next up previous contents
Home: Berkeley Seismological Laboratory
Next: Data Archive and Distribution: Up: Operations Previous: Data Acquisition and Network

Seismic Data Analysis


Analysis of the data produced by the BDSN and HFN begins as soon as the waveforms are acquired by BSL computers and ranges from automatic processing for earthquake response to analyst review for earthquake catalogs and quality control.

The BSL collaborates with the USGS Menlo Park to provide automatic earthquake information for northern and central California. Event processing at UC Berkeley is performed by the Rapid Earthquake Data Integration (REDI) system and includes the determination of local and energy magnitude, the seismic moment tensor and moment magnitude, and ground motions such as peak ground acceleration, velocity, and displacement. Together with Caltech, UC Berkeley and the USGS provide earthquake information to the State of California and a number of public and private agencies.

On a daily basis, analysts at UC Berkeley review and read events, making phase and amplitude readings from the broadband data, and determine the mechanism and moment. In addition to reading local and regional earthquakes, BSL staff routinely read larger teleseisms and report these observations to the NEIC and ISC.

REDI System

Over the last 7 years, the BSL has invested in the development of the hardware and software necessary for an automated earthquake notification system (Gee et al., 2000). The REDI project is a research program at the BSL for the rapid determination of earthquake parameters with three major objectives: to provide near real-time locations and magnitudes of northern and central California earthquakes; to provide estimates of the rupture characteristics and the distribution of ground shaking following significant earthquakes, and to develop better tools for the rapid assessment of damage and estimation of loss. A long-term goal of the project is the development of a system to warn of imminent ground shaking in the seconds after an earthquake has initiated but before strong motions begin at sites that may be damaged.

Current Status

On April 18, 1996, the BSL and the USGS announced the formation of a joint notification system for northern and central California earthquakes (Gee et al., 1996b). The system merges the programs in Menlo Park and Berkeley into a single earthquake notification system, combining data from the NCSN and the BDSN. On the USGS side, incoming analog data from the NCSN are digitized, picked, and associated as part of the Earthworm system (Johnson et al., 1995). Preliminary locations, based on phase picks from the NCSN, are available within seconds, based on the association of a few arrivals, while final locations and preliminary coda magnitudes are available within 2-4 minutes. Earthworm reports events - both the "quick-look" 25 station hypocenters (without magnitudes) and the more final solutions (with magnitudes) to the Earlybird alarm module in Menlo Park. This system sends the Hypoinverse archive file to the BSL for additional processing, generates pages to USGS and UC Berkeley personnel, and updates the northern California earthquake WWW server.

Figure 9.1: Schematic diagram illustrating the connectivity between the real-time processing systems at the USGS Menlo Park and UC Berkeley. The combined system forms the joint earthquake notification project in Northern California.
\epsfig{, width=7cm}\end{center}\end{figure}

On the UC Berkeley side, the Hypoinverse archive file is used to drive the REDI processing system. This is a modification of the original REDI design, which identified and located events using raw phase data from the BDSN and NCSN (Gee et al., 1996a). In the revised REDI processing, the magnitude of each hypocenter is assessed. If the coda magnitude is greater than or equal to 3.0, waveforms from the BDSN are analyzed to estimate local magnitude. Once an updated magnitude is obtained, the event information is paged to USGS and UC Berkeley personnel, notification is sent to emergency response agencies, and the revised magnitude is transmitted to the Earlybird system in Menlo Park. The earthquake is then evaluated for the next level of REDI processing. If the local magnitude is greater than 3.5, waveforms from the BDSN strong-motion instruments are analyzed to determine peak ground acceleration, velocity, and displacement and to estimate the duration of strong shaking. When the strong-ground motion processing is complete, these values are distributed by email and pager and the event is scheduled for moment tensor estimation. In this stage of REDI processing, both the waveform modeling method of Dreger and Romanowicz (1994) and the surface wave inversion technique of Romanowicz et al. (1993) are run for every qualifying event (earthquakes with ML greater than 3.5). Each algorithm produces an estimate of the seismic moment, the moment tensor solution, the centroid depth, and solution quality. The REDI system uses the individual solution qualities to compute a weighted average of moment magnitude, to compare the mechanisms using normalized root-mean-square of the moment tensor elements (Pasyanos et al., 1996), and to determine a "total" mechanism quality.

At present, two Earthworm-Earlybird systems in Menlo Park feed two REDI processing systems at UC Berkeley (Figure 9.1). One of these systems is the production or paging system; the other is set up as a hot backup. The second system is frequently used to test new software developments before migrating them to the production environment. In addition, the BSL operates a third system, which uses BDSN picks to form an independent list of associated events. This third system provides redundancy in case the communication links with the USGS Menlo Park are disrupted. A fourth system is installed in Sacramento in order to provide a redundant notification facility outside of the Bay Area.

