A plan for seismic location calibration of 30 IMS stations in Eastern Asia

Begin summary:

From March 2000 to September 2003 we developed and carried to completion a consortium effort to improve the capability to locate seismic events based on data acquired by 30 International Monitoring System (IMS) stations in East Asia.

We developed and tested Source Specific Station Corrections (SSSCs) for Pn and Sn travel times at these 30 IMS stations (or suitable surrogates), and for 127 other stations used for validation testing. The SSSCs were initially computed by the method of Bondár (1999), using regionalized 1-D travel-time curves established after extensive review of published studies including many from the Russian literature. Subsequently we developed a 3-D model of the P-wave velocity for East Asia (using a set of 36 different regions in each of which we obtained velocity as a function of depth), and used 3-D ray tracing in the latter model to compute SSSCs. These model-based SSSCs were refined empirically by applying a kriging algorithm to travel-time residuals for ground-truth (GT) events. Off-line validation tests were performed by evaluating travel-time residuals and by relocating GT events, with and without using SSSCs. To test the validity of the model directly, relocation tests were first performed using model-based SSSCs without kriging. Tests were then performed to evaluate the kriged SSSCs, using a leave-one-out approach so that events were not simultaneously used to both compute and test the SSSCs.

Nuclear explosions dominated our ground-truth datasets in the first two years of this project. In particular we used source parameters for Soviet-era Peaceful Nuclear Explosions (PNEs). But this approach, while quite satisfactory for calibrating stations in much of Russia and Central Asia (which made up approximately half the IMS stations we studied) could not be extended to the remaining stations, for which it was necessary to develop GT information on significant numbers of earthquakes. By use of the double-difference method and detailed fault maps, we obtained 64 GT5 earthquakes by re-analyzing the Annual Bulletin of Chinese Earthquakes (ABCE) for a 15-year period (1985 to 1999). It contains phase picks for approximately 1000 earthquakes in and near China, each year. A preliminary examination of digital waveforms for about 14,000 events, in and near China, shows that approximately 9% of them (1301 events) have the property that any one event has almost the same Lg waveform as at least one other event. These events are grouped into 494 sets of events, each of which has essentially the same short-period waveform and thus the events of each set must be within about 1 km of each other. These event sets provide a good method for assessing the quality of standard event catalogs. When combined with other information, they can provide high-quality absolute locations.

Using Pn and Sn arrival times for our GT data sets, we relocated 525 events recorded by various combinations of 140 regional stations. Mislocations were reduced for 66% of the events using the model-based SSSCs, and for 85% of the events using model-based SSSCs refined by kriging. Median mislocation improved from 16.9 km to 11.4 km and 6.5 km, respectively. Median error ellipse area was reduced from 2,616 sq km to 1,633 sq km and 722 sq km, respectively. Error ellipse coverage (percentage of GT locations within 90% error ellipses) was 89% without using SSSCs, 91% using model-based SSSCs, and 92% using kriged SSSCs. These results were obtained for source locations, stations, and paths that sample very extensive and diverse geological provinces throughout much of Asia. The SSSCs are expected to perform, on average, as well as the test results using the model-based SSSCs, and substantially better for areas where GT calibration data were utilized to refine the SSSCs.

End summary

In March 2000, a collaborative academic-industry research consortium comprised of five institutions started an integrated series of projects, all with the goal of improving the capability to locate seismic events based on data acquired by International Monitoring System (IMS) stations in Eastern Asia.

The focus of this effort was to develop and deliver validated high-resolution travel time grids for operational use, in support of the location estimates made by the International Data Centre (IDC) of the Comprehensive Test Ban Treaty Organization, for on the order of a hundred events per day at locations around the world. For background information on IMS and IDC, see the paper by Paul G. Richards, Building the Global Seismographic Network for Test Ban Monitoring, EARTHmatters, pp 37 -- 40, Fall 1999 [ text and graphics, or link to EARTHmatters layout].

In the first project we found obtained new ground truth locations in Eastern Asia, whose errors are thought to be of the order of five km or better (so-called GT5 events). We obtained empirical travel times for many of these events, at stations operated by the IMS or near such stations. These events in Central Asia, China, Mongolia, Korea, the Indian subcontinent, and Russia, in some cases have been located principally on the basis of large numbers of signals recorded by regional networks. See further details, for GT5 events in China, in the BSSA paper by Waldhauser and Richards (2004).

