Home

an adaptive, automatic phase- picking and epicenter locating

1. 37 TT T T 36 J 4 RR a A age ml V a ae N Pa eg Nu we KT IS 347 Ri S 4 gt Ss i i 33 bool Fa H m T 32 a a 4 poe eee l boundary ene man e2e 31 i i i I 110 109 108 107 106 105 104 103 2002 09 REF 146 37 r r ii 36 te a N x N i 4 1 T i r it 34 i 7 4 33 i i H 44 l l 32 l ere See See Se Se gt eee 4 er eee i l boundary na ne man e2e 31 L f i i jl i 110 109 108 107 106 105 104 103 Figure 3 8 The comparison of adaptive automatic location results of two months The upper plot is the result of April 2002 and the plot at the bottom is the result of September 2002 The cross symbols are location results given by PLRR and the star symbols are manual location results The line between each star cross symbol pair is the discrepancy between manual and automatic location _43 In Figure 3 8 the seismic event location results by PLRR over two months are shown The left plot is the earlier result processed by event pair cross correlation matching method with 81 reference events and the right plot is a recent result processed by spectrogram event pair cross correlation matching method with 146 reference events It is very clear that increasing the number of reference eve
2. seismic arrival times were commonly read and picked from analog paper records The original digital data acquisition system used an algorithm distributed by IASPEI called XDECTECT to detect seismic events XDECTECT s algorithm was based on the common STA LTA the ratio between short term average STA and long term average LTA seismic phase detection method The XPICK picking program and the Seismos location program Hartse 1991 were used to locate earthquakes in New Mexico Withers 1997 developed a method called 299 LWCEDS Local Waveform Correlation Event Detection System and used it to detect eatthquakes and also estimate coarse location magnitude and original time based on processed waveform correlation The seismic data was saved in CSS 3 0 Center for Seismic Studies Version 3 format Zhang 1999 used Matseis to process seismic data of New Mexico from July 1997 to February 1998 Young et al 2001 used a dendrogram based hierarchical tool to classify regional events of New Mexico into several sources regions In 1999 the Earthworm system was setup to maintain seismic data from the SC network by with assistance from the US Geolgical Survey From the middle of 2000 all possible seismic events automatically triggered by Earthworm were saved on hard disk in AH or SAC format Only the raw waveforms were saved without any phase information available and more than half of them were not earthquakes or mining blasts but were instead noi
3. subdirectory with the name mmddhhmmss We do not yyyy ty yyyy enerate new site and sitechan tables every day because they are typically stationary g ry day y y ry 239 dailycss adjusts another Perl function named foreach pl programmed by Andreas Gohr a gohr web de and a shell script named sac2css_sc with the following contents bin csh set echo set prefix 2 sac lt lt END read 1 writecss 1 prefix END Reference seismic dataset The reference dataset consists of many well located seismic events with complete seismic phases picked and high signal to noise ratio waveforms It is essentially a knowledge base of historical seismic events for the New Mexico region with corresponding phase information Since earthquakes tend to reoccur in the same source region over the years the waveform similarity between a new event and old events reference events occurring in the same source region can be high By applying the cross correlation technique and supposing the reference seismic dataset is completed we can find the location and phases of an unknown seismic event far mote precisely and robustly than the results given by traditional STA LTA algorithm As shown in Figure 3 2 the reference seismic dataset currently used in PLRR has 153 events by June 2004 from all over the state of New Mexico representing typical seismicity But the contribution from each station i
4. Wavetorn Mag Reference MaSels match 4 Database Comperalion r Duration selup MagTod DE Current Ongn by Evert ee ty match setup aap da Print nn Next Ongn Exil nu ata tokete data Figure 2 Gives the menu structure within Matseis PLRR 73 PLRR uses three CSS database references in its analysis the Input Reference and Output databases The input database represents the data for the new events to be processed using PLRR methods The reference database as its name implies is the database holding the set of cataloged events that the new events are being compared to using WCC The last database is the output database in which the processed events are placed with locations obtained through the PLRR processing method The setup windows for all of three databases are similar and are shown in Figure 3 It is most convenient to save all information into input files and use Reconfigure to read it Directory fraid data Eworm Triagers refbase Database File Prefix File Names affiliation amplitude mal at redid ogn O instrument netmad network origerr origin sensor sitechan wfdisc ziz aug 1 Babe egz 88 98 gr 216 8 Figure 3 Database Setup GUI window for the Reference database Set the directory path and database prefix for the database Once the database setup is ready you can use Read under Orig sub menu to read the Origin table from the input database into the main G
5. arrange the CC matrix 27 from random patterns into connected groups In our application we would like each group to only contain waveforms generated from the same soutce region 28 Chapter 3 THE DEVELOPMENT OF PLRR MatSeis Matlab is not only a very popular tool for numerical computations but it is also an advanced programming language comparable to Visual Basic and Visual C It provides many Graphic User Interface GUI functions that users can use to build interactive windows for analyzing and visualizing data The Matlab package includes a signal processing toolbox library of algorithms and because of Matlab s design structure users ate easily able to construct new functions for use by writing Matlab code MatSeis was first developed as a seismic data processing package under Matlab by Mark Harris at Sandia Laboratory and it relies on using Matlab s core signal processing libraries and programming environment to process seismic data It is currently maintained and updated by a team at Sandia Laboratory in New Mexico 18 MatSeis has a GUI window style interface in which users execute algorithms by selecting pop up menus or by clicking buttons Since Matlab is a command line language any operation inside Matlab can be executed by inputting command line statements Therefore users can write an M script file to finish a batch of jobs automatically in MatSeis in addition to GUI mouse operation As a platform for
6. from the data around M 1 5 which tells us that earthquakes with a magnitude less than 1 5 are incomplete in 3 Figure 1 3b shows that the b value of the Socorro Seismic Anomaly SSA is 0 77 The fitting residuals were very small which means the catalog of 3 is complete for SSA region down to approximately M 1 0 The SC seismic Network and nearby Seismic Stations 110W 108W 106 w 104w 102W 100W 98 W 112 W ea 38 N 38 N SDco Trinidad 36 N 36 N 34 N ione 34N Phoenix Morenci CRRX Santa Rita guc prone u N 32 N Tucson Las Cruces GDL2 cl m 100 200 n a 112 w 110 W 108 W 106 W 104 W 102 W 100 W Figure 1 4 Seismic stations around New Mexico The green area is the Socorro Seismic Anomaly SSA the red points are seismic stations and the cyan diamonds are major mines 11 Table 1 1 Seismic Station list 1997 March 2004 Station Network Component Latitude Longitude Altitude m BAR SC ZEHZ 34 150 106 628 2121 BMT SC ZEHZ 34 275 107 260 1987 CAR SC ZEHZ 33 953 106 735 1658 CBET SC ZEHZ 32 421 103 990 1042 CL2B SC ZEHZ 32 264 103 879 1045 CL7 SC ZEHZ 32 413 103 807 1033 CPRX SC ZEHZ 33 031 103 867 1356 GDL2 SC ZEHZ 32 200 104 364 1213 HTMS SC ZEHZ 32 473 103 634 1192 LAZ SC ZEHZ 34 402 107 139 1878 LEM SC ZEH Z N E 34 166 106 974 1698 LPM SC ZEHZ 34 312 106 632 1737 MLM SC ZEHZ 34 814 107 145
7. it Duration magnitude has been used to calculate seismic magnitude in New Mexico for several decades You have to find a duration magnitude equation for your region if you want to apply this tool to calculate the magnitude There is only one function Update Monthly Tables under Update sub menu Once you have finished processing a new event you can run Update Monthly Tables to add this event into your output database As shown in figure 4 the Update Monthly Tables GUI window shows the contents of three CSS tables Origin Arrival and Wfdisc You can modify the information in Origin table and select which waveforms and arrivals to save into the output database You have to click Change before you finally click Update to save these CSS tables The button Restore is empty now 75 Figure 4 Update Monthly tables GUI window 2 Reference set The quality of the reference set dominates the accuracy of the automatic phase picking Based on our limited experience we put forward the following criteria to select events for the reference set 1 Only include waveforms with high signal to noise ratio The maximum absolute amplitude ratio between signal and noise should large than 3 0 2 Only include waveforms with distinguishable phase structure 3 Remove glitches from the waveform 76 There is an express way to build reference set 1 Set the output database to be the reference set 2 Read in an event into the main GUI window of
8. pick and locate parts within this package Our future goal is to integrate the automatic initial phase picking and locating with phase repicking Rowe et al 2000 and relocating Haijiang and Thurber et al 2003 for groups of similar events The phase pick processing is called automatic waveform cross correlation phase picking method because it is based on a reference set Before running this process a user will have to compile a reference database that contains typical seismic data for their special research region The reference database should include events with high signal noise ratio waveforms and good phase picks e g P and S or any particular characteristic phases The reference database should be analyst reviewed and contain events that are most typical for that region In the New Mexico area the three types of activity of interest are earthquake swarms recurrent earthquakes and mining blasts Typical mining activity up 400 events annually is recorded coming from the Morenci Tyrone and Santa Rita open pit mines to the southwest of the New Mexico Tech Seismic Network NMTSN Natural seismic events are less likely to occur in the same relative location unless their part of a long duration earthquake swarm event or a recurrent seismic source Because of this our main focus is to recognize and locate the mining swarm and recurrent events 69 However an adaptive tool is envisioned which could be used to insert short duration s