This structure has greatly expedited automatic earthquake processing in northern California (Table 9.1). The dense network and Earthworm-Earlybird processing environment of the NCSN provides rapid and accurate earthquake locations, low magnitude detection thresholds, and first-motion mechanisms for smaller quakes. The high dynamic range dataloggers, digital telemetry, and broadband and strong-motion sensors of the BDSN and REDI analysis software provide reliable magnitude determination, moment tensor estimation, peak ground motions, and source rupture characteristics. Robust preliminary hypocenters are available about 25 seconds after the origin time, while preliminary code magnitudes follow within 2-4 minutes. Estimates of local magnitude are generally available 30-120 seconds later, and other parameters, such as the peak ground acceleration and moment magnitude, follow within 1-4 minutes. Earthquake information from the joint notification system is distributed by pager, e-mail, and the WWW. The first two mechanisms "push" the information to recipients, while the current Web interface requires interested parties to actively seek the information. Consequently, paging and, to a lesser extent, e-mail are the preferred methods for emergency response notification. The Northern California Web site has enjoyed enormous popularity since its introduction and provides a valuable resource for information whose bandwidth exceeds the limits of wireless systems and for access to information which is useful not only in the seconds immediately after an earthquake, but in the following hours and days as well.

Table 9.1: Current REDI processing stages. The location procedures in Stage 1 are skipped in normal operation of the joint system, but are utilized if problems prevent the flow of hypocenter information from the USGS. The items in italics are recent extensions to the REDI system.
Stage Qualification Processing
1 All events Location
2 M > 3.0 Local magnitude, Energy magnitude
3 M > 3.5 Peak ground motion
4 M > 3.5 Moment tensor
5 M > 6.0 Line source finite fault
6 M > 6.0 2D finite fault

System Monitoring

In order to ensure that automatic systems are operating correctly, BSL staff and students participate in alarm response. Each week, two people are on duty in order to respond to earthquakes - or computer problems. One person is designated as primary and is responsible for earthquake-related issues. The second individual serves as a backup to the second and addresses more operational problems. For earthquake notification, the alarm response team receives pages from several sources including an automatic monitoring of the ground velocity at BKS (known as the SEismic ALarm or SEAL), a human-initiated alarm from the UCB Police Department, and separate notifications from REDI and the USGS. For operational monitoring, we have developed a component of the REDI system for tracking the heartbeats and data flow from critical systems and processes. Because this is inherently a distributed system, it is critical to monitor the "health" of every component. In the REDI system, the monitor program is a master scheduler that can perform several types of monitoring. As presently implemented, the monitor program watches critical computers and network components, disk resources, specified processes, and particular subsystems, such as data acquisition.

Updates and Recent Developments

In the past year, we have made progress on a number of fronts.

Year 2000 compliance

The REDI system did not experience any major problems due to Y2K issues. Modifications to the Earthworm components of the REDI system had been implemented and thoroughly tested by the end of the year. Only small changes were required in the REDI software, primarily in the moment tensor codes. One small bug slipped through the cracks, and turned up during the first event which qualified for moment tensor processing in 2000. This bug was quickly identified and fixed.

Three-Component Picker

In the last year, we have worked to stabilize the picker for an operational environment. This has included numerous bug fixes and development of error handling conditions (for example, how to treat telemetry outages). The code has been put into a software distribution package and is ready to be exported to potential users.

At present, the picker is running on a test system, feeding a stand-alone REDI system. Our remaining effort is focusing on testing the picker in conjunction with the associator. This work is being done in real time and by running historical events through an off-line version. We are also working on the issue of phase identification and upgrading the associator to take advantage of azimuth and phase id.

As an illustration, we show the results for the M4.8 earthquake near Bolinas.

Figure 9.2: Distribution of P travel-time residuals for an analyst, the picker, and the MHH detector. The residuals for BH, HH, and HL channels are computed from the phase readings from the BSL catalog (typically made on the HH data stream).

Figure 9.2 shows three sets of P travel-time residuals for a human, the picker, and the Murdock-Hutt (MHH) detector computed from the BSL catalog readings. Picks from the BH, HH, and HL channels are included. The scatter shown in the three plots is similar (around 1 to 1.5 sec), although the distribution of MHH is more compact. The scatter in the analysts' picks is somewhat surprising, but is primarily limited to the BH picks (HH and HL picks show better agreement). As this event illustrates, the picker is performing reasonably well, although it seems to pick slightly later than either the MHH or the analyst. The tendency to pick later is correlated with distance, as arrivals become more emergent. The picker, which was tuned on BH data, behaves more reliably on the BH and HH channels than on the HL channels. At near distances to an event, better picks are obtained from the HH data as the higher sampling rate reduces problems associated with the acausal FIR filters.

The picker can identify S as well as P arrivals, although it has some difficulty at short distances.

Figure 9.3: Map comparing the different locations obtained from combining the picker with the associator. The star is the location from the BSL catalog. The downward triangle is the location based on the BH picks, the upward triangle is the location based on the HH picks, and the square is the location based on the HL picks. The differences between the BSL catalog location and the BH and HH picks are only a few km. The HL location is shifted significantly west.