In the second project, the University of Wyoming contributed observed travel times for about 3000 three-component recordings at stations widely deployed in the Soviet era to detect regional waves from 21 nuclear explosions carried out during the Deep Seismic Sounding program. This dataset is an invaluable resource for thorough calibration of major aseismic regions in Russia and Central Asia. Our personnel at the University of Wyoming were Scott Smithson, Elena Morozova, and Igor Morozov.

In the third project, Mission Research Corporation derived and tested travel time surfaces, for IMS stations, that fit the GT data and Calibration Event Bulletin data. The lead person at MRC for us was Mark Fisk and much of the work was done by Relu Burlacu.

URS Greiner Woodward Clyde contributed some ground truth data for India, Nepal, Pakistan, and much 1D and 2D modeling experience; the University of Connecticut contributed 3D modeling experience. Both these organizations, and Wyoming and Lamont, worked together in Project 4 to provide expected travel times to 30 IMS station locations in Eastern Asia. In this fourth project, detailed studies of a small percentage of our claimed GT5 events were carried out for purposes of validation of their location quality. The lead person at URS Greiner is was Chandan Saikia, and at the University of Connecticut was Vernon Cormier assisted by Anastasia Stroujkova.

Mission Research Corporation packaged all of the products of the consortium for delivery to the Center for Monitoring Research in the fifth project, including quantitative evaluation of location improvements.

An Experts Group Review provided initial guidance. It included Bob Engdahl and Chuck Langston.

[download a pdf file for the two-day program of the consortium's first "Experts Group Review" meeting, held at Lamont Feb 15 and 16, 2001]

This was a three-year program of work. It was completed on time, and extensively documented. An eleven-page summary paper is downloadable (it was delivered as part of the Seismic Research Review of September 2003). A 280-page final report was prepared, and it and all associated models and data were included on a CD that was distributed to about 50 addresses in September --- October 2003. Much of this material is still available on request to the PI (e-mail to:richards@LDEO.columbia.edu), who is still willing to distribute copies of the CD, and/or to make relevant information accessible from the Lamont ftp site.



Major users of seismic data include:

  1. the national and international groups now being organized to monitor compliance with the Comprehensive Nuclear Test-Ban Treaty (CTBT);
  2. researchers who improve our knowledge of Earth's internal structure and the physics of earthquake processes; and
  3. those engaged in earthquake engineering and earthquake hazard mitigation.

Although the most basic data in seismology for all these users are seismograms, in practice the great majority of those who work with seismic data do not use seismograms directly. Instead they mostly use data products derived from seismograms.

The most important of these products, are bulletins of seismicity.

In the last 20 years there have been enormous improvements in the quality and quantity of seismograms, associated with the development of broadband feedback sensors and techniques of digital recording to permit high dynamic range across wide bands of frequency.

There is ongoing revolutionary improvement in access to seismogram data, as satellite communications and the internet spread even to remote locations.

It has therefore been frustrating to find that the quality of the principal data product derived from seismograms acquired internationally, the global bulletin of seismicity, has not yet seen the types of radical improvement needed by any of the user communities, 1 through 3, above.

The US Geological Survey (USGS) and the International Seismological Centre (ISC) publish their bulletins months to years in arrears, using volunteered data, and methods of analysis that essentially have not changed for sixty years. These are very useful bulletins, and their quality has greatly improved because of the increased number of reported signal detections.

The Reviewed Event Bulletin (REB) of the CTBT monitoring community, produced since 1995 January 01 by GSETT-3 and the PIDC and now by the IDC in Vienna, is vastly improved over the other global bulletins in its timeliness of publication. However, both the REB location estimates, and the estimates of their uncertainty (error ellipses), require improvement.

It appears that the principal reason for inaccuracies in the REB locations is lack of a sufficiently good model of Earth structure, and specifically of travel time information. It is desirable to calibrate each IMS station so that in effect the location of a new event can be located with reference to another event, whose location is known accurately and which, preferably, is not far from the new event. By using a sufficiently large number of calibration events, whose location is accurately known and whose signals are detected reliably at IMS stations, it is possible to generate a station-based travel time surface (a function of distance and azimuth), for each seismic phase.

As noted in the report of the first Oslo Workshop on IMS Location Calibration (January 1999, which led to the paper CTBT/WGB/TL - 2/18):

"such calibration is necessary in order to significantly improve the location precision of internationally reporting earthquake agencies,"


"no attempt has so far been made to include such corrections in routine location processing on a global scale."

Our consortium project carried out such an approach to calibration for 30 IMS stations.