9. seismic data processing Matseis is good for data visualization and handling multiple events It provides GUI windows that allow users to pick seismic phases map the location of earthquakes and cluster waveforms into different seismic source regions 29 Seismic data when loaded into Matseis are organized in the matrix form and managed by four basic functions Waveform handles all the information related to the waveform data e g station channel start time sample rate Origin handles the event origin information e g location original time magnitude and location error of event Arrival stores the phase information Traveltime stores the theoretical traveling time information PLRR Seismic data from the SC network is maintained by the Earthworm system Every day the trig2disk module of Earthworm produces seismic event data files on disk based on the STA LTA algorithm Due to the layout of seismic stations and telemetric spike like noise at least half of these automatically generated files do not contain any seismic signals Moreover we could not find a good way to combine the seismic phase information generated by the Earthworm modules with the SAC format seismic event data files and ensure that the seismic phases generated by Earthworm modules were of good enough quality to be used as the initial seismic phases for earthquake relocation We plan on answering the following two questions How to select real seismic events from these a
10. to form a PLRR Sis readable CSS database are Origin Wfdisc Arrival Assoc Site and Sitechan All of these tables are in ASCII format and can thus be edited using vi or emacs The seismic event files automatically generated by the Earthworm system are in SAC or AH format Therefore the first step of PLRR is to convert the data format of Earthworm output into the format which Matseis can read A Perl script was written to perform this work daily The source code follows with filename dailycss see reference II The output files from Earthworm are saved each month under the taid data Eworm Triggers incoming yyyymm directory and in SAC format The file name has the format of yyyymmdd_hhmmss_quakeid in which the quakeid is assigned by Earthworm The saved section duration of each waveform is set to be 180 seconds All of the above sets could be modified in the trig2disk d config file of Earthworm With the help of the UNIX command crontab the dailycss could be executed once every day The following sentence is set in crontab to run dailycss at 20 00 everyday 03 20 raid users wyang outgoing dailycss The outputs of dailycss are in CSS 3 0 format and saved under the taid data Eworm Triggers outgoing directory Each day two new tables are built yyyymmdd origin and yyyymmdd wfdisc and the corresponding waveforms of one event are saved under the mm
11. 0 50 60 70 80 90 100 time second Figure 3 6 Waveform bottom of a mining blast from the Tyrone mining field and recorded by station LAZ 222 km to NE of the epicenter and corresponding sonogram upper plot The energy is concentrated 15 Hz In the band between 0 to 5 Hz the background noise dominates the signal making it difficult to apply for the cross correlation technique In the band between 6 and 10 Hz the signal energy is very clean and both P and S phase arrival energy is very strong and easily distinguishable In the band between 11 15 Hz the energy again decays rapidly because of attenuation and the limited frequency response of the short period station telemetry _40 In the frequency domain the method selects a suitable frequency band 6 10Hz and calculates the CC curve of each frequency Then it stacks these curves and takes the mean of them as the spectrogram CC curve Compared with waveform CC curves in the time domain the spectrogram CC curve has less sample points sacrificing the accuracy for better similarity robustness But for the initial phase picking in this case the new event is initially collocated with the most similar reference event the accuracy given by the spectrogram cross correlation matching method is sufficient Ground truth test Starting at the end of 2003 I began to use the above methods to process the monthly seismic data of New Mexico and compare the location results given by PLRR w
12. 03 00 19 03 10 19 03 20 MatSeis PLRR Figure 3 10 An example of double event that occurred in the Socorro area in November 2002 This type of earthquake together with ripple fires several mining blasts in a short period makes it difficult for PLRR to find the true location because matching reference events may not be available or appropriate _45 Chapter 4 EARTHQUAKE LOCATION IN NEW MEXICO 2000 2004 As discussed in chapter one local seismic activity in New Mexico comprises natural earthquakes and man made mining explorations concentrated in several source regions These two types of seismic activity form the bulk of monthly seismicity of New Mexico In this project the PLRR program had been used to pick seismic phases and locate seismic events 2000 2004 in New Mexico automatically and then we repick the phases and relocate the epicenters manually As the products of this research the distribution of seismic event in this period is demonstrated in this chapter In this chapter the epicenter distribution of seismic events around the state of New Mexico will be presented every 4 months in a year from 1997 to 2004 The seismic event database from July 1997 to February 1998 was collected by Xiaobing Zhang 8 and repicked and relocated by Darren Hart In each of the following maps the epicenters of earthquakes and mining blasts are plotted in different colors The criteria we used to differentiate between mining blast from earth
13. 2088 SBY SC ZEHZ 33 975 107 181 3230 SMC SC ZEHZ 33 779 107 019 1560 SSS SC ZEHZ 32 355 103 397 1073 WTX SC ZEHZ 34 072 106 946 1555 ANMO IU ZBHZ 34 946 106 457 1853 SDCO US ZBHZ 37 746 105 501 2569 WMOK US ZBHZ 34 738 98 781 486 TUC US ZBHZ 32 310 110 784 906 ISCO US ZBHZ 39 800 105 613 2743 WUAZ US ZBHZ 35 517 111 374 1592 LTX US ZBHZ 29 334 103 667 1013 CBKS US ZBHZ 38 814 99 737 677 DAG SC ZEHZ 32 591 104 692 1396 7 ps After May 2004 station HTMS was replaced by SRH Digitized data from 26 seismic stations in and outside New Mexico has been collected for processing at the New Mexico Seismological Observatory since 1992 In 1999 the Earthworm system was set up at NMSN to manage data flow and produce automatic triggered waveforms files from automatic triggers Since the beginning of 2003 the digitized seismic data i e waveform data from the NMSC network has been archived at IRIRS DMC AD in Seattle This was done so that researchers all over the world could download the data from the IRIS DMC website and to provide an independent long term backup archive Out of the 26 seismic stations in the region 18 of them are equipped with short period seismometers maintained by the New Mexico Tech and are known as the SC the associated Federation of Digital Seismic Networks code network The SC network is concentrated in two areas Socorro area and Carlsbad Ten seismic stations are loc
14. 30 seconds Figure 1 8 h There is more than one kind of shooting type in Morenic area that affect the waveforms beside single shot blasts double shot and ripple fire blasts are sometimes used The waveforms from the Carlsbad WIPP area have obvious P and S phases The S P time is about 35 seconds Figure 1 8 e The amplitudes of P and S phases are nearly equal to each other The waveforms from Raton area have P and S phases visible but both of them are relatively weak The S P time is about 40 seconds 19 a Gallup Mining field e WIPP induced events f Santa Rita open pit mine field g Raton events h Morenie open pit mine field Figure 1 8 Some characteristic vertical component waveform patterns from major source regions recorded by station LEM vertical component Pervious research Dr Allan Sanford began research on New Mexico earthquakes and seismicity based on instrumental data in the late of 1950 s Sanford and Holmes 1961 observed strong distinct 20 arrivals other than direct P and S on the seismograms in the Socorro area leading to the discovery of Socorro Magma Body SMB Hans Hartse 1991 systematically summarized former progress on the structure research of the central New Mexico and used direct and reflected micro earthquake phases to estimate hypocenters and to invert for a layered a crustal velocity model This velocity model Figure 1 9 is the standard model used to locate
15. AN ADAPTIVE AUTOMATIC PHASE PICKING AND EPICENTER LOCATING PROGRAM BASED ON WAVEFORM CROSS CORRELATION AND ITS APPLICATION FOR PROCESSING SEISMIC DATA IN NEW MEXICO by Wenzheng Yang An independent study submitted in partial fulfillment of the requirements for the degree of Master of Science in Geophysics New Mexico Institute of Mining and Technology 2004 Approved by Chairperson of Supervisory Committee Program Authorized to Offer Degree Date ABSTRACT It is widely observed that seismic sources occurring very close to each other can share very similar waveforms Using this property an adaptive automatic phase picking and epicenter locating program based on waveform cross correlation has been developed and applied to process daily seismic data of New Mexico The core of this program is a reference seismic event database from which the phase and location information of a new event can be inferred With a sufficiently complete reference seismic dataset this program should produce robust initial seismic phase estimates while greatly improving the handling of false triggers due to telemetry or other transient noise These automatic results are designed to be as precise as the manually processed results 160 well picked earthquakes and mining explosions in New Mexico during 1997 2003 were collected as an initial reference event set covering most of the historically active source regions in the state Each w
16. Events collected by the New Mexico Tech Seismic Network Using a Matseis based Review Tool and a Postgres Database Master degree thesis New Mexico Institute of Mining and Technology 1997 9 Withers Mitchell An Automated Local Regional Seismic Event Detection And Location System Using Waveform Correlation Ph D Dissertation New Mexico Institute of Mining and Technology Socorro New Mexico 1997 10 J Young B J Merchant and Aster R C Comparison of cluster analysis methods for identifying seismic events 23 Annual DTRA NNSA Seismic Research Review 229 238 2001 11 C A Rowe R C Aster B Borchers and C J Young An automatic adaptive algorithm for refining phase picks in large seismic data sets Bull Seism Soc Am 92 1660 1674 2002 12 Sanford A R and P Einarsson Magma chambers in rifts Continental of Ocean Rifts Geodynamics Series Volume 8 147 168 1982 13 Papanicolaou Lecture on Mathematical Geophysics and Uncertainty in Earth Models summer school Golden Colorado June 14 25 2004 14 Song X D P G Richards Observational evidence for differential rotation of the Earth s inner core Nature 382 221 224 1996 66 15 Aster R C P M Shearer and J C Berger Reply to Comment on Quantitative measurements of shear wave polarizations at the Anza Seismic Network Southern California implications for shear wave splitting and earthquake prediction Journal of Geophysical Rese
