As a part of the Working Group on California Earthquake Probability 99 (WGCEP 99) effort to reanalyze the probability of significant earthquakes occurring in the San Francisco Bay Region (SFBR), we have systematically reanalyzed historical earthquakes which have occurred in the SFBR and vicinity (Figure 10.1) since 1951. The goal is to redetermine, using modern analytical algorithms, the location and magnitude, including formal uncertainty estimates, of all SFBR earthquakes at the local magnitude (ML) 3.0 threshold and extending back as far in time as the data allow. The aim is to develop an uniform and internally consistent catalog of SFBR earthquakes for subsequent probabilistic analyses. Variation in the procedures used when the original BSL analysis sheets were generated, in the number of reporting seismic stations and in the seismic instrumentation over time add complexity to the problem of developing an uniform SFBR seismicity catalog. This project was made possible by funding from the USGS-PGE CRADA and by the existence of the original analysis sheet and seismogram archives kept on store at the Berkeley Seismological Laboratory (BSL).
The effort involved transcription of the data on the original BSL analysis sheets to computer readable format and systematic reanalysis of the location and magnitude of all ML 2.9 and larger earthquakes which have occurred from 1951 through 1998 in the SFBR and in a 30 km buffer zone around the SFBR. The 1951 start date was dictated by the requisite data availability on the original BSL analysis sheets. Extension of the reanalysis to pre-1951 SFBR earthquakes will require the manpower intensive task of reading the amplitude data from the original seismograms, kept on store in the BSL seismogram archive, in order to determine the magnitude of the earthquakes.
Once the events were reanalyzed, an interim 1951-1998 SFBR seismicity catalog was created by culling the events that were located outside the SFBR study area or that were below the ML 3 threshold. This interim catalog differs from existing SFBR seismicity catalogs in that it, spans a longer time interval, has uniformly determined event location and magnitude, and in includes formal location and magnitude uncertainty estimates. The uncertainty estimates are crucial for statistical analyses associated with the probabilistic estimation of earthquake occurrence. The interim 1951-1998 SFBR seismicity catalog was presented at the 1999 Seismolgical Society of America Annual Meeting (Uhrhammer, 1999). Also, the event list, along with appropriate caveats regarding its interim status, are available on the BSL home page at URL: http://seismo.berkeley.edu/herp.
A proposal has been submitted to USGS Earthquake Hazards Reduction Program to obtain funding to extend the SFBR catalog back to circa 1928 at the ML 3.0 threshold and back to 1911 at the ML 4.5 threshold. If the proposal is funded, we will undertake a 2 year project to construct a uniform and internally consistent SFBR seismicity catalog which spans most of this century.
Prior to 1984, the data on the original BSL analysis sheets were not routinely transcribed to computer readable format. In order to apply modern analytical algorithms to the reanalysis of the SFBR events, the first task was to transcribe the requisite phase and amplitude data. The transcription effort initially focussed on pre-1968 data since USGS Northern California Seismic Network (NCSN) data from 1968 to the present for the SFBR events already exists in computer readable form. We progressed back in time to 1944 and did not proceed any further back once it was discovered that amplitude data were not listed on the original BSL analysis sheets prior to 1951. We then concentrated on transcribing the 1968 through 1983 BSL analysis sheet data so that all the 1951 through present BSL phase and amplitude data would exist in computer readable format.
The transcription process included transcribing not only the phase and amplitude data but also any location, magnitude, felt reports and comments that were on the original BSL analysis sheets. The process required an average of approximately 24 man-hours to transcribe a year's worth of SFBR earthquakes.