Figure 9.3 illustrates how the association algorithm performs with output from the picker. The locations computed from the BH (downward triangle), HH (upward triangle), and HL (square) are shown, in comparison with the location from the BSL catalog. The BH and HH locations are within a few km of the catalog location, while the HL location is shifted to the west. The picker made fewer detections on the HL data, which reduced the quality of the location.

Finite Fault Parameterization

We have recently expanded the REDI processing environment to include the estimation of finite-fault parameters. This approach is based on the use of theoretical Green's functions and is an extension of the finite-source inversions that are commonly performed on local strong-motion data (for example, Wald et al. (1996) and Cohee and Beroza (1994)). Using a pre-computed set of Green's functions for an appropriate 1-D velocity model, it is possible to consider an arbitrarily oriented source using parameters obtained from the automated moment tensor analysis and a directivity model of an expanding circular rupture with a constant rupture velocity and dislocation rise time.

Based on the development of Kaverina et al. (1997) and Dreger and Kaverina (1999), we have implemented the finite-fault estimation procedure in two REDI stages. In stage 5, BDSN broadband waveform data are prepared for inversion and rough estimates of the fault dimensions are derived using the empirical scaling relationships of Wells and Coppersmith (1994). Using these parameters to constrain the overall dimensions of the extended source, the process tests the two possible fault planes (determined in stage 4) over a range of rupture velocities by performing a series of inversions using a line-source representation.

Each line-source computation is quite rapid. Calculations for the 1992 Landers and 1994 Northridge earthquakes required approximately 3.5 minutes to test 7 rupture velocities for the two different planes (Dreger and Kaverina, 1999). In tests based on the Landers and Northridge earthquakes, the fault plane was clearly defined by the variance reduction from the line-source inversions. In addition to the identification of the fault plane and apparent rupture velocity, this stage yields preliminary estimates of the rupture length, dislocation rise time, and the distribution of slip in one dimension. The second component of the finite-fault parameterization uses the best-fitting fault plane and rupture velocity from Stage 5 to obtain a more refined image of the fault slip through a full two-dimensional inversion. If Stage 5 fails to identify the probable fault (due to insufficient separation in variance reduction), Stage 6 will compute the full inversion for both fault planes. In the present implementation, the full inversion requires an additional 20-30 minutes per plane, depending on the resolution, on a Sun UltraSPARC1/200e. This extension of REDI processing is running on a development system at the BSL, where it is being tested. We anticipate migrating these modules to the production systems in the fall of 2000. At the time of the 1999 Hector Mine earthquake, the REDI implementation was not complete. However, the approach was tested using several regional-distance stations from TriNet. Dreger and Kaverina (2000) first used the line-source inversion to identify the causative fault plane and then obtained the distribution of slip on the NW-trending plane using the best-fitting rupture velocity. The image of the finite rupture obtained by this procedure agreed well with aftershock locations and observed surface rupture. These results are described more completed in the research section on Source Process of the October 16, 1999 Hector Mine Earthquake.

Shake Maps

During the past year, the USGS Menlo Park implemented version 1.0 of the ShakeMap software for northern California earthquakes. This software package, developed as part of the TriNet project in southern California (Wald et al., 1999), combines observed values of ground motion with predicted values (primarily based on attenuation relations) in order provide maps of strong shaking. During the past year, we implemented software to push ground motion data from the BDSN to the USGS for use in ShakeMaps. The northern California maps combine data from the primary earthquake monitoring agencies (USGS, CDMG, BSL) with other sources of strong motion data (such as PG&E) in order to provide the most complete view of ground motion following an event.

The first "official" northern California ShakeMap was produced for the August 18, 1999, earthquake near Bolinas, California (see discussion below). As presently implemented, the REDI system pushes ground motion data to the ShakeMap system in Menlo Park for events of M3.5 and higher at the completion of the stage 2 processing.

The USGS Menlo Park is currently testing ShakeMap version 2.0, which provides improved corrections for geology and greater flexibility for different attenuation corrections and other sources of predicted ground motions. Once version 2.0 is implemented in Menlo Park, we intend to install the ShakeMap codes as part of the REDI system at the BSL.

Data exchange

In order to improve our capabilities on the edges of the network, we have initiated efforts to establish mechanisms of data exchange with neighboring networks.

We have recently completed an agreement with the University of Nevada, Reno, to enhance the earthquake monitoring capabilities in northern California and Nevada. As part of this agreement, we agreed to exchange waveform data. At the present, three-component data from CMB, WDC, MOD, and ORV and vertical component data from YBH, ARC, HOPS, WENL, SAO, and KCC are being sent to UNR. In exchange, the BSL is receiving three-component data from BEK, OMM, PAH, and WCN. In addition, UNR is forwarding data from the NSN stations WVOR, MNV, DAC, and ELK. The UNR sensors are Guralp 40Ts and these stations will enhance the REDI capabilities in eastern California and western Nevada.