The IMS stations in East Asia which were the subject of this project, are listed in Table 1 with station coordinates as originally given in Annex 1 to the CTBT Protocol. Not all of these sites currently have operational IMS stations. However, is some such cases there are non-IMS stations which are operating at or near the IMS site, and in other cases stations have operated in the past, near the IMS site. In general we refer to such non-IMS stations as surrogate stations, and their data can potentially be used to assist in building up the necessary station-based travel-time data set for purposes of obtaining the types of travel-time surface needed at the IDC for every IMS station site.

Table 1. The list of 30 IMS stations for which calibration was carried out in this project.
IMS code Country Station name latitude longitude
PS12 China Hailar 49.27 119.74
PS13 China Lanzhou 36.09 103.84
PS23 Kazakstan Makanchi 46.80 82.00
PS25 Mongolia Javhlant 47.99 106.77
PS29 Pakistan Pari 33.65 73.25
PS31 Republic of Korea Wonju 37.50 127.90
PS33 Russian Federation Zalesovo 53.94 84.81
PS34 Russian Federation Norilsk 69.40 88.10
PS35 Russian Federation Peleduy 59.63 112.70
PS37 Russian Federation Ussuriysk 44.28 132.08
PS41 Thailand Chiang Mai 18.80 99.00
AS7 Bangladesh Chittagong 22.40 91.80
AS20 China Baijiatuan 40.02 116.17
AS21 China Kunming 25.15 102.75
AS22 China Sheshan 31.10 121.19
AS23 China Xi'an 34.04 108.92
AS57 Kazakstan Borovoye 53.06 70.28
AS58 Kazakstan Kurchatov 50.72 78.62
AS59 Kazakstan Aktyubinsk 50.40 58.00
AS60 Kyrgyzstan Ala-Archa 42.64 74.49
AS68 Nepal Everest 27.96 86.82
AS86 Russian Federation Seymchan 62.93 152.37
AS87 Russian Federation Talaya 51.68 103.64
AS88 Russian Federation Yakutsk 62.01 129.43
AS89 Russian Federation Urgal 51.10 132.36
AS90 Russian Federation Bilibino 68.04 166.37
AS91 Russian Federation Tiksi 71.66 128.87
AS92 Russian Federation Yuzhno-Sakhalinsk 46.95 142.75
AS93 Russian Federation Magadan 59.58 150.78
AS100 Sri Lanka Colombo 6.90 79.90

Table 2 lists our knowledge (as of year 2000) of the availability of data at each of the 30 IMS sites in Table 1, whether or not there is an IMS station operating at the site. In the case where no IMs station is operating, we list some appropriate surrogate stations.

Our basic approach, was to acquire lists of reliably located seismic events in Eastern Asia, preferably occurring since the beginning of publication of the REB on January 1, 1995, and large enough to be included in the REB. From such events, of GT5 quality or better, we obtained the picked arrival times at IMS stations and thus built up an empirical set of station-based travel-times for events of accurately known location. We used these to krige on model-based travel times.

Table 2. Summary of 30 IMS sites, operating IMS stations, and surrogate stations, useful for acquiring phase picks and waveforms for the 30 East Asian stations in Table 1.
Code Phase data Digital waveform data Analog waveform data
IMS ISC PIDC Operator Operation Source
LZH LZH CDSN 1986/06- DMC  
MAK KZ/GSN 1994/07- LDEO  
ZAL NVS,ELT ZAL IDC (ZAL), LDEO (ELT) 1995/01 (1998/08-, ELT) IDC, ZAL (LDEO, ELT)  
NRI NRI NRI GSN (NRIL) 1992/12- DMC/IDC 1964-, Obninsk
PDY BOD PDY IDC 1995/01- IDC  
USK VLA CDSN (MDJ) 1986/10- DMC    
KMI KMI CDSN (KMI) 1986/06- DMC  
SSE SSE CDSN (SSE) 1986/06- DMC  
XAN XAN CDSN (XAN) 1992/11- DMC  
BRVK KZ/GSN 1994/07- LDEO  
KURK KZ/GSN 1994/07- LDEO  
AKTO KZ 1994/09- LDEO  
AAK FRU GSN (AAK) 1990/10- DMC  
EVN DMN,KKN 1991/06-11 ING  
SEY SEY GSRAS/GS (SEY) 1990/09- DMC/GS 1969- Magadan
TLY IRK GSRAS/GSN (TLY) 1990/10- DMC 1964- Irkutsk
URG Sogda 75/01-76/10 CSE
BIL ILT GSRAS/GSN (BIL) 1995/08- DMC Bilibino 64- Magadan
TIXI TIK GSRAS/GSN (TIXI) 1995/08- DMC 1964- Obninsk
MA2 MAG,MGD GSRAS/GSN (MA2) 1993/09- DMC 1964- Magadan

Code - IMS: station code listed on Annex 1, CTBT Protocol, September 1996

Phase data - ISC and PIDC (phase data are available from these stations of the ISC and PIDC).