17. HHAHHHHAHHAHHHHHH incomingpath raid data Eworm Triggers incoming outgoingpath raid data Eworm Triggers outgoing open WOUT date Y om d close WOUT TODAY s s get rid of the end stuff HHSTODAY TODAY 2 2 days before TODAY yesterday open WOUT date Y om subpath lt WOUT gt subpath s s get rid of the end stuff close WOUT If you want to deal the data of last month you have to specific the date as the following TODAY 20030630 print n Daily routine to convert Earthworm SAC incoming into CSS format n TODAY n PAT incomingpath subpath TODAY print The pattern of TODAY TODAY is n PAT n n open WOUT raid users wyang outgoing foreach pl PAT echo newevents lt WOUT gt close WOUT Current date 83 judge new events coming in or not test newevents 0 mark rindex test _ if mark 1 die No events came in today please check earthworm n print New Events coming in processing n if month directory does not exists make it opendir DIR outgoingpath subpath mkdir outgoingpath subpath 0775 readin new events COU 1 loop begin foreach event newevents print New event COU event event s s get rid of the end stuff oo split _ event orid COU o00 2 eventdir TODAY o00 1 print ORIGIN e
18. Matseis PLRR and manually pick phases and locate it 3 Use Update Monthly tables to write this event into the reference set 4 Run sort_wfdisc in Matlab command window to sort and generate wftag table 3 Automatic phase detection Once there is an event selected in the main GUI window you can click Auto pick by Waveform match to find the most similar reference waveform to each waveform of the event station by station You also can find the phases for several waveforms by selecting them before running Auto pick by Waveform match Since Auto pick by Waveform match is specially designed to process seismic data collected in SC network you have to modify it if your data is significantly different from our data The seismic data in New Mexico consists of short period data EHZ and broadband data BHZ The sampling rate of EHZ station is 100 samples per second and the sampling rate of BHZ station is 40 samples per second In the source code of Auto pick by Waveform match plrr_pic_auto m we band pass the EHZ seismic data using butter 3 0 1 0 6 and band pass the BHZ seismic data using butter 3 0 2 0 6 We also default 0 55 to be the cuto for cross correlation in plrr_pic_auto m If the process cannot find the cross correlation value for one waveform larger than the cuto value there is no phase assignment to that waveform Auto pick by Event match has a GUI window figure 5 through which you can select any reference event to match to any st
19. N AUTOMATICALLY PROCESSING OF THE SC EARTHWORM SYSTEM SEISMIC DATA Plrr M file list plrr_a_update m plrr_ms_zoom m plrr_update_menu m cat2emt m plrr_nextid m plrr_w3_rem m cataloglist m plrr_nextorigin m plrr_w_menu m corr_speed_test m plrr_o_comp m plrt_w_read m css_edit m plrr_o_menu m rcotr m csstableitem m plrr_o_popup m ref_waveedit m cutsac m 80 plrr_o_read m tick_example m db_setup_ref m plrr_o_update m shiftdata m demean m plrr_output_setup m sort_wfdisc m epicenter m plrr_pic_auto m specgram1 m event_swcc_match m plrr_auto_dec m plrr_pic_menu m event_wcc_match m plrr_auto_nov m plrr_picbyevent_auto m example_spec m plrr_auto_oct m plrr_r_chan m filterwave m plrr_cc_matrix m plrr_r_site m findmatch_swcc m 81 plrr_config m plrr_r_sitechan m findmatch_wcc m plrr_css_out m plrr_r_sta m load_css m plrr_dendro m plrr_r_wave m ms_draw m plrr_exit m plrr_reconfig m update_ref m mslocator_tool m plrr_f_menu m plrr_ref_setup m waveform2sacm m mslocator_tool_aedit m plrr_input_setup m plrr_ref_sort m waveform_edit m o_popup m plrr_loc_menu m plrr_setup m whole_shift_phase m 82 plrr m plrr_monthly_update m plrr_tool_menu m write_css m plrr_a_menu m plrr_ms_fig m plrr_tt_menu m Source Perl script of dailycss usr local bin perl w HEHHHHHHHHHHHHHHHEHHEHEEHEE daily routine to convert new coming event of this day to CSS format Wenzheng Y 2003 HHHHHHHHHHA
20. Seis Figure 3 4 Illustration of waveform pair cross correlation matching method This method calculates the CC of a waveform pair one from the new event and one from a reference event recorded by the same station and finds the waveform pair with the highest CC value Then it assigns the phase information of the reference waveform to the new waveform to provide initial picks After finding phases for each waveform of the new event each station s waveforms may possibly be correlated with a different reference event a Matseis locating function MS Locator Too is used to find the initial epicenter This method preprocesses the waveforms in two steps 36 1 All waveforms are high pass filtered with the frequency cutoff equaling 1 0 Hz The waveforms of teleseismic events will become significantly attenuated after this step and will not match any waveforms in the reference dataset with a high CC value e g 0 2 thus making them easily discounted 2 Filtered waveforms are converted into envelope based on the Hilbert transform The Hilbert transform performs a 90 degree phase shift of opposite sign at positive and negative frequencies to a time series without changing its amplitude and is defined as the following convolution _1 fx Os i 3 1 The envelope is given by EM 4x O y A 3 2 By taking the envelope of waveforms prior to cross correlation the signal noise ratio SNR and associated co
21. UI window Comparation is a tool to plot simple dendrograms cross correlation matrices and waveforms between one waveform in the main GUI window and selected reference waveforms for the same station You cannot run Comparation if you do not select orid in the main GUI window The functions of Current Origin and Next Origin are the sames as they are in Matseis Wfm Arr and TT are kept as same as they are in Matseis 1 7 you can find a more detailed explanation in the Matseis user manual The sub menu Pick provides three types 74 of picking methods Manual is the Measure Tool in Matseis a tool to assist in detecting arrival phases manually I use it to help find accurate phases for the reference set Auto pick by Waveform match and Auto pick by Event match are two automatic phase detecting tools using the WCC technique The former one is more elementary than the latter It is based on the comparison of waveform pair The latter one is based on the comparison between event pair and it is more robust than Auto pick by Waveform match These methods were developed to process the seismic data of SC network and you might have to modify them if you have a problem processing seismic data from your research region All tools under Locn are the same as they are in Matseis Auto pick by Waveform match does not provide the location information and you have to use MS locator Tool under Locn sub menu to locate the selected event after you have applied
22. a phone line ot microwave transmission less than half of the triggered events generated by the trig2disk module each month are real seismic events Because this false positive event detection rate we are looking into at a method to adaptively winnow out the noise events from the real seismic events Major Seismic Source Regions in New Mexico Although the northwestern area of the New Mexico is part of the Colorado Plateau and the Rio Grande rift bisects New Mexico down the middle there is no obvious tectonic province association connected with seismicity Major seismic source regions in New Mexico are relatively independent from each other One major source region is the Socorro Seismic Anomaly SSA which is located in the central Rio Grande rift of New Mexico and overlies the Socorro Magma Body SMB The SSA has been estimated to have an area of about 5 000 km 4 green area in Figure 1 4 Nearly 36 of the naturally occurring earthquakes occurring in New Mexico are concentrated in this area even though it covers only 1 6 of the total area of New Mexico 4 Some events occurring in the SSA region generate reflected P and S phases PzP SzP and SzS phase arrivals allowing researchers 4 to invert and to estimate the depth and shape of the Socorro magma body and also to get more accurate estimations of event depth Seismic swatms and repeat earthquakes often occur inside the SSA possibly being generated by the injection of
23. arch 96 6415 6419 1991 16 Bergthora S T and J C Pechmann Constraints of relative earthquake locations from cross correlation of waveforms Bulletin of the Seismological Society of America 77 1626 1634 1987 17 Young C J B J Merchant and Aster R C Comparison of cluster analysis methods for identifying regional seismic events 23 4 Annual DIRA NNSA seismic Research Review 229 238 2001 18 Homepage of Matseis https www nemre nnsa doe gov cgi bin prod nemre matseis cgi 19 Webpage of Earthworm http folkworm ceri memphis edu ew doc modules htm 20 L Carmel Y Koren and D Harel Visualizing and Classifying Odors Using a Similarity Matrix Proceedings of the 9th International Symposium on Olfaction and Electronic Nose ISOEN O2 Aracne pp 141 146 2003 67 Appendix I THE USER S MANUAL OF PLRR MATSEIS MatSeis PLRR 0 1 User s Manual Waveform Cross Correlation WCC Phase Picking based on Analyst Reviewed Reference Data Set Wenzheng Yang Rick Aster and Darren Hart February 16 2004 68 Objective The main goal of this project was to develop and test an automated method for phase picking based on the new events similarity to events archived in a reference database The measurement of similarity used to compare events is waveform cross correlation WCC The typical processing is envisioned as follows 1 Typical seismic activity manmade and natural for a region is cataloge
24. ated seismic events manually All in all 3 023 events around New Mexico were phase picked and located most located in the SMB and major mining fields Some events from the SMB had clearly defined magma body reflection phases enabling us to find the depths of these events more precisely During the data procesings seismic events occurring in new source regions were added in the reference database so that the phase picking and epicenter locating ability of PLRR could adapt to future occurrences of the new waveforms Suggestions During the last two and a half years development I made mistakes along with my progress Although I got everything done this work is far from perfect Here are some suggestions which might be good for future development 2635 1 Whenever a new type of seismic event is found it is supposed to be added to the reference set When the number of reference events becomes very large the processing time to find the phases and location for one event becomes intolerable A cluster algorithm e g dendrogram tree might be needed to organize the reference set and a new event will be compared with a sub group of events or a master event list instead of whole events in the reference set This will decrease the calculation time dramatically 2 Most events in the reference set do not have waveforms for all seismic stations around New Mexico It would increase the event detection ability if some synthetic waveforms were a