The UNR data exchange is implemented using the Earthworm import/export mechanism, over the Internet. This is the same protocol we use for waveform exchange with the USGS Menlo Park. Figure 8.4 illustrates the current dataflow in the REDI environment. Data for export via Earthworm is fed into the Trace Ring, using a comserv client. An Earthworm export process picks data of the ring and transmits it to the UNR client and a separate process does the same for the USGS Menlo Park client. Data imported via Earthworm is brought into the Tracein Ring. A module converts from Earthworm to MiniSEED packets (ew2m) and the data is distributed to the two acquisition systems. Each acquisition system "logs" the data using mserv.

The exchange with UNR has just gotten off the ground. Based on the experience of a few weeks, it seems as if there are a number of issues with reliability. We do not see as many problems with the Earthworm feed to Menlo Park, which may be attributed to the robustness of the frame-relay connection (as opposed to the Internet connection with UNR). We will be exploring other options in the coming months.

In addition to the exchange with UNR, the BSL exchanges waveform data with UCSD via a public domain Orb. We are currently working out the details of an exchange with Caltech.

In the coming year, we anticipate improving our capabilities within the network by expanding our data exchange with the USGS Menlo Park.


We installed the "Quake Data Delivery Service" or QDDS software at the BSL in the last year. This software was developed by the USGS to allow for the exchange of parametric earthquake data, such as locations and magnitudes. QDDS provides the earthquake information for the "recenteqs" maps as well as for other applications. We are currently running QDDS and the "recenteqs" software.

At present, the REDI system is not contributing data to QDDS. The USGS Menlo Park pushes information from the joint notification system to the two QDDS hubs. In the long run, we intend to modify the REDI system so that it provides information to one hub and the USGS provides information to the other. However, this modification requires restructuring some aspects of the way REDI tracks information (essentially, the use of version numbers) and we intend to implement this when we transition to a database environment.


With the recent developments in the finite-fault processing and the planned expansion of the ShakeMap computations, it has become clear that the current use of flat files in REDI for information flow is not adequate to meet future needs. We have been considering two models for incorporating a database system in the REDI environment. The first is the database schema developed as part of the Northern California Earthquake Data Center in cooperation with Caltech. This database has been used by Caltech and the USGS in Pasadena in the TriNet realtime processing system as well as in the SCEC Data Center environment. In contrast, the USGS Menlo Park has been working with the Earthworm database schema, focusing primarily on the realtime environment.

We have begun exploring the issues of interfacing one database schema to another. The effort to incorporate a database within the REDI software is expected to be a major project for the coming year. We anticipate this revision of the REDI software will significantly expand and enhance our capabilities.

System Performance

From 07/01/1999 through 06/30/2000, the northern California joint notification system processed nearly 10,000 events. Approximately 3,000 of these events were distributed by the REDI system and just over 350 qualified for higher-level REDI processing (Figure 9.4).

Most of the 10,000 events were small to moderate earthquakes in northern California. In addition, the joint notification system processed the M7.1 Hector Mine, M6.3 Furnace Creek, and M5.8 earthquakes in southern California, Nevada, and offshore of Cape Mendocino respectively. The largest earthquake in the San Francisco Bay Area was the M4.8 earthquake near Bolinas.

Event quality

Unfortunately, not all of the events in Figure 9.4 are actually local earthquakes. Some of the events (< 2%) were the result of problems in the USGS analog network. For example, problems with microwave glitches plagued the USGS telemetry in north-central California, leading to numerous false reports of events in the region of Mt. Shasta. Other false events were generated by deep teleseisms. Large, deep earthquakes produce impulsive P-arrivals which can trigger the NCSN and BDSN, producing spurious events.

These two sources of false events have different repercussions for the performance of the monitoring system. Microwave glitches often have large coda magnitudes, but small local magnitudes. For example, a microwave glitch on June 5, 2000 had a coda magnitude of 7.0. Fortunately, the REDI processing of the BDSN data resulted in an ML of 2.4, which is essentially a measure of the background noise in the network. The plot in the lower right of Figure 9.4 illustrates difference between ML and Md as a function of the number of stations in the solution. The cluster of events with magnitude differences of 1 or greater are microwave glitches. This source of potential problems will gradually be eliminated as the NCSN is upgraded to a digital network.

In contrast, deep teleseisms present an ongoing challenge. A Mw 7.0 earthquake on 4/23/2000 created 3 false events. In this case, however, the coda magnitudes were not revised down by the local or moment magnitude. The USGS uses the rms of the hypocentral solution as a simple discriminate, and rejects events with an rms > 0.4 (Figure 9.4). Unfortunately, this discriminate can lead to the rejection of events outside the network, such as earthquakes in the Cape Mendocino/Gorda Plate region. Although the USGS uses a higher rms cutoff outside of the network, this is not a particularly satisfactory solution to the problem.

In conjunction with our work on the three-component picker, we have been experimenting with an association algorithm to detect earthquakes and discriminate among local, regional, and teleseismic events based on the curvature of the arriving wavefront. This is an ongoing effort.

Figure 9.4: Map showing the events which met the REDI criteria for distribution (number of stations > 10, rms < 0.8) during 1999-2000. Shaded circles are events with rms greater than 0.4. Lower plots show the distribution of solution quality (as measured by the rms) and the difference between the local and coda magnitudes as a function of the number of stations in the location. The thick solid line indicates the "normal" paging boundary currently used by REDI, which will be revised this coming year to reflect collaborative efforts with UNR.

