U.S. Department of the Interior U.S. Geological Survey General Earthquake Observation System


DEPARTMENT OF THE INTERIOR
U. S. GEOLOGICAL SURVEY

Digital Recordings of Aftershocks of the
17 October 1989 Loma Prieta, California, Earthquake

by
Charles Mueller and Gary Glassmoyer

Open-File Report 90-503

This report is preliminary and has not been reviewed for conformity with U. S. Geological Survey editorial standards. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U. S. Government.

Menlo Park, California 94025

July 1990


Contents

Title

Contents

Introduction

Waveform Dataset

Field Procedure

Site Selection and Instrumentation

Sensor Orientations

Clock Corrections

Data Availability

Acknowledgements

References

Figures

Tables

Appendix


Introduction

After the 17 October 1989 Loma Prieta, central California, earthquake (291 00:04:15.25 UTC, MS=7.1), the U. S. Geological Survey recorded aftershocks at sites of seismologic and engineering interest. This report describes a seismic-waveform dataset collected from 17 October 1989 (291 UTC) to 14 March 1990 (073 UTC) with GEOS digital seismographs deployed from USGS offices in Menlo Park, California. It is a summary of field and data-playback information that is intended to facilitate use of the waveform data in seismologic and engineering studies. It includes station locations, instrumentation histories (trigger parameters, sensor parameters, clock corrections, etc.), listings of waveform records, preliminary seismicity listings, and information about data availability.

Our principal research goal is to understand the influence of seismic-source and wave-propagation phenomena on damaging ground motions, and to use this knowledge to predict hazards from future earthquakes. Source processes encompass the spatial and temporal variations in the excitation of seismic waves at the fault. Wave-propagation processes include amplification and deamplification as seismic energy radiates away from the fault through rocks and soils of variable seismic properties. Since shallow geology at a site (upper tens to hundreds of meters) can exert a relatively great influence on damaging ground motions, it is useful to separate wave-propagation phenomena into whole-path and site parts. In studying these phenomena, aftershock waveform recordings can play an important role. Compared to main shocks, aftershocks are simple earthquakes and aftershock records tend to be relatively dominated by wave-propagation effects. Geometrically, since aftershocks are distributed over the fault zone, aftershock records should resemble part of the main-shock ground motion contributed by localized rupture. This observation forms the basis of the empirical-Green's-function method wherein main-shock ground motion is modeled by summing aftershock records.

The Loma Prieta earthquake provided a rare opportunity that motivated the intensive aftershock-recording effort described in this report. In a seismically noisy urban environment like the San Francisco Bay area, such a large set of high quality seismic data would otherwise be unobtainable.


Waveform Dataset

The waveform dataset described in this report consists of 2615 three-component records (5207 if co-recorded acceleration and velocity records are counted separately) collected from 17 October 1989 (291 UTC) to 14 March 1990 (073 UTC) at 94 sites in central California. Records have been collected using portable, autonomous, ground-motion-triggered, 16-bit digital GEOS seismographs (Borcherdt and others, 1985), typically recording three components of ground acceleration and three components of ground velocity at each site. A typical record is 10 to 40 seconds in duration, including approximately 2 seconds of pre-trigger signal. (Field procedure and instrumentation are discussed in more detail below.)

Data tapes are played into a DEC PDP 11/70 computer using RDGEOS software. Records are stored one-component-per-file in a compact block-binary format. Each file consists of 512-byte header and data blocks: one integer header block (256 two-byte integer), followed by one real header block (128 four-byte real), followed by one-or-more integer data blocks (256 two-byte integer). In this report, records are identified by computer filename, a unique 13-character string constructed from the start-of-record time (UTC), component, and station name: characters 1-3 = day of year (001-366), characters 4-5 = hour (00-23), characters 6-7 = minute (00-59), character 8 = second code (A-T, where A = 0.000-2.999, B = 3.000-5.999, ..., T = 57.000-59.999), character 9 = component code (1 for vertical-component acceleration, 2 and 3 for horizontal-component acceleration, 4 for vertical-component velocity, 5 and 6 for horizontal-component velocity), character 10 = '.', and characters 11-13 = station name.

