectrum, the assumed forms of S inand S ds , the representation of S nl and whether the integrationis carried out in natural characteristic coordinatesalong individual rays or in terms of a discretized advectionoperator in a grid-point system common to all wavecomponents.The most difficult term to model is the non-linearsource term, S nl , and it is in its specification that thedifferences between the categories are to be found.In the energy-balance Equation 5.1, the interactiveterm S nl couples the components. Models based on discretespectral components with a non-linear term whichis formulated in terms of several (if not all) componentsare called coupled discrete (CD) models. In such modelsestimates of all components are needed just to be able tocompute the evolution of any one component.Computations for these models are often timeconsuming, and some modellers prefer to dispense withthe coupling term, and include the weak non-linear interactionsimplicitly in their formulation for S in + S ds . Suchmodels are decoupled propagation (DP) models. Eachcomponent can then be calculated independently.Advanced models in this class may include a simpleparametric form for S nl , but they are nevertheless distinguishedby the pre-eminence of S in and S ds in thesource term.The third type of model uses the evidence thatspectra of growing seas are shaped by the non-linearinteractions to conform to a self-similar spectrum (e.g.JONSWAP — see Section 1.3.9). The spectral shape ischaracterized by some small number of parameters andthe energy-balance equation can then be written in theseterms. This gives an evolution equation for each of asmall number of parameters rather than one for each of alarge number of components. However, this parametricrepresentation is only valid for the self-similar form ofthe wind-sea spectrum, and waves outside the generatingarea (swell) require special treatment. This is usuallyachieved by interfacing the parametric model for windsea with a decoupled propagation model for “swell”through a set of algorithms by which wind-sea energyand swell energy are interchanged, hence the naming ofthis class as coupled hybrid (CH) models.Table 6.2 in Chapter 6 gives details and referencesfor a variety of numerical wave models. A thoroughdescription and discussion of the model classes can befound in SWAMP Group (1985).5.5.1 Decoupled propagation (DP) modelsModels of this class generally represent the wavespectrum as a two-dimensional discretized array offrequency-direction cells in which each cell or componentpropagates at its appropriate group velocity alongits own ray path. The components are grown accordingto a source function of the formS = A + BE(f,θ) .As non-linear energy transfer is basically neglected, thefactors A and B are usually empirically determined.Each component is grown independently of all theother components up to a saturation limit, which is alsoindependent of the other spectral components and isrepresented by a universal equilibrium distribution. Ifnon-linear coupling is considered at all, it is parameterizedin a simple way, e.g. by one or two spectralparameters. The saturation limit may be given by theenergy of a fully developed sea, often represented by thePierson-Moskowitz spectrum (see Section 1.3.9). Let thefully developed sea spectrum be given by E ∞ . A modificationof S in may then appear asδES =δt⎡2 ⎤2⎛ E ⎞ ⎡ E= ⎢A⎜ ⎟ + BE⎥⎛ ⎞ ⎤1– ⎢1–⎢ ⎝ E ⎠ ⎥ ⎜ ⎟ ⎥∞⎝ E ⎠⎣⎦ ⎣⎢ ∞⎦⎥(Pierson et al., 1966; Lazanoff and Stevenson, 1975), orδES = = A + BEδt(Ewing, 1971). It is also possible to use the Phillips’saturation range,2αg– 5E4f2π∞ = ( )E1–⎟E ⎠( ) ⎛ ⎝ ⎜ ⎞as the saturation limit (Cavaleri and Rizzoli, 1981).The introduction of a saturation limit also works asan implicit representation of wave-energy dissipation,except for dissipation due to bottom friction and dissipationof swell. None of these effects is specific to DPmodels and may vary from model to model.For strictly decoupled models, and for only weaklycoupled models, the differential time and space scales dtand ds are related through the group velocity c g for awave component, ds = c g dt. From this it follows that forDP models the laws for fetch-limited waves underuniform stationary wind conditions are immediatelytranslated into the corresponding duration-limitedgrowth laws by replacing the fetch, X, with c g t for eachwave component.Another feature of DP models that can be traced tothe decoupling of wave components is that the spectrumgenerally develops a finer structure both in frequencyand direction than coupled models, which continuallyredistribute energy and smooth the spectrum.5.5.2 Coupled hybrid (CH) modelsThe independent evolution of individual wave componentsis effectively prevented by the non-linear energytransfer. Unless the wind field is strongly non-uniform,the non-linear transfer is sufficiently rapid relativeto advection and other source functions that a∞2

64quasi-equilibrium spectral distribution is established.The distributions appear to be of the same shape for awide variety of generation conditions and differ onlywith respect to the energy and frequency scales. Thequasi-self-similarity was confirmed theoretically byHasselmann et al. (1973 and 1976).Further, a universal relationship appears to existbetween the non-dimensional total energy and frequencyparameters, ε and ν, respectively. The non-dimensionalscalings incorporate g and some wind-speed measure,e.g. the wind speed at 10 m, U 10 , or the friction speed,u * . Hence ε = E g 2 /u 4 , and ν p = f p g/u, where u = U 10or u * , and E is the total energy (from the integratedspectrum).Since the evolution of the developing wind-seaspectrum is so strongly controlled by the shape stabilizingnon-linear transfer, it appears reasonable to expressthe growth of the wind-sea spectrum in terms of one or afew parameters, for example ε, the non-dimensionalwave energy. In such a one-parameter, first-order representation,all other non-dimensional variables (e.g. ν p ,the non-dimensional peak frequency) are uniquely determined,and hence diagnosed.Thus, in one extreme the parametric model mayprognose as few as one parameter (e.g. the total spectralenergy), the wind-sea spectrum being diagnosed fromthat. For such a model the evolution equation is obtainedby integrating Equation 5.1 over all frequencies anddirections:δEδt+∇• c g E( ) = S EGUIDE TO WAVE ANALYSIS AND FORECASTINGin which c – g is the effective propagation velocity of thetotal energy∫ c g E( f,θ)df dθc g =∫ E( f,θ)and S E is the projection of the net source function S ontothe parameter ESE = ∫ S ( f,θ ) df dθc – g is uniquely determined in terms of E by a prescribedspectral shape and S E must be described as a function ofE and U 10 or u * . This function is usually determinedempirically.If additional parameters are introduced, e.g. thepeak frequency, f p , the Phillips’ parameter, α, or themean propagation direction, – θ, the growth of the windseaspectrum is expressed by a small set of coupledtransport equations, one for each parameter. A generalmethod for projecting the transport equation in thecomplete (f,θ) representation on to an approximateparameter space representation is given in Hasselmann etal. (1976).In slowly varying and weakly non-uniform wind,parametric wave models appear to give qualitatively thesame results. The more parameters used, the more variedare the spectral shapes obtained and, in particular, if themean wave direction, θ, – is used, directional lag effectsbecome noticeable in rapidly turning winds.The fetch-duration relation for a parametric wavemodel will differ from that of a DP model in that a meanpropagation speed takes the place of the group speed foreach frequency band, i.e.X = Ac – gt ,where: X = the fetch, t = duration and A is a constant(typically A = 2/3). Thus, it is not possible to tune thetwo types of model for both fetch- and duration-limitedcases.Once the non-linear energy transfer ceases todominate the evolution of the wave spectrum, the parametricrepresentation breaks down. This is the case forthe low-frequency part of the wave spectrum that is nolonger actively generated by the wind, i.e. the swell part.The evolution of swell is controlled primarily by advectionand perhaps some weak damping. It is thereforerepresented in parametric wave models in the frameworkof discrete decoupled propagation. The combination of aparametric wind-sea model and a decoupled propagationswell model is termed a coupled hybrid model.CH models may be expected to encounter problemswhen sea and swell interact. Typical transition regimesarise:• In decreasing wind speed or when the wind directionturns, in which cases wind sea is transformedto swell;• When swell enters areas where the wind speed issufficiently high that the Pierson-Moskowitz peakfrequency f p = 0.13g/U 10 is lower than the swellfrequency, in which case the swell suddenly comesinto the active wave growth regime.These transitions are modelled very simply in CHmodels. For turning winds it is common that the windsea loses some energy to swell. The loss may be acontinuous function of the rate of change of wind directionor take place only when the change is above acertain angle.When the wind decreases, the CH models generallytransfer frequency bands that travel faster than the windto swell. Some models also transfer the energy thatexceeds the appropriate value for fully developed windsea into swell.Swell may be reabsorbed as wind sea when thewind increases and the wind-sea peak frequencybecomes equal to or less than the swell frequency. SomeCH models only allow reabsorption if the angle betweenwind-sea and swell propagation directions fulfil certaincriteria.Some models allow swell to propagate unaffectedby local winds to destination points. Interaction takesplace only at the destinations. If wind sea exceeds swellat a point, the swell is completely destroyed. Thus, thereabsorption of swell into wind sea is non-conservative.CH models generally use characteristics or rays topropagate swell.

INTRODUCTION TO NUMERICAL WAVE MODELLING 65The CH class may include many semi-manualmethods. The parametric approach allows empirical relationshipsfor the evolution of spectral parameters to beused. These may often be evaluated without the assistanceof a computer, as may the characteristics of theswell.5.5.3 Coupled discrete (CD) modelsThe problem of swell/wind sea interaction in CH modelsmay be circumvented by retaining the discrete spectralrepresentation for the entire spectrum and introducingnon-linear energy transfers. In the models in operationaluse today these interactions are parameterized in differentways. The number of parameters are, however, oftenlimited, creating a mismatch between the degrees offreedom used in the description of the spectrum (say forexample 24 directions and 15 frequencies) and thedegrees of freedom in the representation of the nonlineartransfer (e.g. 10 parameters).In CD models a source function of the Miles type,S in = B.E is very common, as in DP models. However,the factor B is strongly exaggerated in the DP models tocompensate for the lack of explicit S nl . A Phillips’forcing term may also be included so that S in = A + B.E,but the value of A is usually significant only in the initialspin-up of the model.The difference between present CH and CDmodels may not be as distinct as the classificationsuggests. In CD models the non-linear transfer is sometimesmodelled by a limited set of parameters. Themain difference can be found in the number of degreesof freedom. It should also be noted that the CD modelsusually parameterize the high-frequency part of thespectrum.The non-linear source term, S nl , may be introducedin the form of simple redistribution of energy accordingto a parameterized spectral shape, e.g. the JONSWAPspectrum. Another solution can be to parameterize S nlin a similar way to the spectrum. This approach isgenerally limited by the fact that each spectral formwill lead to different forms of S nl . This problem may beavoided by using an S nl parameterized for a limitednumber of selected spectral shapes. The shape mostresembling the actual spectrum is chosen. Furtherapproaches include quite sophisticated calculations ofS nl , such as the discrete interaction approximation ofHasselmann and Hasselmann (1985) and the two-scaleapproximation of Resio et al. (1992), and the near exactcalculations which result from numerical integration ofEquation 3.4.The individual treatment of growth for eachfrequency-direction band in CD models provides acertain inertia in the directional distribution. This allowsthe mean wind-sea direction to lag the wind directionand makes the models more sensitive to lateral limitationsof the wind field or asymmetric boundarycondition. The CD models also develop more directionalfine structure in the spectra than CH models.5.5.4 Third generation modelsA classification of wave models into first, second andthird generation wave models is also used, which takesinto account the method of handling the non-linearsource term S nl :• First generation models do not have an explicit S nlterm. Non-linear energy transfers are implicitlyexpressed through the S in and S ds terms;• Second generation models handle the S nl term byparametric methods, for example by applying areference spectrum (for example the JONSWAP orthe Pierson-Moskowitz spectrum) to reorganize theenergy (after wave growth and dissipation) over thefrequencies;• Third generation models calculate the non-linearenergy transfers explicitly, although it is necessaryto make both analytic and numerical approximationsto expedite the calculations.Results from many of the operational first andsecond generation models were intercompared in theSWAMP (1985) study. Although the first and secondgeneration wave models can be calibrated to give reasonableresults in most wind situations, the intercomparisonstudy identified a number of shortcomings, particularlyin extreme wind and wave situations for which reliablewave forecasts are most important. The differencesbetween the models were most pronounced when themodels were driven by identical wind fields from ahurricane. The models gave maximum significant waveheights in the range from 8 to 25 m.As a consequence of the variable results from theSWAMP study, and with the advent of more powerfulcomputers, scientists began to develop a new, thirdgeneration of wave models which explicitly calculatedeach of the identified mechanisms in wave evolution.One such group was the international group known asthe WAM (Wave Modelling) Group.The main difference between the second and thirdgeneration wave models is that in the latter the waveenergy-balance equation is solved without constraints onthe shape of the wave spectrum; this is achieved byattempting to make an accurate calculation of the S nlterm. As mentioned in Section 3.5 a simplified integrationtechnique to compute the non-linear source term,S nl , was developed by Klaus Hasselmann at the MaxPlanck Institute in Hamburg. Resio et al. (1992) has alsoderived a new method for exact computation of thisterm. The efficient computation of the non-linear sourceterm together with more powerful computers made itpossible to develop third generation spectral waveprediction models (e.g. the WAM model, WAMDIGroup, 1988).Third generation wave models are similar in structure,representing the state-of-the-art knowledge of thephysics of the wave evolution. For the WAM model thewind input term, S in , for the initial formulation wasadopted from Snyder et al. (1981) with a u * scaling

66GUIDE TO WAVE ANALYSIS AND FORECASTING90% confidence limitsSignificant wave height (m)3G WAMDay of the month with hour of the dayFigure 5.4 — A comparison of calculated and observed wave heights during Hurricane Camille (1969). At theheight of the storm the wave sensor failed (WAMDI Group, 1988)instead of U 5 (see Section 3.2). This has been supersededby a new quasi-linear formulation by Janssen (1991) (seealso Komen et al., 1994) which includes the effect of thegrowing waves on the mean flow. The dissipation sourcefunction, S ds , corresponds to the form proposed byKomen et al. (1984), in which the dissipation has beentuned to reproduce the observed fetch-limited wavegrowth and to eventually generate the fully developedPierson-Moskowitz spectrum. The non-linear waveinteractions, S nl , are calculated using the discrete interactionapproximation of Hasselmann et al. (1985). Themodel can be used both as a deep water and a shallowwater model. Details are described by the WAMDIGroup (1988) and a comprehensive description of themodel, its physical basis, its formulation and its variousapplications are given in Komen et al. (1994).Other models may differ in the propagationschemes used, in the method for calculating the nonlinearsource term, S nl , and in the manner in which theydeal with shallow water effects and the influence ofocean currents on wave evolution.The WAM model has shown good results inextreme wind and wave conditions. Figure 5.4 shows acomparison between observed significant wave heightsand significant wave heights from the WAM modelduring Hurricane Camille, which occurred in the Gulf ofMexico in 1969. The grid spacing was a 1 /4° in latitudeand longitude. The comparison shows a good performanceof the model in a complicated turning windsituation.The WAM model is run operationally at theEuropean Centre for Medium-range Weather Forecasts(ECMWF) on a global grid with 1.5° resolution. It isalso run operationally at a number of other weatherservices including the National Weather Service (USA)and the Australian Bureau of Meteorology. Further, itmay be run at higher resolution as a nested regionalmodel, and once again a number of national MeteorologicalServices have adopted it in this mode (see alsoTable 6.2 in Chapter 6). It has also been used in wavedata assimilation studies using data from satellites (e.g.Lionello et al., 1992).

CHAPTER 6OPERATIONAL WAVE MODELSM. Khandekar: editor6.1 Introductory remarksSince the pioneering development of wave forecastingrelations by Sverdrup and Munk (1947), operationalwave analysis and forecasting has reached quite asophisticated level. An introduction to modern numericalwave models has been provided in Chapter 5.National Meteorological Services of many maritimecountries now operationally use numerical wave modelswhich provide detailed sea-state information at givenlocations. Often this information is modelled in the formof two-dimensional (frequency-direction) spectra. Thetwo-dimensional spectrum, which is the basic output ofall spectral wave models, is not by itself of great operationalinterest; however, many wave products which canbe derived from this spectrum are of varying operationalutility depending upon the type of coastal and offshoreactivity. An example of a schematic representation of atwo-dimensional spectrum is given in Figure 6.1. Thespectra were generated by the Canadian Spectral OceanWave Model (CSOWM), developed by the AtmosphericEnvironment Service (AES) of Canada. The four spectrashown refer to four overpass times of the satelliteERS-1 during the wave spectra validation field experimentconducted on the Grand Banks of Newfoundlandin the Canadian Atlantic from 10 to 25 November 1991.The wind-sea regions of the wave field are the elongatedenergy maxima located opposite to the wind directionand around the wind-sea frequency of about 0.15 Hz.The wave energy maxima outside of the wind-sea regionare the swells generated by the model during the twoweekperiod of the field experiment (for additionaldetails, see Khandekar et al., 1994).Significant wave height may be regarded as themost useful sea-state parameter. As defined earlier (inSection 1.3.3), the significant wave height describes thesea state in a statistical sense and is therefore of universalinterest to most offshore and coastal activities. Thesignificant wave height can be easily calculated from thetwo-dimensional spectrum, using a simple formula.Besides significant wave height, two other parameters,which are of operational interest, are the peak period (or,15 UTC 14 Nov. 1991 15 UTC 20 Nov.00 UTC 21 Nov.00 UTC 24 Nov.Figure 6.1 —Normalized wave directional spectragenerated by the AES CSOWM at a gridpoint location in the Canadian Atlantic.The spectra are presented in polar plotswith concentric circles representingfrequencies, linearly increasing from0.075 Hz (inner circle) to 0.30 Hz (outercircle). The isopleths of wave energy arein normalized units of m 2 /Hz/rad and areshown in the direction to which wavesare travelling in relative units from 0.05to 0.95. Model generated wind speed anddirection are also shown (fromKhandekar et al., 1994)

68GUIDE TO WAVE ANALYSIS AND FORECASTINGFigure 6.2 —A simplified output from the AESCSOWM in the form of a four-panelchart of significant wave heightcontours. The wave parameters areusually plotted according to a waveplot model shown below. Only thewind parameters and swell directionsare displayed here. The panels show:the diagnostic (0-hour) wave heightchart for 14 March 1993, 12 UTC (topleft), the 12-hour forecast valid for 15March, 00 UTC (bottom left), the 24-hour forecast valid for 15 March, 12UTC (top right), and the 36-hour forecastvalid for 16 March, 00 UTC(bottom right) (courtesy CMC,Canada; after Khandekar and Swail,1995)Full plotting schemeWind WavePeriod(s)Wind Speed Swell Wave10 13and Direction Period(s)Wind wave Swell WaveHeight (m)8 2Height (m)Swell Direction. . . . . . . . . . . .Ice valid 14 March 1993,00 UTC

OPERATIONAL WAVE MODELS 69in some applications, the zero up/downcrossing period)and the direction in which waves are moving. It shouldbe noted that the convention of wave direction fromwave models varies (i.e. “from which” direction or “towhich” direction), whereas measured data are invariablypresented as the direction “from which” waves travel,consistent with the meteorological convention for winddirection.The parameters for significant height, peak periodand direction can be further separated into their wind-seaand swell components giving a total of six additionalwave parameters of operational interest. For the sake ofcompleteness, these six wave parameters are oftenaccompanied by three related atmospheric parameters,namely surface pressure, wind speed and wind direction.Either all, or a selected few of these nine parameters canbe suitably plotted on a map and disseminated to users.The algorithms for separating wave trains may be quitesimple (see Section 3.6) or they may fully partition thespectrum into the wind sea and primary and secondaryswell features (Gerling, 1992). In this case for eachfeature integral parameters for the significant waveheight, mean period and mean direction are computed.6.2 Wave chartsA map (or chart) showing spatial distribution of aselected number of wind and wave parameters is called awave chart. For efficient transmission, a wave chartshould be simple and uncluttered. Almost all wavecharts show isopleths of significant wave height suitablylabelled and a few additional parameters like peakperiod, wave direction, etc. The chart may provide seastateinformation in a diagnostic (analysed) or in aprognostic (forecast) form.Examples of typical output from operational centresThe AES operational wave model CSOWM is run twicedaily at the Canadian Meteorological Center (CMC) inMontreal and is driven by 10 m level winds generated bythe CMC regional weather prediction model. Figure 6.2provides an example of a four-panel wave chart showingwave fields at analysis time (zero-hour forecast) plusforecast fields out to 12, 24 and 36 hours, respectively.The wave height contours refer to the total wave height,H, which is defined as:H 2 = H 2 2wi + H sw (6.1)In Equation 6.1, H wi and H sw are wind-wave and swellwaveheights, respectively. The wave charts cover thenorth-west Atlantic region extending from the CanadianAtlantic provinces to about 20°W and refer to the “stormof the century” in mid-March 1993 when significantwave heights of 15 m and higher were recorded in theScotian Shelf region. A similar four-panel chart coveringthe Canadian Pacific extending from the west coast ofCanada out to about the International Date Line is generatedby the Pacific version of the CSOWM, which is alsorun twice a day at the CMC.Wave charts prepared and disseminated by othernational Services typically include wave height contoursand a few other selected parameters. For example:• The GSOWM (Global Spectral Ocean Wave Modelof the US Navy), which was run operationally atthe Fleet Numerical Meteorology and OceanographyCenter in Monterey, California, until May1994 (when it was replaced by an implementationof the WAM model, see Section 5.5.4), producedwave charts depicting contours of significant waveheight and primary wave direction by arrows (seeClancy et al., 1986);• The National Oceanographic and AtmosphericAdministration (NOAA) in Washington, D.C.(USA) has been running a second generation deepwaterglobal spectral wave model since 1985; themain output from the NOAA Ocean Wave (NOW)model is a wave chart depicting contours of significantwave height and arrows showing primarywave direction. In addition to the global wavemodel, NOAA also operates a regional wave modelfor the Gulf of Mexico, which is also a secondgeneration spectral wave model but with shallowwaterphysics included. A typical output from theregional model is given in Figure 6.3;• The Japan Meteorological Agency (JMA) in Tokyohas been operating a second generation coupleddiscrete spectral wave model (Uji, 1984) for thenorth-west Pacific and a hybrid model for thecoastal regions of Japan. A typical output from thecoupled discrete model is the wave chart inFigure 6.4.In addition, directional wave spectrum outputcharts for selected locations in the north-westPacific and in the seas adjacent to Japan areprepared and disseminated — see Figure 6.5 whichshows two-dimensional spectra at 12 selected locationsnear Japan. The wave spectra are displayed inpolar diagrams. Such detailed spectral informationat selected locations can be very useful for offshoreexploration and related activities.For a general discussion on the various types(classes) of wave models see Section 5.5.6.3 Coded wave productsMeasured wave data (perhaps from a Waverider buoy ora directional wave buoy) can be reported or disseminatedusing a code called WAVEOB. This code is constructedto allow for wave parameters and one-dimensional ortwo-dimensional spectral information.For a one-dimensional spectrum (perhaps measuredby a heave sensor), the report consists of a maximumspectral density value followed by ratios of individualspectral densities to the maximum value. For a twodimensionalspectrum, the report consists of directionalwave functions in the form of mean and principal wavedirections together with first and second normalizedpolar Fourier coefficients.

70GUIDE TO WAVE ANALYSIS AND FORECASTINGFigure 6.3 — A sample output (12-h forecast, valid 12 UTC, 13 September 1992) from the regional Gulf of Mexico shallowwaterwave model operated by NOAA. The chart shows wave height values (in feet) and primary wave directionsplotted at alternate grid points of the model (source: D. B. Rao, Marine Products Branch, NOAA)In addition to the spectral information, theWAVEOB code includes reports of derived wave parameterslike significant wave height, spectral peak period,etc. The details of various sections of the WAVEOB codecan be found in Annex II and in the WMO Manual oncodes (WMO-No. 306, code form FM 65-IX; WMO,1995). A binary format for reporting these data is alsoavailable, namely BUFR (code form FM 94-X Ext.).This code has been used experimentally for several yearsto transmit ERS-1 satellite data over the Global TelecommunicationSystem.For reporting wave model forecast products ina gridded format, the approved code is called GRID(FM 47-IX Ext.) (see WMO-No. 306) in which the locationof a grid point is reported first, followed by otherinformation as defined above. A binary version, calledGRIB, is now widely used.6.4 Verification of wave modelsWith ocean wave models being used in operationalmode, appropriate verification of a wave model againstobserved wind and wave data is necessary and important.The performance of a wave model must be continuallyassessed to determine its strengths and weaknessesso that it can be improved through adjustment or modifications.It is also necessary to develop sufficientconfidence in the model products for operational use.There are a number of levels of model testing andverification. In wave model development a number ofidealized test cases are usually modelled. The basicoutput of spectral wave models is the two-dimensionalwave spectrum and a suitable test would be to use themodel to simulate the evolution of a wave spectrum withfetch or duration for stationary uniform wind fields. Datafrom field experiments, such as the JONSWAP experiment(see Section 1.3.9), can be used to assess themodel’s performance.Ideally, a model would be verified by comparisonswith measured directional wave spectra. However, suchmeasurements are relatively uncommon and there areproblems with interpreting the results of comparisons ofindividual directional wave spectra. For example, inrapidly changing conditions, such as in storms, smallerrors in the prediction of the time of the storm peak can

OPERATIONAL WAVE MODELS 71Figure 6.4 — A sample wave height chart from the wave model MRI-II operated by JMA. The chart shows contours of significant wave heights with thicksolid lines for every 4 m, dashed lines for every metre and dotted lines for every half metre. Prevailing wind-wave directions are shown byarrows, while representative wave periods are shown by a small numeral attached to the arrow (source: Tadashi Asoh, JMA)

72GUIDE TO WAVE ANALYSIS AND FORECASTINGFigure 6.5 — Wave directional spectra at 12 selected locations in the seas adjacent to Japan for the 24-hour forecast valid at 00 UTC, 27 September 1991. The directional spectra arepresented in polar plots with concentric circles representing wave frequencies (in Hz) and isopleths of spectral energy in units of m 2 /Hz/rad. Isopleths are labelled byorder of magnitude (–1 means 10 –1 m 2 /Hz/rad) and indicate the direction from which waves are coming. Also depicted on the perimeters are the wind speed and directionat each location. The values of peak period (s) and significant wave height (m) are shown at the centre of each spectrum plot (source: Tadashi Asoh, JMA)

OPERATIONAL WAVE MODELS 73give large differences in the comparisons. Further, itmust be remembered that in situ measurements are onlyestimates and individual spectral components may havequite large uncertainty due to high sampling variabilityin the estimates.For these reasons, model validation studies areusually performed in a statistical sense using all availabledata, and are based on comparisons of derivedparameters. A common approach is to perform a regressionanalysis on a set of important parameters such asH s , T – z (or T p ) and, if directional wave data are available,the mean direction (for different frequency bands). Inaddition, it may be useful to compare mean wave spectraor cumulative distributions (of, for example, H s ) forsimultaneous model results and measurements. Since thewind field that drives the model is intimately related tothe model wave field, most evaluation studies alsoinclude verification of wind speed and direction.An important requirement for evaluation of a wavemodel is the availability of reliable sea-state measurementsand related weather data. Most of the evaluationstudies reported in the last 15 years have used buoy datafor wind and wave measurements and have occasionallyused analysed weather maps for additional windinformation. A few earlier studies (e.g. Feldhausen et al.,1973) have used visual ship reports for wave height tomake a qualitative evaluation of several wave hindcastingprocedures. More recently, satellite altimeter datahave been used to validate wave model results. Forexample, Romeiser (1993) validated the global WAMmodel (WAMDI Group, 1988; Komen et al., 1994) for aone-year period using global data from the GEOSATaltimeter.Most verification studies have attempted to calculateseveral statistical parameters and analysed themagnitude and variation of these parameters to determinethe skill of a wave model. Among the parametersthat are most commonly used are:• The mean error (ME) or bias;• The root-mean-square error (RMSE);• The Scatter Index (SI) defined as the ratio of RMSEto the mean observed value of the parameter; and• r, the sample linear correlation coefficient betweenthe model and the observed value.A few studies have considered other statistical measuresto evaluate wave model performance such as the slope ofthe regression line between model and observed valuesor the intercept of the regression line on the y axis. Thefour parameters as listed above are the best indicators ofthe performance of a wave model.Verification of operational wave models atnational Services of many countries is an ongoingactivity. At most Services, the initial verification of awave model is performed during the implementationphase and the verification statistics are updated periodicallyso as to monitor the model’s performance.Table 6.1 shows the performance of a few wave modelswhich are presently in operational use at variousnational (international) organizations in different partsof the world. The table presents verification statisticsfor wind speed, wave height and peak period (whereavailable) in terms of four statistics, namely: ME,RMSE, SI and r as defined above. The table includes avariety of models developed over the last twenty years,ranging from first generation (1G) to the recent thirdgeneration (3G) models (see Section 5.5 for a discussionof model classifications).Verification statistics for many more models havebeen summarized elsewhere by Khandekar (1989).Based on these statistics, it may be concluded that a firstgenerationwave model driven by winds obtained from aweather prediction model can provide wave heightsimulations with an RMS error of about 1.0 m and ascatter index of around 35 per cent. The second- andthird-generation wave models, driven by winds fromoperational weather prediction models, can provide waveheight simulations with RMS errors of about 0.5 m andscatter indices of about 25 per cent. The third-generationWAM model (see Section 5.5.4), which has runoperationally since early 1992 at the European Centrefor Medium-Range Weather Forecasts (ECMWF) in theUnited Kingdom, generates wave height values withRMS errors ranging from 0.4 m to 0.7 m in differentparts of world oceans, while the bias (or mean error)ranges from 0.05 m to 0.2 m. Since these statistics weregenerated further improvements have been made both tothe determination of wind fields and the wave modelsthemselves, so it is expected that the results of the verificationswill also be continually improving.A sample wave height plot generated by the USNavy’s GSOWM at a buoy location in the Gulf ofAlaska is shown in Figure 6.6. The figure shows waveheight variation in the Gulf of Alaska as measured by thebuoy and as modelled by the GSOWM. The GSOWMwas driven by winds from the GSCLI (Global SurfaceContact Layer Interface) boundary-layer model. Thefigure also shows wave height variation simulated by theWAM model, which was configured to run on a 1° x 1°grid and was driven by surface wind stress fields generatedby the Navy Operational Global Atmospheric PredictionSystem (NOGAPS). Also shown along with thewave plots are wind and wave error statistics coveringthe one-month period from 20 February to 20 March1992. The GSOWM has a high bias of about 1 m inits wave height simulation on the US west coast. Thisbias is due to the model’s tendency to retain excesslow-frequency energy from swell waves which oftentravel from the central Pacific to the US west coast.Figure 6.7 shows the verification of the global waveheight field generated by the WAM model at ECMWFagainst wave heights measured by the radar altimeteraboard the satellite ERS-1 (see Section 8.5.2). The altimeterwave heights plotted along the sub-satellite trackscover a six-hour window centred around the time forwhich the model wave height field is valid. In general, themodel wave heights show an excellent agreement with

74GUIDE TO WAVE ANALYSIS AND FORECASTINGTABLE 6.1Performance of operational wave models (the statistics quoted in the last columns are defined in the text)Model Organization Location of Verifying data and verification period Wind/wave parameter Verification statistics — analysisbuoys (0-hour forecast unless otherwise indicated)ME RMSE SI r N*CSOWM Environment Canada NW Atlantic 10 buoys on US/Canadian east coast Wind speed (m/s) 1.7 3.2 59 0.68 871(May–July 1990) Wave height (m) 0.19 0.56 42 0.80 974Peak period (s) –1.0 2.4 32 0.30 918NE Pacific 12 buoys on US/Canadian west coast Wind speed 0.0 2.4 41 0.67 575(May–July 1990) Wave height –0.18 0.55 31 0.81 629Peak period –1.4 3.8 42 0.12 630GSOWM US Navy US east coast US/Canada east coast buoys Wind speed –0.1 3.0 37 0.72 967(20 February–20 March 1992) Wave height 0.31 0.86 41 0.83 1016US west coast US/Canada west coast buoys Wind speed –1.2 3.1 55 0.54 644(20 February–20 March 1992) Wave height 0.95 1.16 56 0.58 710Hawaii Buoys around Hawaiian Islands Wind speed –1.3 1.7 23 0.92 203(20 February–20 March 1992) Wave height 0.09 0.63 23 0.59 209NW Pacific Buoys around Japan Wind speed –0.4 2.6 37 0.76 158(20 February–20 March 1992) Wave height 0.25 0.67 39 0.77 160MRI-II Japan Meteorological NW Pacific Buoys around Japan Wave height 0.05 0.58 34 0.83 –Agency (entire year 1991) (24-hour forecast)Coastal wave Coastal Japan 11 buoys off the coast of Japan Wave height –0.06 0.46 39 0.82 –model (hybrid) (entire year 1991) (24-hour forecast)WAM European Centre for NW Atlantic US/Canada east coast buoys Wind speed 0.06 2.0 42 0.64 434Medium-Range Weather (August 1992) Wave height –0.06 0.36 34 0.73 549Forecasts (ECMWF)NE Pacific US/Canada west coast buoys Wind speed –0.01 1.5 23 0.89 369(August 1992) Wave height 0.16 0.40 20 0.88 371Hawaii Buoys around Hawaiian Islands Wind speed –0.42 1.5 20 0.46 351(August 1992) Wave height 0.0 0.35 17 0.50 352NW Pacific Buoys around Japan Wind speed –0.78 2.6 34 0.86 204(August 1992) Wave height –0.08 0.68 31 0.88 194* Number of data points

OPERATIONAL WAVE MODELS 75Waves (m)Wind (m/s)Buoy/GSOWM WavesRMS 1.08Mean error 0.7Scatter Index 0.35Buoy/GSCLI WindsRMS 3.39Mean error 1.5Scatter Index 0.48Buoy/WAM WavesRMS 0.91Mean error 0.1Scatter Index 0.30Buoy/NOGAPS WindRMS 3.90Mean error 2.0Scatter Index 0.55Figure 6.6 — Wind and wave verification at a buoy location (lat. 50°54'N, long. 135°48'W) in the Gulf of Alaska. The plotshows wind speed and significant wave height values for the period 20 February to 20 March 1992 as measuredby the buoy (* **) and as simulated by the wave models GSOWM (– – –) and WAM (——). Also shown arewind and wave verification statistics for GSOWM and WAM (source: Paul Wittman, Fleet NumericalOceanography Centre, Monterey, USA)altimeter wave heights. The WAM model wave height fieldis compared daily against the altimeter fast delivery productfrom the ERS-1 which is received in real time. (Notethat care should be taken in using altimeter data providedby the space agencies, as H s may be significantly biasedand require correction. Further, the bias for data from aparticular instrument may change as algorithms change,and it may differ from one originating source to another.)A list of wave models operated by various nationalMeteorological (and Oceanographic) Services is given inTable 6.2. Most of these wave models are verifiedroutinely against available wind and wave data. Besidesthese wave models, there are other wave models whichhave been developed and are being used either experimentallyor operationally by other organizations likeuniversities, private companies, etc. Additional informationon these wave models and their verification programmesis available elsewhere (see WMO, 1985, 1991,1994(a)).6.5 Wave model hindcastsA wave model in operational use will usually be forcedby forecast winds to produce wave forecasts. However,the model may also be driven by analysed winds pertainingto past events, such as the passage of hurricanes(tropical cyclones) or rapid cyclogenesis over the sea. Insuch cases the wave field generated by the wave modelis called the hindcast wave field. Wave hindcasting is anon-real-time application of numerical wave modelswhich has become an important marine application inmany national weather services.