25. ated to study seismicity associated to the SSA and 8 seismic stations are set to monitor the seismicity localized around the Waste Isolation Pilot Plant WIPP near Carlsbad The sampling rates of these short period seismometers are 100 samples per second sps Station LEM is the only three component short period station used in this study There are 8 broadband seismic stations operated by other organizations around New Mexico All except ANMO Albuquerque New Mexico are located outside of the state of New Mexico These stations compensate for the short coverage of the SC network in the boundary area of New Mexico to some degree All the broadband stations data used has a sampling rate of 40 sps except for TUC Tucson which is 20 sps NMSN uses the Earthworm v6 2 system http folkworm ceri memphis edu ew doc developed by the USGS to digitize analog signals and store the continuous data for the last seven days period on a local hard disk IRIGE signal from GPS is used to time stamp the digitized trace data In Earthworm system the module pzcker_ew picks the P arrivals using STA LTA method and the module binder_ew associates these P arrivals with events The module Zrig2disk saves a user defined section of seismic data usually 180 seconds of to 413 disk as an event Due to the existence of electric noise within the seismic signal transmission system i e signal telemetry from field site location to the data storage facility vi
26. ation available to that event Since the GUI window is still being developed only the Match Method option selection is available for use The other option selections are fixed and the button Matrix is empty in version 0 1 Te Figure 5 Event Match GUI window eventmatch m is the Matlab code to setup the event match GUI window and findmatch m is the Matlab code adjusted by eventmatch m to calculate the cross correlation value and stack the CCc curves based on different types of match methods In findmatch m we set the default sampling rate to be 100 samples per second and resample BHZ data from 40 or 20 samples per second to be the default sampling rate We also set the default waveform window to be the first 100 seconds of the waveform before carrying out cross correlation calculation If the length of waveform data is less than 10000 Default sampling rate x Default window we zero pad the data If the length of waveform data is large than 10000 we only take the first 10000 samples After cutting the data to fit the window we band pass the data band pass parameters can be set in Event Match GUI window BP MIN is the low cuto and BP MAX is the high cuto We also use butter and filtfilt functions in Matlab to finish band pass filtering Waveform data after processing are converted to Hilbert envelope in order to increase signal noise ratio In Match Mathod option selection median takes the median of all waveform pair cross
27. aveform in the reference event set has a high signal to noise ratio and accurately picked P and S phases Two methods were developed to automatically calculate waveform cross correlation phase picking and the location of epicenter The first waveform pair cross correlation matching method correlates the most similar waveform from the reference dataset to each waveform of a new event station by station and assigns the corresponding seismic phases to the new waveform according to cross correlation lag Note that a new event may be correlated with different reference events at different stations The second event pair cross correlation matching method correlates the most similar event from the reference dataset to the new event by stacking the cross correlation curves of all common station waveform paits The second method provides more robust results than the first method Due to the concentration of seismic stations in the Socorro area an attracting problem developed when the event bair cross correlation matching method was applied to process seismic data of New Mexico This problem happened when the above method construed distant events to be local ones because seismic events occurring more than 200 km from Socorro usually have weak S phases and are therefore identified as seismic events from the Socorro area by the event pair cross correlation matching method To nullify this problem the spectrogram event pair cross correlation matching m
28. band whereas local and regional seismic signals contain high frequency energy up to 10s of Hz Figure 1 7 16 1000 2000 5000 6000 7000 0 1000 2000 3000 4000 5000 6000 Figure 1 7 Spectrograms of a local upper and a teleseismic event bottom recorded by WTX The vertical axis is labeled by normalized frequency 0 to 50 Hz Besides natural earthquakes manmade seismic events e g mining explosions are often recorded by the SC network Four major regions where routine mining blasting occurs are Morenci southeastern corner of Arizona cyan point in Figure 1 4 Tyrone southwest corner of New Mexico Santa Rita to the east of Tyrone and Mt Taylor to the northwest of Albuquerque Repeated earthquakes have often been observed around the station DAG since 2003 and they are likely induced seismic events 2 Characteristic Waveform Patterns of different source regions When waveforms from different source regions are recorded at a signal station they show different patterns or signatures due to differing Green s functions associated with their propagation paths Generally waveforms that come from the same source region share similar characteristics In the following I will describe the characteristic waveform characteristics for seismograms from different source regions by comparing waveforms recorded at station LEM Ta The vertical component waveforms recorded by LEM from several major source regions are sh
29. correlation coefficient curves maximum takes the maximum of all waveform pair cross correlation coefficient curves mean takes their mean and weighted weights the curves one by one over epicenter station distance Auto pick by Event match finds the most similar event in the reference set to the selected event in the main GUI window database or lots of small CSS databases you can write a Matlab code to finish the job plrr_auto_nov m and plrr_auto2_nov m are two examples which we used to process _78 daily CSS databases in November 2003 recorded by Earthworm system of SC network using Auto pick by Waveform match and Auto pick by Event match methods respectively 5 Example In the example subdirectory of this package there is one CSS database named ref_ex ref_ex tables wf which is an example of a reference set and there is another CSS database named newdata under newdata sub subdirectory which is an example of a new event You can use these data to check the two automatic phase picking methods but you have to modify the dir set in wfdisc tables to both of these two CSS databases You can use emacs or dbset in Datascope to do the job Due to limited time we could not include as many details as we wanted If you encounter any problems in Matseis PLRR contact Wenzheng Yang wyang ees nmt edu for more detailed help Any and all comments or suggestions are welcome _79 Appendix II PLRR SOURCE CODE LIST AND RELATED CODES I
30. d in a reference database 2 A new event database is compiled based on a real time triggering process e g Earthworm is used at New Mexico Tech 3 WCC is used to determine if new event triggers are found in the reference database If the correlation between the reference waveform and the new event waveforms is above a set threshold then phase picking is done When four or more phases are picked for a new event then the new event can be run through a location program to determine its hypocentral location Introduction MatSeis PLRR or PLRR is a modified version of MatSeis v1 7 with some special functions Currently the PLRR package could not be run without the underlining support of MatSeis v1 7 PLRR was developed and tested on the Sun Unix system Before using PLRR users should be sure that they have already installed Matlab v6 0 R12 or v6 5 R13 and MatSeis v1 7 on their system Currently PLRR only supports one database structure CSS3 0 Flatfile format which one of the database formats supported by MatSeis To find out more about the CSS3 0 flatfile schema download a document from the MatSeis website https www nemre nnsa doe gov cgi bin prod nemre matseis cgi PLRR represents a four step process Pick Locate Repick Relocate Pick represents the initial phase picking process or phase identification problem The second step Locate will assign the event as originating from a region At this stage in development we only implement the
31. dded to stations missing waveforms in the reference set or if new reference events with all stations were used to replace them 3 PLRR could have new future capabilities The phase repicking program developed by C Rowe and R Aster Rowe s Ph D work could to be incorporated into it A double difference relocation program e g hypoDD or tomoDD also could be integrated In the long run PLRR could be implanted into a real time system to detect seismic signals and produce very high resolution seismic event catalogues automatically 4 Some seismic stations need to be set up in the three corners of New Mexico besides Carlsbad to get better seismic event coverage thereby increasing the accuracy and effectiveness of PLRR for the study region Conclusions In this independent study I have presented an adaptive program that automatically detects seismic phases and locates epicenters by using the waveform cross correlation technique I 64 have shown how to combine it with the Earthworm system to process seismic data in New Mexico automatically Two waveform cross cortelation matching methods were presented and the event pair cross correlation matching method was found to be better than the waveform pair cross correlation matching method Event pair cross correlation matching method in the frequency domain can prevent the seismic events far away from the seismic network from being misidentified as events occurring locally I also pres
32. e Socorro Seismic Anomaly SSA 1962 1995 data from 3 Figure 1 2 illustrates the seismicity of natural earthquakes in New Mexico and bordering regions from 1962 to 1995 Most of them were recorded by the Socorro seismic network of New Mexico Tech NMSN 3 but some of the earthquakes that occurred in the 1970s were recorded by the Los Alamos Laboratory the University of Texas at El Paso and the USGS From Figure 1 2 we observe that the NMSN has documented numerous eatthquakes with duration magnitude M above 2 0 inside the state of New Mexico However the NMSN may not be sensitive enough to detect all earthquakes during this time with less than 2 0 M occurring in the NE NW and SW corners of New Mexico hence the incomplete seismic record in those regions According to the historical earthquake catalog B only 8 of all 2000 earthquakes that occurred in New Mexico from 1962 to 1995 were larger than 3 0 M Most earthquakes occurring in New Mexico are small events Their M typically falls within the range of 1 3 to 2 5 and they are usually located in isolated source regions The magnitude frequency curve is usually used to check the completeness of eatthquake catalogue and the slope of this curve b value usually ranges from 0 6 to 1 3 10 around the world However as shown in Figure 1 3a for New Mexico and bordering regions the b value of earthquakes was 0 76 and the fitting residuals were large The best fit line departs