The Loma Prieta dataset is summarized in Figures 1 through 9 and Tables 1 through 7. Figure 1 shows all instrument locations and an estimate of the Loma Prieta main-shock rupture area superimposed on a map of central California. Figures 2 through 8 show station locations in finer detail at progressively larger scale. Figure 9 shows seismicity (events listed in Appendix A and B - see below) at the same scale as Figure 1. Table 1 lists records from identified seismic events by time and station - Table 1a lists all records and Table 1b lists records from events recorded at five-or-more sites. During initial processing, probable seismic events (aftershocks, earthquakes unrelated to Loma Prieta, quarry blasts, etc.) are identified by a computer algorithm that finds multiple triggers in a sliding time window; the window duration is selected just long enough to include all plausible triggers from an event (accounting roughly for traveltimes). Then, seismic events that were excluded (for example, single-station recordings) and false events that were included (for example, coincidental unrelated false triggers) are identified by inspection of seismogram plots. Multiple events (two or more events included in one file) are also identified in this way. Table 1 is complete - all seismic events have been included and all false events have been eliminated to the best of our ability. Tables 2, 3, and 4 list instrument locations, instrument parameters for a typical deployment, and detailed instrument histories, respectively. Tables 5 and 6 list clock-correction data. Table 7 is a list of the data tapes recovered from each station. At the time of writing this report in July 1990 the USGS had computed preliminary origin times and locations for many of the events listed in Table 1 - this information is summarized in Table 1, Figure 9, and Appendix A and B.


Field Procedure

Site Selection and Instrumentation

In the light of the research goals discussed above, analysis of seismologic and engineering data from the Loma Prieta main shock suggested three initial criteria for choosing aftershock recording sites. First, recorders were placed in damaged areas - the Marina district in San Francisco, for example. At sites where no main-shock strong-motion records are available, aftershock data can be used to evaluate the relative influence of source and wave propagation on main-shock ground motion and damage. Second, recorders were co-sited with strong-motion accelerographs that recorded the main shock. Aftershock records from these sites, especially near the fault zone, can be used as empirical Green's functions to model the main-shock strong motions. While the strong-motion accelerographs typically trigger on only the largest aftershocks, the more sensitive portable seismographs assure a much richer empirical-Green's-function dataset. Third, to achieve uniform spatial coverage of the recording zone, recorders were sited in areas that were conspicuously underinstrumented after the first two criteria were satisfied. Instruments were rearranged as enough data were collected to meet these initial goals; later configurations included a north-south line of recorders along the San Francisco peninsula from the source region to San Francisco, an east-west line across San Francisco Bay from Palo Alto to Fremont, an east-west line through San Jose, and several small arrays.

In a typical deployment, each GEOS recorder was independently triggered and recorded six components of ground motion at 200 samples/second/channel: three components of ground acceleration with a Kinemetrics FBA-13 triaxial force-balance accelerometer (fba) and three components of ground velocity with a Mark Products L-22D 2-Hz triaxial geophone. Typical gain settings were 6-18 dB for acceleration and 36-48 dB for velocity. This provided a good tradeoff between numbers of triggers and dynamic range: the largest ground motions clipped the geophone channels, but were well recorded on the fba channels, while small motions were well recorded on the geophone channels, but below noise on the fba channels. Instruments were deployed at ground or basement level at quiet, secure sites. Sensors were buried in dirt or fixed in place with silicone caulk to prevent sliding during strong shaking. Table 3 summarizes GEOS parameters for a typical deployment and sensor parameters. Table 4 lists detailed station histories, including exceptions to the standard deployment.

