Double Difference Earthquake Location(hypoDD) - Geophysics at ...
Double Difference Earthquake Location(hypoDD) - Geophysics at ...
Double Difference Earthquake Location(hypoDD) - Geophysics at ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Loc<strong>at</strong>ing <strong>Earthquake</strong>s Using HypoDD<br />
Hongfeng Yang<br />
October 25, 2013<br />
1 Equ<strong>at</strong>ions<br />
Loc<strong>at</strong>ing earthquake is a classic but still active research subject. Accuracy of loc<strong>at</strong>ion results depends<br />
on several factors, such as the st<strong>at</strong>ion coverage, the number of available phases, and the<br />
velocity model used [Pavlis, 1986]. Rel<strong>at</strong>ive loc<strong>at</strong>ion methods can effectively reduce the uncertainties<br />
due to structure and thus give more reliable results [Pavlis, 1992; Waldhauser and Ellsworth,<br />
2000].<br />
The arrival time difference ∆t of two events <strong>at</strong> one st<strong>at</strong>ion can be written as<br />
∆t = − −→ p · ∆ −→ x + ∆o, (1)<br />
where −→ p is the slowness vector, ∆ −→ x represents the separ<strong>at</strong>ion of the events, and ∆o stands for the<br />
origin time difference between the two events. Here we assume th<strong>at</strong> the sp<strong>at</strong>ial separ<strong>at</strong>ion between<br />
the two events is much smaller than the event-st<strong>at</strong>ion distance, so th<strong>at</strong> the ray p<strong>at</strong>hs between the<br />
sources and the st<strong>at</strong>ion are similar and the effect of l<strong>at</strong>eral structural vari<strong>at</strong>ion on ∆t can be ignored<br />
[Waldhauser and Ellsworth, 2000]. In other words, the −→ p is same.<br />
If we have many observ<strong>at</strong>ions, equ<strong>at</strong>ion (1) can be expressed as<br />
∆t i = − −→ p i · ∆ −→ x + ∆o<br />
= −(p xi ∆x + p yi ∆y + p zi ∆z) + ∆o,<br />
(2)<br />
where i stands for the ith st<strong>at</strong>ion. We re-write equ<strong>at</strong>ion (2) into m<strong>at</strong>rix form<strong>at</strong><br />
⎛ ⎞ ⎛<br />
⎞<br />
∆t 1 −p x1 −p y1 −p z1 1 ⎛ ⎞<br />
∆t 2<br />
⎜ ...<br />
⎟<br />
⎝ ... ⎠<br />
= −p x2 −p y2 −p z2 1<br />
∆x<br />
⎜ ... ... ... ...<br />
⎜∆y<br />
⎟<br />
⎟ ⎝<br />
⎝ ... ... ... ... ⎠ ∆z⎠ , (3)<br />
∆o<br />
∆t n −p xn −p yn −p zn 1<br />
or<br />
d = Gm. (4)<br />
In most cases the number of observ<strong>at</strong>ions n will be gre<strong>at</strong>er than the number of unknown parameters<br />
(4 in this case) so th<strong>at</strong> equ<strong>at</strong>ion 3 is overdetermined. If the number of differential times is not too<br />
large, we use the singular value decomposition to obtain the solution<br />
m = V Λ −1 U T d, (5)<br />
where U is the m<strong>at</strong>rix of eigenvectors th<strong>at</strong> spans the d<strong>at</strong>a space, V is the m<strong>at</strong>rix of eigenvectors<br />
th<strong>at</strong> spans the model parameter space, Λ is a diagonal m<strong>at</strong>rix whose elements are non-neg<strong>at</strong>ive<br />
singular values of G.<br />
1
2 HypoDD<br />
HypoDD is a Fortran77 computer program package for reloc<strong>at</strong>ing earthquakes with the doubledifference<br />
algorithm of Waldhauser and Ellsworth [2000]. The l<strong>at</strong>est version of HypoDD is version<br />
1.3 and is available from http://www.ldeo.columbia.edu/˜felixw/<strong>hypoDD</strong>.html.<br />
This tutorial follows instructions on http://geophysics.eas.g<strong>at</strong>ech.edu/people/cwu/teaching/<strong>hypoDD</strong>/<strong>hypoDD</strong>.<br />
2.1 Input files<br />
<strong>hypoDD</strong> reads input parameters and files from the <strong>hypoDD</strong>.inp, which contains general inform<strong>at</strong>ion<br />
such st<strong>at</strong>ion, DD time, output file, and etc. It also includes a few important parameters th<strong>at</strong> are<br />
used during inversion.<br />
1. OBSCC and OBSCT<br />
OBSCC and OBSCT define the minimum number of links between event pair for cross correl<strong>at</strong>ion<br />
differetial time (CC) and c<strong>at</strong>alog differential time (CT). This parameters determine the<br />
strength of links between event pair, which is crucial for reloc<strong>at</strong>ion process. Below is quoted<br />
from the <strong>hypoDD</strong> manual.<br />
“To prevent an ill-conditioned system of DD-equ<strong>at</strong>ions, <strong>hypoDD</strong> ensures connectedness between<br />
events by grouping events into clusters, with each cluster having a chain of links from<br />
any event to any other event in the cluster. The strength of this chain is defined by a minimum<br />
number of observ<strong>at</strong>ions per event pair th<strong>at</strong> build up the chain (OBSCC, OBSCT for<br />
cross-correl<strong>at</strong>ion and c<strong>at</strong>alog d<strong>at</strong>a, respectively). Typical thresholds are similar to the number<br />
of degrees of freedom for an event pair, which is 8 (3 sp<strong>at</strong>ial and 1 time for each event). If P-<br />
and S-wave d<strong>at</strong>a is used, the threshold has to be higher to actually reach 8 st<strong>at</strong>ions per event<br />