76GUIDE TO WAVE ANALYSIS AND FORECASTINGFigure 6.7 — Verification of global wave height field generated by the WAM model against wave heights measured by the radar altimeter aboard the satellite ERS-1 (radar altimeter waveheights from 25 September, 21 UTC, to 26 September 1992, 03 UTC; model field 26 September 1992, 00 UTC) (source: A. Guillaume, while at ECMWF, UK)

OPERATIONAL WAVE MODELS 77TABLE 6.2Numerical wave models operated by national Meteorological ServicesCountry Name of model Area Grid Type of model Products Source of wind informationAUSTRALIA WAM Global 3° x 3° Deep water, Forecasts (+12 to +48 h), significant wave, Australian global NWP model, GASPlat./long. coupled spectral swell and wind-sea height, period, directionWAM Australian region 1° x 1° Forecasts (+12 to +36 h) Australian regional NWP model RASP80°E–180°, 0°–60°S lat./long.CANADA Canadian Spectral North Atlantic Coarse mesh Deep water, 4-panel charts (t + 0, 12, 24, 36) plots of swell Regional Finite Element (RFE) modelOcean Wave Model (1.08° long.) 1st generation height and period, wind-wave height and period, surface wind applicable at 10 m level(CSOWM) transverse spectral surface wind speed and direction, contours ofMercator significant wave heightCSOWM North Pacific Coarse mesh Deep water, 4-panel charts (t + 0, 12, 24, 36) plots of swell Global spectral model 1 000 hPa winds(1.08° long.) 1st generation height and period, wind-wave height and period,transverse spectral surface wind speed and direction, contours ofMercator significant wave heightEUROPE WAM Global 1.5° x 1.5° Deep/shallow water 2-D spectra, significant wave height, mean direction, Surface winds (10 m) from ECMWF(EU) lat./long. modes, coupled peak period of 1-D spectra, mean wave period. T+0 analyses and forecastsspectral to T+120 at 6-h intervals; T+132 to T+240 at 12-hintervals. Outputs coded into FM 92-X Ext. GRIBWAM Baltic and 0.25° x 0.25° Same as global model except for T+0 to T+120 atMediterranean lat./long. 6-h intervalsFRANCE VAGMED Western 35 km polar Deep water, coupled Forecast every 3 h up to 48 h. Maps of wave Wind fields from the fine mesh modelMediterranean stereographic at discrete model height contours with swell and wind-sea directions. PERIDOTSea 60°N Directional spectra (telex). Archive of analysed2-D spectra and of forecast height fieldsVAGATLA North Atlantic 150 km polar Deep water, coupled Forecasts every 6 h up to 48 h. Maps of wave Wind fields from EMERAUDE modelstereographic at discrete model, height contours with swell and wind-sea directions.60°N 2nd generation Directional spectra (telex). Archive of analysed2-D spectra and of forecast height fieldsGERMANY Deutscher Atlantic, north of Deep water, coupled Prognoses for 6 h, then at 12-h intervals to 96 h BKF model, 950 hPa level, diagnosticallyWetterdienst 15°N hybrid run twice daily, verification only interpolated for about 20 m above seaAMT für North-east Atlantic, 50 km Continental shelf (1) Maps of wave and swell direction for south Reduced wind components of geopotentialWehrgeophysik southern Norwegian of North Sea, Norwegian Sea and North Sea – 24 h prognoses fields at 1 000 hPa from meteorologicalSea, North Sea coupled hybrid (2) Wind direction and velocity, significant wave forecast model 7-LPEheight, swell height, wave period, swell period –24 h hindcast and forecast for nine geographicpositions

78GUIDE TO WAVE ANALYSIS AND FORECASTINGTABLE 6.2 (cont.)Country Name of model Area Grid Type of model Products Source of wind informationGREECE Mediterranean Central and eastern 100 km polar Deep water, decoupled Daily wave forecasts for 6-h intervals to 48 h Operational atmospheric numericalmodel Mediterranean Sea stereographic propagation (radio-facsimile broadcasts) model of the ECMWF at 1 000 hPaHONG MRI-II 5°–35°N, 2.5° x 2.5° Deep water, coupled Hindcast and 24- and 48-h forecasts of significant Hindcast: 2.5° winds – objective analysisKONG 105°–135°E Mercator discrete waves and swell at grid points every 12 h, charts of 6-h wind observations and incorporationof tropical cyclone wind modelForecast: Prognostic surface wind fieldsfrom NWP modelHK coast Coastal waters 4.4 x 4.4 km Shallow water SMB 6-, 12-, 18- and 24-h forecasts of significant Surface winds at selected coastalaround Hong Kong model with swells wave height and swells every 12 h, charts stations in Hong Kongfrom deep watermodel outputINDIA Sverdrup-Munk Indian seas Deep water, simple Wave forecastsBretschneider SMB modelIRELAND Adapted NOWAMO North Atlantic Hybrid 12-h sea, swell and combined wave heights(Norway)JAPAN MRI-II Western North 381 km 36 x 27 Deep water, spectral Fax chart of analysis and 24-h forecast, daily Surface winds from numerical weatherPacific model, coupled discrete prediction modelMRI-II Seas adjacent to 127 km 37 x 31 Deep water, spectral Fax chart of 24-, 36- and 48-h forecast, Surface winds from numerical weatherJapan model, coupled discrete twice a day prediction modelCoastal wave model Coastal waters of 10 km inshore Hybrid of significant Fax chart of 24-, 36-, and 48-h forecast, Surface winds from numerical weatherJapan of 100 km wave and spectral twice a day prediction modelmodelMALAYSIA GONO Equator to 18°N, 2° x 2° Mercator Deep and shallow Analysis and forecast of 6-h intervals for Analysis: Grid point wind from subjec-110°–118°E coupled hybrid pre-selected points tively analysed chartsForecast: PersistenceNETHER- GONO North Sea and 75 km Cartesian Deep and shallow, Analysis and 6-, 12-, 18-, 24-h forecasts of Numerical atmospheric 4-layer baroclinicLANDS Norwegian Sea stereographic coupled hybrid wind-sea charts every 6 h model based on air pressure analysisNEW – South-west Pacific (in- Polar stereographic Deep water, Daily wave combined height contours, primary Regional 10-layer primitive equation NWPZEALAND cluding Tasman Sea 40 x 30 Cartesian coupled spectral and secondary direction arrows, analysis and 24-, model coupled to surface winds withand Southern Ocean) 190 km at 60°S 48-h prognoses. Swell also at 24 h 2-layer diagnostic boundary layer modelNORWAY WINCH North Sea, Norwegian 75 km polar Deep water, coupled Forecast charts of wave parameters at 6-h intervals, 10 m winds from Norwegian Meteoro-Sea, Barents Sea, stereographic discrete, 2nd time series and wave spectra for selected points logical Institute (DNMI) limited areanorthern parts of north- generation model run up to +48 h. Also 10 meastern Atlantic winds from ECMWF to +168 h

OPERATIONAL WAVE MODELS 79TABLE 6.2 (cont.)Country Name of model Area Grid Type of model Products Source of wind informationSAUDI Advective singular 5°N, 45°E; 20°N, 70°E; 28.1 km Shallow water, Charts of significant wave height and period at From numerical analysis model – aARABIA wave model 30°N, 30°E; advective singular 6-h intervals numerical sea breeze simulation is used43°N, 55°W; Red Sea near coast and a gradient wind modifiedand Arabian Gulf from friction is computedSWEDEN NORSWAM North Sea 100 km Deep water, coupled Wave forecasthybridUNITED European model North Atlantic, east 0.25° lat. x To 200 m depth 12-h hindcast and 36-h forecasts of wind sea and Forecast winds from lowest level ofKINGDOM of 14°W; 30°45'N– 0.4° long. with 2 m resolution, swell sea (height, direction and period), also operational, fine mesh, limited area66°45'N including coupled discrete significant wave height. Fields output on charts or numerical weather prediction models,Mediterranean for selected grid points as meteograms or by every hourand Black Seas printer. Tabulated results of coastal areasGlobal model Global oceans 1.25° lat. x Deep water, 12-h hindcasts and 120-h forecasts of wind sea Forecast winds from lowest level of0.8333° long. coupled discrete and swell sea (height, direction and period), also operational, coarse-mesh, globalsignificant wave height. Fields output on charts for numerical weather prediction model, everylocal use or in digitally-coded bulletins (GRID two hoursor GRIB) on GTSUSA NOAA/WAM Global 2.5° x 2.5° Deep water, coupled Significant wave height charts for T+12, T+24, Lowest layer winds corrected to 10 mlat./long. spectral T+48 and T+72. Charts of direction of peak height from the operational Mediumenergy for Atlantic and Pacific, and period of peak Range Forecast (MRF) modelenergy for Pacific. Alpha-numeric bulletins ofwave spectra, special charts for Gulf of Alaska,Hawaii and Puerto Rico regionsGWAM Global 1° x 1° Deep water, coupled Significant wave height, sea height, swell height, Surface winds stress for Navy Operationallat./long. spectral max. wave height, mean wave direction, sea Global Atmospheric Prediction Systemdirection, swell direction, mean period, sea period, (NOGAPS)swell period, whitecap coverage, T+0 to T+144IOWAM Indian Ocean 0.25° x 0.25° Shallow water, As above, T+0 to T+48 Navy Operational Regional Atmosphericlat./long. coupled spectral, Prediction System (NORAPS)nested in GWAMMEDWAM Mediterranean Sea 0.25° x 0.25° Shallow water, As above, T+0 to T+72 NORAPSlat./long. coupled spectralKORWAM Korean area 0.2° x 0.2° Shallow water, As above, T+0 to T+36 NORAPSlat./long. coupled spectral,nested in GWAM

80GUIDE TO WAVE ANALYSIS AND FORECASTINGThe purpose of a wave hindcast is to generate wavedata that will help describe the temporal and spatial distributionof important wave parameters. The existingnetwork of wave observing buoys is very limited and quiteexpensive to maintain. Consequently, a meaningful waveclimatology describing temporal and spatial distribution ofwave parameters cannot be developed solely from thebuoy data. An operational wave model can be used in ahindcast mode to create a valuable database over historicaltime periods for which very limited wave information maybe available. A reliable and an extensive database can helpdevelop a variety of wave products which can be used indesign studies for harbours, coastal structures, offshorestructures such as oil-drilling platforms and in planningmany other socio-economic activities such as fishing,offshore development, etc.Several wave models, including some of thoselisted in Table 6.2, have been used in the hindcast modeto create wave databases and related wave climatologies.Among the well-known studies reported in last 15 yearsare the following:(a) A 20-year wind and wave climatology for about1 600 ocean points in the northern hemispherebased on the US Navy’s wave model SOWM (USNavy, 1983);(b) A wave hindcast study for the north Atlantic Oceanby the Waterways Experiment Station of the USArmy Corps of Engineers (Corson et al., 1981);and(c)A 30-year hindcast study for the Norwegian Sea,the North Sea and the Barents Sea using a versionof the wave model SAIL, initiated by theNorwegian Meteorological Institute (Eide, Reistadand Guddal, 1985).In Chapter 9, Table 9.3 gives a more extensive listof databases generated from wave model hindcasts.In recent years, wave hindcasting efforts have beenextended to simulate wave fields associated with intensestorms that particularly affect certain regions of theworld. Among the regions that are frequently affected bysuch storms are the North Sea and adjacent northEuropean countries, the north-west Atlantic and the eastcoast of Canada, the north-east Pacific region along theUS-Canadian west coast, the Gulf of Mexico and theBay of Bengal in the north Indian Ocean. Severalmeteorological studies initiated in Canada, Europe andthe USA have developed historical catalogues of theseintense storms and the associated weather patterns. Tworecent studies have simulated wind and wave conditionsassociated with these historical storm events in order todevelop extreme wind and wave statistics. One of thestudies is the North European Storm Studies (NESS)reported by Francis (1987) and the other is EnvironmentCanada’s study on wind-wave hindcast extremes for theeast coast of Canada (Canadian Climate Center, 1991).A similar study for the storms in the north-east Pacific isin progress at Environment Canada and is expected to becompleted shortly. More details on wave climatologyand its applications will be found in Chapter 9.6.6 New developmentsUntil 1991 most operational wave models were purelydiagnostic, both in forecast and hindcast modes. That is,they used wind fields as the only input from which todiagnose wave conditions from these winds. Thesemodels are initialized with similarly diagnosed wavehindcasts. Wave data had been too sparse to considerobjective analyses in the same sense as those performedfor initializing atmospheric models. However, widespreadwave and wind data from satellites have becomea reality and there are opportunities for enhancing thewave modelling cycle with the injection of such data.One of the goals of the WAM Group (see Komen etal., 1994) was to develop data assimilation techniques sothat wave and wind data, especially those from satellites,could be routinely used in wave modelling. An algorithmwas developed to incorporate ERS-1 altimeter wind andwave data (see Section 8.5.2), and the ECMWF model wasmodified to use ERS-1 scatterometer (see Sections 2.2.4and 8.5.4) data. Further, some techniques for retrievingwind information from scatterometers use wave modeloutput. Similar efforts have been made at several otherinstitutes, including the UK Meteorological Office inBracknell, Environment Canada in Downsview, Ontario,the Australian Bureau of Meteorology in Melbourne, theNorwegian Meteorological Institute in Oslo, and the USNavy’s Fleet Numerical Meteorology and OceanographyCenter, in Monterey, California. A list of projects isprovided in Chapter 8, Table 8.1. It is expected that in thenext few years, satellite wind and wave data will be assimilatedroutinely in increasing numbers of operational wavemodels.

CHAPTER 7WAVES IN SHALLOW WATERL. Holthuijsen: editor7.1 IntroductionThe evolution of waves in deep water, as treated inChapter 3, is dominated by wind and by propagationalong straight lines (or great circles on the globe). Whenwaves approach the coast, they are affected by thebottom, currents and, very close to shore, also by obstacles,such as headlands, breakwaters, etc., the effects ofwhich usually dominate — surpassing the effects of thelocal wind — and the resulting wave propagation is nolonger along straight lines.When approaching the continental shelf from theocean the initial effects of the bottom on the waves arenot dramatic. In fact, they will hardly be noticeable untilthe waves reach a depth of less than about 100 m (orrather, when the depth is about one-quarter of the wavelength).The first effect is that the forward speed of thewaves is reduced. This generally leads to a slight turningof the wave direction (refraction) and to a shortening ofthe wavelength (shoaling) which in turn may lead to aslight increase or decrease in wave height. Wind generationmay be enhanced somewhat as the ratio of windspeed over wave speed increases when the waves slowdown. However, this is generally masked by energy lossdue to bottom friction. These effects will be relativelymild in the intermediate depths of around 100 m but theywill accumulate so that, if nothing else happens, theywill become noticeable as the distances increase.When the waves approach the coast from intermediatewater depth and enter shallow water of 25 m or less,bottom effects are generally so strong (refraction anddissipation) that they dominate any wind generation. Theabove effects of refraction and shoaling will intensify andenergy loss due to bottom friction will increase. All thissuggests that the wave height tends to decrease but propagationeffects may focus energy in certain regions,resulting in higher rather than lower waves. However, thesame propagation effects may also defocus wave energy,resulting in lower waves. In short, the waves may varyconsiderably as they approach the coast.In the near-shore zone, obstacles in the shape ofheadlands, small islands, rocks and reefs and breakwatersare fairly common. These obviously interrupt thepropagation of waves and sheltered areas are thuscreated. The sheltering is not perfect. Waves will penetratesuch areas from the sides. This is due to theshort-crestedness of the waves and also due to refractionwhich is generally strong in near-shore regions. Whenthe sheltering is very effective (e.g. behind breakwaters)waves will also turn into these sheltered regions by radiationfrom the areas with higher waves (diffraction).When finally the waves reach the coast, all shallowwater effects intensify further with the waves ending upin the surf zone or crashing against rocks or reefs.Very often near the coast the currents become appreciable(more than 1 m/s, say). These currents may begenerated by tides or by the discharge from rivers enteringthe sea. In these cases the currents may affect waves inroughly the same sense as the bottom (i.e. shoaling,refraction, diffraction, wave breaking). Indeed, wavesthemselves may generate currents and sea-level changes.This is due to the fact that the loss of energy from thewaves creates a force on the ambient water mass, particularlyin the breaker zone near a beach where long-shorecurrents and rip-currents may thus be generated.7.2 ShoalingShoaling is the effect of the bottom on waves propagatinginto shallower water without changing direction.Generally this results in higher waves and is best demonstratedwhen the wave crests are parallel with the bottomcontours as described below.When waves enter shallow water, both the phasevelocity (the velocity of the wave profile) and the groupvelocity (the velocity of wave energy propagation) change.This is obvious from the linear wave theory for a sinusoidalwave with small amplitude (see also Section 1.2.5):gcphase = ω(7.1)k= tanh( kh)kwith wavenumber k = 2π/λ (with λ as wavelength),frequency ω = 2π/T (with T as wave period), localdepth, h, and gravitational acceleration, g. Thesewaves are called “dispersive” as their phase speeddepends on the frequency. The propagation speed ofthe wave energy (group speed c g ) is c g = βc phase withβ = 1 /2 + kh/sinh (2kh). For very shallow water (depthless than λ/25) both the phase speed and the group speedreduce to c phase = c g = √(gh), independent of frequency.These waves are therefore called “non-dispersive”.The change in wave height due to shoaling (withoutrefraction) can be readily obtained from an energy balance.In the absence of wave dissipation, the total transport ofwave energy is not affected, so that the rate of changealong the path of the wave is zero (stationary conditions):d(ds c g E ) = 0,(7.2)where c g E is the energy flux per unit crest length (energyE = ρ w gH 2 /8, for wave height H) and s is the coordinate

82GUIDE TO WAVE ANALYSIS AND FORECASTINGin the direction of wave propagation. Any decrease in c gis therefore accompanied by a corresponding increase inwave height (and vice versa, any increase in c g by adecrease in wave height). This is readily illustrated witha wave perpendicular to a straight coast. Its wave heightcan be determined from this conservation of energytransport (i.e. assuming no dissipation), with:deeper waterCoastwave rayc g,1H 2 = ⋅H 1 ,(7.3)c g,2where H is the wave height and the subscripts 1 and 2refer to any two different locations along the (straight)path of the wave.The inclusion of shoaling in an Eulerian discretespectral wave model as described in Section 5.3 is relativelysimple. Only the determination of c g,x and c g,y inEquation 5.4 needs to be adapted as described above byc g = βc phase .7.3 RefractionIn addition to changing the wave height, the change inphase velocity will turn the wave direction (when thecrests are not parallel to the bottom contours). This isreadily illustrated with a long-crested harmonic waveapproaching a straight coastline at an angle. In this casethe crest of the wave is infinitely long in deep water.When the crest enters shallow water, its velocity isreduced, in the first instance at the shallow water part ofthe crest (see Figure 7.1), which will therefore slowdown, whereas the deeper part of the crest retains (moreor less) its speed. This results in a turning of the cresttowards shallow water (the deep water part keepsrunning whereas the shallow water part delays). In thefinal situation, when the wave runs up the beach, thewave crest is parallel to the beach irrespective of its deepwater direction.In the situation considered here of long-crested,harmonic waves approaching a straight beach (withstraight bottom contours), the direction of the wave isgoverned by the well-known Snell’s law, by which,along the wave ray (the orthogonal to the wave crests):sinθc phase= constant.(7.4)The angle θ is taken between the ray and thenormal to the depth contour. At two different locations,with subscripts 1 and 2, the wave direction can then bereadily determined from sin θ 1 /c phase,1 = sin θ 2 /c phase,2 .When c phase approaches zero (at the beach), the angle θapproaches zero and the crest is parallel with the beach.Again, the change in wave height can be readilyobtained from an energy balance. In the absence of dissipation,the energy transport between two wave rays isnot affected (note the addition of “between two waverays” compared with the shoaling case of Section 7.2 inwhich crests are parallel to the bottom contours). Thismeans that the energy transport between two adjacentwave rays ∆b is constant (stationary conditions):wave crestSea(7.5)An increase in the distance between the orthogonals istherefore accompanied by a corresponding decrease inwave height.In the situation considered here the change in waveray separation can be readily obtained from Snell’s law:∆b 1 /∆b 2 = cos θ 1 /cos θ 2 . It follows then from the energytransport being constant thatH 2 =d( c g E ⋅∆b)= 0.dscosθ 1cosθ 2⋅(7.6)When the waves approach the coast at normal incidencethe crests are parallel to the bottom contours. No refractionoccurs and the wave rays remain straight (at rightangles with the straight depth contours) so that the ratiocos θ 1 /cos θ 2 is equal to 1. The change in wave height isthen entirely due to the change in group velocity as inthe case of shoaling described above (Section 7.2). Theextra ratio √(cos θ 1 /cos θ 2 ) therefore represents the effectof refraction on the wave height.In situations where the coast is not straight, thewave rays are computed with a generalization of Snell’slaw which says that the rate of turning of the wave directiondepends on the rate of change of the phase speedalong the crest (due to depth variation):dθds = – 1 cc g,1c g,2⋅H 1 .∂c∂n ,shallower waterFigure 7.1 — Waves turning towards shallow water due tovariations in phase speed along the crest(refraction)(7.7)where s is the distance along the wave ray and n is thedistance along the wave crest. Manual methods to obtaina wave ray by integrating this wave ray equation havebeen developed (e.g. CERC, 1973) but computers havemostly taken over the task. In this traditional approachone usually calculates refraction effects with a set ofinitially parallel wave rays which propagate from a deep

WAVES IN SHALLOW WATER 83water boundary into shallow water. The changes in wavedirection thus obtained may focus energy in areas wherethe rays converge, or defocus energy in areas where therays diverge. An example of a ray pattern in an area ofnon-uniform depth is given in Figure 7.2.From the energy balance (Equation 7.5) betweentwo adjacent wave rays, it is relatively simple to estimatethe wave height. If no generation or dissipation occursthe wave height at a location 2 is computed from thewave height at a location 1 with the energy balancebetween two adjacent wave rays:∆bH 2 = 1c g,1⋅ ⋅H 1 , (7.8)∆b 2 c g,2where the ray separations and the group speeds are ∆b 1 ,∆b 2 , and c g,1 , c g,2 , respectively.In areas with a very smooth bathometry, theapproach based on the local ray separation generallypresents no problem. However, in the case of an irregularbathometry, the ray pattern may become chaotic (seeFigure 7.2 for example). In such cases a straightforwardtransformation of local ray separation into local waveheights, if at all possible, yields highly erratic values.Quite often adjacent rays will cross and fundamentalproblems arise. The ray separation becomes zero and thewave height would be infinite (as diffraction effects areignored, see Section 7.4). A continuous line of suchcrossing points (a caustic) can be created in classicalacademic examples. A spatial averaging technique tosmooth such erratic results has been proposed by Bouwsand Battjes (1982). An alternative ray method whichavoids the problem of crossing wave rays is to backtrack(a)depthFigure 7.2 —(a) The bathometry of the Haringvliet branch of the Rhineestuary (the Netherlands), (b) The wave ray pattern for theindicated harmonic wave entering the Haringvliet, and (c) Thesignificant wave height computed with the shallow waterspectral wave model SWAN (no triad interactions; Ris et al.,1994) (courtesy Delft University of Technology)(b)(c)wave rays

84GUIDE TO WAVE ANALYSIS AND FORECASTINGthe rays from a given location in shallow water to deeperwater in a fan of directions. The corresponding relationshipbetween the shallow water direction of the ray andits deep water direction provides the wave height estimatein shallow water (Dorrestein, 1960). This techniqueavoids the caustic problem by an inherent smoothingover spectral directions.With either of the above techniques it is relativelysimple to calculate the change in wave height for aharmonic wave. A proper calculation for a random,short-crested sea then proceeds with the combination ofmany such harmonic waves and the inclusion of wavegeneration and dissipation. The spectrum of such adiscrete spectral model is more conveniently formulatedin wavenumber (k) space. This has the advantage that theenergy density, moving along a wave ray withthe group velocity, is constant (Dorrestein, 1960).Generation and dissipation can be added as sourcesso that the energy balance moving along the wave raywith the group velocity c g is (for stationary and nonstationaryconditions):d E(k)= S(k,u,h) . (7.9)d tThis is exploited in the shallow water model ofCavaleri and Rizzoli (1981). However, non-linear effectssuch as bottom friction and non-linear wave-wave interactionsare not readily included in such a Lagrangiantechnique. An alternative is the Eulerian approach whereit is relatively simple to include these effects. This techniqueis fairly conventional in deep water, as described inSection 5.3. However, in shallow water the propagationof waves needs to be supplemented with refraction. Theinclusion of refraction in an Eulerian discrete spectralmodel is conceptually not trivial. In the above wave rayapproach, a curving wave ray implies that the direction ofwave propagation changes while travelling along the ray.In other words, the energy transport continuouslychanges direction while travelling towards the coast. Thiscan be conceived as the energy travelling through thegeographic area and (simultaneously) from one directionto another. The speed of directional change c θ , whiletravelling along the wave ray with the group velocity, isobtained from the above generalized Snell’s law:(7.10)To include refraction in the Eulerian model, propagationthrough geographic x — and y — space(accounting for rectilinear propagation with shoaling) issupplemented with propagation through directionalspace:∂ E( ω,θ)∂ t( )[ ]+ ∂ c θ E ω,θ∂θc θ = – c gc[ ( ) ]+ ∂ c x E ω,θ∂ x∂c∂n .= S( ω,θ,u,h).[ ( ) ]+ ∂ c y E ω,θ∂ y(7.11)In addition to the above propagation adaptations, thesource term in the energy balance requires some adaptationsin shallow water too. The deep water expressions ofwind generation, whitecapping and quadruplet wave-waveinteractions need to be adapted to account for the depthdependence of the phase speed of the waves (WAMDIGroup, 1988; Komen et al., 1994) and some physicalprocesses need to be added. The most important of theseare bottom friction (e.g. Shemdin et al., 1980), bottominduced wave breaking (e.g. Battjes and Janssen, 1978)and triad wave-wave interactions (e.g. Madsen andSorensen, 1993). When such adaptations are implemented,the above example of a confused wave ray patternbecomes quite manageable (Figure 7.2).7.4 DiffractionThe above representation of wave propagation is basedon the assumption that locally (i.e., within a small regionof a few wavelengths) the individual wave componentsbehave as if the wave field is constant. This is usually agood approximation in the open sea. However, near thecoast, that is not always the case. Across the edges ofsheltered areas (i.e. across the “shadow” line behindobstacles such as islands, headlands, rocks, reefs andbreakwaters), rapid changes in wave height occur andthe assumption of a locally constant wave field no longerholds. Such large variations may also occur in theabsence of obstacles. For instance, the cumulative effectsof refraction in areas with irregular bathometry may alsocause locally large variations.In the open sea, diffraction effects are usuallyignored, even if caustics occur. This is usually permissiblebecause the randomness and short-crestedness ofthe waves will spatially mix any caustics all over thegeographic area thus diffusing the diffraction effects.This is also true for many situations near the coast, evenbehind obstacles (Booij et al., 1992).However, close behind obstacles (i.e. within a fewwavelengths) the randomness and short-crestedness ofthe waves do not dominate. Moreover, swell is fairlyregular and long-crested, so the short-crestedness and therandomness are less effective in diffusing diffractioneffects. The need for including diffraction in wavemodels is therefore limited to small regions behindobstacles and to swell-type conditions and mostlyconcerns the wave field inside sheltered areas suchas behind breakwaters and inside harbours (see, forexample, Figure 7.3).For harmonic, unidirectional incident waves, a varietyof diffraction models is available. The mostillustrative model is due to Sommerfeld (1896) whocomputed the penetration of unidirectional harmonicwaves into the area behind a semi-infinite screen in aconstant medium (see also Chapter 5 of Mei, 1989).Translated to water waves, this means that the waterdepth should be constant and the screen is interpreted asa narrow breakwater (less than a wavelength in width).Sommerfeld considers both incoming waves and

WAVES IN SHALLOW WATER 85Figure 7.3 —The diffraction of aharmonic wave behind abreakwater (courtesy DanishHydraulic Institute)reflected waves, although the latter are usually negligible.In that case the technique can be readilyillustrated with Huygens’ principle. Each point on a crestis the source of a new wave on the down-wave side. Ifthe waves are undisturbed, the next down-wave crest canbe constructed from the point sources on the seawardcrest. When the wave propagation is interrupted by anobstacle, the down-wave crests are constructed onlyfrom those seaward point sources that can radiatebeyond the obstacle (along straight lines). This techniqueof reconstructing diffraction has been transferred into agraphical technique (the Cornu spiral, Lacombe, 1965).Graphs for standard situations are available to use forrough estimates (e.g. CERC, 1973; Wiegel, 1964).A joint theory for refraction and diffraction (forcases with an uneven sea-bed) is available in the mildslopeequation of Berkhoff (1972). The key assumptionfor mild slope is that depth variations are gradual buthorizontal variations in wave characteristics may berapid, as in refraction (see for example, Section 3.5 ofMei, 1989). The Berkhoff equation is a generalization ofthe Helmholtz equation which is the basis of the aboveSommerfeld solution. However, the computationaldemands are rather severe in terms of computer capacityand an approach based on the full mild-slope equation isof practical use only for small areas with dimensions of afew wavelengths (Booij et al., 1992).The mild-slope equation includes the effects ofwave reflections (e.g. from steep bottom slopes andobstacles). For waves that approach a coast with mildslopes, the reflection may often be neglected, usuallyimplying that only the wave height variations along thewave crest are relevant (and not the variations in the wavedirection). This provides a computational option (theparabolic approach) in which the solution is obtainedsimultaneously for all points along a line more or lessaligned with the wave crests. The solution then movesdown-wave through a succession of such lines. Thisapproach saves considerable computer capacity, so it issuitable for substantially larger areas than the full mildslopeequation. It has been implemented in numericalmodels in which also the effects of currents, wind anddissipation due to bottom friction and bottom inducedbreaking have been included (Vogel et al., 1988).The mild-slope equation, and its parabolic simplification,is expressed in terms of wave height variations. Ithas not been formulated in terms of energy density and,until such a formulation is developed, diffraction cannot beincorporated into the spectral energy-balance equation.A somewhat more general approach than the mildslopeequation is provided by the Boussinesq equation(e.g. Abbott et al., 1978). The basic assumption of a harmonicwave is not required and the random motion of thesea surface can be reproduced with high accuracy (e.g.Schäffer et al., 1992) except for plunging breakers. However,this approach has the same limitations as the mildslopeequation (suitable only for small areas) and it too isnot suitable for solving the spectral energy balance.7.5 Wave growth in shallow watersIn situations with moderate variations in the wave field(therefore at some distance from the coast, or in theidealized case of a constant wind blowing perpendicularlyoff a long, straight coast over water with uniformdepth), the frequency spectrum of the waves seems tohave a universal shape in shallow water in the samesense as it seems to have in deep water (where thePierson-Moskowitz and the JONSWAP spectra havebeen proposed, see Section 1.3.9). The assumption of ak –3 spectral tail in the wavenumber spectrum (Phillips,1958) leads, in deep water, to the corresponding f –5 tailin the Pierson-Moskowitz and JONSWAP spectra. Thesame assumption in shallow water leads to another shapeof the frequency tail, as the dispersion relationship is

86GUIDE TO WAVE ANALYSIS AND FORECASTINGdifferent. For very shallow water the result is an f –3frequency tail. This assumption has lead Bouws et al.(1985) to propose a universal shape of the spectrum inshallow water that is very similar to the JONSWAP spectrumin deep water (with the f –5 tail replaced with thetransformed k –3 tail). It is called the TMA spectrum.The evolution of the significant wave height and thesignificant wave period in the described idealized situationis parameterized from observations in deep andshallow water with the following formulations:* * m33H Atanhκ h= ( )⎡ * m1Xtanh ⎢κ1tanh*⎢ κ h⎣4m4( )⎡ * m2* * mXT = Btanh( h ) tanh ⎢κ⎤422π κ⎥4, (7.12)tanh* m⎢4( κ h⎣4 ) ⎥⎦where dimensionless parameters for significant waveheight, H * , significant wave period, T * , fetch, X * , anddepth, h * , are, respectively: H * = gH s /u 2 , T * = gT s /u,⎤⎥⎥⎦X * = gX/u 2 , h * = gh/u 2 (fetch is the distance to the upwindshore). The values of the coefficients have been estimatedby many investigators. Those of CERC (1973) are:A = 0.283 B = 1.2κ 1 = 0.0125 κ 2 = 0.077 κ 3 = 0.520 κ 4 = 0.833m 1 = 0.42 m 2 = 0.25 m 3 = 0.75 m 4 = 0.375Corresponding growth curves are plotted in Figure 7.4.7.6 Bottom frictionThe bottom may induce wave energy dissipation in variousways, e.g., friction, percolation (water penetratingthe bottom) wave induced bottom motion and breaking(see below). Outside the surf zone, bottom friction isusually the most relevant. It is essentially nothing but theeffort of the waves to maintain a turbulent boundarylayer just above the bottom.Several formulations have been suggested for thebottom friction. A fairly simple expression, in terms ofthe energy balance, is due to Hasselmann et al. (1973) inthe JONSWAP project:Dimensionless significant wave height H*0.30.20.10.020.01h* = ∞10 0 10 1 10 2 10 3 10 4 10 5 10 6 10 7Dimensionless fetch X*510.50.20.10.050.02Dimensionless period T*832h* = ∞0.021Figure 7.4 —Shallow water growth curves fordimensionless significant waveheight (upper) and period(lower) as functions of fetch,10 0 10 1 10 2 10 3 10 4 10 5 10 6 10 7plotted for a range of depths Dimensionless fetch X*510.50.20.10.05

WAVES IN SHALLOW WATER 87T p = 1.8 sH s = 0.2 mStations1 2 3 4 5 60.8 m1:200,2 m1:104.5 mStation 1 Station 2 Station 3Energy (cm 2 /Hz)––– H s = 0.22 m––––– H s = 0.22 m––– H s = 0.23 m––––– H s = 0.22 m––– H s = 0.23 m––––– H s = 0.21 mStation 4 Station 5 Station 6Energy (cm 2 /Hz)––– H s = 0.19 m––––– H s = 0.19 m––– H s = 0.13 m––––– H s = 0.13 m––– H s = 0.09 m––––– H s = 0.11 mFrequency (Hz) Frequency (Hz) Frequency (Hz)Figure 7.5 — Comparison of spectral observations and computations of wave breaking over a bar in laboratory conditions.Solid lines indicate the computations and dashed lines the experiment (Battjes et al., 1993) (courtesy: DelftUniversity of Technology)Sbottom ωθ, – Γgωkh E ωθ , ,sinh( ) = ( )(7.13)where Γ is an empirically determined coefficient.Tolman (1994) shows that this expression is very similarin its effects to more complex expressions that have beenproposed.7.7 Wave breaking in the surf zoneWhen a wave progresses into very shallow water (withdepth of the order of the wave height), the upper part of thewave tends to increase its speed relative to the lower part.At some point the crest attains a speed sufficiently high toovertake the preceding trough. The face of the wavebecomes unstable and water from the crest “falls” alongthe forward face of the wave (spilling). In extreme casesthe crest falls freely into the trough (plunging). In all cases22a high-velocity jet of water is at some point injected intothe area preceding the crest. This jet creates a submergedwhirl and in severe breaking it forces the water up again togenerate another wave (often seen as a continuation of thebreaking wave). This wave may break again, resulting inan intermittent character of the breaker (Jansen, 1986).Recent investigations have shown that the overalleffect of very shallow water on the wave spectrum canbe described with two processes: bottom induced breakingand triad interaction. The latter is the non-linearinteraction between three wave components rather thanfour as in deep water where it is represented by thequadruplet interactions of Section 3.4.The breaking of the waves is of course very visiblein the white water generated in the surf zone. It appearsto be possible to model this by treating each breakeras a bore with a height equal to the wave height. The

88GUIDE TO WAVE ANALYSIS AND FORECASTINGdissipation in such a bore can be determined analyticallyand, by assuming a certain random distributionof the wave heights in the surf zone, the total rate ofdissipation can be estimated. This model (e.g. Battjesand Janssen, 1978) has been very successful in predictingthe decrease of the significant wave height in thesurf zone.The effect of triad interactions is not as obviousbut, when wave records are analysed to show the spectralevolution of the waves in the surf zone, it turns outthat, in the case of mild breaking, a second (highfrequency)peak evolves in the spectrum. In that casethe waves transport energy to higher frequencies withlittle associated dissipation. This second peak is notnecessarily the second harmonic of the peak frequency.In cases with more severe breaking, the transport tohigher frequencies seems to be balanced by dissipationat those frequencies as no second peak evolves (whilethe low frequency part of the spectrum continues todecay). In either case numerical experiments haveshown that these effects can be obtained by assumingthat the dissipation is proportional to the spectralenergy density itself.The evolution of the waves in the surf zone isextremely non-linear and the notion of a spectral modelmay seem far-fetched. However, Battjes et al. (1993)have shown that a spectral triad model (Madsen andSorensen, 1993), supplemented with the dissipationmodel of Battjes and Janssen (1978) and the assumptionthat the dissipation is proportional to the energydensity, does produce excellent results in laboratoryconditions, even in high-intensity breaking conditions(Figure 7.5). The source term for breaking in the spectralenergy balance, based on Battjes and Janssen(1978), is then:Sbreaking( )2Qbω HmE ωθ ,( ωθ , )= – α,8πEtotal(7.14)in which α is an empirical coefficient of the order one,–ω is the mean wave frequency, Q b is the fraction ofbreaking waves determined with1–QbEtotal= – 8 ,2(7.15)lnQbH mH m is the maximum possible wave height (determined asa fixed fraction of the local depth) and E total is the totalwave energy.The triad interactions indicated above have not (yet)been cast in a formulation that can be used in a spectralenergy balance. Abreu et al. (1992) have made an attemptfor the non-dispersive part of the spectrum. Another moregeneral spectral formulation is due to Madsen andSorensen (1993) but it requires phase information.7.8 Currents, set-up and set-downWaves propagate energy and momentum towards thecoast and the processes of refraction, diffraction, generationand dissipation cause a horizontal variation in thistransport. This variation is obvious in the variation of thesignificant wave height. The corresponding variation inmomentum transport is less obvious. Its main manifestationsare gradients in the mean sea surface and thegeneration of wave-driven currents. In deep water theseeffects are usually not noticeable, but in shallow waterthe effects are larger, particularly in the surf zone. Thisnotion of spatially varying momentum transport in awave field (called “radiation stress”; strictly speaking,restricted to the horizontal transport of horizontalmomentum) has been introduced by Longuet-Higginsand Stewart (1962) with related pioneering work byDorrestein (1962).To introduce the subject, consider a normally incidentharmonic wave propagating towards a beach withstraight and parallel depth contours. In this case thewaves induce only a gradient in the mean sea surface(i.e. no currents):dηd Mxx=dx– 1ρwg( h+η)dx,(7.16)where h is the still-water depth, – η is the mean surfaceelevation above still-water level and M xx is the shorewardtransport (i.e. in x-direction) of the shoreward componentof horizontal momentum:M xx = (2 β – 0.5) E , (7.17)where E = ρ w gH 2 /8 and β = 1 /2 + kh/sinh(2kh).Outside the surf zone the momentum transporttends to increase slightly with decreasing depth (due toshoaling). The result is a slight lowering of the meanwater level (set-down). Inside the surf zone the dissipationis strong and the momentum transport decreasesrapidly with decreasing depth, causing the mean seasurface to slope upward towards the shore (set-up). Themaximum set-up (near the water line) is typically 15– 20per cent of the incident RMS wave height, H rms .In general waves cause not only a shoreward transportof shoreward momentum M xx but also a shorewardtransport of long-shore momentum M xy . Energy dissipationin the surf zone causes a shoreward decrease of M xy whichmanifests itself as input of long-shore momentum into themean flow. That input may be interpreted as a force thatdrives a long-shore current. For a numerical model andobservations, reference is made to Visser (1984). In arbitrarysituations the wave-induced set-up and currents canbe calculated with a two-dimensional current model drivenby the wave-induced radiation stress gradients (e.g.Dingemans et al., 1986).