Operators of digital stations, often part of joint programs, are:

GSN = Global Seismographic Network;

CDSN = Chinese Digital Seismographic Network;

KZ = Kazakstan Broadband Seismographic Network (NNC-RK/LDEO);

GS = GeoScope;

GSRAS = Geophysical Survey, Russian Academy of Sciences.

Data sources are:

IDC = International Data Centre for IMS;

DMC = IRIS Data Management Center

ING = The National Institute of Geophysics, Italy.

The AWRE operation of GBA has ended, with this station handed over to local operation, but much relevant data for this array is easily available.

From phase picks and waveform data, we made our own location estimates, including waveform studies of the depth, in order to validate our conclusions as to the quality of the locations.

For the major aseismic areas of Eastern Asia (for example, for much of the northern part of this region, which is in Russia), such an approach cannot be used. However, we were fortunate in that major reflection/refraction profiles were carried out in this region during the Soviet era, in the Deep Seismic Sounding program (DSS). Table 3 lists the DSS data which were analysed in our consortium project by the University of Wyoming -- though some of the RUBY profiles were not used. From such arrival time picks, and searches of the literature, it was possible to generate 2D and 3D regional models and hence travel-times to the IMS station sites.

Table 3. 19 PNEs and two weapons test site explosions in the DSS database; numbers of 3-component recordings; and data delivery schedule. Column 2 for the four RUBY shots, gives the numbers of 3C stations in each of the two RUBY profiles. These data were analysed by the University of Wyoming. Digital data are now in hand, with the exception of profiles RUBY-1 and RUBY-2. Much of the work of checking station locations and arrival-time picking was be done in the first 12 months of the project.

Explosion name

Number of 3-component recordings after editing Delivery date of travel-time data, after start of the project Delivery date of signal-to-noise ratios, velocity, amplitude data, spectral ratios after start of the project
QUARTZ-2 316 3 months 21 months
QUARTZ-3 290
QUARTZ-4 245
CRATON-1 126 7 months 24 months
CRATON-2 218
CRATON-3 198
CRATON-4 194
KIMBERLITE-1 162 11 months
RIFT-1 192 15 months 27 months
RIFT-3 122
RIFT-4 154
METEORITE-2 130 18 months
RUBY-1 218+123 21 months 30 months
RUBY-2 87+151
KAZ 1 (RUBY) 103+29
KAZ 2 (RUBY) 25+54

We note that a sophisticated modeling effort is the only way to set up the required travel time grid for each IMS station, in two important cases: from large aseismic regions; and to IMS station locations where no station or nearby surrogate now exists. But we are well aware that it would be inappropriate for the IDC to rely for its monitoring operations on purely 3D calculations in a 3D model. Therefore we made great efforts to search for appropriate validation, to the extent possible, of any predicted effects on travel times caused by 3D structure. In some cases we used data from analog stations, which operated all over the former Soviet Union during the period 1965 to1990 at locations that were selected in 1996 to become the sites of modern instrumentation (IMS). In addition to the CD, the 280-page final report, and the summary report mentioned above, we make available the following files that describe our project as it evolved:

[download a pdf file for a paper submitted July 2000 to the New Orleans CTBT monitoring symposium]
[download a pdf file for the two-day program of the consortium's first "Experts Group Review" meeting, held at Lamont Feb 15 and 16, 2001]
[download a pdf file for the abstract submitted July 2001 for the Jackson Hole monitoring symposium]
[download a pdf file for the paper submitted July 2001 for the Jackson Hole monitoring symposium of Sept 2001]
[download a 7 Mb pdf file (100 pages) giving the basis for SSSCs presented to the CCB in May 2002]
[download the CCB proposal approved May 23 2002, for Pn SSSCs of 14 IMS stations in East Asia]
[download a pdf file for the paper submitted July 2002 for the Ponte Vedra monitoring symposium held September 2002]

[download a 3Mb pdf file (zoomable) of our poster (shown here as a jpeg) for the Ponte Vedra meeting (the poster was composed by David Schaff)]

[download a pdf file for Richards' invited talk "A subjective overview of 24 papers on Seismic Detection and Location" given Sept 19, 2002 at Ponte Vedra, FL]
[download a pdf file for Richards' invited colloquium "Earthquake Location Location Location" given Sept 27, 2002 at Lamont-Doherty Earth Observatory]