33. earthquakes in New Mexico since the early 1990s Horzontal Offset km o 30 T Vp 5 95 0 05 km s Vs 3 41 0 04 km s Possion ratio 0 256 10 Vp 5 80 0 08 km s Vs 3 44 0 07 km s Possion ratio 0 228 a 4 o A 20 Magma Body Depth 18 75 0 28 km Vp 6 4 km s Mantle Vp 8 1km s Figure 1 9 Velocity model in Socorro area modified from Hartse 1991 The solid circle is a hypothetical earthquake Triangles indicate seismic stations The reflecting layer at 19 km is the nearly flat top of the Socorro Magma body The dashed lines between the hypocenter and stations are a direct P and S waves the reflected lines between the hypocenter and stations are PzP or SzS green incident angle is equal to reflect angle and SzP red reflected angle is larger than the incident angle 21 With this crustal velocity model regional theoretical travel time tables for seismic events in New Mexico can be built to help researchers to pick seismic phases The travel time curves for several often detected seismic phases are shown in Figure 1 10 Travel time with focal depth 1 km 120 T T T T Time second 0 l l l l l l l 0 0 5 1 1 5 2 2 5 3 3 5 4 Distance degree Figure 1 10 Theoretical travel time curves for several often detectable seismic phases in New Mexico for earthquakes with depth 1km Before the digital data acquisition system was installed in 1991 Zhang 1997
34. eismic activity into the reference database or a subset reference database There are two automatic picking methods available in PLRR besides the manual picking method provided in Matseis i e Measure Tool The first automatic picking method we call the waveform matching method For each waveform available for a new event this method finds the most similar waveform from the same station in the reference database set and assigns all its phase pick information to the new event waveform If more than 4 phases are picked by this method then the Matseis locator tool mslocator_tool is used to locate the new event The second method for automatic picking method we call the event matching method For a new event this method finds the most similar event to it from the reference set and assigns the corresponding phases and location of the reference event to the new event _70 This user s manual provides the following information Installation Start up procedure of PLRR PLRR main window and menus How to construct a reference set How to use waveform matching method How to use event matching method The collection of an automatic picked event Example a ONO NE SS We assume that users who want to use PLRR have some experience using MatSeis New users should read through the MatSeis manual as we will not explain any MatSeis features in this manual A downloadable copy can be obtained from the MatSeis web page https www 1 Instal
35. ense Threat Reduction Agency and Air Force Research Laboratory contract number D TRA01 01 C 0085 US Department of Defense the content does not necessarily reflect the position or the policy of the US Government and no official endorsement should be inferred Substantial components of the New Mexico Tech Earthworm system were funded by the US Geological Survey TABLE OF CONTENTS PL DSUBACE ee ie neg au teenth mand eet a ee 1 ALE Knowledsenem AS ange dei a ee cua hae deity EHE 3 Table or Contents se RR ee 5 Chapter 1 Introduction Background Seismicity in New Mexico ccecece cece cence ee rrer 7 The SC seismic Network and nearby Seismic Stations 10 Major Seismic Source Regions in New Mexico 00000eeee 13 Characteristic Waveform Patterns in different source regions 17 Pervio us teseatch u EI 20 Chapter 2 Waveform cross cortelation technique Waveform Cross correlation ash 23 Cross correlation coefficient Mate aeg 26 Chapter 3 The development of PLRR Masse neu E EE E O E R 29 PERRusE Sense 30 Data Format Conversion en 31 Reference dataset nenne ers 33 Strategy of phase detection and epicenter location 044 35 Waveform pair cross correlation matching 0 cece eee 36 Event pair cross correlation matching unseren 37 Spectrogram event pair cross correlation matching 39 Gr unddata
36. ent CC When considering N waveforms from different sources and received at the same station N N 7 2 cross correlation coefficients are calculated which forms a symmetric cross correlation coefficient 26 matrix with NXN elements e g Lower left plot in Figure 2 2 The CC matrix is a way to visualize the similarity between each waveform pair By incorporating the CC matrix with cluster analysis techniques e g dendrograms upper left plot of Figure 2 2 waveforms from different source regions could be connected in such a way so as to graphically depict the degree to which the waveforms are similar Figure 2 2 19 reference P phase aligned vertical component waveforms recorded by station LEM right the CC matrix of each waveform pair left bottom and the corresponding dendrogram left top The index of the coefficient matrix and the waveform plot are sorted according to the dendrogram result Waveforms are classified into four groups as marked by red squares in the matrix plot and are labeled with the source region name in the waveform plot The CC matrix can be used to illustrate the waveforms grouped results calculated from the cluster analysis e g Bergthora et al 1987 One method proposed by Young et al 2001 is to group the CC matrix using a dendrogram cluster analysis technique left top plot of Figure 2 2 The fundamental idea behind these clustering methods is to re
37. ented the seismic event location results in New Mexico from 2000 to 2004 This study will be helpful for people in the future because it may lead to a system that will vastly decrease the amount of time researchers spend on manual data processing 65 REFERENCE 1 http www ees nmt edu Geop Museum Posters NMseismology html 2 J Schlue and Jaksha L A Report On the SEISMICITY OF THE WIPP SITE for the period 1 July 2003 through 20 September 2003 2003 3 A R Sanford K Lin I Tsai and Jaksha L H Earthquake catalogs for New Mexico and bordering areas 1869 1998 New Mexico Bureau of Geology Mineral Researces Circular 210 2002 4 R S Balch H E Hartse A R Sanford and Lin K A New Map of the Geographic Extent of the Socorro Mid Crustal Magma Body Bull Seism Soc Am Vol 87 174 182 1997 5 Sanford A R and C R Holmes Note on the July 1960 earthquakes in central New Mexico Bull Seism Soc Am 51 311 314 1961 6 Hans E Hartse Simultaneous hypocenter and velocity model estimation using direct and reflected phases from microearthquakes recorded within the center Rio Grande rift New Mexico Ph D Dissertation New Mexico Institute of Mining and Technology Socorro New Mexico 1991 7 H E Hartse Sanford A R and Knapp J S Incorporating Socorro Magma Body reflections into the earthquake location process Bull Seism Soc Am 82 2511 2532 1992 8 Xiaobing Zhang Processing of Seismic
38. ethod was developed This method matches event pair in the frequency domain decreasing the effect of attracting phenomenon This works because within the frequency band of 6 10Hz the energy of S phase for local events is prominent and the background noise energy is very small The computer program PLRR incorporating these methods has been developed under Matseis Matlab developed at the Sandia National Laboratories in the UNIX environment It is used to detect initial seismic phases and identify locations of local seismic events in New Mexico from 2000 to 2004 Manual phase repicking and epicenter relocating are applied to the results given by this program to produce a final reviewed catalogue If there is a new high signal to noise event that is incorrectly located by PLRR as determined by the review process it will be added into the reference dataset The accuracy of phase picking and epicenter locating thus improves as more data are processed ACKNOWLEDGMENTS I have been very privileged to have undoubtedly the most intuitive smart and supportive research and academic advisor anyone could ask for Dr Richard C Aster I have been stimulated and excited by his constant flow of good ideas I will always admire Rick s to cut through reams of algebra with a single visual explanation In addition he has taught me a great deal of new computer skill Last but certainly not least he has always known how to give me a little push f
39. fixed threshold e g 0 3 it will be added into the reference dataset if it has high quality waveforms thus improving the reference dataset over time Strategy of phase detection and epicenter location For a new seismic event there may exist several very similar events in the reference dataset there are many cluster analysis methods available for use to address this situation based on different strategies Here we prefer to use the K most similar algorithm proposed by Carmel et al 2003 It is stated as the follows 1 Define a reference set ie a set of representative measurements for which the class association is known in advance 2 Given an unknown sample calculate its similarity index with respect to each of the measurements in the reference set 3 Find the K most similar measurements from within the reference set 4 Associate the unknown sample with the cluster that contains the majority of the K most similar measurement In PLRR we use the default K 1 which simplifies the strategy to simply be the most similar waveform 35 Waveform pair cross correlation matching There are two waveform cross correlation methods tested in PLRR to detect seismic phases and find locations for new events The first one is waveform pair cross correlation matching The schematic of this method is shown in Figure 3 4 003 09 2006 37 30 06 38 00 06 38 30 06 39 00 06 39 30 06 40 00 06 40 Mat
40. gram is a package of Matlab m files and depends on the support of Matseis The only acceptable seismic data format for PLRR is the CSS3 0 format A perl script was written to convert the daily triggered events of Earthworm from SAC format to CSS3 0 flatfile format At the beginning of each new month analysts can run PLRR to process the daily CSS3 0 data of the previous month and use it to detect seismic events and pick phases 62 automatically There are two cross correlation methods available with PLRR One is the waveform pair cross correlation matching method and another is the event pair cross correlation matching method The latter one was found to be more robust than the former Due to the layout of Socorro network seismic events located about 200 km away from Socorro are often mistakenly identified as occurring in Socorro area by the event pair cross correlation matching method This is because the amplitudes of S phases in these events decrease to the background level when their waveforms arrive at the seismic stations from 200 km away But in the frequency domain the amplitudes of their S phases are very strong in the band of 6 10 Hz We developed the spectrogram event pair cross correlation matching method to increase signal to noise ratios and thereby avoid these types of gross location errors Using PLRR we processed the seismic data of New Mexico from 2000 to 2004 automatically then went back and repicked seismic phases and reloc