CHAPTER 8WAVE DATA: OBSERVED, MEASURED AND HINDCASTJ. Ewing with D. Carter: editors8.1 IntroductionWave data are often required by meteorologists for realtimeoperational use and also for climatologicalpurposes. This chapter discusses the three types of wavedata that are available, namely, observed, measured andhindcast. Sections 8.2 and 8.3 deal with visual wavedata. Section 8.4 is a brief description of instruments forwave measurement and includes a discussion of themeasurement of wave direction. Section 8.5 contains areview of remotely-sensed data which are becoming ofincreasing importance in real-time and climatologicalapplications. These new data sources have made possiblethe assimilation of real data into wave modelling procedures,and an introduction to this subject is given inSection 8.6. The analysis of measured data records isbriefly considered in Section 8.7 and some of the repositoriesfor wave data are indicated in Section 8.8.8.2 Differences between visual andinstrumental dataAlthough the simplest method to characterize waves is tomake visual observations of height and period, thisproduces data which are not necessarily compatible withthose from instrumental measurements.It is generally accepted that visual observations ofwave height tend to approximate to the significant waveheight (see definitions in Section 1.3.3). Although thereare several formulae which have been used to convertvisual data to significant wave height more accurately,for almost all practical meteorological purposes it isunlikely to be worth the transformation.Visually observed wave periods are much less reliablethan instrumentally observed ones, as the eye tendsto concentrate on the nearer and steeper short-periodwaves, thereby ignoring the longer-period and moregently sloping waves, even though the latter may be ofgreater height and energy. This can be seen by examinationof joint probability plots (scatter diagrams) ofvisually observed wave period and height. In many ofthese the reported wave period is so short that the steepness(height to length ratio) is very much higher than isphysically possible for water waves. It is almost certainlythe wave period which is in error, not the height.8.3 Visual observationsWaves are generally described as either sea (wind sea) orswell; in this context, sea refers to the waves producedby the local wind at the time of observation, whereasswell refers either to waves which have arrived fromelsewhere or were generated locally but which havesubsequently been changed by the wind.Useful visual observations of wave heights can bemade at sea from ships. Visual observations from land ismeaningful only at the observation site because thewaves change dramatically over the last few hundredmetres as they approach the shore, and the observer istoo far away from the unmodified waves to assess theircharacteristics. Shore-based observations normally applyonly to that particular location and although relevant to astudy of local climatology they are rarely meaningful forany other meteorological purpose.Mariners, by the very nature of their work, can beregarded as trained observers. The observation of wavesis part of their daily routine, and a knowledge of changesin sea and swell is vitally important for them, since suchchanges affect the motion of the ship (pitching, rollingand heaving) and can be the cause of late arrivals andstructural damage.The observer on a ship can usually distinguish morethan one wave train, make an estimate of the height andperiod of each train and give the directions of wavetravel. Waves travelling in the same direction as the windare reported as sea; all other trains are, by definition,swell (although mariners often refer to well-developedseas in long fetches, such as in the “trade winds”, asswell). To a coastal observer, waves usually appear toapproach almost normal to the shore because of refraction.8.3.1 Methods of visual observationVisual observations should include measurement orestimation of the following characteristics of the wavemotion of the sea surface in respect of each distinguishablesystem of waves, i.e. sea and swell:(a) Height in metres;(b) Period in seconds;(c) Direction from which the waves come.There may be several different trains of swellwaves. Where swell comes from the same direction asthe sea, it may sometimes be necessary to combine thetwo and report them as sea. The following methods ofobserving characteristics of separate wave systemsshould be used as a guide.Figure 8.1 is a typical record drawn by a waverecorder and is representative of waves observed on thesea (see also Figures 1.14 and 1.15). However, it cannotindicate that there are two or more wave trains or giveany information on direction. It shows the height of the

90GUIDE TO WAVE ANALYSIS AND FORECASTINGInterval 6 secondsFigure 8.1 — Example of a wave record. In an analysis thewaves marked with arrows are ignored(b)(c)lower when running with the current. Observers maynot necessarily be aware of this in a given case;Refraction effects due to bottom topography inshallow water may also cause an increase ordecrease in wave height;Waves observed from a large ship seem smallerthan those same waves observed from a small ship.sea surface above a fixed point plotted against time, i.e.it represents the up-and-down movement of a floatingbody on the sea surface as it is seen by the observer, andis included to illustrate the main difficulty in visuallyobserving height and period — the irregularity of thewaves. The pattern shown is typical of the sea surface,because waves invariably travel in irregular groups ofperhaps five to 20 waves with relatively calm areas inbetween groups.It is essential that the observer should note theheight and period of the higher waves in the centre ofeach group; the flat and badly formed waves (markedwith arrows) in the area between the groups must beentirely omitted. The analysis should therefore include:• Height and period: that is the mean height andperiod of about 15–20 well-formed waves from thecentres of the groups; of course all these wavescannot be consecutive;• Wave direction: the direction from which the wavescome should be noted, and reported to the nearest10° on the scale 01–36, as for wind direction. Forexample, waves arriving from the west (270°)should be reported as direction 27. Where morethan one wave train is clearly identifiable, the directionas well as an observed height and periodshould be reported for each such train.The observer must bear in mind that only goodestimates are to be reported. Rough guesses will havelittle value, and can even be worse than no estimate atall. The quality of observations must have priority overtheir quantity. If only two, or even one, of the threeelements (height, period, direction) can be measured, orreally well estimated (e.g. at night), the other element(s)could be omitted and the report would still be of value.When different wave trains — for example sea andswell or several swells — are merged, the heights do notcombine linearly. Wave energy is related to the square ofthe wave height and it is the energy which is additive.Consequently, when two or more wave trains arecombined, the resultant height is determined from thesquare root of the sum of the squares of the separatetrains:2 2H = H + H . (8.1)combined sea swellMore than one swell train can be combined if necessary.The following possible systematic errors in theobservation of waves should be borne in mind by anobserver:(a) Waves running against a current are steeper andusually higher than when in still water, while they are8.3.1.1 Observations from merchant shipsA traditional source of wave information has been fromso-called ships of opportunity. Given the lack of surfacebasedmeasurements of waves on the open ocean, thecoverage offered by merchant shipping has been utilizedin the WMO Voluntary Observing Ship (VOS)Programme. The participating ships report weatherinformation including visual observations of waves. Theguidelines for making these observations are set down inthe WMO Guide to meteorological instruments andmethods of observation (WMO, 1996, see Part II,Chapter 4).HEIGHTWith some experience, fairly reliable estimates of heightcan be made. To estimate heights of waves which haverelatively short lengths with respect to the length of theship, the observer should take up a position as low downin the ship as possible, preferably amidships, where thepitching is least violent, and on the side of the ship fromwhich the waves are coming. Observations should bemade during the intervals, which occur every now andthen, when the rolling of the ship temporarily ceases.In the case of waves longer than the ship, thepreceding method fails because the ship as a whole risesover the wave. Under these circumstances, the bestresults are obtained when the observer moves up or downin the ship until, when the ship is in the wave trough, theoncoming waves appear just level with the horizon. Thewave height is then equal to the height of the eye of theobserver above the level of the water beneath him. By farthe most difficult case is that in which the wave height issmall and the wavelength exceeds the length of the ship.The best estimate of height can be obtained by going asnear to the water as possible, but even then the observationcan only be approximate.An observer on a ship can often distinguish wavescoming from more than one direction. Wave characteristicsfrom each direction must be reported separately;only the waves under the influence of the local wind (i.e.of the same direction as the local wind) are termed sea,the others are, by definition, swell and should bereported as such. Sometimes it is possible to distinguishmore than one swell train, and each such wave trainshould be reported separately as swell.PERIOD AND DIRECTIONPeriod and direction can be reported as outlined inSection 8.3.1. The period of waves can often be estimatedby watching a patch of foam or other floating

WAVE DATA: OBSERVED, MEASURED AND HINDCAST 91matter and timing the intervals between successive crestspassing by. It is necessary to do this for a number ofwaves, as many as is convenient and preferably at leastten, to form a good average value.8.3.1.2 Observations from coastal stationsHEIGHT AND PERIODAt coastal stations, it is important to observe the heightand period of waves at a spot where they are notdeformed either by the wave being in very shallow water(i.e. of a depth only a low multiple of the wave height),or by the phenomenon of reflection. This means that thespot chosen for observations should be well outside thebreaker zone; not on a shoal or in an area where there isa steep bottom gradient, nor in the immediate vicinity ofa jetty or steep rocks which could reflect waves back onto the observation point. The observation point should befully exposed to seaward, i.e. not sheltered by headlandsor shoals.Observations are generally more accurate if a fixedvertical staff with some form of graduated markings canbe used to judge the height of passing waves. Observingan object on the surface, such as a floating buoy, mayalso improve estimation.If the observations are to be of use for waveresearch it is important that:(a) They should always be taken at the same place, sothat correction for refraction, etc. can be applied(b)later;The exact mean depth of water at the place andtime of observation should be known, so thatcorrections for change of height with depth can beapplied later.It is worth repeating here that, as for observationsfrom ships, only the well-developed waves in the centresof the groups should be chosen for observation. The flatand badly formed waves between the groups should beentirely omitted, both from height and period observations.The mean height and period of at least 20 waves,chosen as above and hence not necessarily consecutive,should be noted.DIRECTIONCoastal observations of wave direction are only meaningfulat that particular location. If the user does notrealize that the data were obtained at a site whereshallow water has a major effect, i.e. refraction, his interpretationof the report could be erroneous. This problemis described more extensively in Chapters 1 and 7 (seealso Figures 1.6, 1.7 and 1.8 which illustrate someeffects in the coastal zone).8.4 Instruments for wave measurementsMany different techniques are used for the measurementof sea waves. There is no universal instrument suitablefor all wave measurements. An instrument used at onelocation may be entirely inappropriate at another position.Furthermore, the type of instrument to be deployeddepends on the application for which the wave data isrequired. For example, the design of breakwaters at thecoast depends on the properties of the longer waves orswell while the motions of small ships are influenced byshort, steep seas. Care must therefore be taken in thechoice of instrument. Finally, the chosen instrumentmust be capable of being readily deployed and have agood prospect of giving a high return of wave data forclimatological purposes over at least one year.8.4.1 Types of instrumentWave measurement techniques can be grouped into threecategories:(a) Measurements from below the sea surface;(b) Measurements at the sea surface;(c) Measurements from above the sea surface.8.4.1.1 Wave measurements from below the surfaceSystems to measure waves from below the surface havean advantage in that they are not as vulnerable todamage as systems on the surface. However, there areproblems in bringing the data ashore as cable is expensiveand can be damaged. An alternative technique is totransmit the information by radio from a buoy moorednearby.Pressure transducers are most often used in shallowwater (< 15 m) but have also been mounted below thesurface on offshore platforms in deep water. Here thechange in pressure at the sensor is a measure of waveheight. A pressure spectrum at the measurement depthcan be derived from the pressure signal using spectralanalysis. The measured pressures must be corrected forhydrodynamic attenuation with depth. For this purposelinear wave theory (see Section 1.2) is used (althoughthere is evidence that the correction it gives is too small).One-dimensional wave spectra and associated parameterssuch as significant wave height can then becalculated. The attenuation has the effect of filtering outthe shorter wavelengths, but for most practical applicationsthe loss of high frequency wave information is nota disadvantage. However, if the water depth is greaterthan about 10–15 m, the attenuation affects too much ofthe frequency range and the correction factors becomevery large, diminishing the value of the data.Inverted echo sounders on the sea-bed can also beused in shallow water. The travel time of the narrowbeam of sound is directly related to wave elevation andgives a measurement without depth attenuation. Insevere seas, however, the sound beam is scattered bybubbles from breaking waves making the measurementsunreliable.8.4.1.2 Wave measurements at the sea surfaceIn shallow water, where a platform or structure is available,it is possible to obtain wave measurements at thesea surface by using resistance or capacitance wavestaffs. The wave elevation is then directly related to the

92GUIDE TO WAVE ANALYSIS AND FORECASTINGchange in resistance or capacity of the wave staff. Wavestaffs can however be easily damaged by floating objectsand are subject to fouling by marine growth.In deep water, surface-following wave buoys areused. In these systems, the vertical acceleration ismeasured by means of an accelerometer mounted on agyroscope (or partially stabilized platform). Wave heightcan then be inferred from the measured acceleration bystandard methods. (Since vertical acceleration is thedouble time differential of the surface elevation, theessence of the processing is a double time integration ofthe acceleration.) The buoys are carefully moored so thattheir motions are not greatly influenced by the mooring.The most widely used system of this type is the Datawell“Waverider”. Data can be recorded internally on boardthe buoy and transmitted to a nearby shore station or,over larger distances, transmitted by satellite.A Shipborne Wave Recorder (SBWR) has beendeveloped to obtain wave information from measurementsof the motion of a stationary vessel or lightship. Inthese systems pressure recorders mounted below thesurface in the hull on both sides of the ship givemeasurements of the waves relative to the ship.Accelerometers in the ship measure the vertical motionof the vessel. After calibration, the sum of the signalsfrom the pressure sensors and the accelerometers give anestimate of the motion of the sea surface, from which thewave height and period are derived. The SBWR is arobust system which is not as vulnerable as a buoy and isalso able to give wave information in extreme sea conditions,although the measurement accuracy is generallynot quite as good as a Waverider buoy.For more information on analysis of such data seeSection 8.7.8.4.1.3 Measurements from above, but near, thesurfaceWaves can be measured from above the surface bydownward-pointing laser, infra-red, radar or acousticinstruments if a suitable platform is available. However,some offshore platforms can significantly modify thewave field by refraction, diffraction and sheltering. Caremust be taken in siting the instruments to minimize theseeffects. For “jacket” type structures, a rough guide is thatthe measurement “footprint” on the sea surface shouldbe located more than ten leg-radii away from the platformleg. Advantages with this type of instrument arethat they are non-intrusive (that is, they do not disturbthe flow), and can be easily mounted and maintained.Comprehensive intercomparisons between thesesame instruments and buoys have shown that for significantwave height, maximum wave height and variouswave periods, there is no detectable difference (Allenderet al., 1989). However, in recent years, much interest hasbeen shown by the offshore industry in accurately specifyingthe observed asymmetry of high waves, since thisaffects the design of offshore structures. The extremecrest height (distance from mean water level to the wavecrest) turns out to be significantly greater than half themaximum wave height (Barstow, 1995; Vartdal et al.,1989). Generally, downlooking instruments seem tomeasure higher crests than buoys do, but the reasons forthis are not yet resolved.Commonly used instruments of the downlooking,non-intrusive type include the EMI and Schwartz lasersand Saab and Marex (Plessey) radar. The radar instrumentsgenerally measure over a larger footprint on theocean surface, but are less able to measure highfrequency (short) waves.The subject of remote sensing from large distancesis discussed in Section 8.5.8.4.2 Wave directionVarious methods are available for measuring the directionalwave spectra at sea, on offshore installations andat the coast. The most common systems for routine datacollection are various types of wave buoy (Section8.4.2.1) which are still the major source of data in deepwater, away from offshore platforms. Satellite syntheticaperture radar (SAR) measurements (see Section 8.5.5)show some promise, but the data are collected along thesatellite orbit and are therefore only sporadically availableat any fixed location.If a platform is available several other options arepossible including direct sub-surface measurements andremote sensing (Sections 8.4.2.2, 8.4.2.3 and 8.5.3).8.4.2.1 Directional buoysAt present the most commonly used directional wavemeasuring system is Datawell’s Directional Waverider.This buoy measures the acceleration (doubly integratedto give displacement) in the vertical and two horizontaldirections. Time series data over typically 30 minutes arethen analysed on board the buoy to compute directionalwave spectra using similar techniques to theheave/pitch/roll buoy. The latter measures the verticalacceleration and two components of the surface slope.The Directional Waverider buoy has been available since1989 and various intercomparison studies have sincebeen made against conventional heave/pitch/roll buoysand platform instruments (O’Reilly et al., 1995; Barstowand Kollstad, 1991). The consensus is that, at least indeep water, the Directional Waverider provides moreaccurate estimates of the directional wave spectrum thanconventional buoys.A comprehensive intercomparison of commerciallyavailable heave/pitch/roll buoys andplatform-based directional wave systems was carriedout in the mid-1980s (Allender et al., 1989). However,only one of these systems remains on the market today,Seatex’s Wavescan buoy (which also includes a meteorologicalstation). Datawell’s Wavec is still apparentlyavailable but has largely been superseded by theDirectional Waverider. An alternative to the Wavescanis Oceanor’s Seawatch buoy, a multi-sensor marineenvironmental monitoring buoy which can incorporate

WAVE DATA: OBSERVED, MEASURED AND HINDCAST 93a Directional Waverider (Barstow et al., 1994(a)), andof which there are about 25–30 operational worldwide.In the USA, NOAA operates a number of directionalbuoys in its extensive measurement network, includinglarge 6–20 m heave/pitch/roll buoys and a smaller 3 msystem operated closer to the coast. The large buoysfrom which the UKOOA (United Kingdom OffshoreOperators Association) acquired directional wave dataalso used the heave/pitch/roll system. For further informationon systems presently in use refer to Hamilton(1990, presently under revision) and WMO (1985,1991, 1994(a)). For more on analysis of these data seeSection 8.7.2.8.4.2.2 Arrays of wave recordersArrays of recorders can provide directional detail andhave been used in several oceanographic and engineeringstudies. The degree of directional resolution dependson the number of wave recorders and their spacing. Thetypes of instruments which can be used include wavestaffs (Donelan et al., 1985) and pressure transducers(O’Reilly et al., 1995). Wave directional informationcollected from such systems are usually site specific asthe recorded waves are influenced by refraction anddissipation in shallow water. The arrays can be eithermounted on an offshore platform or bottom mounted inshallow water if the depth does not exceed 10–15 m.Similarly, pressure transducer arrays on offshore platformsshould not be located deeper than 10–15 m belowthe surface because of depth attenuation of the surfacewave signature.8.4.2.3 Current metersUltrasonic or electromagnetic current meters whichmeasure the two horizontal components of wave orbitalvelocity in conjunction with a pressure recorder or wavestaff can provide useful directional information. Thesesystems are usually deployed in shallow water but canalso be mounted on offshore platforms, provided theinfluence of the platform is not too great. These systemsare directly analogous to the Directional Waverider orpitch/roll buoy system discussed in Section 8.4.2.1.Perhaps the most well-known, self-contained instrumentof this type is the “S4” current meter. Also commonlyused is the “UCM-40” current meter.8.5 Remote sensing from large distancesIn general, conventional wave recorders measure thedisplacement with respect to time of the water surface ata fixed point. Remote-sensing techniques (E. D. R.Shearman, 1983) are not able to measure in this way.Instead they interrogate an area or footprint which, in thecase of satellite-borne sensors, has typical dimensions ofseveral kilometres (except synthetic aperture radar,which can have a footprint of a few tens of metres) andreport some measure of the average wave conditionsover the whole area. Surface-based radars also can havea range resolution of a few metres, smaller than thewavelengths of the gravity waves being measured,although the footprint is still broad in azimuth. In consequence,low-resolution sensors are rarely used forcoastal or shallow water sites, because the wave fieldthere can vary considerably over distances of much lessthan one kilometre. The strength of remote sensing isthat equipment can usually be kept in a safe place wellaway from the ravages of the sea and be readily accessiblefor maintenance and testing.The satellite is the most spectacular mounting for asea-wave sensor. It is an extremely powerful tooland can obtain vast quantities of oceanic wave dataworldwide.Wave measurements made using remote sensinggenerally use active microwave sensors (radar), whichsend out electromagnetic waves. Deductions about waveconditions are made from the return signal. Beforeconsidering individual sensors, some features of theinteraction between electromagnetic waves and the seasurface will be discussed.8.5.1 Active sensing of the ocean surface withelectromagnetic wavesIf we consider a radar sending energy directly downtowards a level, glassy, calm sea surface, the energy willbe reflected from a small region directly below the radaras if from a mirror. If now the surface is perturbed bywaves, a number of specular reflections (glints) willarise from facets within the radar beam and perpendicularto it — essentially from horizontal facets for anarrow beam radar (see Figure 8.2).Now consider a radio beam pointing obliquely atthe sea surface at a grazing angle ∆ as in Figure 8.3. Ifthe sea wavelength, λ s , is scaled to the radar wavelength,λ r , as shown, backscattered echoes returning to theradar from successive crests will all be in phase andreinforced. The condition for this “Bragg resonantscatter” is:λrλscos ∆ = .(8.2)2In the real sea, there will be a mix of many wavelengthsand wave directions, and the Bragg resonancemechanism will select only those waves approaching orFigure 8.2 — A radar altimeter normally incident on a roughsea, illustrating the different path lengths forenergy reflected from facets (from Tucker,1991)Incident wave frontReflections from facets

94GUIDE TO WAVE ANALYSIS AND FORECASTING8.5.2 Radar altimeterFigure 8.3 — Bragg diffraction of radio waves from a sinusoidalsea-wave. Conditions for constructivelyinterfering backscatter from successive crests(after E. D. R. Shearman, 1983)receding from the radar and of the correct wavelength.For a microwave radar these would be ripples of a fewcentimetres. However, the mechanisms illustrated inFigure 8.4 ensure that the longer waves influence (modulate)the ripples, by:(a) “Straining” (stretching and compressing the ripplesby the orbital motion of the long waves), accompaniedby velocity modulation;(b) Tilting the surface on which the ripples move andso modifying the Bragg resonance conditionthrough ∆ in Equation 8.2; and(c)Hydrodynamic interaction, which concentrates onthe leading edge of long waves, or through theperturbed wind flow over the long-wave crests,both of which produce a roughness modulation.A radar observing backscatter from short waves by thesemechanisms sees exhibited, in the spatial variation ofbackscatter intensity and in the Doppler spectrum, arepresentation of the long waves.Figure 8.4 — Mechanisms of spatial modulation by longgravity waves of radar scattering from capillarywaves (a) Straining and velocity modulation ofshort wind waves by longer wind waves andswell; (b) Tilt modulation; and (c) Roughnessmodulation (after E. D. R. Shearman, 1983)(a)(b)(c)wind waveswellλ sλ r /2∆The higher the waves then, as indicated in Figure 8.2, thegreater the time between the arrivals of returns from thecrests and from the troughs of the waves, and the morespread out is the return pulse. From our knowledge ofthe statistics of the sea surface, this stretching of theshape of the return pulse can be related quantitatively tothe variance of the sea surface, and hence to the significantwave height H m0 (using the spectral definition — seeSection 1.3.8). Satellite altimeters usually have a pulserate of 1 000 Hz; estimates of H m0 are derived on boardand values averaged over one second are transmitted.Theoretically, there should be no need to calibrate thesevalues, but in practice it has been found necessary to doso, comparing them with buoy measurements (Carter etal., 1992).Radar altimeters give near-global coverage andhence provide wave height data for almost anywhere onthe world’s oceans, including areas where data werepreviously very scarce. Four long-term satellite altimetermissions have so far made global measurements. Theseare GEOSAT (1985–89), ERS-1 (1991–), ERS-2(launched April 1995) and Topex/Poseidon (1992–).Raw data from each of these missions can be obtainedfrom the responsible space agencies. Fast delivery datafrom the ERS satellites are also available in near-realtime (see also Section 6.4). Note that these data can besignificantly biased compared to reference buoy data andshould therefore be corrected before use. From thesedata world maps of significant wave height can beproduced, see, for example, Chelton et al. (1981), andChallenor et al. (1990) and the example in Figure 8.5. Inaddition, specialized products are available from anumber of commercial companies (e.g. Oceanor inNorway, Satellite Observing Systems in the UK).The strength of the return pulse is also dependentupon the statistics of the sea surface which are affectedby the wind speed over the surface. So there is a relationshipbetween the strength of the return pulse and thewind speed, but this relationship is complex and notfully understood. In general the stronger the wind, thelower the strength of the return. Various algorithms havebeen derived from observations which give estimates ofwind speed at 10 m above the sea surface from thestrength of the altimeter return, but further work needs tobe done to determine which is the most accurate, see, forexample, Witter and Chelton (1991) and the recent studiesbased on the Topex/Poseidon data (Freilich andChallenor, 1994; Lefevre et al., 1994).8.5.3 Oblique platform-mounted sensorsAn interesting sensor for measurements at fixed sites,using the phenomena shown in Figure 8.4 to measurewave characteristics, is the Norwegian MIROSmicrowave wave radar (Grønlie et al., 1984), which usesan oblique 6 GHz (5 cm wavelength) beam and a radarrange resolution of 7.5 m. The velocity of ripples is

WAVE DATA: OBSERVED, MEASURED AND HINDCAST 9580°N150°W 100°W 50°W 0° 50°E 100°E 150°E80°N60°N60°N40°N20°N4343340°N20°N0°-20°S-40°S-60°S324232420°-20°S-40°S-60°S42-80°S150°W 100°W 50°W 0° 50°E 100°E 150°E-80°S0Figure 8.5 — Mean significant wave height (in metres) for the period January–March 1996 from the Topex altimeter (courtesyD. Cotton, Southampton Oceanography Centre)small compared with that of the long wavelength ofinterest, so that the radial velocity measured by the radarDoppler shift is dominantly due to the orbital motion ofthe long waves. By analysis of the fluctuation of theobserved Doppler shift, the spectrum of the waves travellingalong the radar beam is deduced. Directional wavespectrum measurements are achieved by makingmeasurements in 6 different directions (using a steerablebeam) for a claimed wave-height range 0.1–40 m andperiod 3–90 s.8.5.4 The satellite scatterometerThe satellite-borne scatterometer (Jones et al., 1982) isanother oblique-looking radar sensor. The total echopower from its radar beam footprint is used to estimatewind speed, and the relative return power from differentlook directions gives an estimate of the wind direction,because the small-scale roughness of the sea surface,seen by the radar, is modulated by the longer windwaves. Calibration is achieved by comparison with nearsurfacewind measurements. The scatterometer does notgive information about the waves, except for the directionof the wind waves, but its estimates of wind velocityand hence estimates of wind stress on the sea surface areproving to be very useful inputs to wave models, particularlyin the Southern Ocean where few conventionalmeasurements are made.8.5.5 Synthetic aperture radar (SAR)Practical aircraft and satellite-borne antennas have beamwidths too large to permit wave imaging. In the syntheticaperture technique, successive radar observations aremade as the aircraft or satellite travels horizontally.Subsequent optical or digital processing producesnarrow focused beams and high-grade imaging of thelonger waves, as evidenced by the variation of the radarechointensity (“radar brightness”) produced by themechanisms shown in Figure 8.4. Figure 8.6(a) showsan example from SEASAT of wave imagery and Figure8.6(b) shows a wave directional spectrum (with 180°ambiguity) achieved by analysis of an image. Figure 8.7shows a high resolution scene in north-western Spaintaken from the ERS-1 SAR.SAR has the advantage of being a broad-swathinstrument, with swath width and resolution of about100 km and 25 m, respectively. However, the physicalprocesses underlying its imaging of waves are complexand still not universally agreed. The main difficulty ininterpreting the images of ocean waves is that the seasurface is not at rest, as assumed by the synthetic processor,and the orbital velocities of the longer waves,which transport the ripples responsible for backscatteringthe radar waves, are around 1 m/s. This results in ahighly non-linear effect which can lead to a completeloss of information on waves travelling in the alongtrackdirection. Moreover, waves of length less thanabout 100 m travelling in any direction are not imagedby the SAR — because of smearing and decrease in thesignal-to-clutter ratio.Thus SAR is more likely to provide useful data inthe open ocean rather than in enclosed seas such asthe North Sea where wavelengths tend to be less than100 m, but even in mid-ocean the waves can sometimesbe so short that the SAR will fail to “see” them.Given the directional wave spectrum, it is thenpossible to obtain a good estimate of the spectrum from

96GUIDE TO WAVE ANALYSIS AND FORECASTINGFigure 8.6(a) (left) —SEASAT SAR image of thewave field between theIslands of Foula andShetland (ESA photograph,processed at the RoyalAircraft Establishment,United Kingdom)Figure 8.6(b) (below) —Examples of directionalwave-energy spectrumderived by digital processingof SEASAT synthetic apertureradar data. The fivelevels of greyness indicatespectral amplitude, while thedistance from the centrerepresents wavenumber(2π/λ). The circles are identifiedin wavelength. A200 m swell system isshown coming from ESEand broader spread 100 mwaves coming from ENE.The analysis has an 180°ambiguity (after Beale,1981)the SAR image. Our problem is to go the other way: toderive the directional wave spectrum given the SARimage. Research has shown that if a global wave modelis run to provide an initial estimate of the wave spectrum,then the difference between the observed SARspectrum and that estimated from the model spectrumcan be used to correct the model spectrum and hence toimprove the wave model output (Hasselmann et al.,1991).A practical problem with SAR, which tends toinhibit its use, is the vast quantity of data it produces(108 bits per second) and the consequent cost of processingand acquiring the data. Moreover, these data cannotbe stored on board the satellite, so SAR data can only beobtained when the satellite is in sight of a ground receivingstation. An exception to this is the ERS-1 whichobtains a small SAR image, with a footprint of 5 x 5 km,every 200 km. These “wave vignettes” can be stored onboard and transmitted later to ground.8.5.6 Microwave radiometryAs well as reflecting incident radio waves, the sea radiatesthermal radio noise as a function of its temperatureand emissivity; this radiation can be detected by amicrowave radiometer (analogous to an astronomer’sradio telescope). The emissivity varies with surfaceroughness, amount of foam and, to a small extent, withsalinity. Thus the signal received at the antenna is mainlySecondarycomponentOceanwavelength(metres)Spacecraftvelocity vectorPrimary componentImage energydensity (linear scale)

WAVE DATA: OBSERVED, MEASURED AND HINDCAST 97Figure 8.7 — ERS-1 SAR sub-scene acquired on 17 January 1993 by the ESA station in Fucino, Italy. The image showsthe entrance to the Ria de Betanzos (Bay of La Coruña), north-western Spain. The image represents an areaof 12.8 x 12.8 km. An east-southeast swell (i.e. travelling from west-northwest) enters the bay and thewaves are diffracted as they travel through the narrower opening of the bay. However, waves do not enterthe very narrow northern Ria de El Ferrol. The inner part remains wind and wave sheltered and thereforedark. As the waves get closer to the shore their length becomes shorter, as can be observed near the coast tothe north. The long linear feature through the middle of the image is probably linked to a strong currentshear, not usually visible in such a sea state (image copyright: ESA; J. Lichtenegger, ERS UtilisationSection, ESA/ESRIN, Frascati)a combination of sea-surface temperature and windeffects, modified by atmospheric absorption and emissiondue to water vapour and cloud liquid water. Sincethe sensitivity to each of these parameters is frequencydependent, a multichannel radiometer can be used toseparate them.The sea-surface temperature data complementinfra-red values in that they can be obtained throughcloud, though with less accuracy and poorer spatial resolution.Wind speed can be derived over a broader swaththan from the present satellite scatterometer, but noestimate of wind direction is obtained.8.5.7 Ground-wave and sky-wave HF radarHF radar (employing the high frequency band, 3–30 MHz,wavelength 100–10 m), is of value because it is capableof measuring wave parameters from a ground station toranges far beyond the horizon, utilizing the Dopplerspectrum of the sea echo (E. D. R. Shearman 1983,Barrick and Gower 1986, Wyatt and Holden, 1994).