The SFBR events were systematically relocated using a modified version of the Regional Earthquake Location Program (RELP) (Uhrhammer, 1982) which is routinely used by BSL personnel when locating local and regional earthquakes. For the initial effort, a single linear gradient layer over a half-space velocity model and corresponding station travel time adjustments, as determined for the Central Coast Ranges (McKenzie and Uhrhammer, 1987), were used to relocate all the SFBR candidate events. Since the number of recording seismic stations was generally less than seven for a large fraction of the SFBR events, particularly prior to 1968, it was assumed that all observations had equal weight went calculating the event location. Observations with large residuals were checked against the entries on the original analysis sheets to determine whether they were due to either a typo or an incorrectly identified phase. The observation was then either corrected of removed and the analysis repeated.
The local magnitude (ML) of each of the SFBR earthquakes was determined from the maximum trace amplitude recorded on standard Wood-Anderson torsion seismograms (Richter, 1935). Eight sites in the SFBR and vicinity have operated Wood-Anderson instruments, as shown in the Table 10.1. Records from these instruments are particularly important for compiling a uniform and internally consistent catalog of seismicity because the instruments were operated by the BSL continuously for more than 65 years (1927-1994) and the records were used routinely to determine the ML of the earthquakes for most of that time. The station ML adjustments (Uhrhammer et al., 1996) were applied when determining ML.
During the analysis we discovered that the original ML assigned for the larger events (ML 5) recorded prior to circa 1960 is based on Pasadena records. We have evidence that these ML estimates are biased high, most likely by 0.2 ML. This has potentially the most significant influence on the statistical inferences drawn from the interim catalog and it needs to be investigated. We have acquired the requisite Caltech Wood-Anderson amplitude data so that this analysis can proceed as time is available.
The events in the interim SFBR seismicity catalog were selected from the reanalyzed events which were located in the SFBR rectangle (dashed line in figure and defined by the vertices: 39.02,-122.08; 37.19,-120.61; 36.43,-122.09; 38.23,-123.61) and which had a ML of 3.0 or larger. The locations of the 1854 selected SFBR events are shown in Figure 10.2. A few of the events do not have uncertainties listed (due either to inadequate information on the original analysis sheets and/or being in the coda of a previous event) and approximately 120 of the events (about 6.4 percent of the total) have horizontal or depth uncertainties of 10 km or larger and/or magnitude uncertainty of 0.2 ML or larger. Reducing the uncertainty for these events will require pulling and reading the original seismograms in the BSL archive. This is a manpower intensive effort that will not proceed until funding becomes available.
As a check on the completeness of the interim 1951-1998 SFBR seismicity catalog, we plotted the cumulative number of earthquakes as a function of magnitude (the standard Gutenberg-Richter relation) as shown in Figure 10.3. The linear trend of the cumulative distribution at small ML indicates that the catalog is complete at the ML 3 threshold. The annual rate of SFBR earthquake occurrence at a given ML threshold is r = 10logN where logN = 4.105 - 0.89ML. For example, at the ML 5.5 threshold, where damage to weak structures is typically reported, the observed rate of occurrence is 0.16 earthquakes/year. The probability of one or more ML 5.5 or larger earthquakes occurring in any given year in the SFBR is P = 1 - e-r or 15 percent.
We thank Bill Bakun and Al Lindh of the USGS in Menlo Park for their encouragement and enthusiasm and also the people involved with the USGS-PGE CRADA for providing the funding that made this initial effort possible.
McKenzie, M. R. and R. A. Uhrhammer, Bulletin of the Seismographic Stations, 56, Nos. 1 and 2, pp. 1-105, 1987.
Richter, C. F., An instrumental earthquake magnitude scale, Bull. Seism. Soc. Am., 25, 1-32, 1935.
Uhrhammer, R. A., The optimal estimation of earthquake parameters, EARTHQUAKE ALGORITHMS: Phys. Earth Planet. Interiors, 30, no. 2-3, 105-118, 1982.
Uhrhammer, R. A., S. J. Loper, and B. Romanowicz, Determination of local magnitude using BDSN records, Bull. Seism. Soc. Am., 86, 1314-1330, 1996.
Uhrhammer, R. A., J. Fink and S. Ford, San Francisco bay region historical earthquake relocation project, Seism. Res. Lett., 70, 271-272, 1999.