41. his case the seismicity is possibly enhanced by mining and associated injection activity There have been several large earthquakes in the area between Raton and Trinidad Colorado in recent years In the period from August 28 to September 21 2001 twelve earthquakes were recorded by the USGS National Earthquake Information Center The magnitude of the largest event was M 4 6 Figure 1 6 shows three middle size events that occurred in the 15 Raton area between June and August 2003 On March 22 2004 another large earthquake with a magnitude of 4 4 occurred in the Raton area x 10 2008 0815 00 22 15 1 o 1 Tot ANMOBHZ 2008 08 14 00 11 15 ANMOBHZ o 2003 09 13 15 22 35 5 ANMOBHZ p 5 10 Se 4 SS ee o 2 80 100 120 140 160 180 Seconds Figure 1 6 Similar earthquakes that occurred in the Raton area Teleseismic events with large magnitudes often trigger the Earthworm system and may thus be initially improperly identified as local earthquakes For this study of New Mexico seismicity we needed to filter out these events The earth itself acts as a low pass filter since teleseismic events have to travel much longer ray paths compared to local and regional events It is therefore easy to differentiate teleseismic from regional seismic events by compating spectrograms or by applying a high pass filter to the signals The main energy of teleseismic signal at short period stations is concentrated in 1 2 Hz
42. ircle in gray plot and mining blasts are in blue circles gray circles in gray plot Triangles symbols are seismic stations Diamond symbols are mining fields and square symbols are cities 52 110 W 108 W 106 W 104 W 102 W 38 N A 7 38N 36 N 36 N 34 N 32 N 32 N 30 N f 110 W 108 W 106 W ee 104W 102 W Figure 4 7 Epicenter distribution of seismic events around New Mexico September 2001 December 2001 Earthquakes are in red circles dark circle in gray plot and mining blasts are in blue circles gray circles in gray plot Triangles symbols are seismic stations Diamond symbols are mining fields and square symbols are cities 53 110W 108 W 106 W 104 W 102W 38 N 5 e 38 N 36 N 34 N 32 N 32 N 30 N gz a ee re 110 W 108 W 106 W 104 W 102 W Figure 4 8 Epicenter distribution of seismic events around New Mexico January 2002 April 2002 Earthquakes are in red circles dark circle in gray plot and mining blasts are in blue circles gray circles in gray plot Triangles symbols are seismic stations Diamond symbols are mining fields and square symbols are cities 110 W 108 W 106 W 104 W 102 W 38 N A 3 N 36 N 36 N 34 N 34 N 32 N 32 N 30 N LJ Se EI 30 N 110 W 108 W 106 W 104 W 102 W Figure 4 9 Epicenter distribution of seismic events around New Mexico May 2002 August 2002 Earthquakes are in red circles dark circle in gray p
43. ircle in gray plot and mining blasts are in blue circles gray circles in gray plot Triangles symbols are seismic stations Diamond symbols are mining fields and square symbols are cities 58 110 W 108 W 106 W 104 W 102 W 38 N f y 38 N 36 N 36 N 34 N 32 N 32 N 30 N LJ SS 30 N 110 W 108 W 106 W 104 W 102 W Figure 4 13 Epicenter distribution of seismic events around New Mexico September 2003 December 2003 Earthquakes are in red circles dark circle in gray plot and mining blasts are in blue circles gray circles in gray plot Triangles symbols are seismic stations Diamond symbols are mining fields and square symbols are cities 59 110 W 108 W 106 W 104 W 102W 38 N p 38 N 36 N 36 N 32 N 32 N 30 N LJ p le 30 N 110 W 108 W 106 W 104 W 102 W Figure 4 14 Epicenter distribution of seismic events around New Mexico January 2004 April 2004 Earthquakes are in red circles dark circle in gray plot and mining blasts are in blue circles gray circles in gray plot Triangles symbols are seismic stations Diamond symbols are mining fields and square symbols are cities 60 110 W 108 W 106 W 104 W 102 W 38 N A 36 N 34 N 32 N 32 N 30 N LJ ee 30 N 110 W 108 W 106 W 104 W 102 W Figure 4 15 Epicenter distribution of seismic events around New Mexico May 2004 June 2004 Earthquakes are in red circles dark circle in gray pl
44. ith the results of careful manual location 41 37 T p 36 F 6 35 gt 34 4 33 4 32 boundary O manual wave pair 31 L l I 110 109 108 107 106 105 104 103 102 101 37 TT T T T 1 T wg 36 a J k SS cee a a amp of 35 gt 34h 4 33 a oO 32 x A boundary manual event pair 31 L L i L L 4 I 110 109 108 107 106 105 104 103 102 101 Figure 3 7 Comparison of waveform pair cross correlation matching method upper and event pair cross correlation matching method bottom Both methods were used to process the same data in the month of September 2003 The circle symbols are epicenters of manual location and the star symbols are epicenters given by automatic method The line between each star circle symbol pair is the discrepancy between manual and automatic locations In Figure 3 7 seismic data from the same month September 2003 were processed using the waveform pair cross correlation method and the event pair cross correlation method with the same initial condition The comparison between results given by manual location result and the 42 results given by the two automatic methods shows that the event pair cross correlation matching method is more robust than the waveform pair cross correlation matching method the mislocation vectors are smaller and fewer 2002 04 REF 81
45. lation of Matseis PLRR Download the matseis_plrr tar gz file from Wenzheng s web site and save it to the desired location To extract the files use gunzip matseis_plrr tar gz tar xvf This will create three directories plrr example and input The important directory here is plrr The plrr directory contains the core code necessary for running this package Three environmental variables will be set on startup or can be set prior to startup if needed 2 Start up of Matseis PLRR Before you run the package of PLRR please make sure that Matlab and Matseis are already installed on the same Unix computer Plrr is an executable c shell script used to set environmental variables MATSEIS HOME EXAMPLE MS_CONFIG_FILE starts matlab and finally starts the PLRR package The MATSEIS_ HOME EXAMPLE and MS_CONFIG_FILE directory paths will have to be modified to match your system before attempting to use it The following are the contents of plrr lines starting with are considered commented bin csh MatSeis PLRR startup script for unix users set home directory of MatSeis startup executable setenv MATSEIS HOME bin matseis 1 7 set example home directory location e setenv EXAMPLE users wyang pIrr_package example config file from command line setenv MS_CONFIG_FILE users wyang plrr_package example example config start matlab with no desktop add the path to MATSIES HOME and finally start PLRR matlab nodesk
46. lot and mining blasts are in blue circles gray circles in gray plot Triangles symbols are seismic stations Diamond symbols are mining fields and square symbols are cities 55 110 W 108 W 106 W 104 W 102 W 38 N H I 38 N 36 N 36 N 34 N 32 N 32 N 30 N en _ un u 110 W 108 W 106 W 104 W 102 W Figure 4 10 Epicenter distribution of seismic events around New Mexico September 2002 December 2002 Earthquakes are in red circles dark circle in gray plot and mining blasts are in blue circles gray circles in gray plot Triangles symbols are seismic stations Diamond symbols are mining fields and square symbols are cities 56 110 W 108 W 106 W 104 W 102 W 38 N A H 38 N 36 N 34 N 32 N 32 N 30 N m E ee 110 W 108 W 106 W 104 W 102 W Figure 4 11 Epicenter distribution of seismic events around New Mexico January 2003 April 2003 Earthquakes are in red circles dark circle in gray plot and mining blasts are in blue circles gray circles in gray plot Triangles symbols are seismic stations Diamond symbols are mining fields and square symbols are cities 57 110 W 108 W 106 W 104 W 102 W 38 N f en 36 N 36 N 34 N 34 N 32 N 32 N 30 N el a 110 W 108 W 106 W 104 W 102 W Figure 4 12 Epicenter distribution of seismic events around New Mexico May 2003 August 2003 Earthquakes are in red circles dark c
47. lue of fand g respectively The purpose of the summation in the denominator is to normalize the cross cortelation coefficient so that similarities between several time series pairs can be compared irrespective of amplitude 24 Figure 2 1 shows how cross correlation works using two seismograms When two equal length waveforms are aligned and cross correlated the possible sample lags range from N 1 to N 1 400 T 200 counts oO 200 400 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 1000 500 a 500 F 1000 l l l l i l 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 0 5 J o nee dN me a l i 0 0 2 0 4 0 6 0 8 1 1 2 1 4 1 6 1 8 2 Figure 2 1 Cross correlation coefficient curve bottom plot between waveforms of two similar earthquakes recorded by station WTX The upper waveform is an earthquake that occurred at 1997 12 04 02 15 and the middle waveform is an earthquake that occurred at 1997 12 03 21 33 Both waveforms have 100 s of data with a sampling rate of 100 samples per second The maximum cross correlation coefficient is 0 9 at time lag equaling zero the time segments were chosen to align on the P arrivals Since cross correlation between two limited time series can be expressed as convolution between one limited time series and the time reverse of the second limited time series cross correlation in seismolog
48. n this reference dataset varies considerably Figure 3 3 For example the Socorro subnet is located in the center of New Mexico and these stations may receive higher quality waveforms for events occurred around the Socorro Magma Body The subnet in Carlsbad in the southeastern corner of New Mexico is very sensitive to nearby events but are nearly blind to microearthquakes occurring in distant source regions The 33 stations located outside New Mexico are similarly most sensitive to small events occurring nearby For example station TUC typically receives high quality waveforms only from Morenci open pit mine in our area of interest a CPRX a sp a eD CL2B Figure 3 2 Epicenter distribution of reference seismic dataset for New Mexico as of June 2004 153 events selected from the seismic data of 39 months 140 120 100 80 60 number of waveforms 40 20 El I FL FE IESE PE FS IIS EO CE SE SED SL station name Figure 3 3 Station contributions to the reference seismic dataset 153 events 34 As opposed to the traditional automatic seismic phase detection algorithm the PLRR is an adaptive automatic seismic phase detection process Its seismic phase detection capability will ultimately depend on the coverage and quality of the reference seismic dataset When a new type of event occurs that does not correlate well with any reference event with CC over a
49. ngpath TODAY origin die Could not open file flock FILE exclusive_lock print FILE out n close FILE end of append ORIGIN table COU loop end print update wfdisc table n system cat outgoingpath subpath TODAY TODAY wfdisc gt outgoingpath TODAY wfdisc 85