98GUIDE TO WAVE ANALYSIS AND FORECASTINGRange in km67.560.052.545.037.530.022.515.0Figure 8.8 —Time historyof significantwave heightvs range, asmeasured byHF groundwaveradar(after Wyattet al., 1985)December 8 9 10Ground-wave radars use vertically polarized radiowaves in the HF band and must be located on a seacoast, island, platform or ship. Coverage of 0–200 km inrange is achievable. The potential for continuous seastatemonitoring is shown in Figure 8.8. Recent workindicates that two stations some distance apart can triangulateon particular sea areas and can yield maps ofdirectional spectra with a one-hour update.Sky-wave radars are large installations, but can belocated well back from the sea coast as they employradio waves travelling by way of the ionosphere.Coverage extends from 900 to 3 000 km. Surface winddirection measurement is readily achieved, but ionosphericvariability limits wave-height measurement(Barrick and Gower, 1986).8.5.8 Comparison of remote-sensing methodsPlatform-based microwave sensors fulfil a useful role forpoint measurement in environments where buoys are ahazard. HF ground-wave radars promise to provide mapsof spatial-temporal wave spectra (if necessary, at hourlyintervals), up to 200 km from the coast. Sky-wave radarsextend coverage to oceanic areas, particularly for trackingfronts and hurricanes, but suffer ionosphericlimitations (Georges and Harlan, 1994).Satellite altimeters, scatterometers and radiometersprovide worldwide coverage of wave height and surfacewind for modest telemetry and processing overheads. Dataclose to coastlines can be suspect, either due to land in thesensor footprint or to a delay in the sensor switching fromland to ocean mode as the satellite ground track movesfrom land to sea. Spatial and temporal coverage of theEarth is constrained by the satellite orbit. The satelliteusually has an exact repeat cycle of between three and 35days. The spacing between tracks is inversely proportionalto the repeat cycle — from about 900 km (three-day cycle)to 80 km (35-day cycle) at the Equator. Satellite data aretherefore particularly suitable for long-term wave climatestudies, but they can also be incorporated, with otherobservations, into numerical forecast models or used forvalidating hindcast data.Satellite synthetic aperture radars require expensivereal-time telemetry and off-line processing. Difficulty isexperienced in imaging waves travelling parallel to thesatellite track and waves travelling in any direction if thewavelength is less than about 100 m. But research indicatesthat SAR data can be used with wave models to giveimproved estimates of directional wave spectra.8.6 Wave data in numerical modelsWave models are being increasingly used, not only forpredicting wave conditions a few hours or days ahead,but also for deriving wave climate statistics (see Section9.6.2 et seq.). Wave data are needed to validate themodels, and they have recently (since 1992) also beenused in model initializations.The assimilation of wave data into a model, reinitializingthe model at each time step, can improve itsperformance, and considerable activity is under way todevelop techniques for doing this. Satellite data, withtheir global coverage, are especially useful. Table 8.1lists countries at present carrying out research into theassimilation of satellite data into wave models, andindicates those which are using the data in their operationalwave models.Data from the satellite altimeter are most readilyand widely available, but the only wave parameter itmeasures is significant wave height. How best to incorporatethis one value into a model’s directional wavespectrum has been addressed in several studies and workablesolutions found (see, for example, Thomas, 1988).8.7 Analysis of wave recordsNowadays, most wave data are collected digitally forsubsequent computer analysis. Data are commonlyeither transmitted from the measurement site by radiolink or by satellite (e.g. via ARGOS), or recorded locally(e.g. on board a buoy). A number of systems are availablefor real-time presentation of data in the form of timeseries plots or statistical presentations. In addition toproviding information suitable for operations, suchsystems also give an immediate indication of the correctfunctioning of the system. Traditionally, however, analysiswas carried out by a manual technique based onanalogue chart records (see also Section 1.3).

WAVE DATA: OBSERVED, MEASURED AND HINDCAST 99TABLE 8.1On-going projects on the assimilation of remotely sensed data in ocean wave models(R = Research project; O = Operational project (with starting date in brackets). Collated from responses toa WMO survey of Member states)Members/ Project status Data Data origin InvestigatorsorganizationsAustralia R SAR ERS-1 I. JonesO Altimeter G. WarrenR Scatterometer R. SeamanR/O HF sky-wave radar Jindalee T. KeenanCanada R/O (1994) SAR (wave mode) Radarsat L. WilsonERS-1/ERS-2M. L. KhandekarR. LalbeharryECMWF R/O (1993) SAR ERS-1 (Fast delivery products) A. GuillaumeAltimeterFrance R Altimeter ERS-1 J. M. LefèvreTopex-PoseidonGermany R SAR GEOSAT S. HasselmannAltimeterERS-1Netherlands R/O (1992) Altimeter ERS-1 G. BurgersV. MakinJapan R SAR GEOSAT H. KawamuraScatterometerERS-1AltimeterTopex-PoseidonMicrowave-radiometerNew Zealand R Scatterometer GEOSAT A. LaingAltimeterERS-1Norway R Altimeter ERS-1 M. ReistadOSARUnited Kingdom R/O (1992) Altimeter ERS-1 (Fast delivery products) S. ForemanUSA R/O (1993) SAR ERS-1 D. EstevaScatterometerW. GemmillAltimeter8.7.1 Digital analysis of wave recordsWave measurements are analysed over, typically, a 20- to35-minute record with values of wave elevation sampledevery 0.5–2 seconds. Measurements are then taken eithercontinuously (for operational monitoring) or threehourlyfor long-term data collection, althoughcontinuous measurements are becoming more standard.A preferred oil-industry standard for the measurementand analysis of wave data has been given by Tucker(1993).The most commonly used wave analysis nowadaysestimates the (directional) wave spectra (Section 1.3.7)using Fast Fourier Transform techniques and hencecomputes a standard set of wave parameters (e.g. H m0 ,T m02 ).In the case of directional wave measurements, twoadditional parameters are commonly analysed and storedfor future use. These are the mean wave direction, θ 1 , thedirectional spread, σ, at each frequency of the wavespectrum (see, for example, Ewing, 1986) and associateddirectional wave parameters such as the wave directionat the spectral peak period. Full directional waveanalysis in real time is also commonly available fromcommercial wave measuring systems.8.7.2 Manual analysis of chart recordsA rapid analysis of a 10-minute chart record, similar tothat illustrated in Figure 8.1 (or Figure 1.14), can bemade in the following widely-used manner. This wasoriginally presented by Tucker in 1961, and is morewidely available in Draper (1963) and interpreted byDraper (1966) for practical application.Measure the height of the highest crest above themean (undisturbed) water level, label this A. Measure theheight of the second highest crest, B. Similarly, measurethe depths of the lowest, C, and second lowest, D,troughs:A + C = H 1 B + D = H 2 .

100GUIDE TO WAVE ANALYSIS AND FORECASTINGThe simplest estimate of the significant wave height (theaverage value of the height of the highest one-third of allthe waves) is calculated from:H s = 0.625 H 1 .A second estimate can be obtained from:H s = 0.69 H 2 .The two results can be combined by taking the average.The relationships between H s , H 1 and H 2 dependslightly on the number of upcrossing waves. Theseexpressions are for 100 waves or for a 10-minute recordif – T z = 6 s. See Tucker (1991) for further details.8.8 Sources of wave dataWe discuss the three types of wave data in turn.8.8.1 Visual observationsMuch of the visual wave information is derived fromthe observations made by WMO-recruited VoluntaryObserving Ships under the WMO/VOS Programme.Many of the observations are reported in real time aspart of the routine meteorological reports which arecirculated internationally on the Global TelecommunicationSystem. These reports use the WMOSHIP code (see the Manual on codes, WMO, 1995).The observations are also logged and conveyedthrough the Port Meteorological Officers to centralrepositories.The National Oceanographic and MeteorologicalServices of many nations have such information onvisual observations of waves (and winds) and can beapproached in the first instance for data on observationsin their own regions. The largest collections of worldwidevisual observations are however maintained in theUSA and the UK with information available from thefollowing:• National Climatic Data Centre/World Data CentreA – Meteorology,Federal Building, Asheville, NC 28801, USA.• UK Meteorological Office, Marine ConsultancyService,Johnson House, Meteorological Office, LondonRoad, Bracknell, Berkshire RG12 2SY, UKAn atlas, Global wave statistics was produced byBritish Maritime Technology, Ltd., in conjunction withthe UK Meteorological Office, in 1986 (Hogben etal., 1986). This publication gives the annual andseasonal statistics of waves for 104 sea areas worldwide.A recent extension of the atlas has been made toinclude detailed information in European waters. SeeHogben (1990) for further details. A PC version of theatlas is also available.8.8.2 Measured wave dataIn the 1970s, the Intergovernmental OceanographicCommission (IOC) of UNESCO established a centrewith the responsibility of identifying and cataloguinglocations at which instrumental wave data had beenobtained. A catalogue was published from time to timeduring the 1980s containing a summary of the data andits location. Until recently this service was provided bythe UK through the British Oceanographic Data Centre(formerly the MIAS Data Banking Service). However, atpresent there is no centre responsible for maintainingand updating this information.To facilitate reporting, exchange and archiving ofmeasured wave data, WMO has developed the WAVEOBcode. This enables a uniform format for wave spectraand is sufficiently flexible to cater for a variety of bothdirectional and non-directional spectra. The details ofthis code are given in Annex II.8.8.3 Hindcast wave dataHindcast wave data from numerical wave models areproduced operationally and archived by many majormeteorological services. These centres should be firstapproached for wave data in their regions. Modelled dataare also produced for special case studies by public andprivate organizations.Chapter 9 of this Guide includes a useful catalogueof hindcast climatologies available at the present time(Section 9.6.2).8.8.4 Satellite wave dataSatellite wave data with global coverage are now availablefrom various sources. In raw form, data can beobtained from the space agencies. These data requireconsiderable treatment before they can be used. Variousnational space agencies have, however, funded the workrequired to make these data more readily available.High-level altimeter wave data (sorted, quality controlledand corrected) can be provided from the GEOSAT,Topex/Poseidon and ERS-1 missions by the followinginstitutions:• OCEANOR, Pir-Senteret, N7005 Trondheim,NorwayFax: +47 73 52 50 33(Data also available on a PC MS-Windowsapplication, World Wave Atlas)• Satellite Observing Systems, 15 Church St,Godalming, Surrey GU7 1EL, United KingdomFax: +44 1483 428 691On the world wide web:http://www.satobsys.co.uk• MeteoMer, Quartier des Barestes – RN783480 Puget-sur-Agens, FranceFax: +33 94 45 68 23The GEOSAT data have also been presented inaccessible form both as hard copy and on interactiveCD-ROM in the Atlas of oceans: wind and wave climateby Young and Holland (1996).

CHAPTER 9WAVE CLIMATE STATISTICSD. Carter with V. Swail: editors9.1 IntroductionChapter 8 described methods of measuring waves andexplained how to analyse the results to obtain estimatesof values describing the sea state, such as the significantwave height. These estimates are usually obtained fromrecords made routinely over 15–35 minutes at three-hourintervals. This chapter describes ways of analysing andpresenting the results from such records collected overmany months or years in order to give a description ofthe wave climate at the recording location. Other recordlengths or intervals may be used; for example, measurementsare sometimes made at hourly intervals. Some ofthe results may vary with the recording interval and it isadvisable always to specify this interval when givingwave climate statistics.This chapter first describes the relevant sea-stateparameters and defines the term “return value” which iswidely used to specify extreme environmental values fordesigning structures such as offshore oil platforms andsea defences. Methods of plotting wave data and statisticalanalyses of the data to illustrate the wave climate arebriefly explained, together with the use of these plots tocheck for possible errors in the data. Various methodswhich have been developed over the years to estimate50- and 100-year return values of wave heights are thenoutlined. Some publications recommended for furtherreading — in addition to the references cited — arelisted and the chapter concludes with a section on waveclimatologies and a description of the process of wavehindcasting (Section 9.6). Annex III to this Guidecontains formulae for the statistical distributions used inthis chapter.9.2 Definitions9.2.1 Sea-state parametersThe two parameters most widely used to describe seastate are the significant wave height, H – 1/3 , and the meanzero-upcrossing or zero-downcrossing wave period, T – z,or their spectral equivalents H m0 and T m02 definedin Section 1.3.8. The notations H s and T – z are usedthroughout this chapter to represent both pairs ofparameters — since the methods described are applicableto both.High waves are often of particular importance, andanother sea-state parameter which is commonly used isthe height of the highest wave most likely to occurduring a three-hour interval, H max,3h . Its value may beestimated from H s and T – z (see Section 1.3.6) assumingthese values remain constant for three hours. H max,3hvaries only slowly with T – z and is about 1.9 H s . In recentyears the height of the maximum wave crest (relative tomean water level) has been of great interest, particularlyto the oil industry (interested readers can find moredetails in Barstow, 1995).There are numerous measures of ocean waveperiod, but T – z, which was originally chosen because itcould readily be estimated from an analogue wave trace,remains the most popular for many applications.9.2.2 Return value of wave heightDesigners of structures which have to stand for manyyears need an estimate of the severest conditions likelyto be experienced. The usual parameter chosen todescribe such conditions is either the 50- or the 100-yearreturn value of wave height, where the N-year returnvalue is defined as that which is exceeded on averageonce every N years.This definition assumes that wave climate willremain unchanged over 50 or 100 years, which is mostunlikely. An alternative definition is the height with aprobability of two or one per cent, respectively, of beingexceeded by the highest wave during one year. This isequivalent to the above definition, except it does notallow for the very small probability that the rare 50- or100-year event might occur more than once during ayear.Return value is a statistical parameter, and theengineer in his design has to allow for the possibility ofa wave greater than say the 100-year return value, oreven of several such waves, occurring within a fewyears. Nevertheless, the concept of return value as adesign criterion has proved useful (but see, for example,Borgman, 1963, on risk analysis). The wave heightspecified for the return value can be either H s or H max,3h ,or even the height of an individual wave. Much of theeffort given to wave climate studies in recent years hasconcentrated upon methods for estimating return valuesof these parameters — see Section 9.4. Sometimes, thereturn value of wave crest elevation is also required inorder to specify, for example, the base height of an oilplatform. (Elevation is measured from mean sea level;wave height is measured from crest to trough — seeSection 1.2.1.) Note that the maximum crest height isnot half of H max , as would be the case for a sinusoidalwave, because extreme waves tend to be quite asymmetricwith typical values of crest to total wave height ofaround 0.6.

102GUIDE TO WAVE ANALYSIS AND FORECASTING9.3H s (m)DATEJAN 1979DATEFEB 1979H s (m) H s (m)H s (m)DATEMAR 1979DATEAPR 1979Figure 9.1 — Time series of significant wave height measurements at Ocean Weather Ship “Lima” (57°30'N, 20°W), January–April 1979. Gaps indicate calms (from HMSO, 1985)

WAVE CLIMATE STATISTICS 103Presentation of data and wave climatestatistics9.3.1 Plot of the dataHaving obtained a data set of wave parameters over, say,a year, it is important to plot the results to obtain anoverall view of the range of values, the presence of anygaps in the data and any outliers suggesting errors in thedata, etc. Figure 9.1 shows an example, in the form of a“comb” plot, which provides a good visual impact.9.3.2 Plotting statistical distributions ofindividual parametersEstimates from the data set of the parameter’s probabilitydistribution can be obtained by plotting a histogram.Given, for example, T – z measurements for a year (2 920values if a recording interval of three hours is used),then a count of the number of measurements T – z in, say,0.5-s bins (i.e. 0.0–0.5 s, 0.5–1.0 s, ...) is made andestimates of the probability of a value in each bin isobtained by dividing the total in the bin by 2 920. Sucha plot, as shown in Figure 9.2, is called a histogram.The bin size can, of course, be varied to suit the rangeof data — one giving a plot covering 5–15 bins is probablymost informative. Note that a histogram or combplot of the spectral peak period, T p , may give additionalinformation to a T – z plot. Over large expanses of theworld’s oceans long swells commonly coincide withshorter wind seas, leading to wave spectra which arebimodal (double-peaked form). Over a long measurementseries the histogram distribution may also bebimodal.Wave height data may also be presented in ahistogram, but it is more usual to give an estimate of thecumulative probability distribution, i.e. the probabilitythat the wave height from a randomly chosen member ofthe data set will be less than some specified height.Estimates are obtained by adding the bin totals forincreasingly high values and dividing these totals by thenumber of data values. Sometimes, to emphasize theoccurrence of high waves, the probability of wavesgreater than the specified height is plotted — seeFigure 9.3.9.3.3 Plotting the joint distribution of heightand periodA particularly useful way of presenting wave climatedata, combining both height and period data in onefigure, is an estimate from the data of the joint distributionof H s and T – z (often called the joint frequency tableor scatter table). Data are counted into bins specified byheight and period and the totals divided by the grandtotal of the data to give an estimate of the probability ofoccurrence. In practice — see for example Figure 9.4 —the estimates are usually multiplied by 1 000, thusexpressing the probability in parts per thousand (‰), androunded to the nearest whole number, but with a specialnotation to indicate the bins with so few values that theywould be lost by rounding.It is sometimes more convenient for further analysisto plot the actual bin totals. In any case, the grand totalshould be given with the scatter plot together with thenumber of calms recorded. It is also useful to draw onthe scatter plot lines of equal significant steepness(2πH s /gT – z 2 — see Section 1.3.5). The line representing asignificant steepness of one-tenth is particularly useful,since this seems in practice to be about the maximumvalue found in measurements from open waters. Anyrecords indicating steeper waves should therefore bechecked for possible errors.9.3.4 Checks on the data setsAs mentioned above, the various plots of the data andthe estimates from the data of probability distributionsFigure 9.2 —Histogram ofzeroupcrossingperiod measurements(12 520valid observations,includingsix calms), atthree-hourintervals atOWS “Lima”,December 1975to November1981 (fromHMSO, 1985)Percentage occurrence

104GUIDE TO WAVE ANALYSIS AND FORECASTINGFigure 9.3 —Percentageexceedance ofsignificant waveheight, H s , and themost likely highestwave in threehours, H max,3h ,from measurements(12 520valid observations,including sixcalms), at threehourintervals atOWS “Lima”,December 1975 toNovember 1981(xx=H max,3h ;++=H s ) (fromHMSO, 1985)Percentage exceedanceWave height (m)provide useful indications of possible errors in the data.Of course, no data should be totally discarded simplybecause they do not meet some criteria without furtherinvestigation.Before progressing, it is important to examine theresults so far to see whether to analyse the entire dataset as a whole or whether to divide it in any way toseparate values arising from different physical realizations.It is most important when analysing data fromsites affected by tropical storms to separate out thosecaused by such storms — see Section 9.4.3. The occurrenceof mixed distributions has occasionally beenreported from some sites. For example, Resio (1978)found that wave heights off Cape Hatteras, NorthCarolina, hindcast from storm winds could be betterexplained by fitting different distributions dependingupon the storm tracks.In most parts of the world there is a markedseasonal variation in wave conditions. In higher latitudes,for example, generally much higher waves occurin the winter than in the summer and Dattatri (1973)reports higher waves off the west coast of India duringFigure 9.4 — Joint distribution (scatter plot) of significant wave height, H s , and zero-upcrossing period,T – z, from measurements (12 520 valid observations, including six calms), at three-hour intervalsat OWS “Lima”, December 1975 to November 1981 (Key: n – parts per thousand (PPT),n – number of occurrences (< 1 PPT) (from HMSO, 1985)

WAVE CLIMATE STATISTICS 105the monsoon season of June to September than at othertimes of the year. It would be preferable if this seasonalcycle (or intra-annual variation) could be removedbefore further statistical analysis, but wave data recordsare generally too short for this to be done satisfactorily(because of the large inter-annual variation superimposedupon the intra-annual cycle) and the data areusually analysed as a whole — although distributionplots for each season are often produced (see, forexample, Jardine and Latham (1981) and Smith(1984)). The seasonal cycle, which is of fixed length,does not present quite the same problem as that of tropicalstorms which occur with random frequency fromyear to year.When analysing data with a marked seasonal cycle,it is essential to check that the number of data valuesfrom each season is proportional to its length. Clearly,analysis of, say, 18 months’ data would give a poorindication of the wave climate for any site with a markedintra-annual cycle unless this was taken into account.However, even when given complete years of data, it isstill necessary to check that any gaps are uniformlyspread and that, for example, there are not more gapsduring the winter months.It is difficult to make wave measurements — thesea is often a hostile environment for electronicequipment as well as for man — and long series of wavedata often contain gaps. Providing these occur atrandom, they can generally be allowed for in the statisticalanalysis. Stanton (1984), however, found that gapsin measurements made by a Waverider buoy in the NorthAtlantic off Scotland appeared particularly in high-seastates, which poses considerable problems for the statistician(see Stanton (1984) for further details). Theactive avoidance by merchant ships of heavy weathercould also bias visual wave data statistics.Seasonal cycles are not the only ones to be borne inmind. Some sites can be routinely affected by a markedsea or land breeze. Elsewhere, the state of the tide cansignificantly affect the sea state — either by wavecurrentinteraction or, if the site is partially protected, bythe restricted water depth over nearby sandbanks. Theseare largely diurnal or semi-diurnal cycles and hence areof particular importance if the data set has only one ortwo records per day.9.4 Estimating return values of wave heightIn this section we first discuss how to estimate the 50-year return value of significant wave height from waverecords at sites unaffected by tropical storms. Themethods are applicable not only to data sets of significantwave height, – H 1/3 , but also to data sets of thespectral estimate, H m0 . The notation H s50 will be used forthe 50-year return value of either. The same methods canalso be applied to other data sets — such as H max,3h —and to obtain other return values such as the 100-yearvalue, H s100 .Then we consider the estimation of the 50-yearreturn value of individual wave height and, finally,briefly comment upon estimates of return values in areasaffected by tropical storms. The methods can also beused to analyse “hindcast” wave heights (estimated fromobserved wind fields using wave models described inChapter 6). See Section 9.6.2 for further details.9.4.1 Return value of significant wave heightexcluding tropical storms9.4.1.1 IntroductionThe method usually employed to estimate the 50-yearreturn value of significant wave height is to fit somespecified probability distribution to the few years’ dataand to extrapolate to a probability of occurrence of oncein 50 years. This method is commonly used formeasured in situ data and, recently, has been successfullyapplied to satellite altimeter data (Carter, 1993;Barstow, 1995). Sometimes, the distribution is fitted onlyto the higher values observed in the data (i.e. the “uppertail” of the distribution is fitted).An alternative method, given considerably longerdata sets of at least five years, is that of “extreme valueanalysis”, fitting, for example, the highest valueobserved each year to an extreme value distribution. So,if ten years of data are available — i.e. 29 220 estimatesof H s if recorded at three-hour intervals and allowing forleap years — only ten values would be used to fit theextreme-value distribution, which illustrates why longseries of data are required for this method. (The advantageof the method is explained below.) It is widely usedin meteorology, hydrology, and in sea-level analysis inwhich records covering 20–30 years are not uncommon.There are rarely sufficient data for it to be used for waveanalysis, but it has been applied, for example, to theNorwegian hindcast data set (Bjerke and Torsethaugen,1989) which covers 30 years.F(.) will be used to represent a cumulative probabilitydistribution i.e.:F(x) = Prob (X < x) , (9.1)where X is the random variable under consideration.In our case, the random variable is H s which isstrictly positive soF(0) = Prob (H s < 0) = 0. (9.2)Assuming a recording interval of three hours, so thatthere are on average 2 922 values of H s each year, theprobability of not exceeding the 50-year return valueH s50 is given by1F( H s50) = 1 – ≈ 0. 9999932.50 × 2922(9.3)It is important to note that the value of F(H s50 ) dependsupon the recording interval. If 12-hour data are beinganalysed, then:

106GUIDE TO WAVE ANALYSIS AND FORECASTINGTABLE 9.1Distributions for wave height1 Log-normal Two-parameter distribution — widelyused at one time but found often to be apoor fit, for example to wave data in UKwaters. (Possibly a good fit if theseasonal cycle could be removed.)2 Weibull Sometimes gives a good fit — particularlyif the three-parameter version isused (also often fitted to wind speeds) —more often fitted only to the upper-tail.3 Extreme value The Fisher-Tippett Type I (FT-I) distribudistributionstion often seems to give a good fit tothree-hourly data from the North Atlanticand North Sea. The FT-III is boundedabove so should be more appropriate inshallow water but there is no goodevidence for this. (The FT-I and FT-IIIgive probabilities of H s < 0, but whenfitted to wave data the probabilities arefound to be extremely small.)F ( 1H s50)= 1 – 50 × 365 25× 2 ≈ 0 ..9999726 .Therefore, to estimate H s50 we have to select:(a) The distribution to fit; and(b) The method of fitting.9.4.1.2 The choice of distribution(9.4)The choice of distribution to fit all the data is open,providing that F(0) = 0. There is no theoretical justificationfor any particular distribution. This is an enormousweakness in the method, particularly since considerableextrapolation is involved. In practice, various distributions,from among those which have been used withsome success over the years, are tried and the one givingthe best visual fit is accepted. These distributions, whichare defined in Annex III, are listed with comments inTable 9.1.The choice of distribution to fit the observedmaxima is limited by the theory of extreme values (see,for example, Fisher and Tippett, 1928; Gumbel, 1958;and Galambos, 1978) which shows that the distributionof maxima of m values are asymptotic with increasingm to one of three forms (FT-I, II and III)*. These canall be expressed in one three-parameter distribution: thegeneralized extreme value distribution (Jenkinson,1955). This theoretical result is a great help in the analysisof environmental data, such as winds and wavesfor which the distributions are not known, and explainsthe frequent application of extreme value analysis whensufficient data are available. As usual, the theory____* Figure 9.5 shows an example of FT-I scaling: the cumulativeFT-I distribution appears then as a straight line.contains some assumptions and restrictions, but inpractice they do not appear to impose any seriouslimitations. The assumption that the data are identicallydistributed is invalid if there is an annual cycle. Carterand Challenor (1981(a)) suggest reducing the cycle’seffect by analysing monthly maxima separately, but thishas not been generally accepted, partly because itinvolves a reduction by a factor of 12 in the number ofobservations from which each maximum is obtainedand, hence, concern that the asymptotic distributionmight not be appropriate. Estimates of either or boththe seasonal or directional variation of extreme values(e.g. what is the N-year H m0 for northerly waves) maybe possible, but suffer from decreased confidence dueto shorter data sets. This may result in higher returnvalues in certain directions or months than the all-dataanalysis.The choice of distribution to fit the upper tail is ingeneral limited to those given in Table 9.1. Theory givesasymptotic distributions for the upper tails of anydistribution, analogous to the extreme value theory. Forexample, distributions whose maxima are distributedFT-I have an upper tail asymptotic to a negativeexponential distribution (Pickands, 1975). However, inpractice, there is the problem of determining where theupper tail commences — which determines theproximity to the asymptote — and the theory has notproved useful, to date, for analysing wave data.Sometimes, instead of analysing all values abovesome threshold, a “peaks-over-threshold” analysis iscarried out, analysing only the peak values betweensuccessive crossings of the threshold. Usually theupcrossings of the threshold are assumed to be from aPoisson process and the peak values either from a negativeexponential distribution or from a generalized Paretodistribution (see NERC, 1975; Smith, 1984).9.4.1.3 The method of fitting the chosendistributionThe method of fitting the chosen distribution often involvesthe use of probability paper. Alternatives includethe methods of moments and of maximum likelihood.Probability paper is graph paper with non-linearscales on the probability and height axes. The scales arechosen so that if the data came from the selected distributionthen the cumulative distribution plot, such asshown in Figure 9.3, should be distorted to lie along astraight line. For example, if the selected distribution isFT-I given by⎧ h AF( h) =⎡ ⎤⎫exp ⎨– exp – ( – ) ⎬ (9.5)⎩ ⎣⎢ B ⎦⎥⎭then taking logarithms and rearranging givesh = A + B [– log e (– log e F)]. (9.6)So, a plot of h against –log e (–log e F) should give astraight line with intercept A and slope B. (Note in this

WAVE CLIMATE STATISTICS 107Probability of non-exceedanceReturn period (years)Wave height (m)Figure 9.5 — Cumulative probability distribution of significant wave height, H s , from measurements (12 520 validobservations), at three-hour intervals at OWS “Lima”, December 1975 to November 1981 on FT-Iplotting paper (from HMSO, 1985)case that the height axis remains linear.) Figure 9.5shows such an example.Probability paper can be constructed for manydistributions — including those given in Table 9.1, butthere are some notable omissions (e.g. the gammadistribution).In practice, plotting the data raises difficulties. Ifcumulative values are plotted at, say, 0.5 m intervals asin Figure 9.5, then besides any non-linear groupingeffect, the highest few plotted points are determined byconsiderably fewer observations than those with lowerwave heights, a point can even be plotted for a bin whichcontains no data. Alternatively, a point can be plotted foreach observation by ordering the n observations and thenplotting the r th highest at, say, the expected probability ofthe r th highest of n from the specified distribution.Determining this probability is not a trivial problem andit is often approximated by r/(n + 1) — which is correctfor a uniform distribution U(0,1). See Carter andChallenor (1981(b)) for further details.The distribution is then fitted by putting a straightline through the plotted values, either by eye or by leastsquares. Sometimes, if the values are obviously notco-linear, the line to extrapolate for H s50 is obtained byfitting only the higher plotted values. The visual fit canbe highly subjective and it is advisable to obtain independentcorroboration. Least-squares is not strictlycorrect since the errors are not normally distributed;cumulative plots should employ weighted least squares,but it is often felt that, in estimating H s50 , the highermeasurements should have extra weight.This method is immediately applicable to distributionswith two parameters, such as the FT-I. It isextended to fit three-parameter distributions by plottingthe data over a range of values of the third parameter tosee which value gives the best fit to a straight line (eitherby eye or by minimizing the residual variance or bymaximizing the correlation coefficient). For example, thethree-parameter Weibull distribution is given by:C⎡ h AF( h) = −⎛ – ⎞ ⎤1 exp ⎢–h A⎝ B ⎠⎥ ≥⎣ ⎦ (9.7)= 0h 0, i.e.:1loge( h– A) = loge[ – loge( 1– F)]+ logeB.(9.8)CTherefore, A is chosen by trial at, say, 0.5 m intervals sothat a plot of log e (x – A) versus log e [– log e (1 – F)] isnearest to a straight line, the slope of which is 1/C andthe intercept is log e B.The method of moments is an analytic rather thangeometric method of estimating the parameters of adistribution. (So it avoids any problem of plotting positions.)It is based upon the simple idea that, since themoments of the distribution (mean, variance, skewness,etc.) depend upon the parameter values, estimates ofthese values can be obtained using estimates from thedata of the mean, etc. An example will make this clear:for the FT-I distribution, specified in Equation 9.5, themean and variance are given by:mean = A+γBB(9.9)variance = π 2 26

108GUIDE TO WAVE ANALYSIS AND FORECASTINGwhere: γ = Euler’s constant ≈ 0.5772, and π 2 =(3.14159) 2 ≈ 9.8696.For a data set h i (i = 1, ... n), the estimates of populationmean and variance, h – and s 2 , are given by:1h = ∑ hin i(9.10)2 12s = ∑( hi– h).n – 1The moments estimators of A and B, A ~ and B, ~ areobtained from Equations 9.9 and 9.10, which give:A˜ = h – γB˜(9.11)˜ sΒ ≈ 6 . πIf the individual h i are not available, then the mean andvariance can be estimated from grouped data given in ahistogram or a scatter plot.The two-parameter log-normal distribution isreadily fitted by this method (see Johnson and Kotz,1970, for further details). The three-parameter Weibulldistribution can also be fitted by moments, the thirdparameter being estimated by comparing the theoreticaland estimated skewness (Johnson and Kotz, l970,p. 257). In theory, the FT-III could also be fittedby moments but, in practice, it does not appear to bedone.The method of maximum likelihood consists offinding values for the distribution parameters, such as Aand B in the FT-I (Equation 9.5), which maximize the likelihoodthat the observed data come from this distribution.This is probably the most widely used method of estimationin statistics because it generally gives statisticallyoptimal estimates for large samples, but the method isoften numerically difficult and time-consuming, even if acomputer is available. For example, for FT-I, the maximumlikelihood estimates, Â and ˆB, are given by:ˆ – ˆ ⎡1⎛log exp –ˆh ⎞⎤iA = B e ⎢ ∑⎝ B ⎠⎥⎣ni ⎦∑ h exp –ˆhiiBˆ 1 i B= ∑ hi–exp –ˆ.nhii∑Bi(9.12)iThese equations have to be solved numerically.The two-parameter log-normal and the twoparameterWeibull are readily fitted by maximum likelihood— also requiring numerical solution of equations— but problems can arise when used for the threeparameterWeibull (and the three-parameter log-normal).(See Johnson and Kotz, 1970, for details.)Both the method of moments and the maximumlikelihood method assume data are statisticallyindependent and identically distributed. Neither assumptionapplies to wave data but the methods appear to berobust and to give useful results. To meet more nearlythe requirement for independent data, it might be preferableto fit data recorded at, say, 24-hour intervals — thusobtaining separate estimates from all the midnightvalues, another from all the 03.00 values, etc. — but thisis not usually done in practice. The considerable annualcycle in wave height found in many parts of the worldmeans that the data are often not identically distributed.As already mentioned in Section 9.4.1.2, it would bemore satisfactory to remove this cycle before carryingout any statistical analysis but this is rarely attempted.9.4.1.4 Estimating the associated value of T – zAn estimate of the zero-downcrossing or zeroupcrossingwave period, T – z, at the time when thesignificant wave height has an extreme value, such asH s50 , is obtained by assuming a specific value forsignificant wave steepness. Often, a value of oneeighteenthis used, so that T – z is obtained from:2πHs501= ,2gT 18i.e., using metres and seconds:Tzz≈ 34 . Hs50.Measurements indicate that one-eighteenth isappropriate for open ocean sites, but larger values ofabout one-fourteenth apply at sites where waves arefetch-limited. In practice, a value is often obtained by avisual inspection and extrapolation of the H s – T – z scatterplot. These rather crude estimates of T – z are often sufficient,but if the exact value of T – z is important — forinstance if a structure being designed is sensitive tosmall changes in T – z — then a range of values should bespecified. For example, ISSC (1979) recommends arange from approximately 2.6√H s50 to 3.9√H s50 , whereH s50 is in metres. (The precise recommendation is interms not of T – z but of the period corresponding to thespectral peak frequency, which is given as approximately1.41 T – z.)9.4.2 Return value of individual wave heightA value for the highest individual wave which mightoccur is often obtained from the probability distributionof the maximum individual wave height during the time(about 3–6 hours) for which the return value of significantwave height prevails. The mode — the most likelyvalue — of this distribution is generally quoted but othervalues such as its mean may be used (see Sections 1.3.6and 9.4.1.4 for the required T – z). Alternatively, a data setmay be obtained, consisting of the most likely highestwave corresponding to each measurement of significantwave height, and a return value estimated from thisdata set.Both methods are based upon the assumption thatthe highest individual wave, H max , occurs during thetime of highest significant wave height. This is notstrictly correct. H max will most probably occur duringthis time but might occur by chance during a time oflower significant wave height. Values for the maximum