pair. Increasing the threshold might increase the stability of the solution, but it might also<br />
split up a large cluster into subclusters. Lowering the threshold might include more events in<br />
a single cluster, but might decrease the stability of the solution. A high threshold, however,<br />
does not necessarily guarantee stable reloc<strong>at</strong>ions, as the spread of the partial deriv<strong>at</strong>ives is<br />
not taken into account directly.”<br />
2. ISTART<br />
This parameter tells <strong>hypoDD</strong> whether the intial loc<strong>at</strong>ion of events is from a centroid loc<strong>at</strong>ion,<br />
or is taken from network sources (different initial loc<strong>at</strong>ions).<br />
3. ISOLV<br />
For a small number of events, it is easy to contol the performance of <strong>hypoDD</strong> by SVD<br />
(ISOLV=1). However, it is more difficulte to assess the reliability of the loc<strong>at</strong>ions in the case<br />
of large number of events (thousands or more) th<strong>at</strong> can only be reloc<strong>at</strong>ed in LSQR mode<br />
(ISOLV=2).<br />
4. NSET and Weighting<br />
NSET is number of iter<strong>at</strong>ions during inversion. Specific<strong>at</strong>ions during each iter<strong>at</strong>ion can be<br />
defined by values of weighting and re-weighting.<br />
2
2.2 D<strong>at</strong>a form<strong>at</strong><br />
3 Waveform Cross Correl<strong>at</strong>ion<br />
Let f(t) and g(t) represent two time series, conventional non-normalized cross-correl<strong>at</strong>ion of the<br />
two is<br />
C(t) = f(t) ∗ g(t) =<br />
∫ ∞<br />
−∞<br />
f(t + τ)g(τ)dτ. (6)<br />
We obtain differetial times of each event pair <strong>at</strong> one st<strong>at</strong>ion using waveform cross correl<strong>at</strong>ion (CC).<br />
Following are the procedures to perform CC.<br />
The tutorial d<strong>at</strong>aset is loc<strong>at</strong>ed <strong>at</strong> “hyangteachingHYPODDtutsmalld<strong>at</strong>a”.<br />
1. Prepar<strong>at</strong>ion of sac files<br />
All waveform d<strong>at</strong>a are saved in sac form<strong>at</strong> named as r, t, z, where r usually refers to east<br />
component and t stands for north component. Note this is not required, but is recommended<br />
in order to use my following processing scripts. All the sac files are saved under directory<br />
named by conventional event ID (year, month, day, hour, minute, second).<br />
2. P and S arrivals<br />
All P and S arrival times need to be manually picked (usually for a small d<strong>at</strong>aset), extracted<br />
from c<strong>at</strong>alog arrivals, or estim<strong>at</strong>ed from a velocity model. In this tutorial, the P and S arrivals<br />
for all events are estim<strong>at</strong>ed from detection times based on templ<strong>at</strong>e event waveforms. They<br />
are saved in the sac header files as t 1 and t 2 , respectively. So we can simply extract them<br />
using saclst and save them into event/p.arr and event/s.arr, respectively.<br />
Then two arrival time files, sta.parr and sta.sarr, need to be built following the form<strong>at</strong> of<br />
eve/sta.z arr 1/0, where “1” indic<strong>at</strong>es good waveform and “0” represents noisy one. It is<br />
also important to check the waveform quality in this step before running waveform cross<br />
correl<strong>at</strong>ions.<br />
3. check waveform quality<br />
There is a GUI interface written by Dr. Lupei Zhu <strong>at</strong> SLU to read in waveforms and to turn<br />
it on or off using mouse. This program is named “mcc.tcl”. To use it, prepare arrival times<br />
following the above form<strong>at</strong> and type “mcc.tcl sta.parr” and check waveform quality. One can<br />
also adjust arrival times of individual traces by dragging and moving them by mouse.<br />
4. waveform cross correl<strong>at</strong>ion<br />
Once you have the sta.parr and sta.sarr, you can perform waveform cross correl<strong>at</strong>ion using<br />
“crscrl.pl”. Refer to README in the example d<strong>at</strong>aset for the usage of the perl script.<br />
5. Check cross correl<strong>at</strong>ion results<br />
Use “chkCluster.tcl t 1 /t 2 sta.cc” to check the quality of waveform cross correl<strong>at</strong>ion. Adjust<br />
the trace individually if necessary.<br />
6. Obtain differential times<br />
Use “cc2dt.pl” to obtain differential times after waveform cross correl<strong>at</strong>ion. Refer to “README”<br />
in the sample d<strong>at</strong>aset.<br />
3
References<br />
Pavlis, G. L., Appraising earthquake hypocenter loc<strong>at</strong>ion errors: a complete practical approach for<br />
single-event loc<strong>at</strong>ions, Bull. Seismol. Soc. Am., 76, 1699–1717, 1986.<br />
Pavlis, G. L., Appraising rel<strong>at</strong>ive earthquake loc<strong>at</strong>ion errors, Bull. Seismol. Soc. Am., 82, 836–859,<br />
1992.<br />
Waldhauser, F., and W. L. Ellsworth, A double-difference earthquake loc<strong>at</strong>ion algorithm: Method<br />
and applic<strong>at</strong>ion to the northern Hayward fault, California, Bull. Seismol. Soc. Am., 90, 1353–<br />
1368, 2000.<br />
4