50. nts and the use of the spectrogram cross correlation technique greatly improves the accuracy and stability of local seismic event automatic location 100 sol 60 F 407 Average mislocation error km 307 207 Ta Rs 10 70 80 90 100 110 120 130 140 150 160 Number of events in reference dataset Figure 3 9 The relationship between the average mis location error and the number of events in the reference dataset After processing the monthly seismic data for three years I can get a plot of the average location error of PLRR results to manual results against the increase of events in the reference dataset figure 3 9 As the mislocated seismic events of the previous month are absorbed by the reference dataset the mislocation error of the next month will decrease according to the trend in figure 3 9 However this will not apply when the waveforms of the new events do not 44 match any in the reference data set Several point around 150 in the x axis have higher location errors even through the reference dataset has more events This happened because in that months e g October and November 2003 a large earthquake swarm occurred in SMB and there were many cases of double events on one waveform Figure 3 10 making it more difficult for the automatic method to find the correct location since it uses the whole waveform window to carry out cross correlation 88 143 89 143 002 11 04 19 02 50 19
51. orward whenever I needed it My most intense collaboration has been with Darren Hart Darren gave me lots of help and ideas in my quest to understand Matseis and develop PLRR I cannot remember how many weekends we spent working together in MSEC 345 Darren also proofread this independent study very carefully and provided many valuable suggestions Dr Allan Sanford and Dr John Schlue gave me valuable instruction on how to pick seismic phases especially the reflected phases off the Socorro magma body My discussions with them never failed to ignite my interest in using similar waveforms to study crustal structure above the mid crustal magma body The graduate students of the New Mexico Tech Geophysics Department both past and present comprise such a superb research group The opportunity to bounce ideas off so many excellent minds has been priceless The upcoming Dr Daniel Bowman helped me process seismic events in New Mexico in the last month of this project and proofread this independent study work My fascination with the physical world is undoubtedly due to the influence of my parents When I was a young child they often brought me books about famous scientists and the natural world These stories had a lasting impression on me The accumulation of knowledge and inspiration I inherited from them has greatly contributed to my current fascination with earth science research This work has resulted from a project supported in part by the Def
52. ot and mining blasts are in blue circles gray circles in gray plot Triangles symbols are seismic stations Diamond symbols are mining fields and square symbols are cities Gis Chapter 5 SUMMARY SUGGESTIONS AND CONCLUSIONS Summary The intention of this independent study was to find a way to automatically process seismic data for New Mexico to save waveform data with triggered signals to identify seismic events to pick seismic phases and to find their locations and use this method to locate seismic events in New Mexico from 2000 to 2004 The Earthworm provides a front end for the processing of seismic data for the New Mexico network and produces automatic triggered events based on the STA LTA algorithm However less than half of the triggers are typicallyseismic events because of various sources of transient noise By considering the seismic characteristics of New Mexico seismograms source regions are somewhat independent of each other and waveforms of events occurring in the same source region are relatively similar we developed an automatic program PLRR to detect seismic events and find seismic phases and locations PLRR uses a reference event waveform database that includes seismic events from all possible source regions of New Mexico to compare an unknown event with each event in the database with the waveform cross correlation technique and associate new events with similar past events prior to manual review The PLRR pro
53. own in Figure 1 8 Events from the SSA usually have very sharp P and S arrivals and some of them have observable SzS SzS or PzP magma body upper surface reflected phases The S P time is about 2 seconds at LEM for events from the SSA region Figure 1 8 c The waveforms of SSA events decay very quickly into the background noise level since most earthquakes observed from the SSA are microearthquakes with magnitudes of 1 ot less Mining events from the Mt Taylor region have relatively weak S phases and obvious Lg coda caused by the crustal resonance of the seismic waves personal communication with Darren Hart The S P time at LAZ is about 17 seconds Figure 1 8 b Blasting activity has also been observed in the region north of Gallup New Mexico Gallup is further to the northwest of Mt Taylor from station LEM The Pg phase arrivals from these events are as weak as the Sg phase arrivals The S P time is about 29 seconds for events in the Gallup region Figure 1 8 a Pn phases are the first arrivals and can be clearly detected on the waveforms The waveforms from the Tyrone copper mine field have obvious Pn Pg and Sg phases The S Pg time is about 29 seconds Figure 1 8 d The waveforms from the Santa Rita open pit mining field only have obvious Pg and S phases The S P time is about 25 seconds Figure 1 8 f 18 The waveforms from the Morenci open pit mining field have obvious Pn Pg and S phases The S Pg time is about
54. quakes were 1 Mining blasts almost always occur during daytime working hours 8 00 am 6 00 pm 2 The epicenters of mining blasts are associated with known major mining fields _46 110 W 108 W 106 W 104 W 102 W 38 N A 38 N 36 N 36 N 34 N 34 N 32 N 32 N 30 N LJ a ee 30 N 110 W 108 W 106 W 104 W 102 W Figure 4 1 Epicenter distribution of seismic events around New Mexico July 1997 August 1997 Earthquakes are in red circles dark circle in gray plot and mining blasts are in blue circles gray circles in gray plot Triangles symbols are seismic stations Diamond symbols are mining fields and square symbols are cities AT 110 W 108 W 106 W 104 W 102 W 38 N a m 36 N 36 N 34 N 32 N 32 N 30 N Ld ee ee 30 N 110 W 108 W 106 W 104 W 102 W Figure 4 2 Epicenter distribution of seismic events around New Mexico September 1997 December 1997 Earthquakes are in red circles dark circle in gray plot and mining blasts are in blue circles gray circles in gray plot Triangles symbols are seismic stations Diamond symbols are mining fields and square symbols are cities 48 110 W 108 W 106 W 104 W 102 W 38 N f a 36 N 36 N 34 N 34 N 32 N 32 N 30 N f ee 110 W 108 W 106 W 104 W 102 W Figure 4 3 Epicenter distribution of seismic events around New Mexico January 1998 February 1998 Earthquakes are in red circles da
55. rk circle in gray plot and mining blasts are in blue circles gray circles in gray plot Triangles symbols are seismic stations Diamond symbols are mining fields and square symbols are cities 49_ 110 W 108 W 106 W 104 W 102W 38 N A p 38N 36 N 34 N 32 N 32 N 30 N Ld re E N 30 N 110 W 108 W 106 W 104 W 102 W Figure 4 4 Epicenter distribution of seismic events around New Mexico May 2000 August 2000 Earthquakes are in red circles dark circle in gray plot and mining blasts are in blue circles gray circles in gray plot Triangles symbols are seismic stations Diamond symbols are mining fields and square symbols are cities 50 110 W 108 W 106 W 104 W 102 W 38 N f N 36 N 36 N 34 N 34 N 32 N 32 N 30 N 5 ee a ee Se 30 N TH 108 W 106 W 104 W 102 W Figure 4 5 Epicenter distribution of seismic events around New Mexico January 2001 April 2001 Earthquakes are in red circles dark circle in gray plot and mining blasts are in blue circles gray circles in gray plot Triangles symbols are seismic stations Diamond symbols are mining fields and square symbols are cities 51 110 W 108 W 106 W 104 W 102 W 38 N P en N 36 N 36 N 34 N 32 N 32 N 30 N m Fe er en 110 W 108 W 106 W 104 W 102 W Figure 4 6 Epicenter distribution of seismic events around New Mexico May 2001 August 2001 Earthquakes are in red circles dark c
56. rrelation robustness is improved relative to what it would be if just using the waveforms had been used However a drawback of the Waveform pair cross correlation matching method is its instability This technique produced too many false associations leading to the development of the following method Event pair cross correlation matching The second cross cortelation method tested in PLRR is event pair cross correlation matching The schematic of this method is shown in Figure 3 5 This method stacks the CC curves of waveform pairs between the unknown event and waveforms grouped from each event in the reference database Then it finds the reference event that has the highest stacking CC value 137 by taking the mean median of stacking CC curves and defines it to be the most similar event to the unknown event 003 09 2006 37 30 06 38 00 06 38 30 06 3900 06 39 30 06 4000 06 40 eis Figure 3 5 Illustration of event pair cross correlation matching method In this method waveforms with different sampling rates for example the sampling rate of CAR is 100 samples per second and the sampling rate of ANMO is 40 samples per second have to be converted to the same sampling rate because the stacking CC curve needs to stack the CC curves from different stations point by point I resample the low sampling rate waveform to be 100 samples per second Also in this method data windo
57. se Rowe et al 2002 developed an automatic adaptive algorithm to refine the phase picking but this method requires that all waveforms have initial phase picks produced by an analyst or autopicker To use the method developed by Rowe et al to relocate earthquakes that occurring in New Mexico it would be necessary to reprocess the original triggered seismic events generated by Earthworm The following chapters of this report will discuss an adaptive automatic method which uses a cross correlation technique to detect the location of seismic events and identify initial phases 23 Chapter 2 WAVEFORM CROSS CORRELATION TECHNIQUE Waveform Cross correlation In linear vector space the distance between two vector spaces e g two time series with a relative lag of t could be measured by L 2 norm Euclidean distance as dfg f elt ro se Niat 1 0 The definition of cross correlation for two continuous time series f t and g is c f g FOE odr 2 c g is the cross correlation coefficient for the two time series at a particular time lag When f and g f are discrete digitized data the normalized cross correlation expression is DIG mu m a m Seam Where 7 0 1 2 N 1 and N is the number of samples in fand g c is the cross correlation coefficient value at sample lag d and it ranges from 1 0 to 1 0 w and m are the mean of time series fand g respectively f and g are the ir or 7 d th sample va