WAVE CLIMATE STATISTICS 109individual wave height derived using this assumptionare, therefore, likely to be too low. On the other hand, theusual method of estimating the highest individual waveheight (assuming a narrow band sea with crest to troughheight equal to twice the crest elevation) over-estimatesthis height. Errors from these two incorrect assumptionstend to cancel out. For further details see Hogben andCarter (1992).Battjes (1972) gives a method, modified by Tucker(1989), which avoids this assumption. Battjes derivesnumerically the distribution of individual wave heightsfrom the H s – T – z scatter plot, assuming that, for a specifiedsignificant wave height, the individual waves have aRayleigh distribution (Equation 1.16). The number ofindividual waves during the year with this significantwave height is obtained from the T – z distribution in thescatter plot; summation gives the total number of wavesin the year. Extrapolation of the distribution of individualwave heights to a probability of the one wave in the totalnumber in N years — using Weibull probability paper —gives the N-year return value of individual wave height.See Battjes (1972) or Tucker (1989, 1991) for details.9.4.3 Return value of wave height in tropicalstormsThe methods so far described for estimating returnvalues, such as H s50 from a set of measurements, assumethat all the measurements in the set and all the estimatedextreme values come from the same probability distribution,with the implication that they are generated by thesame physical processes. In particular, we have assumedthat very severe storms, which give rise to extremevalues, are essentially the same as other storms, merelymore violent examples. This assumption is obviouslyinvalid in parts of the world where extreme waves areinvariably associated with tropical storms. The analysisof wave records from a site in such areas, using themethods outlined in this chapter, may well describe thegeneral wave climate at the site but sensible estimates ofextreme conditions will not be obtained.Unfortunately, without this assumption, it is notpossible to estimate long-term extremes using only a fewyears’ measurements from one site because the relevantdata are insufficient. We might expect between five and15 tropical storms in an area such as the Caribbean orthe South China Sea during a year, but some might passtoo far from the site to make any impact on the localwave conditions. Severe storms are rare. For example,Ward et al. (1978) estimate that only 48 severe hurricanes— the only ones to “affect extreme wave statisticssignificantly” — occurred in the Gulf of Mexicobetween 1900 and 1974.Estimates of return values of wave heights in areaswhere extremes are dominated by tropical storms maybe obtained by analysing storm data from the whole areaand by constructing a mathematical model to give theprobability distribution of significant wave height. Themodel usually consists of three basic parts:(a)(b)(c)The probability that a storm will occur in the area(a Poisson distribution is assumed with an averageinterval estimated from historical records);The probability that a storm centre will pass withina specified distance of a site (a uniform distributionis assumed over the area or part of it determinedfrom historical track records);The probability of the significant wave heightexceeding a particular value given a tropical stormpassing at a specified distance (from analysis ofwave data throughout the area or hindcast wind/wave model results).Parts (a), (b) and (c) can be put together to give the probabilitydistribution of maximum significant wave heightat any site but (c) may require modification to includefetch-limited constraints.Further details are given, for example, in Ward et al.(1978) and Spillane and Dexter (1976). For estimatingthe height of the highest individual wave in a tropicalstorm, see Borgman (1973).9.5 Further readingStanton (1984) applies many of the procedures andmethods discussed in this chapter to a set of wave dataobtained off north-west Scotland with a clear exposition ofthe methods. Chapter 4 of Carter et al. (1986) expandssomewhat on many of the topics covered here and givesfurther references, see also Goda (1979), which includes adetailed comparison of the various expressions used forwave period. Mathieson et al. (1994) recently published thefindings of an IAHR working group on this subject.Information on fitting distributions can be found,besides that given in Johnson and Kotz (1970), in NERC(1975). A comparison of fitting methods for the FT-I isdescribed in Carter and Challenor (1983) and an applicationof maximum likelihood to fit environmental data toa generalized extreme-value distribution is given inChallenor and Carter (1983).Methods of estimating return wave height andfurther information on fitting distributions are describedby Borgman and Resio (1982), by Isaacson andMackenzie (1981), on pages 529–544 in Sarpkaya andIsaacson (1981), and by Carter and Challenor (1981(b))which includes a summary of some plotting options foruse on probability paper.9.6 Wave climatologiesKnowledge of the wave climatology for a specific location,a region or an entire ocean basin is important for awide range of activities including:(a) Design, planning and operability studies forharbours, coastal structures including fish farms,offshore structures such as oil platforms, andvessels;(b) Coastal erosion and sediment transport;(c) Environmental studies, e.g. the fate of, and clean-upprocedures for, oil spills;(d) Wave energy estimation;

110GUIDE TO WAVE ANALYSIS AND FORECASTING(e) Insurance inquiries, damage or loss of property atsea.The type of wave climate analysis required dependson the particular application, but includes the following:(a) Long-return-period wave heights (e.g. 100 years),associated periods and directions at sites of interestand over regions;(b) Percentage frequency of wave heights or waveperiods by wave direction;(c) Exceedance analyses for wave height and waveperiod;(d) Persistence analysis for wave heights (or waveperiods) greater (or less) than selected thresholds;(e) Joint distribution of significant wave height andwave period;(f) Time series plots of wave heights and wave(g)periods;Relationships between significant wave height andmaximum wave height and crest height.In order to produce many of these statistical analyses along-time series of wave data at one or more locations isrequired.The information used to produce wave climatologiescomes primarily from two sources: (a) wavemeasurements and observations, (b) wave hindcasts.Each of these will be discussed in more detail in thefollowing paragraphs.9.6.1 Climatologies from wave measurementsand observationsFor climatological purposes wave data has traditionallybeen derived from two major sources: (a) visual observationsfrom vessels participating in the VoluntaryObserving Ships scheme; (b) measurements from buoysand ships. Wave data has also in recent years becomeavailable from satellite sensors and marine radar, but thefrequency of observation, length of record and data qualityhave until recently limited their routine use in climatologystudies. Some investigations into the use of satellite radaraltimeter measurements to estimate significant wave heightdistributions and extreme values have been carried out(Carter et al., 1994). Figure 8.5 (from D. Cotton,Southampton Oceanography Centre) shows the globalmean significant wave height distribution from January toMarch 1996 derived from Topex data. Carter et al. (1991)analysed the global variations in monthly mean waveheights using one year of GEOSAT data. At present,significant wave height data are available globally fromGEOSAT (1986–89), Topex/Poseidon (1992–), ERS-1(1991–) and ERS-2 (launched April 1995), see, forexample, Young and Holland (1996), in which detailedclimatologies of the world’s oceans are derived fromGEOSAT data. Already these data are giving us highqualitywave climate information of particular value inFigure 9.6 —Cumulative distribution of H s values fromGEOSAT transects of a 2° x 2° bin south of NewZealand from November 1986 to October 1989plotted on a FT-I scale. The line was fitted bymaximum likelihood and extrapolated to give a50-year return wave height, assuming three-hourvalues, of 16.5 m (courtesy Satellite ObservingSystems, Godalming, UK)0.9999GEOSAT data: 47°S, 167°E50 yrsDetailed location of data cell-38-40-42-44-46Probability (FT-I scale)0.9990.99-48-50-520.9Loc = 3.139Scale = 1.119H s50 = 16.451-540.5-56156 158 160 162 164 166 168 170 172 174 1760.10.010 5 10 15 20Hs (m)

WAVE CLIMATE STATISTICS 111remote, data-sparse areas. However, these data are alsovaluable in giving supplementary information in data-richareas on the spatial variability, in extending measuredseries in time and for investigating their temporal representativenessfor long-term conditions. A review ofapplications of satellite data is given in Barstow et al.(1994 (b)). As further years of data from GEOSAT andlater satellites, such as ERS-1 and Topex-Poseidon,become available, such analyses will provide invaluableinformation on wave climate. Papers on estimatingextreme values from altimeter data have been published byTournardre and Ezraty (1990), Carter (1993) and Barstow(1995). Figure 9.6 shows an estimate of wave height in anarea 46°–48°S, 166°–168°E based on three years ofGEOSAT data, with an estimated 50-year return value of16.5 m. The method used and other figures of the waveclimate around New Zealand are given in Carter et al.(1994). In addition, it is soon expected that SAR data willbecome more important.A list of available databases composed of visualobservations and measurements of waves is shown inTable 9.2. A detailed description of wave observationsand measurements is given in Chapter 8.While the data measured by buoys is considered to beof excellent quality, its utility for wave climatology isseverely limited. Wave data collection is expensive andlogistically difficult. As a result, there are often insufficientdata to adequately define the wave climatology over aregion, or for deriving with confidence any climate statisticwhich requires a long-time series (e.g. extremal analysis).Directional wave data from buoys are not yet common buthave increased considerably in the last decade. Buoy datacan be used for some operability studies, as long as dueconsideration is given to ensuring that the period of instrumentalrecord is representative of the longer term climate.Buoy data are usually inappropriate for extremal analysisdue to short record lengths. Buoy data are particularlyuseful for validating (or calibrating) numerical models andremote sensing algorithms.Shipborne Wave Recorders are more robust thanwave buoys and, fitted in Ocean Weather Ships andmoored Light Vessels since the 1960s, have providedseveral years of data at a few locations (see, for example,Figure 9.5). Shipborne Wave Recorder data, as well asvisual estimates, have been used to investigate an apparentupward trend in average wave heights in parts of theNorth Atlantic by one to two per cent per year over thelast few decades (Neu, 1984; Carter and Draper, 1988;Bacon and Carter, 1991). It is likely that long-term monitoringof wave climate trends can now be carried out bysatellite altimeters provided that the data are validatedand corrections applied to make data from differentsatellites comparable.Visual wave observations from ships have associatedproblems of quality and consistency (Laing, 1985) whichreduce their utility for climatology. In addition, the fact thatmost observations are from transient ships makes their usefor time-series type applications of climate virtually impossible.Ship data are most useful for climatological purposeswhen treated as an ensemble for regional descriptiveclimate applications such as “downtime” estimates foroffshore operations. Transient ship data are inappropriatefor extremal or time-series analysis.In practice, wave climate studies usually use datafrom a combination of sources for specifying offshoreconditions, and employ shallow water modelling if thesite of interest is near the coast.9.6.2 Wave hindcastsThere is no doubt that hindcasts are playing an increasingrole in marine climatology. Most recent wave climatologies,especially regional climatologies, are based onhindcast data. The same applies to design criteriaproduced by offshore oil and gas exploration and productioncompanies, and the regulatory authorities in manycountries around the world. The reason is simple: the costsinvolved in implementing a measuring programme, especiallyon a regional basis, and the period spent waiting fora reasonable amount of data to be collected, are unacceptable.Therefore, given the demonstrated ability of thepresent generation of spectral ocean wave models, thetimeliness and relatively lower cost of hindcast databecomes quite attractive. The 2nd (Canadian ClimateCenter, 1989) and 3rd (Canadian Climate Center, 1992 (a))International Workshops on Wave Hindcasting andForecasting identified no fewer than 12 separate recentlycompletedregional hindcasts around the world. A numberof other such studies are also under way.The quality of hindcast data (or at least confidencein it) can be improved by validating the results againstavailable in situ measurements or satellite altimeter H sdata (for instance, this is currently being carried out in aEuropean Union project, WERATLAS, in which a waveenergy resource atlas is being constructed based onECMWF’s WAM model archive and validated by buoydata and altimeter data, Pontes et al., 1995).Hindcasts can be classified into two basiccategories: continuous hindcasts and discrete storm hindcasts;a third possibility arises as a hybrid of the two.The features of each type are described briefly in thefollowing paragraphs.9.6.2.1 Continuous hindcastsContinuous hindcasts are usually the most useful form ofhindcast data, representing a long-term, uniform distributionin space and time of wind and wave information.The time spacing is typically six or 12 hours. Thedatabase is then suitable for all manner of statisticalanalysis, including frequency analysis and persistence, ata single location or over a region. If the period of thehindcast is sufficiently long, the database can also beused for extremal analysis to long-return periods. Themajor drawback to continuous hindcasts is the time andexpense involved in their production. The generation ofup to 20 years of data at six-hour intervals requires anenormous amount of effort, especially if it involves the

112GUIDE TO WAVE ANALYSIS AND FORECASTINGTABLE 9.2Available databases of long-term records of visual observations and measurementsCountry Source of observations Area covered Time spanned Parameters availableCANADA Visual observations from Voluntary Observing Global, but primarily Canadian Late 19th century to Wind speed and direction; wave and swell height, period,Ship scheme vessels over whole globe; OWS Papa waters present direction; air and sea temperature, humidity, visibility,observations; light vessel observations cloud amount and type, atmospheric pressureMoored buoys – Waverider, WAVEC, NOMAD, Canadian offshore areas Waverider: 1972 to present Wind speed and direction (5 m), (NOMAD and discus3 m discus (Atlantic, Pacific, Arctic) NOMAD, discus: 1987 to only); significant wave height, peak period, direction;present 1-D spectra; 2-D spectra (WAVEC only)HONG KONG Visual wave observations; visual and measured 0°–26°N, 100°–120°E 1971 to 1980 Wind speed and direction; wave height, period, direction;wind observations Marine climatological summary charts for the South ChinaSea 1971–80INDIA Visual wave observations; visual and measured Indian Ocean: north of 15°S, Wave height and period; Marine climatological summarieswind observations between 20° and 100°EIRELAND Visual observations from Voluntary Observing 48°–59°N, 3°–16°W 1854 to present Wind speed and direction; wave height, period, direction; airShip scheme vessels and sea temperature, humidity, visibility, cloud amount andtype, atmospheric pressureBeaufort scale wind estimates from light vessels 5 vessels in area bounded by Variable, from June 1939 Beaufort wind estimates twice daily51°30'–53°12'N, 5°30'–8°12'W to February 1982Observations from lighthouse stations 6 stations in area bounded by Variable, from September Wind speed and direction; sea state, weather, cloud cover,51°24'–53°18'N, 5°54'–9°54'W 1985 to present visibility, air pressure, air temperature at 0600, 0900, 1100,1400, 1700, 2300 hours civil timeWind measurements from anemometer at 65 m 51°22'N, 7°57'W 1 August 1979 to Wind speed and direction; air temperature, sea temperature;ASL; wave measurements from Baylor wave staff 2 February 1990 significant wave height, maximum wave height, wave periodtype D19595, air temperature from platinum and mean water levelresistance thermometer in Stevenson screen onhelideck of Marathon Gas PlatformISRAEL Visual observations 20 years Wind, wave, sea temperatureITALY Datawell Waverider buoys, recording 10 minutes 15 locations in area bounded by Variable, from 1974 All relevant sea state parametersevery 3 hours 36°54'–45°42'N, 8°30'–18°E to 1991JAPAN Visual wave observations, visual and measured North Pacific Ocean 1964 to 1973 Wind speed and direction; wave height, period and direction;wind observations Papers of Ship Research Institute, Supplement No. 3,March 1980

WAVE CLIMATE STATISTICS 113TABLE 9.2 (cont.)Country Source of observations Area covered Time spanned Parameters availableJAPAN (cont.) Visual wave observations, visual and measured North Pacific Ocean 1961 to present Wind speed and direction; wave height, period andwind observations direction; Marine climatological summariesNEW ZEALAND Waverider buoy, Maui ‘A’ 39°30'S, 173°20'E 1977 to 1988 Wave spectra (f)NORWAY Moored buoy data, including directional data ship- 15 buoy and ship locations Variable: late 1949 to Meteorological parameters (including wind), wavesborne recorder (OWS Mike, from 1949) vertical bounded by 56°30'–72°N, presentlasers, microwave radar coastal stations (visual) 2°–30°ESOUTH PACIFIC Waverider Rarotonga, Cook Is. July 87 – Jan. 91 )Waverider Tongatapu, Tonga Is. May 87 – July 92 )Directional Waverider June – Aug. 92 )Waverider Efate, Vanuatu Nov. 90 – Feb. 93 )Directional Waverider Nov. 92 – Feb. 93 ) Wave height, period and energyWaverider Funafuti, Tuvalu May 90 – April 92 ) Olsen et al. (1991)Waverider Western Samoa Sept. 89 – July 92 )Directional Waverider March – June 93 )Waverider Kadavu, Fiji Jan. 91 – Oct. 93 )Directional Waverider April – July 94 )UNITED Visual observations from Voluntary Observing Global Late 19th century to Wind speed and direction; wave and swell height, period,KINGDOM Ship scheme vessels over whole globe; present direction; air and sea temperature, humidity, visibility, cloudSBWR observations at standard OWS locations amount and type, atmospheric pressure; R. Shearman (1983)and light vessels around UKUSA Visual observations from Voluntary Observing Global 1854 to present Wind speed and direction; wave and swell height, period,Ship scheme vessels over whole globe; OWS direction; air and sea temperature, humidity, visibility, cloudobservations at standard locations amount and type, atmospheric pressure; Slutz et al. (1985)Moored buoys US coastline (Atlantic, Pacific, 1972 to present Wind speed and direction; wave height, period, and direction;Gulf of Mexico) 1-D and some 2-D wave spectra; Climate Summaries forNDBC Buoys and Stations, Update No. 1, 1990, NDBC,NOAA

114GUIDE TO WAVE ANALYSIS AND FORECASTINGquality control and re-analysis of historical informationnecessary to produce a high-quality database (particularlyof the winds needed to drive the hindcast wavemodel).One approach, which has been adopted in order tominimize the costs associated with continuous hindcasts,is to archive the analysis portion of operational waveanalysis and forecast programmes. This is a very costeffectivemeans of producing a continuous database,being a by-product of an existing operational programme.The disadvantages are that the operational timeconstraints mean that not all available data are includedin the analysis, that time-consuming techniques such askinematic analysis cannot be performed, that the use ofbackward as well as forward continuity in the developmentof weather patterns is not possible, and that it willtake N years of operation to produce an N-year database.There is also a danger that such archives may suggest“climate change” that is not real but is a result ofchanges in the characteristics of the models used.Nevertheless, this approach represents a viable way todevelop a continuous database of wave information,albeit of lesser quality. Attention also needs to be givento the temporal homogeneity of the model data, especiallywhen different procedures to estimate ocean windsare used at different times.9.6.2.2 Storm hindcastsIn order to perform the extremal analysis necessary forestablishing design criteria for offshore operations, it isnecessary to have wave information for a period of atleast 20 years. In fact, recent experience suggests thateven 20 years may not be sufficient to produce stableestimates of long return period wave heights. The costassociated with producing a continuous hindcast formore than 20 years is prohibitive in most instances. As aresult, an approach has been adopted in many countrieswhereby a selection of the top-ranked wave-producingstorms over a period of 30 or more years is hindcast,with the wave heights analysed using peak-overthresholdtechniques (see Section 9.4.1.2).This approach has several advantages, not least ofwhich is that the cost is a fraction of that for a continuoushindcast. Storms hindcast are typically of about fivedays duration, so for a sample of 50 storms in a period of30 years, for example, the total hindcast period would be250 days. Another major advantage is that the stormscan be hindcast in considerably more detail, with full reanalysisof each storm, including kinematic analysis, andforward and backward continuity. All available data canbe used in the analysis, including those data abstractedfrom ships’ logs as well as that originally obtained fromthe Global Telecommunication System (GTS). There isno doubt that this approach produces the highest qualitydata of any hindcast procedure for the storm periodsselected.On the negative side, the database produced, whilesuitable for extremal analysis, is inappropriate for anyanalysis which requires a time series of data at one ormore locations (such as persistence analysis). It alsocannot give any information on frequency distributionsof waves, since only the extreme conditions are analysed.One additional drawback is that one cannotremove lingering doubts that the storms selected for thedetailed hindcasts are the most severe wave-producingstorms, since these must usually be selected in theabsence of direct wave measurements, by proxy criteriasuch as the time history of pressure gradients over wavegeneratingareas.9.6.2.3 Hybrid hindcastsHybrid hindcasts are being used increasingly to try tocombine the best features of the continuous and stormhindcasts. They start from a continuous hindcast producedas described above, and are then augmented byhindcasts of the most severe storms and of periods wherethe verification against measurements shows that thecontinuous hindcast has produced significant errors. Theperiods from the storm hindcast then replace the archivein the continuous hindcast. This approach has been usedeffectively in the Gulf of Mexico and off the west coastof Canada. It is relatively cost effective, and continuedcorrection to the continuous hindcast can be madedepending on available resources. The databaseproduced is suitable for all types of statistical analysis,including extremal analysis if the hindcast period issufficiently long.9.6.3 Hindcast procedureThe following paragraphs describe a wave hindcaststorm procedure. Most aspects of the hindcast procedure,other than storm selection, are the same whether thecontinuous or storm approach is selected. If the hindcastis produced using the analysis cycle of an operationalwave forecast programme, most of the decisions onmodel domain, input wind fields, etc. will have beenmade; the only remaining decision will involve thearchiving process. A detailed description of the stepsinvolved in a storm hindcast are given in the WMOGuide to the applications of marine climatology (WMO,1994(b)).The application of the hindcast method includes thefollowing main steps:(1) Given the point or area where hindcast wave dataare required, decide the area necessary to beincluded to represent the wave conditions in theregion of interest (to catch distantly generatedswells, etc.).(2) Select the time span for the hindcast; previousexperience with the historical marine meteorologicaldatabases supports selection of storms fromabout the past 30 years. The database for earlierperiods is much less extensive and wind fields maynot be specified as accurately. Therefore, the historicalperiod which should generally consideredextends from about the mid-1950s to the present.

WAVE CLIMATE STATISTICS 115For a storm hindcast, select the most severe waveproducingstorms over the time span, using waveand wind data where available, and archived analysesof surface pressure.(3) Choose a suitable wave model, such as one of thoselisted in Table 6.2; the grid spacing and time stepshould be appropriate for the application.(4) Specify surface wind fields on discrete grids for thetime span selected (or for each selected historicalstorm).(5) Execute the numerical hindcast of the time historyof the sea state on a grid of points representing thebasin, for the time span (or for each storm).(6) Archive input wind fields and hindcast waveparameters and spectra at a large number of sitesfor each model time step.9.6.4 Wind field analysisThe most crucial point is the production of the windfields. Data are sparse over the oceans and it is difficultto find grid data of a sufficiently fine resolution fartherback than the early 1970s. An example of a proceduremay be to:(1) Procure all available surface data from ships, buoys,synoptic and climatic stations, drilling vessels,Ocean Weather Stations, satellite data and sea icecover data (where appropriate).(2) Digitize historical surface pressure analyses (afterre-analysis using the available data if necessary), oracquire digital gridded fields of sea-level pressureor wind speed and direction.(3) Apply a planetary boundary-layer model to theisobaric analysis to approximate the near-surfacewind field (e.g. Cardone, 1969).(4) Construct streamlines and isotachs using all synopticobservations of wind speed and direction fromships and land stations, using forward and backwardcontinuity which defines the movements ofstorm centres and fronts and other significantfeatures of the surface wind field.(5) Extract kinematic winds (speed and direction) fromthe streamline/isotach analyses on the defined grid;where kinematic analysis is performed over part ofthe grid only, the kinematic winds replace thewinds derived from the planetary boundary-layermodel, with blending along the boundaries of thekinematic area.The kinematic winds are by far the most accurateand least biased winds, primarily because themethod allows a thorough re-analysis of the evolutionof the wind fields. Kinematic analysis alsoallows the wind fields to represent effects not wellmodelled by pressure-wind transformation techniques,such as inertial accelerations associatedwith large spatial and temporal variations in surfacepressure gradients and deformation in surfacewinds near and downstream of coasts.(6) Define the resulting wind fields at a specific level.Some modellers have adopted the simple conceptof the “effective neutral” wind speed introduced byCardone (1969) to describe the effects of thermal stratificationin the marine boundary layer on wavegeneration. The effective neutral wind speed is simplythe wind which would produce the same surface stress atthe sea surface in a neutrally stratified boundary layer asthe wind speed in a boundary layer of a given stratification.Calculation of the effective wind at a referenceelevation from measured or modelled winds and air-seatemperature differences requires a model of the marinesurface boundary wind profile which incorporates astability dependence and a surface roughness law.A boundary-layer model is set up to provide theeffective neutral wind speed, for example at 10 m.Reports of wind speed from ships and rigs equipped withanemometers are then transformed into the effectiveneutral 10 m values (see Dobson, 1982; Shearman andZelenko, 1989), using a file of anemometer heights ofships in the merchant fleet. For ships which use estimatedwind speeds, values are adjusted according to thescientific Beaufort scale (see Chapter 2, Table 2.2). Arevised table of wind speed equivalents is used toretrieve the 10 m wind speed and then correct forstability.If winds must be interpolated in time for input tothe wave model, the recommended algorithm is linearinterpolation in time of zonal and meridional windcomponents to compute wind direction, and interpolationof the fourth power of wind speed, because waveenergy scales with this quantity. This scheme has beenfound to provide sufficient resolution in wind fieldsencompassing extra-tropical cyclones off the east coastof North America.9.6.5 Archiving of wind and wave fieldsThe fields archived depend on the user’s needs andshould be as comprehensive as possible. For example,for a full spectral wave model it would be useful toarchive the following hindcast gridded wind and wavefields at each model grid point:• Wind speed in m/s (e.g. effective neutral 10 mwinds);• Wind direction in degrees (meteorologicalnotation);• Wind stress or friction velocity (u * );• Significant wave height (H s ) in metres and tenths ofmetres;• Peak period (T p ) or significant period in secondsand tenths of seconds;• Vector mean direction, or spectral peak direction indegrees;• Directional (2-D) spectral variance in m 2 .9.6.6 Verification of wave hindcastsThe validation of hindcasts against measurements usingcomparisons reveal the skill achievable in the hindcasts.Time series plots, error statistics and correlation should

116GUIDE TO WAVE ANALYSIS AND FORECASTINGTABLE 9.3Hindcast wave databasesCountry Model used Grid and time step Verification Area covered Time spanned Parameters availableARGENTINA DSA V deep water Orthodromic grid, spacing Within box defined by 1976 to present Significant wave height, peak period, peakdecoupled model; 90 n.mi. 28°S, 48°W; 27°S, 26°W; direction; Rivero et al. (1978)Rivero et al. (1974) Time step: 3 hours 57°S, 70°W; 55°S, 15°WCANADA ODGP 1-G deep water Lat./long. nested grid; Unbiased for wind East and west coasts of 1957–1990; 70 Wind speed and direction (20 m), significantspectral wave model; spacing 1.25° x 2.5° in speed and significant Canada discrete storm periods wave height, peak period, vector mean15 frequencies by 24 coarse mesh, 0.625° x 1.25° wave height; scatter on east coast, 51 on direction, at all grid points every 2 h; 2-Ddirections; Cardone et in fine mesh near coasts index 0.14 west coast spectra every 6 h at 118 points; Canadianal. (1976) Time step: 2 hours Climate Center (1991, 1992 (b))ODGP 1-G deep water Wind input on lat./long. grid; Unbiased for wind 69°–76°N; 120°–162°W 1970–1990; 45 Wind speed and direction (20 m), significantspectral wave model; spacing 1° x 3°; waves speed and significant (Canadian Beaufort Sea) discrete storm periods wave height, peak period, vector mean15 frequencies by 24 on transverse Mercator, wave height; scatter direction, at all grid points every hour; 2-Ddirections; Cardone et spacing 37.3 km index 0.13 spectra every 6 h at 51 points; MacLarenal. (1976) Time step: 1 hour Plansearch Ltd. and Oceanweather, Inc. (1992)ODGP 1-G deep water Lat./long. nested grid; Unbiased for wind North Atlantic and North 1 October 1983 – Wind speed and direction (20 m), significantspectral wave model; spacing 1.25 x 2.5 in speed and significant Pacific Oceans 30 September 1986 wave height, peak period, vector mean15 frequencies by 24 coarse mesh, 0.625 x 1.25 in wave height; scatter for North Atlantic; direction, at all grid points every 2 h; 2-Ddirections; Cardone et fine mesh near coasts index 0.30 for 1 January 1987 – spectra every 6 hours at 54 points; Eid et al.al. (1976) Time step: 2 hours Atlantic 31 December 1989 (1989)for North PacificGREECE Deep water model, Polar stereographic grid, Eastern Mediterranean No parameters in database at presentdecoupled propagation; spacing 100 km SeaHellenic National Time step: 6 hoursMeteorological ServiceReports No. 14, 15IRELAND Irish Meteorological 23 x 23 grid, spacing 150 km 7 locations between 48°– September 1984 to Significant wave height, period and directionService model, adapted Time step: 3 hours 55°N and 8°–14°W presentNOWAMO; sea modelplus swell modelJAPAN MRI-II coupled Grid spacing 300 km North Pacific Ocean 1980–1989 Wave height, period and directiondiscrete wave model; Time step: 6 hours Ship and Ocean Foundation (1990, 1991)Uji (1984) wind input 20 m wind derivedfrom 1 000 hPa winds fromECMWF global NWP model

WAVE CLIMATE STATISTICS 117TABLE 9.3 (cont.)Country Model used Grid and time step Verification Area covered Time spanned Parameters availableNEW NIWA 2-G coupled Polar stereographic grid; Scatter index for H s Seas within 800 km of 1980–1989 Wind speed and direction, full spectra at allZEALAND spectral model; spacing 190 km at 60°S 0.29 compared to New Zealand (in progress) grid points in sub-grid around New Zealand15 frequencies by Time step: 2 hours GEOSAT, 0.24 com-18 directions pared to WaveriderNORWAY WINCH wave model Grid spacing 75 km North Sea, Norwegian 1955–1994 Wind speed and direction, significant waveTime step: 2 hours Sea, Barents Sea height, peak period, peak direction every 6 hUNITED Deep water, 2-G wave Lat./long. grid, Global 1 October 1986 – Wind speed and direction, significant waveKINGDOM model; Golding (1983) spacing 1.5° x 1.875° 11 June 1991; new height, 1-D energy spectra, mean waveTime step: 1 hour archive started 12 June direction for each frequency.development, variable 1991 with higheradvection time step resolutionShallow water, 2-G Lat./long. grid, 30°30'–66°45'N; 1 October 1986 – Wind speed and direction, significant wavewave model; Golding spacing 0.25° x 0.4° 14°4'W–35°56'E 11 June 1991; new height, 1-D energy spectra, mean wave(1983) Time step: 1 hour archive started 12 June direction for each frequency, depthdevelopment, variable 1991 with higheradvection time step resolutionUSA US Navy Spectral Ocean Northern hemisphere 1956–1975: Atlantic Wind speed and direction, 2-D wave spectra;Wave Model (SOWM); 1964–1977: Pacific US Navy Hindcast Spectral Ocean Wave15 frequencies by 24 1975–1985: Medi- Model climatic atlas (1983, 1985, 1990)directions; Pierson (1982) terranean SeaUS Navy Global Grid spacing 2.5° Global 1985–1994 Wind speed and direction, 2-D wave spectra;Spectral Ocean Wave NCDC Tape Deck 9782Model (GSOWM);15 frequencies by24 directions; NCDCTape Reference ManualDeep plus shallow water US coastline (Atlantic, 1956–1975 Surface pressure fields, wind speed and2-G wave model; Pacific, Gulf of Mexico) direction, 2-D wave spectra; Wave informationBrooks and Corson studies of the US coastline, Waterways(1984) Experiment Station, Vicksburg, MS, USA

118GUIDE TO WAVE ANALYSIS AND FORECASTINGbe computed; the bias, RMS (root-mean-square) errorand scatter index (ratio of RMS error to mean measuredvalue) are particularly useful as measures of performance.Continuous hindcasts typically show scatterindices for significant wave height in the range of0.25–0.30, while for peak-to-peak comparisons of stormhindcasts they are often in the range of 0.12–0.16.In addition, it is useful to construct 1-D spectralplots of the respective observed and modelled spectra,e.g. at peak wave height or within three hours. Forcontinuous wave measurements, an appropriate movingaverage should be used on the recorded data (e.g. 6 or 7point moving average). It is instructive to represent thedata with error bars indicating appropriate confidencelimits (for instance 95 per cent). See, for example,Figure 5.4, where the 90 per cent confidence limits areshown.If 2-D spectral measurements are available, theyshould be used to evaluate the model predicted values.9.6.7 Extremal analysisFrom the wave hindcast model, the following quantitiesare available at all points and at each time step:• H s — significant wave height;• T p — spectral peak period;• θ d — vector mean wave direction;• W s — average wind speed;• W d — wind direction.The objective of the extreme analysis is to describeextremes at all contiguous grid locations for the followingvariables:• H s versus annual exceedance probability or inversereturn period;• W s versus annual exceedance probability;• H m (maximum individual wave height) versusannual exceedance probability.Techniques for extremal analysis of these quantities aredescribed in preceding sections.9.6.7.1 Extreme wave/crest height distributionFor the points at which a detailed extremal analysis isperformed, the maximum individual wave height maybe estimated in each storm from the hindcast zeroth andfirst spectral moments following Borgman’s (1973)integral expression, which accounts for storm build-upand decay. The integral may be computed fortwo assumed maximum individual wave heightdistributions:(a)Rayleigh (as adapted by Cartwright and Longuet-Higgins, 1956);(b) Forristall (1978).The same approach may be used to estimate themaximum crest height at a site in a storm using theempirical crest-height distribution of Haring andHeideman (1978). The median of the resulting distributionsof H m , H c may be taken as the characteristicmaximum single values in a storm. The mean ratios ofH m /H s and H c /H s should be calculated and used todevelop a mean ratio to provide extremes of H m and H cfrom fields of extreme H s .9.6.7.2 Presentation of extremesFields of extremes of H s , H m , W s should be tabulated anddisplayed as field plots (contour plots if necessary) ofnumerical values.Results of detailed extremal analysis at selectedgrid locations should be presented in tabular form foreach analysed point and in graphical form. The graphicaldisplay of extrapolations shall include the fitted line, theconfidence limits on the fit and the fitted points.9.6.8 Available hindcast databasesSeveral hindcast databases have been created covering awide range of ocean basins. Most of these are continuoushindcasts, a few are storm hindcasts. There are nohybrid hindcast databases presently available. A list ofavailable hindcast databases, with their characteristics, isshown in Table 9.3. These databases cover periods ofthree years or longer (continuous) or 40 storms or more(storm). It is not possible to include the many other hindcasts,related to short periods for verification purposes orto studies of particular events, in this publication.