Sensor Orientations

As Table 4 shows, sensor orientations were problematic at some sites: some orientations were unreported and double-checking revealed that some reported orientations were wrong. We feel obliged to estimate the reliablity of this information for readers who need it, so we have codified our subjective judgement in Table 4 as follows. Orientations listed without question marks are the most reliable - either the orientations were reported explicitly in field notes, or the sensors were installed by workers known to be especially scrupulous in this regard, or both. Some, but not all, of these numbers were double-checked. Orientations listed with question marks are less reliable - either they were only reported anecdotally after the field work was completed (weeks after installation in some cases), or the sensors were installed without a compass, or both. Only a question mark is listed in cases where no orientation information was available. Caution is advised in using the orientations listed in Table 4; readers who must use them should try to devise an independent test of their accuracy (partitioning of the amplitude of a pulse incident from a known azimuth, for example).

Clock Corrections

Data presented in Tables 5 and 6 can be used to correct GEOS clock drifts. Although each GEOS can be programmed to measure clock errors relative to WWVB using an internal time-code receiver, reception was generally unreliable during the Loma Prieta deployment, and WWVB provided significant data at fewer than 10 stations (notably BMT and DMD). Instead, most clock errors were measured with portable masterclocks referenced to a rubidium standard clock in Menlo Park. GEOS clocks were routinely synchronized to an external time standard (synched) at initial deployment, redeployment after a failure, and whenever clock errors exceeded several hundred milliseconds. Otherwise, to allow a pattern of clock behavior to be established, it was generally preferable to measure errors without synching. The following sign convention for clock errors is used in this report: GEOS time - clock error = time standard ("true" time).

Clock-correction data in Tables 5 and 6 were transcribed from tape playback logs and field notes. The information varies in completeness and accuracy, since both data sources can break down in various ways. Most problems are caused by GEOS failure or by incomplete or inaccurate field notes. Synchs (regardless of time standard) and WWVB error measurements are usually written to tape and recovered from tape playback logs. In case of tape failure or synching without an installed tape, however, missing data must be recovered from field notes, or, in the worst case, inferred from apparent discontinuities in clock history. Ideally, the clock error just before a synch can be computed from the difference between the old and new times reported in the tape playback log. In case of GEOS failure, however, missing data must be extrapolated (by assuming a clock drift rate based on the unit's history). Masterclock error measurements are not written to tape - they (as well as the masterclock history) must be recovered from field notes. Four masterclocks were used; the masterclock serial number (5,6,10,11) is listed in Table 5 when it is known.

It was our intention in Tables 5 and 6 to present all the clock data with a minimum of interpretation. Question marks denote missing data and inconsistent, but apparently reliable, data - readers are sure to find other cases that we have missed. In a few cases of gross timing problems the data file headers themselves were adjusted manually - no further correction is needed in these cases. When two or more clock events have the same time, they are listed in order of occurrence. We have added a leap second to the errors of clocks that were operating at 1990 001 00:00 UTC; errors were estimated by linear interpolation or extrapolation from nearby data points (straightforward - except at KOI, MON, and STQ). Time standards such as WWVB and the rubidium clock accommodated the leap second immediately (this will be reflected in hypocenter listings, for example), but several days elapsed before the GEOS clocks could be resynched. Although the masterclock histories (relative to the rubidium standard) in Table 6 are very incomplete, it is worth noting that the policy was to check masterclocks daily - they were generally not allowed to drift more than 20 ms from the rubidium standard clock.


Data Availability

The Loma Prieta aftershock waveform dataset described in this report is available from the U. S. Geological Survey on a single ISO-9660-standard CD-ROM optical disk (compact disk - read only memory). Waveform files have been reformatted to ASCII from the compact block-binary form used at the USGS - a sample file is presented in Appendix C. For information contact:

ES&G Data Project
U. S. Geological Survey, MS-977
345 Middlefield Rd.
Menlo Park, CA 94025

[ Note from Gary Glassmoyer (28 February 1997): The binary data are now available online at https://ca.water.usgs.gov/nsmp/GEOS/LPE ]