58. shallow magmatic materials into the crust at the depths above 12 km 14 12 The largest recent seismic swarms inside the SSA occurred during October 30 and November 2 2002 with about 246 micro earthquakes recorded automatically by the Earthworm real time system Most of them had obvious waveforms recorded at station WTX Figure 1 5 OCT 30 2002 OCT 31 2002 0 15 0 15 1 15 a 41 15 12 15 Pe 2 15 3 15 3 15 4 15 4 15 5 15 5 15 6 15 6 15 7 15 715 s15 8 15 9 15 9 15 410 15 10 15 11565 F 41115 12 155 12 155 13 15 13 15 14 15 14 15 15 15 15 15 16 15 16 15 17 15 17 15 18 15 18 15 19 15 19 15 20 15 20 15 21 15 F 421 15 22 15 22 15 423 15 i i 23 15 o 1 2 3 4 6 7 8 9 10 11 12 13 14 15 0 2 3 4 6 7 8 9 10 t 12 13 14 15 TIME MINUTES TIME MINUTES Nov 01 2002 Nov 02 2002 T 0 15 F T T 30 15 1 15 1 15 216 p 12 15 3 15 3 15 4 15 F 4 15 5 15 5 15 6 15 F 6 15 7 15 7 15 8 15 8 15 19 15 9 15 1015 10 15 11 15 5 11 15 12 155 12 155 13 15 13 15 14 15 14 15 15 15 15 15 16 15 16 15 17 15 17 15 18 15 18 15 1915 19 15 20 15 20 15 21 15 21 15 22 15 22 15 23315 t Ll Dr 23 15 o 1 2 3 4 11 12 13 14 15 0 1 2 3 4 71 12 13 14 15 ae Soe RD TIME MINUTES Figure 1 5 WTX pseudo helicorder plots show the SSA swarm during Oct 30 Noy 2 2002 Each line represents one event The Raton area is another relatively active seismic region inside the New Mexico In t
59. test sspe near ea 41 Chapter 4 Earthquake location in New Mexico 1997 2004 Epicenter distribution of seismic events in New Mexico 1997 2004 ee einer 46 Chapter 5 Summary Suggestion and Conclusions SUMMAE een 62 DUSCCS IONS SS Ya deie ean ge death an ce tae iene cage A Sas 63 FON Se cigs et Monae tant tel alee aot ee at Oh Monee cela ti tah 64 References 66 Appendix I THE USER S MANUAL OF PLRR Appendix II PLRR source code list and related codes in automatically processing of the SC Earthworm system seismic data Chapter 1 INTRODUCTION Background Seismicity in New Mexico Due to its long distance from major tectonic plate boundaries and weakness of internal tectonic activity the seismicity in the state of New Mexico is relatively quiet compared to very active seismically regions such as California Figure 1 1 However historic activity shows a moderate level of earthquake activity Figure 1 2 Earthquake epicenter distribution in New Mexico and surrounding regions 1962 1995 using the data from 3 Dashed line outlines the boundary of the state of New Mexico 9 NM regional Seismicity SSA Seismicity 4 T T T 3 5 T T T 3 5 N in log yy Number of event fog 49 Number of event 5 y 076 45 y 077x 1 6 o 1 1 4 3 5 2 3 Magnitude Magnitude Figure 1 3 a b b value in for New Mexico regional seismicity and New Mexico seismicity associated to th
60. top r addpath SMATSEIS HOME pirr Once the above changes have been made to the plrr file check that it is found in the current search path by typing Which plrr should produce a similar results as below on our system uset wyang plrr_package There are three database used by PLRR input reference and output The directory locations of these databases are set from the main GUI window under the File pulldown menu The contents inside input file are nearly the same as the contents inside input file used in Matseis except that we ve added several lines to set the directory for reference set Matseis only has the input and output directories setting After finishing setting the directories you can launch the GUI window of PLRR by executing plrr script at the UNIX command prompt Figure shows the main GUI window of PLRR _72 File i Orig Wim Ar TT 5 Pick Locn Updat Tool Help ZOOM undof ful faring data N yf xf yf OUT yf xf y ORAS yf xf y MOVE up iow tet right Figure The main GUI window of PLRR is shown here The origin station and waveform windows are used to display data just like in MatSeis The menu selection structure of PLRR is shown in figure 2 MatSeis PLRR File Orig Wfm Arr TT Pick Locn Update Help 4 Reconigure gis N l Manual MS locator Update boite oo Monthy varsi Input Database a eked Version hes Same Auto pkk _ tables Information setup in by
61. utomatic generated files automatically How to find the more accurate initial phase information and locations for the real seismic events using an automated system To address these problems we developed an adaptive automatic phase picking and epicenter locating program The program is built upon the Matlab environment under UNIX and is supported by Matseis 1 7 The package is named PLRR Pick Locate Repick Relocate The fundamental idea in PLRR is to use cross correlation instead of the traditional STA LTA 30 algorithm to identify seismic events then associate phase arrivals for certain phases that we are interested in e g P S Pn The most important function of PLRR is to identify the most similar event from a reference database and then assign phases and location information using spectrogram cross cortelation No preliminary information is needed for the unknown event The diagrammatic data processing of PLRR is illustrated in figure 3 1 Perl Eart Scripts Dai or high quality waveform Finished by PLRR Figure 3 1 Flowchart of PLRR The following sections will discuss the realization of PLRR step by step Data Format conversion The seismic event set format currently accepted in MatSeis 1 7 is the CSS3 0 format which consists of waveform data and several tables containing origin time arrival time relationship between waveforms and identifying numbers Several of the tables necessary
62. ventdir print ORID orid COU n copy new event of today to yearmonth directory print cp rf event outgoingpath subpath TODAY 001 n system cp rf event outgoingpath subpath STODAY oo 1 system rm outgoingpath subpath TODAY oo 1 init system rm outgoingpath subpath TODAYS oo 1 quicklook system rm outgoingpath subpath TODAYS oo 1 repick system rm outgoingpath subpath TODAYS oo 1 saclist convert SAC into CSS format using SAC2000 print raid users wyang outgoing sac2css_sc outgoingpath subpath TODAY oo 1 in processing n system raid users wyang outgoing sac2css_sc outgoingpath subpath TODAY o00 1 STODAY print n n stack the CSS tables and fix the problem of ORIGIN table generate a CSS table for events of today open WOUT outgoingpath subpath TODAY 00 1 TODAY origin originline lt WOUT gt close WOUT temp1 substr originline 0 47 temp2 substr originline 48 7 temp3 substr originline 57 180 format output the origin number between 1 1000 if Gorid COU lt 10 84 temp2 orid COU if Gorid COU lt 100 amp amp orid COU gt 10 temp2 orid COU if Gorid COU lt 1000 amp amp orid COU gt 100 temp2 orid COU if Gorid COU gt 1000 temp2 _ orid COU out temp1 temp2 temp3 append the public origin file open FILE gt gt outgoi
63. vents in the Mt Taylor and Tyrone areas Figure 1 8 From the point of view of waveform cross correlation this type of waveform only one strong phase is very similar to waveform of earthquake occurring very close to the station because local earthquakes have a very short S P time In the time domain the amplitude of S phases might be nearly as low as the background noise but in certain frequency bands the amplitude of S phases stands out from the background noise So one way to solve the attracting problem is to convert the waveform data from the time domain into the frequency domain and only keep the frequency bands with both strong P and S energy for the cross correlation Using this knowledge I developed the spectrogram event pair cross correlation matching method to calculate the stacking cross correlation curve in several frequency bands 39 The distance from major mining fields around New Mexico to the center of Socorro sub network is more than 200 km Waveforms coming from such sources usually have very weak S phases As shown in Figure 3 6 a waveform from Tyrone recorded by station LAZ has distinct Pn and P phases but a weak S phase The normal waveform cross cortelation technique has a strong tendency to detect Pn as P and P as S thereby potentially misidentifying this event as an earthquake occurring around the SMB frequency o 1000 Oo 2 0 Msn enon lt 1000 l i I 0 10 20 30 4
64. w lengths for the waveform cross correlation calculation have to be taken as the same because waveforms recorded by different stations may not be of the same length in a given event Event pair cross correlation matching is more robust than the waveform pair cross correlation matching It considers the cross correlation curves of all waveform pairs of events as a whole _38 instead of treating them separately requiring the waveforms to be similar at all stations for a good match It must match the time order of seismic phase arrival times of an event across the network The contribution of each waveform pair ideally enhances the stacking waveform cross correlation curve Spectrogram event pair cross correlation matching Spectrogram event pair cross correlation matching method is an enhanced method of the event pair cross correlation matching When I first processed the seismic data of New Mexico using the event pair cross correlation matching method I found that there exists an attracting phenomenon seismic events occurring further away from the center of the SC network are sometimes identified as eatthquakes occurring around the center of SC network One fundamental reason for this phenomenon is that the reference dataset is incomplete and more events should be added to it Another reason is that the waveforms of earthquakes occurring further away from the station usually have a strong P and weak S phase or vice versa e g seismic e
65. y study has also been documented as Papanicolaou 2004 _25 IOS RP ER 4 Ku Where y is the station and x and x are two sources I y is the maximum cross correlation coefficient of two time series waveform data P x and time reverse P x at time lag t x y T x y Cross correlation techniques have been widely used in seismological research Song 1996 used the cross correlation method to compare seismic waveforms passing through the inner and outer core during the last 30 years and found a systematic time shift between the P P DF and P amp P BC phases Aster et al 1991 used the waveform cross correlation method to evaluate possible temporal anisotropic variations between similar waveforms and ruled out previous claims of using shear wave splitting for earthquake prediction When we are going to use the cross correlation method to study seismic waveforms there are several necessaty conditions First the geological structure background of ray paths between several epicenters and the seismic station does not vary greatly i e earth structure along commonly traveled ray paths are similar Second the source radiation patterns of closely spaced earthquakes are similar i e source characteristics are compatable Cross correlation coefficient Matrix The similarity of two waveforms is measured by the cross correlation maximum and is quantitatively expressed in the cross correlation coeffici

an adaptive, automatic phase- picking and epicenter locating

Contents

Download Pdf Manuals

Related Search

Related Contents