ANNEX IABBREVIATIONS AND KEY TO SYMBOLS(Preferable units are shown with alternatives in brackets)Abbreviation/SymbolDefinitionaWave amplitude — m (ft)ASCEAmerican Society of Civil Engineers∆bDistance between two wave raysCERCCoastal Engineering Research Center (US)c; c phase Phase speed; the speed of propagation of a wave — m/s (kn)c g ; c gGroup speed, velocity. The speed (velocity) of propagation of a wave group — m/s (kn)–c g Effective group velocity for total energy — m/s (kn)c θSpeed of directional change — rad/s (deg/s)C d ; C 10Drag coefficient; drag coefficient at the 10 m level.CDCoupled discrete — model classCHCoupled hybrid — model classCMCCanadian Meteorological CenterdegMeasure of angle, degreeD pDuration of wave generation — hDNMIThe Norwegian Meteorological InstituteDPDecoupled propagation — model classe, exp Exponential constant ≅ 2.71828E, E total Wave energy (variance) per unit area — m 2 per areaE ∞Energy of fully developed wave field — m 2 per areaE(f)Wave spectral energy density as a function of frequency — m 2 /HzE(f,θ)Wave spectral energy density as function of frequency and direction — m 2 /Hz / radE[ ] Expected value of variable in bracketsECMWF European Centre for Medium-range Weather ForecastsESAEuropean Space AgencyfWave frequency — HzCoriolis parameter — s–fMean wave frequency — Hzf pWave frequency corresponding to the peak of the spectrum — HzftUnit of length, footFFTFast-Fourier transformF(H)Probability of not exceeding the wave height HFNMOC Fleet Numerical Meteorology and Oceanography Centre (USA)FTFisher-Tippett distributiong Acceleration due to gravity — m/s 2G, (u g , v g ); G Geostrophic wind velocity; speed — m/s (kn)GDGroen and DorresteinGr; GrGradient wind velocity; speed — m/s (kn)GTSGlobal Telecommunications SystemhUnit of time, hourhWater depth — m

120ANNEX IAbbreviation/SymbolDefinitionh bBreaking depth; water depth at which a wave of given height breaks — mhPahectoPascal — unit of pressureHHigh pressure centreHWave height — m (ft)–HAverage wave height in a record — m (ft)H bBreaker wave height — m (ft)H cCharacteristic wave height (average height of the larger well-formed waves, observedvisually) — m (ft)H combined Wave height for total wave field — m (ft)H max , H m , H max,3h Maximum wave height for specified period of time — m (ft)H m0 Wave height derived from energy spectrum, approximately equal to H – 1/3 ; H m0 = 4 m 1/2 0 —m (ft)H rmsRoot-mean-square wave height = (8 E/ρ w g) 1/2 — m (ft)H sSignificant wave height = H m0 or H – 1/3 , according to specification — m (ft)H sN , H s50 N-year, 50-year return value of H s — m (ft)H seaWind wave height — m (ft)H swellSwell height — m (ft)H zZero-crossing wave height (the vertical distance between the highest and lowest value of thewave record between two zero-downcrossings or upcrossings)–H 1/nThe average height of the 1/n highest waves — m (ft)–H 1/3Significant wave height: average of the 1/3 highest waves in a record (approximates to thecharacteristic wave height) — m (ft)Hz Hertz, measure of frequency = 1 cycle per second — s –1HFHigh FrequencyHMSOHer Majesty's Stationary Office (UK)I, (u i , v i ) Isallobaric wind vectorIAHRInternational Association for Hydraulic Research (Delft, Netherlands)JWave power — kW/m (of wavefront)JMAJapan Meteorological AgencyJONSWAP Joint North Sea Wave Proj ectk; k Wavenumber (2π/λ); vector wavenumber — m –1kmUnit of length, kilometreknUnit of speed, knot (1.94 kn = 1 m/s)kWUnit of power, kilowattKMixing coefficientK rRefraction factorK sShoaling factorKNMIRoyal Netherlands Meteorological InstituteLLow pressure centreLMonin-Obukhov mixing length — mmUnit of length, metremksMetric system of units based on the metre, kilogram and second as the units of length, massand time; it forms the basis of the International System of Units (SI)m nThe n-th order moment of the wave spectrumm 0Moment of zero order; represents the integral of a wave spectrumm/sUnit of speed, metre per secondMEMean error

ANNEX I 121Abbreviation/SymbolDefinitionMOSnn(f,θ)nn.mi.NNDBCNMCNMSNOAApPMPNJQ(H)Q pQ bModel output statisticsA given number of waves, years, etc.Wave action densitySpatial variable normal to direction of travelUnit of distance, nautical mile (one second of arc of latitude)A given number (of waves, years, etc.)National Data Buoy Center (USA)National Meteorological CentreNational Meteorological ServiceNational Oceanic and Atmospheric Administration (USA)Atmospheric pressure — hPaPierson-MoskovitzPierson, Neumann and James (wave forecasting method)Probability of exceeding the wave height HSpectral peakedness parameterFraction of breaking wavesr Sample correlation (see also ρ)radR pRMSRMSEsssS(f); S(ω)SS inS dsS nlS bottomS breakingSARSBWRSISSM/ISWAMPtTT cT – H1/nT – H1/3T m01T m02Radius of curvature of an isobarUnit of plane angle, radian (1 radian = 57.296 deg)Range or distance to point "p" from storm front — m, km (n.mi.)Root-mean-squareRoot-mean-square errorUnit of time, secondSpatial variable in direction of motionStandard deviation of a distributionWave-spectral (variance) density as function of frequency; as a function of angularfrequency S(ω) = S(f)/(2π)Energy-balance equation: combined source termsEnergy-balance equation: input source termEnergy-balance equation: dissipation source termEnergy-balance equation: non-linear interaction source termEnergy-balance equation: bottom friction termEnergy-balance equation: wave breaking in surf zoneSynthetic Aperture RadarShipborne Wave RecorderScatter index (ratio of rmse to mean)Special Sensor Microwave/ImagerSea Wave Modelling ProjectTime variable — s, hWave period — sTemperatureCharacteristic wave period — s (average period of the larger well-formed waves, observedvisually)Average period of the 1/n highest waves — sSignificant wave period: average period of 1/3 highest waves — sA mean period corresponding to inverse of mean frequency = m 0 /m 1 , — sA wave period from spectral moments = (m 0 /m 2 ) 1/2 ; approximates T – z — s

122ANNEX IAbbreviation/SymbolDefinitionT m–10T pT sEnergy wave period = m –1 /m 0 , — sPeak period: wave period at the peak of the spectrum (modal period) — sSignificant or peak period according to specificationT – zAverage zero-crossing wave period in a recordu; u Wind speed; wind velocity — m/s, (kn)U 10Wind speed at the 10 m level — m/s, (kn)Wind speed level z — m/s, (kn)U zu *; u *Friction speed / velocity = (τ/ρ a ) 1/2(u g , v g ), G; G Geostrophic wind velocity; speed — m/s, (kn)(u i , v i ), I Isallobaric wind velocityUKMOUnited Kingdom Meteorological OfficeUTCUniversal Time CoordinatesVOSVoluntary Observing ShipsWAMWAve Model (a specific third generation wave model)WAVEOB Code for interchange of wave observationsXFetch length — km (n.mi.)x; x Space variable; space vectorySpace variablezHeight variablez 0Roughness lengthαScaling (Phillips’) parameter in model spectrum (= 0.0081 in Pierson-Moskowitz spectrum)Angle between wave front (crest) and isobathCharnock’s constantγPeak-enhancement parameter in JONSWAP spectrumεSpectral width parameterDimensionless wave energyη(x,t)Elevation of water surface at position x and time t– η Mean water surface elevationθDirection of wave propagationθ dVector mean wave directionκ von Karman constant (≅ 0 41)λ, λ s Wavelength, wavelength of sea — mλ rRadar wavelengthνNon-dimensional frequencyπ A constant; ≅ 3.14159ρCorrelation coeffficient (r = sample correlation coefficient)ρ; ρ w ; ρ a Density; density of water; density of airσWave frequency associated with a (current modified) wavenumberξWave steepness (various definitions, see text)ωAngular frequency = 2πfτMean stress magnitudeφPhase of a sinusoidal waveψStability function for atmospheric boundary layerWind direction

ANNEX IIFM-65 IX WAVEOBREPORT OF SPECTRAL WAVE INFORMATION FROM A SEA STATION ORFROM A REMOTE PLATFORM (AIRCRAFT OR SATELLITE)CODE FORM:D . . . . DorIIiii*SECTION 0 M i M i M j M j A 1 b w n b n b n b ** YYMMJ GGgg/ oror Q c L a L a L a L a L o L o L o L o L o **I 1 I 2 I 2 //00I a I m I p 1hhhh 2H s H s H s H s 3P p P p P p P p (4H m H m H m H m ) (5P a P a P a P a )(6H se H se H se H se ) (7P sp P sp P sp P sp ) (8P sa P sa P sa P sa ) (9d d d d d s d s )SECTION 1 (111B T B T SSSS/ D´D´D´D´/ BB/// 1f 1 f 1 f 1 x 1f d f d f d x . . . . .BB/// nf n f n f n x nf d f d f d x)SECTION 2 (2222x C m C m C m n m n m 1c 1 c 1 c 2 c 2 3c 3 c 3 c 4 c 4 . . . . . n–1c n–1 c n–1 c n c n(or nc n c n //))SECTION 3 (3333x C sm C sm C sm n sm n sm 1c s1 c s1 c s2 c s2 3c s3 c s3 c s4 c s4 . . . . .n–1c sn–1 c sn–1 c sn c sn (or nc sn c sn //))SECTION 4 (4444 1d a1 d a1 d a2 d a2 1r 1 r 1 r 2 r 2 2d a1 d a1 d a2 d a2 2r 1 r 1 r 2 r 2 . . . . .nd a1 d a1 d a2 d a2 nr 1 r 1 r 2 r 2 )SECTION 5 (5555I b 1A 1 A 1 A 1 x (1d 1 d 1 d s d s ) 2A 2 A 2 A 2 x (2d 2 d 2 d s d s ) . . . . .nA n A n A n x (nd n d n d s d s ))NOTES:(1) WAVEOB is the name of the code for reporting spectral wave data from a sea station, or from an aircraft orsatellite platform.(2) A WAVEOB report is identified by M i M i M j M j = MMXX.(3) The code form is divided into six sections (Sections 1 to 5 are optional):SectionSymbolic figure groupContentsnumber0 — Data for reporting identification (type, buoy identifier, date, time,location), indication of frequency or wave number, method ofcalculation, type of station, water depth, significant wave heightand spectral peak period, or wave length, and optional waveparameters1 111 Sampling interval and duration (or length) of record, and description ofmeasurement system bands––––––––––* Included in a fixed sea station report only.** Included in a sea station or remote platform report only.

124GUIDE TO WAVE ANALYSIS AND FORECASTINGSectionnumberSymbolic figure groupContents2 2222 Maximum non-directional spectral density from heave sensor, and ratiosof individual spectral densities to the maximum value3 3333 Maximum non-directional spectral density from slope sensor, and ratiosof individual spectral densities to the maximum value4 4444 Directional wave functions. Mean and principle wave directions and firstand second normalized polar Fourier coefficients, for bands describedin Section 15 5555 Directional or non-directional spectral estimates by frequency or wavenumber, as indicated, and direction with directional spreadREGULATIONS:65.1 General65.1.1 The code name WAVEOB shall not be included in the report.D . . . . D65.1.2 Use of groups or IIiii*M i M i M j M j A 1 b w n b n b n b ** YYMMJ GGgg/ oror Q c L a L a L a L a L o L o L o L o L o **I 1 I 2 I 2 //NOTE: See Regulation 18.2.3, Notes (1), (2) and (3)***.65.1.2.1 Each individual WAVEOB report, whether or not included in a bulletin of such reports, shallcontain as the first group the identification group M i M i M j M j .65.1.2.2 A sea station shall be indicated by either the group D . . . . D or A 1 b w n b n b n b . The position of asea station shall be indicated by the groups Q c L a L a L a L a L o L o L o L o L o . A satellite shall beindicated by the group I 1 I 2 I 2 // and an aircraft shall report ///// for I 1 I 2 I 2 //. A fixed seastation (other than an ocean weather station and a moored buoy), which is considered by theMember operating it to be in the same category as a land station, shall report its identificationand position by means of the group IIiii.NOTE:Data may be transmitted from a sea station or from a remote platform (aircraft or satellite).65.1.2.3 In a report from a sea station (including an ocean weather station and a moored buoy), thelatitude and longitude shall be encoded with the actual location of the station. In a satellite oraircraft report, the latitude and longitude shall indicate the (approximate) centre of the areaobserved.65.1.3 Use of Sections 0 and 165.1.3.1 The first three data groups in Section 0, after the location, shall contain indicators showing ifdata are expressed as frequency or wave number, the method of calculation of data and typeof platform, data on the water depth in metres, significant wave height in centimetres (or tenthsof a metre) and spectral peak period in tenths of a second or spectral peak wave length inmetres. Optional groups, when included, shall contain data on the maximum wave height,average wave period or average wave length, estimate of significant wave height from slope––––––––––* Included in a fixed sea station report only.** Included in a sea station or remote platform report only.*** Notes to Regulation 18.2.3:(1) A 1 b w normally corresponds to the maritime zone in which the buoy was deployed. The WMO Secretariat allocates toMembers, who request and indicate the maritime zone(s) of interest, a block or blocks of serial numbers (n b n b n b ) to beused by their environmental buoy stations.(2) The Member concerned registers with the WMO Secretariat the serial numbers actually assigned to individual stationstogether with their geographical positions of deployment.(3) The Secretariat informs all concerned of the allocation of serial numbers and registrations made by individual Members.

ANNEX II 125sensors, spectral peak wave period or peak wave length derived from slope sensors, averagewave period or average wave length derived from slope sensors, and dominant wave directionand directional spread.65.1.3.2 When used, Section 1 shall contain the section identifier, the total number of bands described inthe section, the sampling interval (in tenths of a second or in metres), the duration in seconds ofrecord of the wave or the length in tens of metres, the number (BB) of bands described in thenext two groups, the first centre frequency (Hz) or first centre wave number (metres) –1 , and theincrement added to obtain the next centre frequency (Hz) or the next centre wave number(metres) –1 and their associated exponents.NOTE: In deriving the value of the first centre frequency or wave number and increment from the groupsnf n f n f n x nf d f d f d x, decimal points are assumed at the left of the numeric values. For example, for centrefrequency, the groups 13004 11004 would be interpreted as a first centre frequency of0.300 x 10 –1 Hz and an increment of 0.100 x 10 –1 Hz. (The maximum spectral density value C m C m C m inSection 2, or C sm C sm C sm in Section 3, is coded in a similar fashion.)65.1.3.3 Except when BB = 00, the two groups for the first centre frequency or first centre wave number,and the increment added to obtain the next centre frequency or the next centre wave number(each time preceded by BB) shall be repeated (n) times as required to describe band distribution.NOTE: If sets of data groups are greater than 9, the group identifier (n) for the tenth set will be 0, the groupidentifier for the eleventh will be 1, etc.65.1.3.4 BB shall be encoded BB = 00 when no increments are given and the following (n) groups areactual centre frequencies or actual centre wave numbers.NOTE: The note under Regulation 65.1.3.3 applies if data groups are greater than 9.65.1.4 Use of Sections 2 and 365.1.4.1 When used, Section 2 shall contain the section identifier, an exponent associated with thefirst data group on the maximum value for non-directional spectra (C m C m C m ) in m 2 Hz –1 forfrequencies or m 3 for wave numbers from wave heave sensors, given as a 3-digit number.The band number (n m n m ) in which the maximum value for non-directional spectra occursshall be included in the same group as the value. Subsequent groups shall contain ratiosof individual spectra to the maximum (c 1 c 1 to c n c n ) as a percentage (00–99), with 00meaning either zero or 100 per cent.NOTES:(1) See note under Regulation 65.1.3.2.(2) Confusion between a zero ratio and the maximum ratio (100 per cent) should not arise since the bandnumber (n m n m ) for the maximum has already been identified.65.1.4.2 Each group containing ratios shall begin with an odd number representing the unit value of thefirst band in the group. Thus, the number 1 shall identify values for the first andsecond or eleventh and twelfth or twenty-first and twenty-second, etc., bands. The last groupshall contain two ratios for even numbers of bands and one ratio for odd numbers of bands. Inthe case of odd numbers of bands, the last two characters in the group shall be encoded as //.65.1.4.3 When used, Section 3 shall contain the section identifier, and non-directional spectral dataderived from wave slope sensors, analogous to Section 2. Regulations 65.1.4.1, with the exceptionof the section identifier, and 65.1.4.2 shall apply.65.1.5 Use of Section 4When used, Section 4 shall contain the section identifier and pairs of data groups of meandirection and principal direction from which waves are coming for the band indicated,relative to true north, in units of 4 degrees, and the first and second normalized polar coordinatesderived from Fourier coefficients. The pairs of groups shall be repeated (n) times asrequired to describe the total number of bands given in Section 1.NOTES:(1) The note under Regulation 65.1.3.3 applies if pairs of data groups are greater than 9.(2) The mean direction and principal direction from which waves are coming will range from 00(actual value 358° to less than 2°) to 89 (actual value from 354° to less than 358°). A value of 99 indicatesthe energy for the band is below a given threshold.(3) Placing d a1 d a1 and d a2 d a2 for each band in the same group, with r 1 r 1 and r 2 r 2 for the same band inthe next group, allows a quick visual check of the state of the sea.

126GUIDE TO WAVE ANALYSIS AND FORECASTING(4) If d a1 d a1 ≈ d a2 d a2 and r 1 r 1 > r 2 r 2 , there is a single wave train in the direction given by the commonvalue of d a1 d a1 and d a2 d a2 .(5) If the coded value of | d a1 d a1 – d a2 d a2 | > 2 and r 1 r 1 < r 2 r 2 , a confused sea exists and nosimple assumption can be made about the direction of the wave energy.65.1.6 Use of Section 5When used, this section shall contain the section identifier, an indicator (I b ) indicating whetherthe section includes directional or non-directional data, pairs of data groups of spectralestimates of the first to the n th frequencies or wave numbers and the direction from which wavesare coming in units of 4 degrees for spectral estimates (1) to (n) and their directional spread inwhole degrees.NOTES:(1) When non-directional spectra are transmitted, the group containing direction and directional spreadmay be omitted.(2) Complete directional spectra may be coded by repeating as many duplets as needed to define the entirespectrum. A partial directional spectrum may be coded by selecting the largest spectral estimate for anyone frequency or wave number band over all directions and coding this for each frequency or wavenumber band. Secondary peaks may not be coded unless the full directional spectrum is transmitted.(3) For non-directional frequency spectra, the spectral estimates are in m 2 Hz –1 . For non-directional wavenumber spectra, the spectral estimates are in m 3 . For a complete directional frequency spectrum,spectral estimates are in m 2 Hz –1 radian –1 . For a complete directional wave number spectrum, thespectral estimates are in m 4 . For incomplete directional spectra, whether in frequency or wave number,the units of the spectral estimates should be m 2 Hz –1 or m 3 . That is, the total integrated energy withina frequency band is given rather than just that of the peak. If the spectral estimate is less than 0.100 x10 –5 , the value of 0 must be used. The exception to this occurs when all subsequent estimates at higherfrequencies are also 0, in which case only the zero immediately after the last non-zero spectral estimateneed be included; all others need not be coded.(4) There may be cases when spectral estimates are given in integrated units, such as m 2 , and it is necessaryto convert these to the units of the code. This is done by calculating the bandwidth at a frequencyby determining the frequency difference between midpoints on either side of the frequency in question.The integrated spectral estimate is then divided by this computed bandwidth.SPECIFICATIONS OF SYMBOLIC LETTERS USED IN WAVEOB (and not already defined)A 1WMO Regional Association area in which buoy, drilling rig or oil- or gas-production platform hasbeen deployed (1 – Region I; 2 – Region II, etc.). (Code table 0161)(FM 13, FM 18, FM 22, FM 63, FM 64, FM 65)A 1 A 1 A 1A 2 A 2 A 2 Spectral estimates of the first to n th frequencies (or wave numbers if so indicated).. . . (FM 65)A n A n A n(1) The use of frequency or wave number is indicated by symbolic letter I a .BBB T B TNumber of bands described by the next two groups, except that BB = 00 indicates each of thefollowing groups represents only a centre frequency or wave number.(FM 65)Total number of bands described.(FM 65)b w Sub-area belonging to the area indicated by A 1 . (Code table 0161)(FM 13, FM 18, FM 22, FM 63, FM 64, FM 65)C m C m C mC sm C sm C smMaximum non-directional spectral density derived from heave sensors, in m 2 Hz –1 forfrequencies and m 3 for wave numbers.(FM 65)Maximum non-directional spectral density derived from slope sensors, in m 2 Hz –1 forfrequencies and m 3 for wave numbers.(FM 65)

ANNEX II 127c s1 c s1 The ratio of the spectral density derived from slope sensors for a given band, to thec s2 c s2 maximum spectral density given by Csm C sm C sm .. . .(FM 65)c sn c sn(1) A coded value of 00 may indicate either zero, or that the band contains the maximum spectraldensity. Since the band containing the maximum value will have been identified, it will beobvious which meaning should be assigned.c 1 c 1The ratio of the spectral density derived from heave sensors for a given band, to thec 2 c 2 maximum spectral density given by Cm C m C m .. . .(FM 65)c n c n(1) See Note (1) under c s1 c s1 , c s2 c s2 , . . . c sn c sn .D´D´D´D´D . . . . Dd a1 d a1d a2 d a2Duration of record of wave, in seconds, or length of record of wave, in tens of metres.(FM 65)(1) The use of frequency or wave number is indicated by symbolic letter I a .Ship’s call sign consisting of three or more alphanumeric characters.(FM 13, FM 20, FM 33, FM 36, FM 62, FM 63, FM 64, FM 65, FM 85)Mean direction, in units of 4 degrees, from which waves are coming for the band indicated, relativeto true north. (Code table 0880)(FM 65)(1) A value of 99 indicates the energy for that band is below a given threshold.Principal direction, in units of 4 degrees, from which waves are coming for the bandindicated, relative to true north. (Code table 0880)(FM 65)(1) See Note (1) under d a1 d a1 .d d d d True direction, in units of 4 degrees, from which dominant wave is coming. (Code table 0880)(FM 65)d s d sDirectional spread, in whole degrees, of the dominant wave.(FM 65)(1) The value of the directional spread is normally less than one radian (about 57°).d 1 d 1d 2 d 2 True direction, in units of 4 degrees, from which waves are coming. (Code table 0880). . . (FM 65)d n d nf d f d f dThe increment to be added to the previous centre frequency or previous centre wave number, toobtain the next centre frequency (Hz) or the next centre wave number (m –1 ), in the series, theexponent being given by symbolic letter x.(FM 65)f 1 f 1 f 1The first centre frequency (Hz) in a series, or the first centre wave number (m –1), the exponentf 2 f 2 f 2 being given by symbolic letter x.. . .(FM 65)f n f n f nGGggH m H m H m H mTime of observation, in hours and minutes UTC.(FM 12, FM 13, FM 14, FM 15, FM 16, FM 18, FM 22, FM 35, FM 36, FM 37, FM 38, FM 42,FM 62, FM 63, FM 64, FM 65, FM 67, FM 88)Maximum wave height, in centimetres.(FM 65)(1) In the event wave height can only be reported in tenths of a metre, the final digit in thegroup shall be encoded as /.

128GUIDE TO WAVE ANALYSIS AND FORECASTINGH s H s H s H sH se H se H se H sehhhhSignificant wave height, in centimetres.(FM 65)(1) See Note (1) under H m H m H m H m .Estimate of significant wave height from slope sensors, in centimetres.(FM 65)(1) See Note (1) under H m H m H m H m .Water depth, in metres.(FM 65)I a Indicator for frequency or wave number. (Code table 1731)(FM 65)I b Indicator for directional or non-directional spectral wave data. (Code table 1732)(FM 65)I m Indicator for method of calculation of spectral data. (Code table 1744)(FM 65)I p Indicator for type of platform. (Code table 1747)(FM 65)I 1 Name of country or international agency which operates the satellite. (Code table 1761)(FM 65, FM 86, FM 87, FM 88)IIBlock number.(FM 12, FM 20, FM 22, FM 32, FM 35, FM 39, FM 57, FM 65, FM 71, FM 75, FM 81, FM 83,FM 85)(1) The block numbers define the area in which the reporting station is situated. They are allocated toone country or a part of it or more countries in the same Region. The list of block numbers for allcountries is given in Volume A of publication WMO–No. 9.I 2 I 2 Indicator figure for satellite name (supplied by operator I 1 ).(FM 65, FM 86, FM 87, FM 88)(1) Even deciles for geostationary satellites.(2) Odd deciles for polar-orbiting satellites.iiiStation number.(FM 12, FM 20, FM 22, FM 32, FM 35, FM 39, FM 57, FM 65, FM 71, FM 75, FM 81, FM 83,FM 85)(1) See Section D of this volume. (Editorial note: WMO-No. 306, Volume I.1, Part A)J Units digit of the year (UTC), i.e. 1974 = 4.(FM 18, FM 62, FM 63, FM 64, FM 65, FM 88)L a L a L a L aL o L o L o L o L oMMLatitude, in degrees and minutes.(FM 22, FM 42, FM 44, FM 57, FM 62, FM 63, FM 64, FM 65)Longitude, in degrees and minutes.(FM 22, FM 42, FM 44, FM 57, FM 62, FM 63, FM 64, FM 65)Month of the year (UTC), i.e. 01 = January; 02 = February, etc.(FM 18, FM 22, FM 39, FM 40, FM 47, FM 49, FM 57, FM 62, FM 63, FM 64, FM 65, FM 71,FM 72, FM 73, FM 75, FM 76, FM 88)M i M i Identification letters of the report. (Code table 2582)(FM 12, FM 13, FM 14, FM 20, FM 32, FM 33, FM 34, FM 35, FM 36, FM 37, FM 38, FM 39,FM 40, FM 41, FM 62, FM 63, FM 64, FM 65, FM 67, FM 85, FM 86, FM 87,FM 88)

ANNEX II 129M j M jIdentification letters of the part of the report or the version of the code form. (Codetable 2582)(FM 12, FM 13, FM 14, FM 20, FM 32, FM 33, FM 34, FM 35, FM 36, FM 37, FM 38, FM 39,FM 40, FM 41, FM 62, FM 63, FM 64, FM 65, FM 67, FM 85, FM 86, FM 87,FM 88)n m n mn sm n smn b n b n bP a P a P a P aP p P p P p P pP sa P sa P sa P saP sp P sp P sp P spNumber of the band in which the maximum non-directional spectral density determined byheave sensors lies.(FM 65)Number of the band in which the maximum non-directional spectral density determined by slopesensors lies.(FM 65)Type and serial number of buoy.(FM 13, FM 18, FM 22, FM 63, FM 64, FM 65)Average wave period, in tenths of a second, or average wave length, in metres.(FM 65)Spectral peak period derived from heave sensors, in tenths of a second, or spectral peak wavelength, in metres.(FM 65)Average period derived from slope sensors, in tenths of a second, or average wave length, inmetres.(FM 65)Spectral peak period derived from slope sensors, in tenths of a second, or spectral peak wavelength, in metres.(FM 65)Q c Quadrant of the globe. (Code table 3333)(FM 13, FM 14, FM 18, FM 20, FM 33, FM 34, FM 36, FM 37, FM 38, FM 40, FM 41, FM 44,FM 47, FM 62, FM 63, FM 64, FM 65, FM 72, FM 76, FM 85)r 1 r 1r 2 r 2SSSSFirst normalized polar coordinate derived from Fourier coefficients.(FM 65)Second normalized polar coordinate derived from Fourier coefficients.(FM 65)Sampling interval (in tenths of a second or in metres).(FM 65)x Exponent for spectral wave data. (Code table 4800)(FM 65)YYDay of the month (UTC), with 01 indicating the first day, 02 the second day, etc.:(a) On which the actual time of observation falls;(FM 12, FM 13, FM 14, FM 15, FM 16, FM 18, FM 20, FM 32, FM 33, FM 34, FM 35, FM 36,FM 37, FM 38, FM 39, FM 40, FM 41, FM 42, FM 62, FM 63, FM 64, FM 65,FM 67, FM 85, FM 86, FM 87, FM 88)

130GUIDE TO WAVE ANALYSIS AND FORECASTINGWAVEOB CODE TABLES0161A 1WMO Regional Association area in which buoy, drilling rig or oil- or gas-production platformhas been deployed (1 – Region I; 2 – Region II, etc.)b w Sub-area belonging to the area indicated by A 1180160 140 120 100 80 60 40 20 0 20 40 60 80 100 120 140 160 180804865646326 25806040204645 44 6243 42 41 13611266242221604020051323115111423535202020544034 33 17 1656 5540607271747360180160 140 120 100 80 60 4020 0 20 40 60 80 100 120 140 160 1800880d a1 d a1d a2 d a2d d d dd 1 d 1d 2 d 2. . .d n d nMean direction, in units of 4 degrees, from which waves are coming for the band indicated,relative to true northPrincipal direction, in units of 4 degrees, from which waves are coming for the bandindicated, relative to true northTrue direction, in units of 4 degrees, from which the dominant wave is comingTrue direction, in units of 4 degrees, from which waves are comingCodeCodefigurefigure00 358° to less than 2° .01 2° to less than 6° 89 354° to less than 358°02 6° to less than 10° 90–98 Not used. 99 Ratio of the spectral density for the band to the. maximum is less than 0.0051731I a Indicator for frequency or wave numberCodeCodefigurefigure0 Frequency (Hz) 1 Wave number (m –1 )1732I b Indicator for directional or non-directional spectral wave dataCodeCodefigurefigure0 Non-directional 1 Directional

ANNEX II 1311744I m Indicator for method of calculation of spectral dataCodeCodefigurefigure1 Longuet-Higgins (1964) 4 Maximum entropy method2 Longuet-Higgins (F 3 method) 5–9 Reserved3 Maximum likelihood method1747I p Indicator for type of platformCodeCodefigurefigure0 Sea station 2 Aircraft1 Automatic data buoy 3 Satellite1761I 1 Name of country or international agency which operates the satelliteCodeCodefigurefigure0 European Community 3 Russian Federation1 Japan 4 India2 USA 5–9 Reserved2582M i M iM j M jIdentification letters of the reportIdentification letters of the part of the report or the version of the code formCode form M i M i M j M jLand SeaAircraft SatellitePart Part Part Part Nostation station A B C D distinctionFM 12–X Ext. SYNOP AA XXFM 13–X SHIP BB XXFM 14–X Ext. SYNOP MOBIL OO XXFM 18–X BUOY ZZ YYFM 20–VIII RADOB FF GG AA BBFM 32–IX PILOT PP AA BB CC DDFM 33–IX PILOT SHIP QQ AA BB CC DDFM 34–IX PILOT MOBIL EE AA BB CC DDFM 35–X Ext. TEMP TT AA BB CC DDFM 36–X Ext. TEMP SHIP UU AA BB CC DDFM 37–X Ext. TEMP DROP XX AA BB CC DDFM 38–X Ext. TEMP MOBIL II AA BB CC DDFM 39–VI ROCOB RR XXFM 40–VI ROCOB SHIP SS XXFM 41–IV CODAR LL XXFM 62–VIII Ext. TRACKOB NN XXFM 63–IX BATHY JJ XXFM 63–X Ext. BATHY JJ YYFM 64–IX TESAC KK XXFM 65–IX WAVEOB MM XXFM 67–VI HYDRA HH XXFM 85–IX SAREP CC DD AA BBFM 86–VIII Ext. SATEM VV AA BB CC DDFM 87–VIII Ext. SARAD WW XXFM 88–X SATOB YY XX

132GUIDE TO WAVE ANALYSIS AND FORECASTING3333Q cQuadrant of the globeCodefigureLatitudeLongitude1 North East3 South East5 South West7 North WestQ c = 7 N Q c = 1EquatorWGreenwich meridianEQ c = 5 S Q c = 3Note: The choice is left to the observer in the following cases:— When the ship is on the Greenwich meridian or the 180th meridian (L o L o L o L o = 0000 or 1800 respectively):Q c = 1 or 7 (northern hemisphere) orQ c = 3 or 5 (southern hemisphere);— When the ship is on the Equator (L a L a L a = 000):Q c = 1 or 3 (eastern longitude) orQ c = 5 or 7 (western longitude).4800xExponent for spectral wave dataCodeCodefigurefigure0 10 –5 5 10 01 10 –4 6 10 12 10 –3 7 10 23 10 –2 8 10 34 10 –1 9 10 4

ANNEX IIIPROBABILITY DISTRIBUTIONS FOR WAVE HEIGHTSThe following distributions are briefly described in the order shown (see Johnson and Kotz, 1970, inter alia, forfurther details):• Normal or Gaussian• Log-normal• Gamma• Weibull• Exponential• Rayleigh• Generalized extreme value (GEV)• Fisher-Tippett Type I or Gumbel• Fisher-Tippett Type III• Generalized ParetoThe symbols – x and s are used for the mean and standard deviation of the series x i (i = 1, n), i.e.:xs2x= ∑ inx x= ∑ ( i – )n – 12where Σ indicates summation over i (i = 1, n).The symbol Γ has its usual meaning as the gamma function:z– 1 – tΓ() z = t e dt; Γ′ () z =t = 0Values are widely tabulated.∫∞dΓdz1. Normal or Gaussian distributionf (x) =1 ⎡ (x – α )2exp ⎢– 2π β ⎣ 2β 2mean = αvariance = β 2⎤⎥ – ∞

134GUIDE TO WAVE ANALYSIS AND FORECASTING2. Log-normal distributionThe two-parameter log-normal distribution is given by:1 ⎡ xf( x) = exp – (log – α⎢)2xβ2π⎣ 2β= 0⎡ ⎤⎢αβ 2mean = exp + ⎥⎣ 2 ⎦2 2[ ] [ ]( [ ] )variance= exp 2α exp β exp β – 12⎤⎥ x > 0⎦otherwiseProbability paper⎡–x ⎛ x ⎞x = Φ ( p) ⎢Φ( x) = ∫ exp∞ ⎜– 211–⎟⎣⎢2π⎝ 2 ⎠y = log( h)⎤dx⎥⎦⎥Maximum likelihood estimatorsˆα = 1 n ∑ log x iˆβ =1n ∑(log x i – ˆα ) 2The three-parameter log-normal is produced by replacing x by x – θ and is defined on [θ, ∞). Estimation is morecomplex.3. Gamma distributionProbability paper cannot be used for estimation with the gamma distribution.Moments estimatorsf( x)xx=αβ Γ ( α) exp(– β) α , β , x > 0= 0otherwisemean = αβ2variance = αβxα˜=2s2˜ sβ =xMaximum likelihood estimators2α – 11∑ log x = log ˆ + ( ˆi β Ψ α)nx = αβ ˆ ˆwhere Ψ(x) = Γ′(x)/Γ(x). These equations have to be solved numerically.The three-parameter gamma is produced by replacing x by x – θ and is defined on [θ, ∞). Estimation is morecomplex.