Acknowledgements

The Loma Prieta aftershock investigation was a large, complex undertaking - we are grateful to our USGS colleagues who made it a success. The field work was accomplished by M. Andrews, J. Bicknell, J. Boatwright, R. Borcherdt, L. Carroll, M. Celibi, C. Dietel, J. Fletcher, G. Glassmoyer, J. Gibbs, T. Hanks, A. Lindh, A. McGarr, C. Mueller, T. Noce, L. Seekins, E. Sembera, J. Sena, P. Spudich, R. Warrick, and L. Wennerberg, all under the joint supervision of Hanks and Spudich. A. Shakal of CSMIP and R. Maley and E. Etheridge of USGS assisted in co-siting GEOS with strong-motion recorders. D. Oppenheimer supplied the hypocenter data that is summarized in Figure 9, Table 1, and Appendix A and B. M. Celebi provided the photograph of the damaged Marina district building reproduced on the cover. The maps in Figures 1 through 6, 8, and 9 were created with F. Klein's QPLOT program. D. Boore provided the digitized main-shock rupture-zone file used in Figure 1, and H. Bundock provided the digitized coastline and fault files used in Figures 1 through 6 and 9. The Marina map ( Figure 7) was conceived by T. Hanks and drafted by R. Eis and L. Hollis. We are grateful to P. Spudich and L. Wennerberg for thorough reviews of this report.


References

Borcherdt, R. D., J. P. Fletcher, E. G. Jensen, G. L. Maxwell, J. R. VanSchaak, R. E. Warrick, E. Cranswick, M. J. S. Johnston, and R. McClearn (1985). A General Earthquake Observation System (GEOS), Bull. Seism. Soc. Am. 75, 1783-1825.

Maley, R., A. Acosta, F. Ellis, E. Etheredge, L. Foote, D. Johnson, R. Porcella, M. Salsman, and J. Switzer (1989). U. S. Geological Survey strong-motion records from the northern California (Loma Prieta) earthquake of October 17, 1989, U. S. Geological Survey Open-File Report 89-568, 85 pages.

Shakal, A., M. Huang, M. Reichle, C. Ventura, T. Cao, R. Sherburne, M. Savage, R. Darragh, and C. Petersen (1989). CSMIP strong-motion records from the Santa Cruz Mountains (Loma Prieta), California earthquake of 17 October 1989, CSMIP Report No. OSMS 89-06, 196 pages.


Figures

Figure 1: Station map - all stations

Figure 2: Station map - south subregion

Figure 3: Station map - south-central subregion

Figure 4: Station map - north-central subregion

Figure 5: Station map - north subregion

Figure 6: Station map - San Francisco area

Figure 7: Station map - Marina area

Figure 8: Station map - small arrays at Agnews, Duveneck, Robinwood, Stanford

Figure 9: Seismicity map


Tables

Table 1: Waveform record tables - description and example

Table 1a: Records listed by time and station (all events)

Table 1b: Records listed by time and station (events recorded at five-or-more sites)

Table 2: Station locations

Table 3: GEOS parameters for typical deployment and sensor parameters

Table 4: Station-instrument histories

Table 5: GEOS clock errors

Table 6: Masterclock clock errors

Table 7: Tapes


Appendix

Appendix A: Hypocenter listing, sorted by time

Appendix B: Hypocenter listing, sorted by location

Appendix C: Sample file reformatted to ASCII

Figure C1: Plot of trace 3472355P4.GAO


U.S. Geological Survey

The USGS Home Page is at http://www.usgs.gov
The USGS Geologic Division Home Page is at http://geology.usgs.gov
The USGS National Strong Motion Program Home Page is at http://earthquake.usgs.gov/monitoring/nsmp

The GEOS Home Page is at /GEOS/geos.html
The URL of this page is /GEOS/LPE/OFR_90-503/Loma_Prieta.html
e-mail: glassmoyer@usgs.gov
or address:
Gary Glassmoyer
345 Middlefield Road M/S 977
Menlo Park, California 94025