Local Magnitude

Of the 350+ events that qualified for REDI processing, 214 had ML of 3.0 and greater, with 97 of ML of 3.5 or greater. Figure 9.5 compares the locations obtained by the joint notification system (open circles) with those from the analyst-reviewed BSL earthquake catalog. Very little difference between the two locations is observed for most events inside the network, although some differences can be seen to the northwest and the southeast. For earthquakes outside the network, the automatic solutions generally try to pull events in toward the network, while the use of S-waves in the reviewed solutions pushes them out. Events far offshore are actually pushed away from the network, which is probably due to the velocity model. Within the network, small differences between the automatic and catalog solutions may be attributed to differences in location algorithm and station corrections.

Figure 9.5: Map comparing the automatic locations from the joint notification system with the final locations in the UC Berkeley earthquake catalog for events of magnitude 3.0 and higher. Lower plots compare the automatic magnitudes (Md and ML) (left) and the automatic versus reviewed estimates of ML (right).

The lower plots in Figure 9.5 compare the automatic magnitudes and the automatic ML with the analyst-reviewed catalog ML. For these earthquakes (which exclude the microwave glitches and teleseisms in Figure 9.4), the two automatic magnitude estimates compare well. In general, the ML is larger than the Md and this trend becomes more pronounced at higher magnitudes. This is primarily due to saturation of the coda magnitude scale (for example, the coda magnitude for the Hector Mine mainshock was 5.7 while the local magnitude was 7.3), although this is also an issue for smaller events on the edge of the network. The coda magnitude appears to systematically underestimate event size in the Cape Mendocino and southern Nevada area. 7 events in Figure 9.5) have ML - Md in excess of 0.5 magnitude units. 4 of these are located in the Cape Mendocino/Gorda plate while the remaining 3 are in southern Nevada.

The automatic and reviewed estimates of ML agree well. Over 83% of the automatic magnitude estimates were within 0.2 magnitude units of the final ML. In general, the automatic system appears to underestimate the final ML. This may be due in part to the fact that the analysts include more observations than does REDI (REDI minimizes the number of stations used in order to accelerate processing). Only 3 events had differences of more than 0.5 magnitude units, with a maximum of 0.93. For this event, an earthquake in the California/Nevada border region, the REDI processing was complicated by the occurrence of two events within 30 seconds. The first was an ML 4.2, while the second was ML 5.4. The automatic processing included some of the second event and thus overestimated the magnitude.

The previous section discussed the problem with microwave glitches where the ML computed by the REDI system corrects estimates of Md that have been biased by analog noise. There are also situations where the ML is significantly larger than the Md

Figure 9.6: Figure displaying the REDI ground motion data. Ratio of observed values to predicted values are plotted for WAS (bottom), PGA (middle) and PGV (top). These data are from the 25 events processed from 1/1/2000 to 6/30/2000 (show by +) and three recent events of approximately M5 (8/12/1998 San Juan Bautista, 8/18/2000 Bolinas, and 9/3/2000 Yountville shown by o).

Ground Motions

Although the REDI system has been processing data from the BDSN strong-motion sensors for several years, the threshold for processing was lowered this year in order to facilitate the generation of Shake Maps for smaller earthquakes. By extending the analysis of ground motions to smaller events, we have identified some problems with the strong-motion sensors. Two sites (ARC and BRIB) have been removed from REDI processing because of frequent spikes in the waveforms. The spikes bias the estimates of PGA and PGV, which can create significant problems in the generation of Shake Maps.

Since the lower threshold was introduced on January 1, 2000, the REDI system has processed 31 events for strong motion data. 6 of these were either teleseisms or regional events. Figure 9.6 shows the results for the 25 local earthquakes. In order to show data from events of different magnitudes, we have normalized the observed values with values predicted from the attenuation relationships of Joyner and Boore (1982). The top figure shows PGV and the middle PGA, plotted as a function of hypocentral distance. For comparison, the bottom plot shows Wood-Anderson synthetic amplitudes (WAS), normalized by the predicted values of Bakun and Joyner (1984).

The results are intriguing. The attenuation relationships overpredict the observed values between 1 and 2 orders of magnitude. As distance increases, this becomes less pronounced. This attenuation relationship, like most others, has been derived with data from large earthquakes. Not much attention has been given to events in the small to moderate magnitude range. Evidence of some magnitude effect is clear when data from three  M5 events are included (circles). In contrast, the WAS results show a better fit, with a less apparent dependence on distance. The level of scatter is still high, but this figure does not include station corrections.

We have examined plots like this in order to see whether any site-specific effects are visible. However, any such effects are second order to the magnitude and distance effects.

Moment Tensor

Automatic moment tensor solutions were computed for 59 earthquakes during 1999-2000. As described in the section on the current status, two methods are employed by REDI, one based on the inversion of surface waves and one based on the inversion of the complete waveform.