ANNEX III 1354. Weibull distribution–xxf( x) = ⎛ β 1⎞ ⎡β1exp –⎛ ⎞ ⎤β , , x⎝ ⎠⎢⎝ ⎠⎥ α β > 0α α ⎣⎢α ⎦⎥= 0otherwise⎡βxFx–⎛ ⎞ ⎤( ) =1– exp ⎢x⎝ ⎠⎥> 0⎣⎢α ⎦⎥= 0 x < 0⎛ 1 ⎞mean = αΓ⎜ + 1 ⎟⎝ β ⎠2 1 1 12⎡2 ⎛ ⎞ ⎧= ⎜ + –⎝ ⎠⎟ ⎛⎜⎝+ ⎞⎫⎤variance α ⎢Γ ⎨Γ ⎟⎬⎥⎢ β ⎩ β ⎠⎣⎭ ⎥⎦Probability paperx = log −log1−py = log( h)Maximum likelihood estimatorsαˆ= ⎛ ⎜∑⎝[ ( )]xnβi⎞⎟⎠1βˆ ⎛βlog( ) log( ) –β = ∑ xixi∑ x ⎞⎜ β –i⎟⎝ ∑ xn ⎠These equations have to be solved numerically.The three-parameter Weibull is produced by replacing x by x – θ and is defined on [θ, ∞). Estimation is morecomplex.5. Exponential distributionif( x) 1 x= exp⎛–⎞α ⎝ α⎠α , x > 0= 0otherwiseFx ( ) = 0 x 0mean = α2variance = α1Probability paperx = – log( 1– p)y = h(Any line must go through the origin.)Moments and maximum likelihood estimatorsα˜= αˆ= xThe two-parameter exponential is produced by replacing x by x – θ and is defined on [θ, ∞). Estimation is morecomplex. The exponential distribution is a special case of the Gamma (α = 1) and Weibull (β = 1) distributions.

136GUIDE TO WAVE ANALYSIS AND FORECASTING6. Rayleigh distributionf( x) 2x ⎛ x ⎞=2exp ⎜ –2 ⎟α ⎝ 2α⎠α,x > 0= 0otherwise2=⎛ πvariance 2α1−⎞⎝ 4 ⎠Fx ( ) = 0 x 0mean = απ2Probability paper[ ( )]x = log −log1−py = log( h)All lines have slope O.5.Moments estimatorsxα˜=π / 2Maximum likelihood estimators˜α =∑ x i 22nThe two-parameter Rayleigh is produced by replacing x by x – θ and is defined on [θ, ∞). Estimation is morecomplex. The Rayleigh distribution is a special case of the Weibull distribution (β = 2, α = √2α).7. Generalized extreme value (GEV) distributionFx ( )= 0 x< α + β and θ < 0θ⎡1⎤x –= exp⎢ ⎛ α ⎞ θ⎢– ⎜1–θ⎥⎟⎝ β ⎠ ⎥⎣⎢⎦⎥= 1 x > α + and θ > 0θmean = α + [ 1–Γ ( 1+θ)]θθ ≠ 0= α+ γ β ( γ = Euler’sconstant = 0. 5772...)θ = 02= ⎛ ⎝ ⎜ β ⎞2variance ⎟ Γ ( 1+2θ)+ Γ ( 1+θ)θ 0θ ⎠2[ ] ≠2= β πθ = 06For methods of estimation, see NERC (1975).

ANNEX III 1378. Fisher-Tippett Type I (FT-I) (or Gumbel) distributionf( x) 1 ⎡ xxexp – – α ⎛– exp – – α ⎞⎤= ⎢⎜ ⎟⎥ β ⎣ β ⎝ β ⎠⎦β > 0Fx ( )⎡ ⎛ x= exp – exp ⎜ – – α ⎞⎤⎢⎟⎥⎣ ⎝ β ⎠⎦mean = α+ γ β ( γ = Euler’sconstant = 0. 5772...)22variance = β π6Probability paper[ ]x = – log −log( p)y = hMoments estimatorsα = x – γ βsβ = 6πMaximum likelihood estimators⎡1⎛ x ⎞⎤iα = – βlog ⎢ ∑exp ⎜ – ⎟⎥⎣n⎝ β ⎠⎦⎛ xi⎞∑ xiexp ⎜ – ⎟1⎝ β ⎠β = ∑xi–n⎛ xi⎞∑ exp ⎜ – ⎟⎝ β ⎠These equations have to be solved numerically.A detailed description and comparison of the various methods of estimation for the FT-I distribution are givenin Carter and Challenor (1983). The FT-I is a special case of the GEV (θ = 0).9. Fisher-Tippett Type III (FT-III) distributionf( x)– x= β ⎛θ ⎞⎝ α ⎠= 0β–11 ⎡ – xexp –⎛θ⎞⎢α ⎝ ⎠⎣⎢αβ⎤⎥ x < θ⎦⎥otherwiseFx ( )⎡β– x= exp –⎛θ⎞ ⎤⎢⎝ ⎠⎥⎣⎢α ⎦⎥= 1x < θx > θmean⎛ 1 ⎞= θ–αΓ⎜ + 1 ⎟⎝ β ⎠⎡22 ⎛ 2 ⎞ ⎧α ⎜ + 1 –⎝ β ⎠⎛ 1variance= Γ Γ ⎜β+ ⎞⎫⎤⎢ ⎨ 1⎟⎬ ⎥⎢ ⎩ ⎝ ⎠⎣⎭ ⎥⎦

138GUIDE TO WAVE ANALYSIS AND FORECASTINGProbability paper[ ( )]x = log −log1−py = log( θ – h)Probability plotting can only be used to estimate θ by plotting several lines and using some goodness-of-fit criterion(see text for details).The FT-III distribution is a special case of the GEV (θ < 0). For this reason, it is advisable to use the estimation techniquesreferenced for the GEV. If the transformation X = –X is applied to the FT-III, a three-parameter Weibulldistribution is obtained.10. Generalized Pareto distribution1 xf( x) =⎛–⎞ β −11 βα ⎝ α ⎠x > 0= 0otherwise1This distribution is the asymptotic distribution for exceedances over a level (in a similar way to the GEVfor extremes). For details of its use and estimators, see Smith (1984). The exponential distribution is a special casewith β = 0.

ANNEX IVTHE PNJ (PIERSON-NEUMANN-JAMES, 1955) WAVE GROWTH CURVESDuration graph. Distorted co-cumulative spectra for wind speeds from 10 to 44 knots as a function of the duration

140GUIDE TO WAVE ANALYSIS AND FORECASTINGFetch graph. Distorted co-cumulative spectra for wind speeds from 10 to 44 knots as a function of the fetch

ANNEX IV 141Duration graph. Distorted co-cumulative spectra for wind speeds from 36 to 56 knots as a function of the duration

142GUIDE TO WAVE ANALYSIS AND FORECASTINGFetch graph. Distorted co-cumulative spectra for wind speeds from 36 to 56 knots as a function of the fetch

REFERENCESAbbott, M. B., H. H. Petersen and O. Skovgaard, 1978: Computations of short waves in shallow water. Proc. 16thCoastal Engineering Conference, Hamburg, Germany, 414–433.Abreu, M., A. Larazza and E. B. Thornton, 1992: Non-linear transformation of directional wave spectra in shallowwater. J. Geophys. Res., 97(C10), 15,579–15,589.Allender, J., T. Audunson, S. F. Barstow, S. Bjerken, H. E. Krogstad, P. Steinbakke, L. Vartdal, L. Borgman andC. Graham, 1989: The WADIC Project: A comprehensive field evaluation of directional wave instrumentation.Ocean Engineering, 161, 505–536.AWS, 1995: Forecasters guide to tropical meteorology. Air Weather Service, Illinois, USA, AWS TR 240, Updatedby Colin S. Ramage. August 1995, 462 pp.Bacon, S. and D. J. T. Carter, 1991: Wave climate changes in the North Atlantic and North Sea. Int. J. of Climatology,11, 545–558.Barnett, T. P., 1968: On the generation, dissipation and prediction of ocean wind-waves. J. Geophys. Res., 73,513–529.Barrick, D. E. and J. F. Gower (Eds.), 1986: Special issue on HF radar oceanography. IEEE Jour. Oceanic Eng., Aprill986.Barstow, S. F., 1995: Wave climate assessment by satellite remote sensing. Proc. Fifth International Offshore andPolar Engineering Conf. (ISOPE-1995), The Hague, Netherlands, June 1995.Barstow, S. F. and T. Kollstad, 1991: Field trials of the Directional Waverider. Proc. First International Offshore andPolar Engineering Conf. (ISOPE-1991), Edinburgh, Scotland, August 1991.Barstow, S. F., O. Haug and T. van der Vlugt, 1994(a): A field validation of a Directional Waverider in a SEAWATCHbuoy. Proc. of the Oceans '94 Conf., Brest, France, Sept. 1993, Vol 2, 32–37.Barstow, S. F., T. I. Bern, S. Bjerken, T. I. Brate, O. Haug, O. G. Houmb and H. E. Krogstad, 1994(b): World waveclimatologies from satellite altimeters. Proc. of the OCEANS '94 Conf., Brest, France, Sept. 1993, Vol. 2., 64–68.Battjes, J. A., 1972: Long-term wave height distributions at seven stations around the British Isles. Deut. Hydrogr. Z.,25(4), 179–189.Battjes, J. A. and J. P. F. M. Janssen, 1978: Energy loss and set-up due to breaking in random waves. Proc. 16thCoastal Engineering Conference, Hamburg, Germany, 569–587.Battjes, J. A., Y. Eldeberky and Y. Won, 1993: Spectral Boussinesq modelling of breaking waves. Proc. 2nd Int.Symp. on Ocean Wave Measurements and Analysis, New Orleans, USA, 813–820.Beale, R. C., 1981: Spatial evolution of ocean wave spectra. In: Spaceborne synthetic aperture radar for oceanography,R. Beale, P. De Leonibus, E. Katz (Eds), The Johns Hopkins Oceanographic Studies, 7, The JohnsHopkins University Press, 110–127.Bendat, J. S. and A. G. Piersol, 1971: Random data: analysis and measurement procedures. Wiley Interscience, NewYork.Berkhoff, J. C. W., 1972: Computation of combined refraction-diffraction. Proc. 13th Coastal EngineeringConference, 471–490.Bjerke, P. L. and K. Torsethaugen, 1989: Environmental conditions on the Norwegian Continental Shelf, Barents Sea.SINTEF NHL Report No. STF60 A89052, Trondheim, Norway.Blackadar, A. K., 1965: A simplified two-layer model of the baroclinic neutral atmospheric boundary layer. Air ForceCambridge Res. Lab. Report, 65-531, 49–65.Blackman, R. B. and J. W. Tukey, 1959: The measurement of power spectra from the point of view of communicationsengineering. Dover, New York.

144GUIDE TO WAVE ANALYSIS AND FORECASTINGBooij, N., L. H. Holthuijsen and P. H. M. de Lange, 1992: The penetration of short-crested waves through a gap. Proc.23rd Coastal Engineering Conference, 1044–1052.Borgman, L. E., 1963: Risk criteria. J. Waterways, Harbors and Coastal Eng. Div., ASCE, 89, (WW3), 1–35.Borgman, L. E., 1973: Probabilities for highest wave in hurricane. J. Waterways, Harbors and Coastal Eng. Div.,ASCE, 99 (WW2), 185–207.Borgman, L. E. and D. T. Resio, 1982: Extremal statistics in wave climatology. A.R. Osborne and P.M. Rizzoli (Eds.).Topics in ocean physics. North-Holland Publ. Co., 439–471.Bouws, E. and J. A. Battjes, 1982: A Monte Carlo approach to the computation of refraction of water waves. J.Geophys. Res., 87(C8), 5718–5722.Bouws, E., H. Günther, W. Rosenthal and C. L. Vincent, 1985: Similarity of the wind-wave spectrum in finite depthwater. 1 – Spectral form. J. Geophys. Res., 90(C1), 975–986.Breivik, L. A. and M. Reistad, 1992: Use of ERS-1 altimeter wave products at DNMI. Evaluation of wave heights andwind speeds. Assimilation in a numerical wave model. Technical Report No. 101, The NorwegianMeteorological Institute.Brooks, R. M. and W. D. Corson, 1984: Summary of archived Atlantic coast wave information study pressure, wind,wave and water level data. WIS Report 13. Waterways Experiment Station, Vicksburg, MS, USA.Brown, R. A. and W. T. Liu, 1982: An operational large-scale marine planetary boundary layer model. J. Appl.Meteor., 21, 261–269.Burroughs, L. D., 1984: Great Lakes near-shore wind predictions from Great Lakes MOS wind guidance. NOAATech. Memorandum, NWS TDL 74.Businger, J. A., J. C. Wyngaard, Y. Izumi and E. F. Bradley, 1971: Flux-profile measurements in the atmosphericsurface layer. J. Atmos. Sci., 28, 181–189.Canadian Climate Center, 1989: Proc. 2nd International Workshop on Wave Hindcasting and Forecasting, Vancouver,B.C., CCC, Downsview, Ontario.Canadian Climate Center, 1991: Wind/wave hindcast extremes for the east coast of Canada. Volume I. Prepared undercontract no. KM169-7-6678 by MacLaren Plansearch Limited and Oceanweather Inc., 109 pp plus appendices.Canadian Climate Center, 1992(a): Proc. 3rd International Workshop on Wave Hindcasting and Forecasting,Montreal, Que., CCC, Downsview, Ontario.Canadian Climate Center, 1992(b): Wind/wave hindcast extremes for the west coast of Canada. Downsview, Ontario,M3H 5T4. 143 pp plus appendices.Cardone, V. J., 1969: Specification of the wind distribution in the marine boundary layer for wave forecasting.Geophysical Sciences Laboratory, TR-69-1. New York University.Cardone, V. J., 1978: Specification and prediction of the vector wind of the United States continental shelf for applicationto an oil slick trajectory program. Final report, Contract T-34530, The City College of the City Universityof New York.Cardone, V. J., W. J. Pierson and E. G. Ward, 1976: Hindcasting the directional spectrum of hurricane-generatedwaves. J. Petrol. Tech., 28, 385–394.Carter, D. J. T. 1993: Estimating extreme wave heights in the NE Atlantic from GEOSAT data. Health and SafetyExecutive Offshore Tech. Report: OTH 93 396. HMSO London, 24 pp.Carter, D. J. T. and P. G. Challenor, 1981(a): Estimating return values of environmental parameters. Quart. J. RoyalMet. Soc., 107, 259–266.Carter, D. J. T. and P. G. Challenor, 1981(b): Estimating return values of wave height. Inst. of OceanographicSciences, 116.Carter, D. J. T. and P. G. Challenor, 1983: Methods of fitting the Fisher-Tippett Type I extreme value distribution.Ocean Eng., 10, 191–199.Carter, D. J. T., and L. Draper, 1988: Has the north-east Atlantic become rougher? Nature, 332, 494.

REFERENCES 145Carter, D. J. T., P. G. Challenor, and P. D. Cotton, 1994: Estimating extreme values of significant wave heights fromaltimeter data. Proc. WMO/IOC tech. conf. on space-based observations. Bergen, Norway, Sept. 1993. WMO,Geneva, TD-No. 649, 63–70.Carter D. J. T., P. G. Challenor and M. A. Srokosz, 1992: An assessment of GEOSAT wave height and wind speedmeasurements. J. Geophys. Res., 97, 11383–11392.Carter, D. J. T., S. Foale and D. J. Webb, 1991: Variations in global wave climate throughout the year. Int. J. RemoteSensing, 12(8), 1687–1697.Carter, D. J. T., P. G. Challenor, J. A. Ewing, E. G. Pitt, M. A. Srokosz and M. J. Tucker, 1986: Estimating waveclimate parameters for engineering applications. Offshore Tech. Report OTH 86/228, HMSO, London.Cartwright, D. E, and M. S. Longuet-Higgins, 1956: The statistical distribution of the maxima of a random function.Proc. Royal Soc. London, A237, 212–232.Cavaleri, L., 1993: Meteorological requirements for wave modelling. WMO Marine Meteorology and RelatedOceanographic Activities Report, 29, WMO/TD-No. 583.Cavaleri, L. and P. M. Rizzoli, 1981: Wind wave prediction in shallow water: theory and applications. J. Geophys.Res., 86, 10961–10973.CERC: Coastal Engineering Research Center, 1973, 1977, 1984: Shore protection manual. Waterways ExperimentStation, US Army Corps of Engineers.Challenor, P. G. and D. J. T. Carter, 1983: Monthly extreme air temperatures at Rothamsted. J. Climat., 3, 395–404.Challenor P. G., S. Foale and D. J. Webb, 1990: Seasonal changes in the global wave climate measured by theGEOSAT altimeter. Int. J. Remote Sensing, 11, 2205–2213.Charnock, H., 1955: Wind stress on a water surface. Quart. J. Royal Met. Soc., 81, 639.Chelton, D. B., K. J. Hussey and M. E. Parke, 1981: Global satellite measurements of water vapour, wind speed andwave height. Nature, 294, 529–532.Clancy, R. M., J. E. Kaitala and L. F. Zambresky, 1986: The Fleet Numerical Oceanography Centre Global SpectralOcean Wave Model. Bull. Amer. Meteor. Society, 67(5), 498–512.Clarke, R. H. and G. D. Hess, 1974: Geostrophic departures and the functions A and B of Rossby Similarity Theory.Boundary Layer Met., 7(3), 267–287.Clarke, R. H. et al., 1971: The Wangara experiment: boundary layer data. Commonwealth Scientific and IndustrialResearch Organization, Division of Meteorological Physics, Technical Paper 19, Australia.Cooley, J. W. and J. W. Tukey, 1965: An algorithm for machine calculation of complex Fourier series. Math. Comput.,19, 297–301.Corson, W. D. et al., 1981: Atlantic coast hindcast; deep water significant wave information. U.S. Army Corps ofEngineers, Waterways Experiment Station, Wave Info. Study for US Coastlines, WIS, Rep. 2, 856 pp.Crapper, G. D., 1984: Introduction to water waves. Publ. Ellis Harwood.Dattatri, J., 1973: Waves off Mangalore Harbor – west coast of India. J. Waterways, Harbors and Coastal Eng. Div.ASCE, 99(WW1), 39–58.Dingemans, M. W., M. J. F. Stive, J. Bosma, H. J. de Vriend and J. A Vogel, 1986: Directional nearshore wave propagationand induced currents. Proc. 20th Coastal Engineering Conference, 1092–1106.Dobson, F. W., 1982: Review of reference height for and averaging time of surface wind measurements at sea. WMOMarine Meteorology and Related Oceanographic Activities Report, 3.Donelan, M., 1982: The dependence of the aerodynamic drag coefficients on wave parameters. Proc. First Int. Conf. onMeteorology and Air-sea Interaction of the Coastal Zone, The Hague, Netherlands, Amer. Meteor. Soc., 381–387.Donelan, M. A., J. Hamilton and W. H. Hui, 1985: Directional spectra of wind generated waves. Phil. Trans. RoyalSoc., A315, 509–562.Dorrestein, R., 1960: Simplified method of determining refraction coefficients for sea waves. J. Geophys. Res., 65,637–642.

146GUIDE TO WAVE ANALYSIS AND FORECASTINGDorrestein, R., 1962: Wave set-up on a beach. Proc. Second Tech. Conf. on Hurricanes, Miami Beach, WashingtonDC, 230–24l.Draper, L., 1963: Derivation of a ‘design wave’ from instrumental records of sea waves. Proc. Inst. Civ. Eng., London,26, 291–304.Draper, L., 1966: The analysis and presentation of wave data — a plea for uniformity. Proc. 10th Int. Conf. onCoastal Eng., Tokyo. ASCE, 1, 1–11.Eid, B. M., C. M. Morton, V. J. Cardone and J. A. Greenwood, 1989: Development and evaluation of a wave climatedatabase for the east coast of Canada. Proc. 2nd International Workshop on Wave Hindcasting and Forecasting,Vancouver, B.C., 140–150.Eide, L. I., M. Reistad and J. Guddal, 1985: Database for computed wind and wave parameters for the North Sea,Norwegian Sea and Barents Sea, each sixth hour for the years 1955–1981. Norwegian Meteorological Institute,Bergen.Ewing, J. A., 1971: A numerical wave prediction model for the North Atlantic Ocean. Deut. Hydrogr. Z., 24, 241–261.Ewing, J. A. 1986: Presentation and interpretation of directional wave data. Underwater Technology, 12, 17–23.Feit, D. H. and N. A. Pore, 1978: Objective wind forecasting and verification on the Great Lakes. J. of Great LakesRes., 4(1), 10–18.Feldhausen, P. H., S. K. Chakrabarti and B. W. Wilson, 1973: Comparison of wave hindcasts at weather station J forNorth Atlantic storm of December, 1959. Deut. Hydrogr. Z., 26(1), 10–16.Fisher, R. A. and L. H. C. Tippett, 1928: Limiting forms of the frequency distribution of the largest or smallestmember of a sample. Proc. Camb. Phil. Soc., 24, 180–190.Forristall, G. J., 1978: On the statistical distribution of wave heights in a storm. J. Geophys. Res., 83, 2353–2358.Francis, P .E., 1987: The North European Storm Study (NESS). Proc. Int'l Workshop on Wave Hindcasting andForecasting, Halifax, Nova Scotia, 23–26 September 1986. Environmental Studies Revolving Fund. ReportSeries No. 065, Ottawa, 370 pp, 295–302.Freilich, M. H. and P. G. Challenor, 1994: A new approach for determining fully empirical altimeter wind speedmodel function. J. Geophys. Res., 99, 25051–25062.Galambos, J., 1978: The asymptotic theory of extreme order statistics. J. Wiley and Sons.Georges, T. M. and J. A. Harlan, 1994: Military over-the-horizen radars turn to ocean monitoring. MTS Journal, 27(4),31–38.Gerling, T. W., 1992: Partitioning sequences and arrays of directional wave spectra into component wave systems.J. Atmos. Oceanic Technol., 9, 444–458.Gilhousen, D. B., 1987: A field evaluation of NDBC moored buoy winds. J. Atmos. Oceanic Technol., 4, 94–104.Glahn, H. R. and D. A. Lowry, 1972: The use of model output statistics (MOS) in objective weather forecasting.J. Appl. Meteor., 11, 1203–1210.Goda, Y., 1970: Numerical experiments on wave statistics with spectral simulation. Report of the Port and HarbourRes. Inst., 9(3), 3–57.Goda, Y., 1978: The observed joint distribution of periods and heights of sea waves. Proc. 16th Coastal EngineeringConference, Hamburg, Germany, 227–246.Goda, Y., 1979: A review on statistical interpretation of wave data. Report of the Port and Harbour Res. Inst., 18(1),5–32.Golding, B., 1983: A wave prediction system for real-time sea state forecasting. Quart. J. Royal Met. Soc., 109,393–416.Gröen P. and R. Dorrestein, 1976: Zeegolven. KNMI Opstellen op Oceanografisch en Maritiem MeteorologischGebied, 11, 124 pp.Grønlie, Ø., D. C. Brodtkorb and J. Wøien, 1984: MIROS: A microwave remote sensor for the ocean surface.Norwegian Maritime Research Quart. Rev., 12(3), 24–28.

REFERENCES 147Gumbel, E. J., 1958: Statistics of extremes. Columbia Univ. Press, New York.Haltiner, G. J. and F. L. Martin, 1957: Dynamic and physical meteorology. McGraw-Hill Book Company.Haltiner, G. J. and R. T. Williams, 1980: Numerical prediction and dynamic meteorology. John Wiley & Sons.Hamilton, G. D., 1990: Guide to moored buoys and other ocean data acquisition systems. WMO Marine Meteorologyand Related Oceanographic Activities Report, 16, WMO/TD-750.Haring, R. E. and J. C. Heideman, 1978: Gulf of Mexico rare wave return periods. Offshore Technology Conference,Paper 3230.Hasselmann, D. E., M. Dunckel, and J. A. Ewing, 1980: Directional wave spectra observed during JONSWAP (1973).J. Phys. Oceanogr., 10, 1264–1280.Hasselmann, K., 1962: On the non-linear energy transfer in a gravity wave spectrum. 1: General theory. J. FluidMech., 12, 481–500.Hasselmann, K., 1974: On the spectral dissipation of ocean waves due to whitecapping. Boundary Layer Meteor., 6,107–127.Hasselmann, K., S. Hasselmann, C. Brüning and A. Speidal, 1991: Interpretation and application of SAR wave imagespectra in wave models. pp 117–124 in Beal R. C. (Ed.) Directional ocean wave spectra, Johns Hopkins Univ.Press, Baltimore. 218 pp.Hasselmann, K., D. B. Ross, P. Müller and W. Sell, 1976: A parametric wave prediction model. J. Phys. Oceanogr.,6(2), 200–228.Hasselmann, K., T. P. Barnett, E. Bouws, H. Carlson, D. E. Cartwright, K. Enke, J. A. Ewing, H. Gienapp, D. E.Hasselmann, P. Kruseman, A. Meerburg, P. Muller, D. J. Olbers, K. Richter, W. Sell and H. Walden, 1973:Measurements of wind-wave growth and swell decay during the Joint North Sea Wave Project (JONSWAP).Deut. Hydrogr. Z., A8(12).Hasselmann, S., and K. Hasselmann, 1981: A symmetrical method of computing the non-linear transfer in a gravitywavespectrum. Hamburger Geophysikalische Einzelschriften, Reihe A: Wiss. Abh. 52, 163 pp.Hasselmann, S. and K. Hasselmann, 1985: Computations and parameterizations of the non-linear energy transfer ina gravity wave spectrum. Part I. A new method for efficient calculations of the exact non-linear transfer integral.J. Phys. Oceanogr., 15, 1369–1377.Hasselmann, S., K. Hasselmann, J. H. Allender and T. P. Barnett, 1985: Computations and parameterizations of thenon-linear energy transfer in a gravity wave spectrum. Part II: Parameterizations of the non-linear energy transferfor application in wave models. J. Phys. Oceanogr., 15(11), 1378–1391.HMSO, 1985: Waves recorded at ocean weather station Lima, December 1975–November 1981. Offshore Tech.Report, OTH 84 204, Her Majesty’s Stationary Office, London.Hogben, N., 1990: Overview of global wave statistics. Chapter II in Encyclopedia of Fluid Mechanics, Volume 10:Surface and groundwater phenomena, Gulf Publishing Co., Houston.Hogben, N. and D. J. T. Carter, 1992: Estimation of wave heights: a review of guidance issued by the Department ofEnergy. Underwater Technology, 18(1), 15–23.Hogben, N., N. M. C. Dacunha and G. F. Olliver, 1986: Global wave statistics. Unwin Brothers for British MaritimeTech. Ltd.ISSC, 1979: Report of Committee I.1: (Environmental conditions). 7th International Ship Structures Congress, Paris.Isaacson, M. de St.Q. and N. G. Mackenzie, 1981: Long-term distribution of ocean waves: a review. J. Waterways,Port, Coastal and Ocean Divn., ASCE, 107(WW2), 92–109.Jansen, P. C. M., 1986: Laboratory observations of the kinematics in the aerated region of breaking waves. CoastalEngineering, 9, 453–477.Janssen, P. A. E. M, 1991: Quasi-linear theory of wind-wave generation applied to wave forecasting. J. Phys.Oceanogr., 21, 1631–1642.Jardine, T. P. and F. R. Latham, 1981: An analysis of wave height records from the NE Atlantic. Quart. J. Royal Met.Soc., 107, 4l5–426.

148GUIDE TO WAVE ANALYSIS AND FORECASTINGJenkins, A. D., 1992: A quasi-linear eddy-viscosity model for the flux of energy and momentum to wind waves usingconservation-law equations in a curvilinear coordinate system. J. Phys. Oceanogr., 22, 844–858.Jenkins, G. M. and D. G. Watts, 1968: Spectral analysis and its applications. Holden-Day.Jenkinson, A. F., 1955: The frequency distribution of the annual maximum (or minimum) values of meteorologicalelements. Quart. J. Royal Met. Soc., 81, 158–171.Johnson, N. L. and S. Kotz, 1970: Continuous univariate distributions. John Wiley and Sons, New York.Jones, W. L., L. C. Schroeder, D. H. Boggs, E. M. Bracalente, R. A. Brown, G. J. Dame, W. J. Pierson and F. J.Wentz, 1982: The SEASAT-A scatterometer: the geophysical evaluation of remotely sensed wind vectors overthe ocean. J. Geophys. Res., 87(C5), 3297–3317.Kanamitsu, M., J. Alpert, K. Kampana, P. Caplan, D. Deaven, M. Iredell, B. Katz, H.-L. Pan, J. Sela and G. White,1991: Recent changes implemented into the global forecast system at NMC. Weather and Forecasting, 6(3),415–424.Khandekar, M. L., 1989: Operational analysis and prediction of ocean wind waves. Coastal and Estuarine Studies,No. 33, Springer-Verlag, 214 pp.Khandekar, M. L. and V. Swail, 1995: Storm waves in Canadian waters: A major marine hazard. Atmosphere-Ocean,33, 329–357.Khandekar, M. L., R. Lalbeharry and V. Cardone, 1994: The performance of the Canadian Spectral Ocean WaveModel (CSOWM) during the Grand Banks ERS-1 SAR wave spectra validation experiment. Atmosphere-Ocean,32, 31–60.Komen, G. J., S. Hasselmann, and K. Hasselmann, 1984: On the existence of a fully developed wind-sea spectrum.J. Phys. Oceanogr., 14, 1271–1285.Komen, G. J., L. Cavaleri, M. Donelan, K. Hasselmann, S. Hasselmann and P. A. E. M. Janssen, 1994: Dynamics andmodelling of ocean waves. Cambridge University Press, 532 pp.Krishna, K., 1981: A two-layer first order closure model for the study of the baroclinic atmospheric boundary layer.J. Atmos. Sci., 38, 1401–1417.Lacombe, H., 1965: Cours d’oceanographie physique, Gauthier-Villars, Paris, 392 pp.Laing, A. K., 1985: An assessment of wave observations from ships in southern oceans. J. Clim. Appl. Meteorol., 24,481–494.Large, W. G. and S. Pond, 1981: Open ocean momentum flux measurements in moderate to strong winds. J. Phys.Oceanogr., 11(3), 324–336.Lazanoff, S. M. and N. Stevenson, 1975: An evaluation of a hemispherical operational wave spectra model. Tech.Note 75-3, Fleet Numerical Weather Central, Monterey, Calif.LeBlond, P. H. and L. A. Mysak, 1978: Waves in the ocean. Elsevier, Amsterdam.Lefevre, J. M., J. Barckicke and Y. Menard, 1994: A significant wave height dependent function for Topex/Poseidonwind speed retrieval. J. Geophys. Res., 99, 25035–25049.Lighthill, J., 1978: Waves in fluids. Cambridge University Press.Lionello, P., H. Gunther and P. A. E. M. Janssen, 1992: Assimilation of altimeter data in a global third generationwave model, Technical Report No. 67, European Centre for Medium Range Weather Forecasts, ECMWF.Lindau, R., 1994: A New Beaufort Equivalent Scale. Proc. International COADS Wind Workshop, Kiel, Germany,May 1994, 232–249.Liu, W. T., K. B. Katsaros and J. A. Businger, 1979: Bulk parameterization of air-sea exchanges of heat and watervapour including the molecular constraints at the interface. J. Atmos. Sci., 36. 1722–1735.Long, P. E. et al., 1978: The state of the techniques development laboratory’s boundary layer model. NOAA Tech.Memorandum NWS TDL 66.Longuet-Higgins, M. S., 1952: On the statistical distribution of the heights of sea waves. J. Marine Res., 11, 245–266.Longuet-Higgins, M. S. and R. W. Stewart, 1962: Radiation stress and mass transport in gravity waves, with applicationsto “surf beats”. J. Fluid. Mech., 13, 481–504.

REFERENCES 149Maat, N., C. Kraan and W. A. Oost, 1991: The roughness of wind waves. Boundary Layer Meteor., 54, 89–103.MacLaren Plansearch Ltd. and Oceanweather, Inc., 1992: Beaufort Sea extreme waves study. Environmental StudiesResearch Funds Report, ESRF-114. National Energy Board, Calgary, Alberta.Madsen, P. A., and O. R. Sorensen, 1993: Bound waves and triad interactions in shallow water, J. Ocean Engineering,20(4), 359–388.Mathiesen, M., Y. Goda, P. J. Hawkes, E. Mansard, M. J. Martin, E. Peltier, E. F. Thompson, and G. van Vledder,1994: Recommended practice for extreme wave analysis. J. Hydr. Res., 32(6), 803–814.Mei, C. C., 1989: The applied dynamics of ocean surface waves. World Scientific Publ., Advanced Series on OceanEngineering – Volume 1, 740 pp.Mendenhall, B. R., 1967: A statistical study of frictional wind veering in the planetary boundary layer. AtmosphericScience Paper 116, Dept. of Atmos. Sci, Colorado State University.Mesinger, F., Z. I. Janjic, S. Nickovic and D. Gavrilov, 1988: The step-mountain coordinate: Model description andperformance for cases of alpine lee cyclogenesis and for a case of an Applalachian redevelopment. MonthlyWeather Review, 116, 1493–1518.Miles, J. W., 1957: On the generation of surface waves by shear flows. J. Fluid Mech., 3, 185–204.Miles, J. W., 1960: On the generation of surface waves by turbulent shear flows. J. Fluid Mech., 7, 469–478.Mitsuyasu, H., F. Tasai, T. Suhara, S. Mizuno, M. Okhuso, T. Honda, and K. Rikiishi, 1975: Observations of the directionalspectrum of ocean waves using a cloverleaf buoy. J. Phys. Oceanogr., 5, 750–760.Monin, A. S. and A. M. Obukhov, 1954: Basic laws of turbulent mixing in the ground layer of the atmosphere. Akad.Nauk. SSSR Geofiz. Inst. Tr., 151, 163–181.Munk, W. H., 1951: Origin and generation of waves. Proc. First Coastal Engineering Conference, Long Beach, Calif., 1–4.NOAA, 1990: Coastal-marine automated network (C-MAN) NWS users guide. US Dept. of Commerce.Neu, H. J. A., 1984: Interannual variations and longer-term changes in the sea state of the North Atlantic from 1970to 1982. J. Geophys. Res., 89(C4), 6397-6402.NERC, 1975: Flood studies report. Vol. 1 – Hydrological studies. Natural Environment Research Council, London.Olsen, E., S. F. Barstow and K. Selanger, 1991: Wave data collection, Efate, Vanuatu, November–December 1990,Oceanographic Company of Norway Report OCN R-91076.O’Reilly, W. C., T. H. C. Herbers, R. J. Seymour and R. T. Guza, 1995: A comparison of directional buoy and fixedplatform measurements of Pacific swell. Submitted to J. Atmos. Oceanic Technol.Overland, J. E. and W. H. Gemmill, 1977: Prediction of marine winds in the New York bight. Monthly WeatherReview, 105, 1003–1008.Phillips, D. W. and J. G. Irbe, 1978: Lake to land comparison of wind, temperature, and humidity of Lake Ontarioduring the International Field Layer for the Great Lakes. CLI 2-77, Atmospheric Environment Service,Environment Canada.Phillips, O. M., 1957: On the generation of waves by a turbulent wind. J. Fluid Mech., 2, 417–445.Phillips, O. M., 1958: The equilibrium range in the spectrum of wind-generated waves. J. Fluid Mech., 4, 426–434.Pickands, J. III, 1975: Statistical inference using extreme order statistics. Ann. Stat., 3, 119–131.Pierson, W. J., 1979: TWINDX: A planetary boundary layer model for the winds in extratropical and tropical oceanareas. Contract no. N00014-77-C-0206. Naval Environ. Pred. Res. Cen., Monterey, California, USA.Pierson, W. J., 1982: The Spectral Ocean Wave Model (SOWM): a northern hemisphere computer model for specifyingand forecasting ocean wave spectra. David Taylor Naval Research and Development Center,DTNSRDC-82-011, Bethesda, MD, USA.Pierson, W. J., 1990: Examples of, reasons for, and consequences of the poor quality of wind data from ships for themarine boundary layer: implications for remote sensing. J. Geophys. Res., 95, 13313–13340.Pierson, W. J. and L. Moskowitz, 1964: A proposed spectral form for fully developed wind-seas based on thesimilarity theory of S. A. Kitaigorodskii. J. Geophys. Res., 69(24), 5181–5190.