Figure 9.7 presents the results for northern California, comparing the mechanisms, the moment magnitudes, and the correlation between the methods. As seen previously (Pasyanos et al., 1996), the estimates of Mw compare well between the methods, even when the mechanisms disagree.

However, the correlation between the automatic solutions was somewhat disappointing. In this study, X is 1 - $\mu$ - the correlation defined by (Pasyanos et al., 1996). In our definition, two mechanisms with complete agreement have X=1 (and thus total disagreement if X=0). Fewer than half of the events in 1999-2000 had a correlation of 0.5 or higher, which is the level required for automatic distribution of moment tensor solution. Analysis of earlier results has shown that the ratio of well-correlated mechanisms increases with event size, as smaller events are more sensitive to noisy waveforms or other problems related to the signal-to-noise ratio.

Analysis of earlier results has also shown that the poorly correlated events tend to be concentrated offshore of Cape Mendocino and in the Mammoth Lakes/eastern California area (Gee et al., 2000). This was true in the last year as well - of the 12 events larger than M4.5, 5 showed good correlation between the two methods. The 7 that did not were either located in the Cape Mendocino/Gorda Plate region or Mammoth Lakes/eastern California area, with the exception an earthquake in the Geysers on 1/10/2000.

In general, the complete waveform method obtains a better variance reduction than the surface wave method for these events. Yet, the strike of one of the nodal planes obtained by the surface wave methodology is similar to one of the nodal planes by the complete waveform approach. The poorer surface-wave solution may reflect the inherent problem of determining the slip angle in surface-wave inversion for shallow events.

The improvement of models for events in the Cape Mendocino/Gorda Plate region and Mammoth Lakes/eastern California area is an active area of research. Chapter III devotes a couple of sections to this topic.

Figure 9.7: Map comparing the automatic moment tensor solutions from the joint notification system. The solutions plotted in grey are from the surface wave inversion, while the solutions plotted in black are from the complete waveform inversion. Below, the difference in the moment magnitude determined by each method and the correlation between the solutions are plotted as a function of moment magnitude.

Routine Earthquake Analysis

On a daily basis, the BSL continues to locate and determine the magnitude of earthquakes in northern California and adjacent regions. As a general rule, events are analyzed if their magnitude is greater than 2.8 in the Central Coast ranges, greater than 3.0 in all of northern California, or greater than 3.8 in the bordering regions. Traditionally, these events were located using hand-picked arrival times from the BDSN stations in conjunction with P-arrival times from the NCSN using the algorithm st-relp. Over the past several years, the BSL has made a transition in the daily analysis to take advantage of the automatic processing system. As part of this transition, events which have been processed by the automatic system are not generally relocated, although phase arrivals are still hand-picked and the synthetic Wood-Anderson readings are checked. Instead, analysts are focusing on the determination of additional parameters, such as the seismic moment tensor, phase azimuth, and measures of strong ground shaking.

From July 1999 through June 2000, BSL analysts reviewed nearly 200 earthquakes in northern California and adjoining areas, ranging from M2.1 to 7.1. Reviewed moment tensor solutions were obtained for 30 events. Figure 9.8 and Table 9.2 displays the earthquakes located in the BSL catalog and the moment tensor solutions. For contrast with the automatic solutions, the figure also shows the comparison between the two estimates of Mw and the correlation between the two methods when available.

Figure 9.8: Map comparing the reviewed moment tensor solutions from the joint notification system. The solutions plotted in grey are from the surface wave inversion, while the solutions plotted in black are from the complete waveform inversion. Below, the difference in the moment magnitude determined by each method and the correlation between the solutions are plotted as a function of moment magnitude.

Table 9.2: Moment tensor solutions for significant events from July 1 1999 to June 30 2000 using both regional methodologies. Epicentral information from the USGS/UC Berkeley joint earthquake notification system. Moment is in dyne-cm and depth is in km. Key to methods: (1) Complete waveform fitting inversion; (2) Regional surface wave inversion.
Location Date UTC Time Lat. Lon. Depth Ml Mw Mo Str. Dip Rake Method
Petrolia 07/05/1999 03:06:29.0 40.316 -124.528 24.0 3.7 4.1 1.42e22 129 61 115 2
Lake Isab. 07/11/1999 18:20:46.0 35.735 -118.485 8.0 4.3 4.4 3.11e22 345 39 -92 1
          8.0 4.3 4.3 2.76e22 345 39 -93 2
Petrolia 07/24/1999 00:38:41.0 40.391 -125.127 4.0 4.4 4.6 9.40e22 117 62 -97 1
          5.0 4.4 4.5 5.65e22 359 64 -18 2
Geysers 07/29/1999 04:52:26.6 38.796 -122.733 8.0 3.6 4.4 4.00e22 359 47 -100 1
          5.0 3.6 3.7 4.30e21 341 65 -149 2
CA-NV bdr 08/01/1999 16:06:22.0 37.39 -117.07 8.0 5.6 5.9 7.00e24 216 45 -92 1
          8.0 5.6 5.7 3.60e24 226 64 273 2
CA-NV bdr 08/02/1999 06:05:13.4 37.361 -116.949 8.0 5.5 5.1 5.60e23 100 55 49 1
          11.0 5.5 5.1 4.90E23 251 82 -44 2
Bolinas 08/18/1999 01:06:19.0 37.907 -122.686 6.0 5.0 4.8 1.60e23 121 51 86 1
          8.0 5.0 4.5 7.25e22 115 49 69 2
San Simeon 09/21/1999 15:42:48.0 35.799 -121.255 8.0 5.6 3.6 2.70e21 119 72 81 1
          8.0 5.6 3.6 2.70e21 119 72 81 2
Santa Rosa 09/22/1999 22:27:13.1 38.393 -122.633 6.0 4.2 4.3 2.80e22 113 49 100 1
          5.0 4.2 4.0 9.68e21 51 50 27 2
CA-NV bdr 09/26/1999 20:11:22.0 37.43 -117.080 6.0 4.1 4.0 1.00e22 108 61 -83 1
          8.0 4.1 4.2 2.53e22 261 72 -92 2
Somes Bar 09/29/1999 06:22:03.0 41.357 -123.424 40.0 4.0 4.0 1.20e22 9 52 -78 1
          39.0 4.0 4.0 1.07e22 188 41 -95 2
South. CA 10/16/1999 09:46:45.0 34.600 -116.270 8.0 7.0 7.1 4.22e26 236 68 -45 2
CA-NV bdr 10/20/1999 10:15:15.0 37.423 -117.953 28.0 4.4 4.0 1.10e22 190 73 32 1
Lone Pine 10/26/1999 22:59:35.0 36.467 -117.883 8.0 4.0 4.1 1.43e22 142 83 -154 2
Bishop 11/08/1999 01:53:12.4 37.356 -118.580 10.0 4.4 4.0 7.10e21 21 70 -22 1
Portola 11/22/1999 05:38:03.2 40.413 -125.222 4.0 4.5 4.8 1.80e23 97 72 -106 1
          4.0 4.5 4.6 8.50e22 98 83 -159 2
Petrolia 12/26/1999 19:41:53.2 40.273 -124.392 24.0 4.2 4.8 2.00e23 183 87 7 1
          24.0 4.2 4.8 1.81e23 191 72 -2 2
Geysers 01/06/2000 21:38:10.9 38.844 -122.826 6.0 3.5 4.0 1.10e22 165 52 -127 1
          5.0 3.5 4.0 1.02e22 247 59 -48 2
Petrolia 01/08/2000 02:17:25.4 40.386 -126.826 4.0 5.4 5.4 1.20e24 320 66 -91 1
          8.0 5.4 5.5 1.74e24 100 78 131 2
Geysers 01/10/2000 21:41:27.0 38.757 -122.913 6.0 4.3 4.5 5.60e22 346 77 -135 1
          5.0 4.3 4.3 3.80e22 3 65 -155 2
Geysers 01/11/2000 14:19:52.0 38.768 -122.913 5.0 4.3 4.3 9.57e22 6 65 -153 2
Petrolia 01/16/2000 01:51:32.0 40.463 -125.708 20.0 4.3 4.1 1.50e22 99 87 172 1
          21.0 4.3 4.3 2.82e22 187 80 14 2
Geysers 01/18/2000 23:26:01.3 38.750 -122.914 6.0 4.2 4.3 3.20e22 359 80 -128 1
          8.0 4.2 4.2 2.60e22 7 66 -158 2
Ferndale 02/02/2000 23:14:32.0 40.514 -124.447 24.0 3.6 3.8 5.28e21 133 80 169 2
Bishop 03/04/2000 16:13:22.0 37.846 -118.203 14.0 4.5 4.1 1.72e22 234 89 -10 2
Punta Gorda 03/16/2000 15:19:56.0 40.386 -125.279 4.0 5.8 5.9 8.10e24 95 74 -153 1
          8.0 5.8 5.7 4.52e24 5 82 -18 2
Coalinga 03/16/2000 21:42:36.0 36.378 -120.392 8.0 3.8 3.8 5.00e21 325 68 109 1
          8.0 3.8 3.8 6.56e21 325 72 97 2
Los Banos 03/25/2000 16:22:49.7 36.894 -121.049 4.0 3.9 3.7 3.70e21 160 71 -153 1
Geysers 04/05/2000 02:20:31.1 38.785 -122.773 6.0 3.7 3.9 9.00e21 0 59 -93 1
Petrolia 05/08/2000 09:12:35.0 40.324 -124.680 14.0 8.2 4.5 6.38e22 94 66 163 2
Lake Pillsb. 05/17/2000 22:32:07.7 39.392 -123.066 12.0 4.2 4.5 7.30e22 162 61 -130 1
          11.0 4.2 4.5 6.38e22 52 89 -26 2

In addition to the routine analysis of local and regional earthquakes, the BSL also processes teleseismic earthquakes. Taking advantage of the CNSS catalog, analysts review teleseisms of magnitude 5.8 and higher. All events of magnitude 6 and higher are read on the quietest BDSN station, while events of magnitude 6.5 and higher are read on the quietest station and BKS. Earthquakes of magnitude 7 and higher are read on all BDSN stations. The locations and magnitude determined by the BSL are cataloged on the NCEDC. The phase and amplitude data are provided to the NEIC, along with the locations and magnitudes, as contributions to the global catalogs, such as that of the ISC.