150GUIDE TO WAVE ANALYSIS AND FORECASTINGPierson, W. J., G. Neumann, and R. W. James, 1955: Practical methods for observing and forecasting ocean wavesby means of wave spectra and statistics. US Navy Hydrographic Office Pub., 603.Pierson, W. J., L. J. Tick and L. Baer, 1966: Computer-based procedure for preparing global wave forecasts and windfield ananysis capable of using wave data obtained by a spacecraft. Proc. Sixth Naval Hydrodynamics Symp.,Washington, 499–533.Pontes, M. T., Athanssoulis, S. F. Barstow, L. Cavaleri, B. Holmes, D. Mollison and H. Oliveira-Peres, 1995: Proc14th Int. Offshore Mechanics and Arctic Eng. Conf. (OMAE'95), Copenhagen, Denmark, June 1995, 155–161.Resio, D. T., 1978: Some aspects of extreme wave prediction related to climatic variation. Proc. 10th Offshore Tech.Conf., Houston, Texas, III, OTC 3278.Resio, D. T., W. Perrie, S. Thurston and J. Hubbertz, 1992: A generic third-generation wave model: AL. Proc. 3rdInternational Workshop on Wave Hindcasting and Forecasting, Montreal, Que., 102-114.Richards, T. L., H. Dragert and D. R. MacIntyre, 1966: Influence of atmospheric stability and over-water fetch on thewinds over the lower Great Lakes. Monthly Weather Review, 94, 448–453.Richardson, L. F., 1922: Weather prediction by numerical process. Cambridge University Press, Reprinted Dover1965.Ris, R. C., L. H. Holthuijsen and N. Booij, 1994: A spectral model for waves in the near shore zone, Proc. 24thCoastal Engineering Conference, October, 1994, Kobe, Japan, ASCE, 68–78.Rivero, O. R., M. L. Lloret and E. A. Collini, 1974: Pronostico automatico de olas en el Atlantico sur. Boletin delServicio de Hidrografia Naval, Vol X, No. 2Rivero, O. R., C. E. del Franco Ereno and G. and N. Possia, 1978: Aleunos resultados de la aplicacion de unprognostico de olas en el Atlantico sur. Meteorologica, Vol VIII, IX, 331–340.Roll, H. U. 1965: Physics of the marine atmosphere, Academic PressRomeiser, R., 1993: Global validation of the wave model WAM over a one-year period using GEOSAT wave heightdata. J. Geophys. Res., 98(C3), 4713–4726.Rye, H., 1977: The stability of some currently used wave parameters. Coastal Eng., 1, 17–30.Sanders, F. 1990: Surface analysis over the oceans – searching for sea truth. Weather and Forecasting, 5, 596–612.Sarpkaya, T. and M. Isaacson, 1981: Mechanics of wave forces on offshore structures. Van Nostrand Reinhold Co.Schäffer, H. A., R. Deigaard and P. Madsen, 1992: A two-dimensional surf zone model based on the Boussinesqequations, Proc. 23rd Coastal Engineering Conference, 576–589.Schwab, D. J., 1978: Simulation and forecasting of Lake Erie storm surges. Monthly Weather Review, 106,1476–1487.Sela, J. G., 1982: The NMC spectral model. NOAA Technical Report NWS 30.Shearman, E. D. R., 1983: Radio science and oceanography. Radio Sci., 18(3), 299–320.Shearman, R. J., 1983: The Meteorological Office main marine data bank. Marine Observer, 53, 208–217.Shearman, R. J. and A. A. Zelenko 1989: Wind measurements reduction to a standard level. WMO MarineMeteorology and Related Oceanographic Activities Report, 22, WMO/TD-311.Shemdin, O. H., K. Hasselmann, S. V. Hsiao, and K. Herterich, 1978: Non-linear and linear bottom interaction effectsin shallow water. In: Turbulent fluxes through the sea surface; wave dynamics and prediction. Plenum Press,347–372.Shemdin, O. H., S. V. Hsiao, H. E. Carlson, K. Hasselman and K. Schulze, 1980: Mechanisms for wave transformationin finite depth water, J. Geophys. Res., 85(C9), 5012–5018.Ship and Ocean Foundation, 1990, 1991. Study of ocean waves and related surface winds for ship-building. Tokyo,Japan.Simmons, A. J., D. M. Burridge, M. Jarraud, C. Girard and W. Wergen, 1988: The ECMWF medium-range predictionmodels. Development of the numerical formulations and the impact of the increased resolution. Meteor. Atmos.Phys, 40, 28–60.

REFERENCES 151Slutz, R. J., S. J. Lubker, J. D. Hiscox, S. D. Woodruff, R. L. Jenne, D. H. Joseph, P. M. Steurer and J. D. Elms, 1985.Comprehensive Ocean-Atmosphere Data Set (COADS), Release 1, NOAA/ERL, Boulder, Colorado.Smith, R. L., 1984: Threshold methods for sample extremes. In: Statistical extremes and applications. J. Tiago deOliveira (Ed.). Proc. NATO Advanced Study Inst. Conf., Vimeiro, Portugal 1983. D. Reidel, Dordrecht, 621–638.Snodgrass, F. E., G. W. Groves, K. F. Hasselmann, G. R. Miller, W. H. Munk, and W. H. Powers, 1966: Propagationof ocean swell across the Pacific. Phil. Trans. Royal Soc. A, 259, 431–497.Snyder, R. L., and C. S. Cox, 1966: A field study of the wind generation of ocean waves. J. Marine Res., 24(2),141–178.Snyder, R. L., F. W. Dobson, J. A. Elliott and R. B. Long, 1981: Array measurements of atmospheric pressure fluctuationsabove surface gravity waves. J. Fluid Mech., 102, 1–59.Sommerfeld, A., 1896: Mathematische Theorie der Diffraction. Math. Ann., 47, 317-374 (from Wiegel, 1964).Spillane, K. T. and P. E. Dexter, 1976: Design waves and wind in the Australian tropics. Australian Meteor. Mag.,24(2), 37–58.Stanton, B. R., 1984: Return wave heights off South Uist estimated from seven years of data. Institute ofOceanographic Sciences, 164, 54 pp.Stull, R. B. 1988: An introduction to boundary layer meteorology. Kluwer Academic Publishers. Dortrecht, Boston,London, 666 pp.Sverdrup, H. U. and W. H. Munk, 1947: Wind, sea and swell: theory of relations for forecasting. H.O. Pub. 601, USNavy Hydrographic Office, Washington, D.C., 44 pp.SWAMP Group, 1985: Ocean wave modelling. Plenum Press, 256 pp.Taylor, P. K., E. Kent, M. Yelland and B. Moat, 1994: The accuracy of wind observations from ships. Proc.International COADS Wind Workshop, Kiel, Germany, May 1994, 132–151.Thomas, J. P., 1988: Retrieval of energy spectrum from measured data for assimilation into a wave model. Quart. J.Royal Met. Soc., 114, 781–800.Thomasell, A. and J. G. Welsh, 1963: Studies of the specification of surface winds over the ocean. Travelers ResearchCenter, IMC, 7049–88.Tiedtke, M. et al., 1979: ECMWF model parameterization of subgrid scale processes. ECMWF Technical Report, 10.Tolman, H. L., 1990: Wind wave propagation in tidal seas. Doctoral Thesis, Delft Univ. Tech.Tolman, H. L., 1994: Wind-waves and moveable-bed bottom-friction. J. Phys. Oceanogr., 24(5), 994.Tournardre, J. and R. Ezraty 1990: Local climatology of wind and sea state by means of satellite radar altimetermeasurements. J. Geophys. Res., 95, 18255–18268.Tucker M. J., 1956: A shipborne wave recorder. Trans. Inst. Nav. Archt., London, 98(1), 236–250.Tucker, M. J., 1989: An improved “Battjes” method for predicting the probability of extreme waves. Applied OceanResearch, 4, 212–218.Tucker, M. J. 1991: Waves in ocean engineering: measurement, analysis, interpretation. Ellis Horwood, Chichester,UK. 431 pp.Tucker, M. J. 1993: Recommended standard for wave data sampling and near-real-time processing. OceanEngineering, 20, 459–474.Uji, T., 1984: A coupled discrete wave model MRI-II. J. Oceanographical Society of Japan, 40, 303–313.US Navy, 1983: Hindcast spectral ocean wave model climatic atlas: North Atlantic Ocean. Prepared by NavalOceanography Command Detachment, Asheville, NC, USA.US Navy, 1985: Hindcast spectral ocean wave model climatic atlas: Pacific Ocean. Prepared by Naval OceanographyCommand Detachment, Asheville, NC, USA.US Navy 1990: Hindcast spectral ocean wave model climatic atlas: Mediterranean Sea. Prepared by NavalOceanography Command Detachment, Asheville, NC, USA.

152GUIDE TO WAVE ANALYSIS AND FORECASTINGVartdal, L., H. E. Krogstad and S. F. Barstow, 1989: Measurement of wave properties in extreme seas during theWADIC experiment. Proc. 21st Offshore Tech. Conf., Houston, Texas, May 1989. 73–82.Verploegh, G., 1967: Observation and analysis of the surface wind over the ocean. KNMI, Mededelingen enVerhandelingen, 89.Visser, P. J., 1984: Uniform longshore current measurements and calculations. Proc. 19th Coastal EngineeringConference, Houston, 2l92–2207.Vogel, J. A., A. C. Radder and J. H. de Reus, 1988: Verification of numerical wave propagation models in tidal inlets,Proc. 21st Coastal Engineering Conference, 433–447.WAMDI Group (S. Hasselmann, K. Hasselmann, E. Bauer, P. A. E. M. Janssen, G. Komen, L. Bertotti, P. Lionello,A. Guillaume, V. C. Cardone, J. A. Greenwood, M. Reistad, L. Zambresky, J. A. Ewing), 1988: The WAMmodel – a third generation wave prediction model, J. Phys. Oceanogr., 18, 1775–1810.Ward, E. G., L. E. Borgman and V. J. Cardone, 1978: Statistics of hurricane waves in the Gulf of Mexico. Proc. 10thOffshore Tech. Conf., Houston, Texas, III, 1523–1536 (OTC 3229). (Also in J. of Petroleum Technology, 1979,632–642.)Wiegel, R. L., 1964: Oceanographical engineering. Prentice Hall, Englewood Cliffs, N.J.Wilkerson, J. C. and M. D. Earle, 1990: A study of differences between environmental reports by ships in the voluntaryobserving program and measured from NOAA buoys. J. Geophys. Res., 95, 3373–3385.Witter, D. L. and D. B. Chelton, 1991: A GEOSAT altimeter wind speed algorithm and a method for altimeter windspeed algorithm development. J. Geophys. Res., 96, 8853–8860.WMO, 1970: The Beaufort scale of wind force. Marine Science Affairs, 3 (out of print).WMO, 1976: Handbook on wave analysis and forecasting. WMO No. 446 (out of print).WMO, 1985, 1991, 1994(a): National reports on wave measuring techniques, numerical wave models and intercomparison.WMO Marine Meteorology and Related Oceanography Activities Report 12, & Suppls. 3, 4. WMOWave Programme, WMO/TD-No. 35.WMO, 1988: Guide to wave analysis and forecasting. 1st edition. WMO No. 702, Geneva, Switzerland.WMO, 1990: Manual on marine meteorological services. 2nd edition. WMO No. 558, Geneva, Switzerland.WMO, 1994(b): Guide to the applications of marine climatology. WMO No. 781, Geneva, Switzerland.WMO, 1995: Manual on codes. WMO No. 306.WMO, 1996: Guide to meteorological instruments and methods of observation. WMO No. 8.Wu, J., 1980: Wind-stress coefficients over sea surface near neutral conditions: a revisit. J. Phys. Oceanogr., 10,727–740.Wu, J., 1982: Wind-stress coefficients over sea surface from breeze to hurricane. J. Geophys. Res., 87, 9704–9706.Wyatt, L. R. and G. Holden, 1994: Limits in direction and frequency resolution for HF radar ocean wave directionalspectra measurements. The Global Atmosphere-Ocean System, 2, 265–290.Wyatt, L. R., J. Venn, M. D. Moorhead, G. B. Burrows, A. M. Ponsford and J. van Heteren, 1985: HF radar measurementsof significant wave height and mean period during NURWEC. In: Evaluation, comparison andcalibration of oceanographic instruments. Advances in underwater technology and offshore engineering.Graham and Trotman, London, 209–222.Young, I. R. and G. J. Holland, 1996: Atlas of the oceans: wind and wave climate. Pergamon, 246 pp + CD-ROM.

SELECTED BIBLIOGRAPHYCERC, 1984: Coastal Engineering Research Center, 1973, 1977, 1984: Shore protection manual. WaterwaysExperiment Station, US Army Corps of Engineers.Crapper, G. D., 1984: Introduction to water waves. Publ. Ellis Harwood, 224 pp.Hogben, N., N. M. C. Dacunha and G. F. Olliver, 1986: Global wave statistics. Unwin Brothers for British MaritimeTech. Ltd.Kinsman, B., 1965: Wind waves, their generation and propagation on the ocean surface. Prentice-Hall Inc.,Englewood Cliffs, New Jersey. Reprinted by Dover Publ. 1984, 676 pp.Khandekar, M. L., 1989: Operational analysis and prediction of ocean wind waves. Coastal and Estuarine Studies No.33, Springer-Verlag, New York, 214 pp.Kitaigorodskii, S. A., 1970: The physics of air sea interaction. Gidrometeorologicheskoe Izdatel'stvo, Leningrad,(translated from Russian). Israel Programme for Scientific Translations, Jerusalem.Komen, G. J., L. Cavaleri, M. Donelan, K. Hasselmann, S. Hasselmann and P. A. E. M. Janssen, 1994: Dynamics andmodelling of ocean waves. Cambridge University Press, 532 pp.Lighthill, J., 1978: Waves in fluids. Cambridge University Press, 504 pp.Massel, S. R., 1994: Ocean waves: their physics and prediction. World Scientific Publ., Advanced Series in OceanEngineering, 350 pp.Pierson, W. J., G. Neumann, and R. W. James, 1955: Practical methods for observing and forecasting ocean wavesby means of wave spectra and statistics. US Navy Hydrographic Office Pub., 603, 284 pp.Stoker, J. J., 1957: Water waves. Interscience, New York, 567 pp.Tucker, M. J. 1991: Waves in ocean engineering: measurement, analysis, interpretation. Ellis Horwood, Chichester,UK. 431 pp.Young, I. R. and G. J. Holland, 1996: Atlas of the oceans: wind and wave climate. Pergamon, 246 pp + CD-ROM.

INDEXPage(s)Page(s)Accelerometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92Action density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40Amplitude . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2Analysiskinematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115manual . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15numerical . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15of extremes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118of pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16of wave data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59of waves, definition . . . . . . . . . . . . . . . . . . . . . . . . . viiiof wind . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16, 115Angular spread (see alsodirectional spread) . . . . . . . . . . . . . . . . . . . . . . . . . . 37reduction factor . . . . . . . . . . . . . . . . . . . . . . . 37, 38, 50Archives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100Assimilationof remotely-sensed data . . . . . . . . . . . . . . . . . . . . . . 99of wave data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59Atlas of waves . . . . . . . . . . . . . . . . . . . . . . . . . . . 100, 110Beaufort scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17, 19Boundary layer (see also constant fluxlayer, Ekman) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27forecasts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31–32parameterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27Boussinesq equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 85Bragg scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93, 94Breaking waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5laboratory measurements . . . . . . . . . . . . . . . . . . 87, 88surf zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87Buoypitch-roll . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92surface following . . . . . . . . . . . . . . . . . . . . . . . . . . . 92Caustic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83Chartswave analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69weather . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18wind . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17Climate/climatology . . . . . . . . . . . . . . . . . . . .101, 109, 111seasonal variation . . . . . . . . . . . . . . . . . . . . . . . . . . 104wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103, 111, 113applications of . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109CodesGRID . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69WAVEOB . . . . . . . . . . . . . . . . . . . . . . . . . 69, 100, 123Constant flux layer . . . . . . . . . . . . . . . . . . . . . . . 21, 27, 30Coriolis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23, 27Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39long-shore . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88meters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93wave driven . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88Databaseshindcast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116, 118measurements, observations . . . . . . . . . . . . . . 110, 112Data records . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8, 11, 90completeness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105long-term . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112–113Depth effects (see also diffraction, dispersion,dissipation, propagation) . . . . . . . . . . . . . . . . . . . . . . 4deep water definition . . . . . . . . . . . . . . . . . . . . . . . . . 4transitional depth definition . . . . . . . . . . . . . . . . . . . . 4Diffraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5, 6, 84, 97Directional spectrumfrom buoys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92from HF radar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98from models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67, 69from SAR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96Directional spread (see also angular spread)in wave estimation . . . . . . . . . . . . . . . . . . . . . . . . . . 50Dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2factor, in manual wave estimation . . . . . 38–39, 50, 51in finite-depth water . . . . . . . . . . . . . . . . . . . . . . . 4, 36Dissipation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4of wave energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39bottom friction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86wave breaking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87whitecapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39Distribution (see also probabilitydistribution)of wave parameters . . . . . . . . . . . . . . . . . . . . . . . . . 103joint, height and period . . . . . . . . . . . . . . . . . . . . . . 103of extreme height . . . . . . . . . . . . . . . . . . . . . . . . . . 118of spectral estimates . . . . . . . . . . . . . . . . . . . . . . . . . 12Dorrestein’s method for refraction . . . . . . . . . . . . . . . . . 54Drag coefficient . . . . . . . . . . . . . . . . . . . . . . 15, 30, 31, 36Durationcorrected . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49equivalent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46Echo sounders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91ECMWF (European Centre for Medium-Range Weather Forecasts) . . . . . . . . . . . . . . . . . . . . 32Ekmanlayer in atmosphere . . . . . . . . . . . . . . . . . . . . 26, 27, 30spiral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

156GUIDE TO WAVE ANALYSIS AND FORECASTINGPage(s)Page(s)Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3density spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57dissipation by bottom friction . . . . . . . . . . . . . . . . . . 86equipartition of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3kinetic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3, 8potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3, 8speed of propagation . . . . . . . . . . . . . . . . . . . . . . . . . . 8total . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3transport of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81, 82Energy balanceequation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58in refraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82in shoaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81source terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58, 84wave breaking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90Errors in wave data . . . . . . . . . . . . . . . . . . . . . . . . 90, 103Estimators, of distribution parameters . . . . . . . . . 133–138Exceedance plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104Exponential distribution . . . . . . . . . . . . . . . . . . . . . . . . 135Extreme height distribution . . . . . . . . . . . . . . . . . . . . . 118Extreme valuedistribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136extremal analysis . . . . . . . . . . . . . . . . . . . . . . 105, 118Fetch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36effective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46Finite-diffference propagation method . . . . . . . . . . . . . 60Fisher-Tippett distribution . . . . . . . . . . . . . . . . . . 106, 137Forecast of waves, definition . . . . . . . . . . . . . . . . . . . . viiiFrequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2angular . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2peak . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13, 40Frequency spectrum (see also wave spectrum) . . . . . . . 11evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42Friction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27bottom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27of sea-surface, effect on wind . . . . . . . . . . . . . . . . . . 27velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30, 35, 59Fully developed sea . . . . . . . . . . . . . . . . . . . . . . 13, 59, 63Gamma distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . 134Gaussian distribution . . . . . . . . . . . . . . . . . . . . . . . 10, 133Gravity waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1Groen and Dorrestein . . . . . . . . . . . . . . . . . . . . . . . . . . . 43growth curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36Group velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . 8, 36, 53deep water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8finite depth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8, 36Growth (see wave growth)Hindcast . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii, 100, 111databases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116, 118procedures for . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114verification of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115Huygens’ principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85Incompressibility of fluid . . . . . . . . . . . . . . . . . . . . . . . . . 1Interactions (see wave interactions)Inviscid fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1Irrotational fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1Joint distribution, height and period . . . . . . . . . . . . . . 103JONSWAPgrowth formulae . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45, 50Linear waves (see also simple waves) . . . . . . . . . . . . . . . 1Log-normal distribution . . . . . . . . . . . . . . . . 106, 108, 134Long-crested waves . . . . . . . . . . . . . . . . . . . . . . . . . . . 1, 6Maximum likelihood method . . . . . . . . . . . . . . . . . . . . 108Measurements of waves . . . . . . . . . . . . . . . . . . . . . 91, 110from buoys . . . . . . . . . . . . . . . . . . . . . . . . . 20, 92, 111from current meters . . . . . . . . . . . . . . . . . . . . . . . . . 93from platforms . . . . . . . . . . . . . . . . . . . . . . . . . . 92, 94from pressure transducers . . . . . . . . . . . . . . . . . . . . . 91from radar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93from satellites . . . . . . . . . . . . . . . . . . . . . . . . . . . 93, 95from ships . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16, 111from wave staffs, arrays . . . . . . . . . . . . . . . . . . . 92, 93long-term records . . . . . . . . . . . . . . . . . . . . . . . . . . 112Moments, method of . . . . . . . . . . . . . . . . . . . . . . . . . . 107Moments of spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . 12Microwaveradar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94radiometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96Mild-slope equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85MIROS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94Model Output Statistics, method . . . . . . . . . . . . . . . . . . 33Monin-Obukhov . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21, 30mixing length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30Non-linear interactions (see waveinteractions)Normal distribution . . . . . . . . . . . . . . . . . . . . . . . . 10, 133Numerical weather prediction . . . . . . . . . . . . . . . . . . . . 15Observations (see also measurements)from coastal stations . . . . . . . . . . . . . . . . . . . . . . . . . 20from satellites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21from ship . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16, 111of waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110of waves, databases . . . . . . . . . . . . . . . . . . . . . . . . . 112of waves, errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90of waves, visual . . . . . . . . . . . . . . . . . . . . . 43, 89, 111Orbital motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3, 4Overshoot phenomenon . . . . . . . . . . . . . . . . . . . . . . 40, 41Pareto distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138Peaks over threshold analysis . . . . . . . . . . . . . . . . . . . 106Perfect prognosis method . . . . . . . . . . . . . . . . . . . . . . . . 32Phase speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2Phillips’ resonance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35Pierson-Moskowitz spectrum . . . . . . . . . . . . . . . . . . 13, 59Plot (see also charts)of wave data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103PNJ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43, 139Poisson distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . 109Prandtl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30Pressure transducers . . . . . . . . . . . . . . . . . . . . . . . . . 91, 93

INDEX 157Page(s)Probability distribution (see alsodistribution) . . . . . . . . . . . . . . . . . . . . . . 103, 106, 133estimation by maximum likelihood . . . . . . . . . . . . 108estimation by method of moments . . . . . . . . . . . . . 107estimation using probability paper . . . . . . . . . . . . . 106exponential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135Fisher-Tippett . . . . . . . . . . . . . . . . . . . . . . . . . 106, 137Gamma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134Gaussian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10, 133generalized extreme value . . . . . . . . . . . . . . . 106, 136Generalized Pareto . . . . . . . . . . . . . . . . . . . . . . . . . 138log-normal . . . . . . . . . . . . . . . . . . . . . . . 106, 108, 134Normal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10, 133Poisson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109Rayleigh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10, 136wave height . . . . . . . . . . . . . . . . . . . . . . . 106, 118, 133Weibull . . . . . . . . . . . . . . . . . . . . . . . . . . 106–108, 135Probability of storm . . . . . . . . . . . . . . . . . . . . . . . . . . . 109Probability paper . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106Propagationfinite-diffference method . . . . . . . . . . . . . . . . . . . . . 60in shallow water . . . . . . . . . . . . . . . . . . . . . . . . . 62, 84numerical methods . . . . . . . . . . . . . . . . . . . . . . . . . . 60past obstacles . . . . . . . . . . . . . . . . . . . . . . . . . . . 84, 85ray tracing method . . . . . . . . . . . . . . . . . . . . . . . . . . 61sprinkler effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61of wave energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60Radaraltimeter . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21, 93–95Bragg scattering . . . . . . . . . . . . . . . . . . . . . . . . . 93, 94from platforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94ground-wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98high frequency (HF) . . . . . . . . . . . . . . . . . . . . . . . . . 97modulation of backscatter . . . . . . . . . . . . . . . . . . . . . .94SAR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95–97sky-wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98specular reflection . . . . . . . . . . . . . . . . . . . . . . . . . . 93Radiation stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88Ray tracing propagation method . . . . . . . . . . . . . . . . . . 61Rayleigh distribution . . . . . . . . . . . . . . . . . . . 10, 118, 136Refraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5, 82factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54manual computation . . . . . . . . . . . . . . . . . . . . . . . . . 54Remote sensing of waves . . . . . . . . . . . . . . . . . . . . . . . . 93Return value, wave height . . . . . . . . . . .101, 105, 108, 10950-year . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105in tropical storms . . . . . . . . . . . . . . . . . . . . . . . . . . 109Reynolds’ stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27Rossby similarity theory . . . . . . . . . . . . . . . . . . . . . . . . 31Roughness length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30Satellite (see also measurements, observations, radar)coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98wave data . . . . . . . . . . . . . . . . . . . . . . . . . . . 59, 94, 95wind data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21, 95Saturation of wave energy . . . . . . . . . . . . . . . . . . . . . . . 63Scatter index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118Scatter plot . . . . . . . . . . . . . . . . . . . . . . . . . . 103, 104, 109Page(s)Scatterometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21, 95Sea state (see also swell, wave field, wind waves)developing (JONSWAP) . . . . . . . . . . . . . . . . . . . . . . 14fetch-limited . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14fully developed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13Self-similarity in wave spectrum . . . . . . . . . . . . 13, 60, 64Set-down . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88Set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88Shallow water waves . . . . . . . . . . . . . . . . . . . . . . . . . . . 81bottom friction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4factors, manual computation of . . . . . . . . . . . . . . . . 52growth curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86wave growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85Shipborne Wave Recorder . . . . . . . . . . . . . . . . . . . . . . . 92Shoaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5, 81factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52Short-crested waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84Simple sinusoidal waves . . . . . . . . . . . . . . . . . . . . . . . . . 1superposition of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7Snell’s law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54, 55, 84and refraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82Source function in wave growth . . . . . . . . . . . . . . . . . . 58Sprinkler effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61SSM/I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21Stability in atmospheric boundarylayer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15, 21, 27, 29Standard normal distribution . . . . . . . . . . . . . . . . . . . . 133Stationary sea-state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8Statistics of waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101Steepness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2, 5, 9significant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103Stokes’ wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5, 6Stresssurface wind . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15, 30Surf zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87wave evolution in . . . . . . . . . . . . . . . . . . . . . . . . . . . 88Swell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2, 6, 36computation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47shoaling and refraction . . . . . . . . . . . . . . . . . . . . . . . 52spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11, 39Synthetic Aperture Radar (SAR) . . . . . . . . . . . . . . . 95–97Trochoidal wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6Tsunamis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1, 4Tucker-Draper method . . . . . . . . . . . . . . . . . . . . . . . . . . . 8Turbulence in atmospheric boundary layer . . . . . . . . . . 27Uniform distribution . . . . . . . . . . . . . . . . . . . . . . . . . . 109Velocity potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1Voluntary Observing Ships . . . . . . . . . . . . . . . . . . 90, 110Waveaction density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40age . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36breaking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101Wave (contd.)propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

158GUIDE TO WAVE ANALYSIS AND FORECASTINGPage(s)steepness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39Wave chart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69of heights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68, 71Wave climate (see also climate/climatology) . . . . . . . . . .viiiWave dataarchives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100assimilation of satellite data . . . . . . . . . . . . . . . . 80, 98errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90, 103from satellites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59in verification of models . . . . . . . . . . . . . . . . . . . . . . 73Wave direction (see also wave spectrum,directional spectrum)coastal observation of . . . . . . . . . . . . . . . . . . . . . . . . 91from satellites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93observation from ships . . . . . . . . . . . . . . . . . . . . . . . 90primary direction . . . . . . . . . . . . . . . . . . . . . . . . 69, 70visual observations . . . . . . . . . . . . . . . . . . . . . . . . . . 90Wave energy (see also energy)conservation equation . . . . . . . . . . . . . . . . . . . . . . . . 35propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36redistribution in spectrum (wave-waveinteractions) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40saturation of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63Wave forecasting (see also wave model) . . . . . . . . . . . . 35manual methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43Wave frequencypeak downshift . . . . . . . . . . . . . . . . . . . . . . . . . . 40, 41peak . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13range of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48Wave groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8group velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . 36, 53Wave growthcurves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43exponential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35, 58formulation of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35fully developed sea . . . . . . . . . . . . . . . . . . . . . . . . . . 63non-linear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41, 60overshoot phenomenon . . . . . . . . . . . . . . . . . . . . 40, 41Phillips’ resonance . . . . . . . . . . . . . . . . . . . . . . . 35, 59shallow water . . . . . . . . . . . . . . . . . . . . . . . . 56, 85, 86source function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58spectral structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41with fetch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40Wave height . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2, 101average . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9characteristic . . . . . . . . . . . . . . . . . . . . . . . . . 12, 36, 43coastal observation of . . . . . . . . . . . . . . . . . . . . . . . . 91combined . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69, 90distribution of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9exceedance plot of . . . . . . . . . . . . . . . . . . . . . . . . . 104extremes from satellite data . . . . . . . . . . . . . . . . . . 110from HF radar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98from satellites . . . . . . . . . . . . . . . . . . . . . . . 76, 94, 110global distribution . . . . . . . . . . . . . . . . . . . . . . . . . . 110Wave height (contd.)maximum . . . . . . . . . . . . . . . . . . . . . . . . 9, 10, 45, 101Page(s)observation from ships . . . . . . . . . . . . . . . . . . . . . . . 90return value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108root-mean-square . . . . . . . . . . . . . . . . . . . . . . . . . . . 12significant . . . . . . . . . . . . . . . . . . . 9, 12, 43, 67, 69, 99tropical storms, return value . . . . . . . . . . . . . . . . . . 109visual observations . . . . . . . . . . . . . . . . . . . . . . . . . . 89Wave interactionsnon-linear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39non-linear approximations . . . . . . . . . . . . . . . . . . . . 65resonance condition . . . . . . . . . . . . . . . . . . . . . . . . . 40triads in shallow water . . . . . . . . . . . . . . . . . 84, 87, 88wind sea and swell . . . . . . . . . . . . . . . . . . . . . . . . . . 64Wave modelclassification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62coupled discrete . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65coupled hybrid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63decoupled propagation . . . . . . . . . . . . . . . . . . . . . . . 63elements of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58in shallow water . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58measures of skill . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73operational . . . . . . . . . . . . . . . . . . . . . . . . . . 67, 77–79parametric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64performance of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74spectral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62, 69third generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65tuning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57use of wave data . . . . . . . . . . . . . . . . . . . . . . . . . 80, 98verification of . . . . . . . . . . . . . . . . . . . . . . . 66, 70, 115WAM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65Wave observations (see measurements, observations)Wave peakedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13Wave period . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1, 2average, mean . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9, 13characteristic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43coastal observation of . . . . . . . . . . . . . . . . . . . . . . . . 91estimation of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108observation from ships . . . . . . . . . . . . . . . . . . . . . . . 90peak . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13significant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9, 43visual observations . . . . . . . . . . . . . . . . . . . . . . . . . . 89zero-crossing . . . . . . . . . . . . . . . . . . . . . . . . . 9, 13, 103Wave power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13Wave rayequation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82refraction pattern . . . . . . . . . . . . . . . . . . . . . . . . . . . 83Wave record (see also data records) . . . . . . . . . . . . . . 8, 9analysis of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8, 98duration of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9Wave spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10, 11component in . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10depth-limited (TMA) . . . . . . . . . . . . . . . . . . . . . 14, 86directional . . . . . . . . . . . . . . . . . . . . . . . . . . . 14, 67, 72directional distribution . . . . . . . . . . . . . . . . . . . . . . . 37energy density . . . . . . . . . . . . . . . . . . . . . . . . . . . 12, 57Wave spectrum (contd.)frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

INDEX 159Page(s)from SAR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96JONSWAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13moments of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12peak of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9peak enhancement . . . . . . . . . . . . . . . . . . . . . . . . . . 14Phillips’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13Pierson-Moskowitz . . . . . . . . . . . . . . . . . . . . . . . 13, 59sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12self-similar shape of . . . . . . . . . . . . . . . . . . . . . . 64, 85swell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11variance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11, 12width of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13wind-sea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11Wave speed (see also group velocity,propagation)deep water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4finite depth water . . . . . . . . . . . . . . . . . . . . . . 4, 62, 81group velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36shallow water . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4, 81Wave staffs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91, 93Wave trains, multiple observations of . . . . . . . . . . . . . . 90Wavelength . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2Wavenumber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2, 53WAVEOB code . . . . . . . . . . . . . . . . . . . . . . . . 69, 100, 123Waverider buoy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92Weather prediction, numerical models . . . . . . . . . . . . . 15Weibull distribution . . . . . . . . . . . . . . . . . . . 106–110, 135Windaccuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19, 20, 21at sea surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15Wind (contd.)Page(s)confluence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26, 27difluence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26, 27direction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19, 45effect of errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15effective neutral . . . . . . . . . . . . . . . . . . . . . . . . . . . 115equivalent neutral . . . . . . . . . . . . . . . . . . . . . . . . . . . 31error in measurement . . . . . . . . . . . . . . . . . . . . . . . . 20estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15extrapolation from land to water . . . . . . . . . . . . . . . 32for wave models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59freshening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45, 47friction velocity . . . . . . . . . . . . . . . . . . . . . . . . . 35, 59from microwave radiometer . . . . . . . . . . . . . . . . . . . 96from satellites . . . . . . . . . . . . . . . . . . . . . 21, 80, 95, 96geostrophic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23, 24gradient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23, 25height adjustment . . . . . . . . . . . . . . . . . . . . . . . . . . . 21isallobaric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .26kinematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115profile . . . . . . . . . . . . . . . . . . . . . . . . . . . 21, 27, 29, 30regional models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32scatterometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21, 95shear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23, 26slackening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45SSM/I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21, 22statistical models for . . . . . . . . . . . . . . . . . . . . . . 32, 33thermal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23, 27, 28turning, effect on wave growth . . . . . . . . . . . . . . . . . 62Wind waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1computation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45Zero-crossing period . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9Zero-downcrossings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8