Bolinas Earthquake

On August 18, 1999, just one day after the devastating Turkey earthquake, a small M4.8 event struck the San Francisco Bay Area, near the town of Bolinas. This event occurred within the San Andreas Fault zone, along a segment that ruptured during the 1906 San Francisco earthquake. This section of the San Andreas fault, which experienced very large displacements during 1906, had almost been devoid of earthquakes since that earthquake. In contrast, other segments of the San Andreas fault involved in the 1906 earthquake show more seismic activity, such as the 1957 Daly City earthquake on the Peninsula segment and the 1989 Loma Prieta earthquake on the Santa Cruz Mtns. segment.

Most earthquakes in the San Andreas Fault Zone show strike-slip motion. However, this event showed primarily reverse motion and is likely to have occurred on a reverse fault that is subparallel to the main trace of the San Andreas. The Daly City earthquake had a very similar mechanism to the Bolinas event, and the Loma Prieta earthquake was a mixture of strike-slip and reverse-slip motion.

Although a relatively small event, the processing times associated with the REDI notification system provide some measure of system performance. The "SEAL" page was issued 25 seconds after the origin time and the first REDI page with the location and ML was issued 258 seconds after the origin. This notification was followed by the distribution of ground motion data at 310 seconds and the issuing of a moment tensor solution and the Mw at 572 seconds after the origin.

At the time of the Bolinas earthquake, the system to push ground motion data to the USGS had not be automated and these data were shipped to Menlo Park via email. However, the first "Northern California" ShakeMap was produced for this event, several hours after it occurred. This earthquake was also the debut for the Northern California Community Internet Intensity Map.

Contribution to IASPEI

As part of the BSL's contribution to the IASPEI Handbook of Earthquake and Engineering Seismology, we have submitted a description of the BDSN, HFN, and HRSN. We have also submitted the UC Berkeley earthquake catalogs for local/regional events and for teleseisms. The handbook is scheduled for publication in late 2000 and will include catalog and phase data from numerous seismic networks.


During the past year, we have worked toward coordinating our earthquake monitoring activities with southern California and on improving the collaboration in northern California. These efforts span the range of activities from realtime earthquake information to catalog generation to long-term archive and distribution.

This effort to coordinate is in part an outgrowth from the interest of the State of California in uniform earthquake reporting and in part a response of the development of the Advanced National Seismic System or ANSS. As the ANSS moves forward, national earthquake monitoring will be coordinated on a regional basis. California is evolving as one of those regions through the development of the California Integrated Seismic Network (CISN). Participating organizations are the Berkeley Seismological Laboratory, the Caltech Seismological Laboratory, the California Division of Mines and Geology, the USGS Pasadena, and the USGS Menlo Park.

In practical terms, the development of the CISN is leading to better coordination among the participating agencies. Progress is being made in terms of sharing data and software. One of the most important components of the CISN is the Standards committee, which meets monthly to work on issues of coordination.


Under Barbara Romanowicz's general supervision, and with Lind Gee as head guru, Doug Neuhauser, Steve Fulton, Rick McKenzie, Asya Kaverina, Doug Dreger, and Hrvoje Tkalcic contribute to the REDI project and/or the analysis of earthquakes. Lind Gee, Doug Neuhauser, Steve Fulton, and Hrvoje Tkalcic contributed to the preparation of this chapter.


Bakun, W., and W. Joyner, The ML scale in central California, Bull. Seis. Soc. Am., 74, 1827-1843, 1984.

Cohee, B., and G. Beroza, A comparison of two methods for earthquake source inversion using strong motion seismograms, Ann. Geophys., 37, 1515-1538, 1994.

Dreger, D., and A. Kaverina, Development of procedures for the rapid estimation of ground shaking, PGE-PEER Final Report, 1999.

Dreger, D., and B. Romanowicz, Source characteristics of events in the San Francisco Bay region, USGS Open-File-Report 94-176 , 301-309, 1994.

Gee, L., D. Neuhauser, D. Dreger, M. Pasyanos, R. Uhrhammer, and B. Romanowicz, The Rapid Earthquake Data Integration Project, Handbook of Earthquake and Engineering Seismology, IASPEI, in press, 2000.

Gee, L., D. Neuhauser, D. Dreger, M. Pasyanos, B. Romanowicz, and R. Uhrhammer, The Rapid Earthquake Data Integration System, Bull. Seis. Soc. Am., 86, 936-945,1996a. Gee, L., A. Bittenbinder, B. Bogaert, L. Dietz, D. Dreger, B. Julian, W. Kohler, A. Michael, D. Neuhauser, D. Oppenheimer, M. Pasyanos, B. Romanowicz, and R. Uhrhammer, Collaborative Earthquake Notification in Northern California (abstract), EOS Trans. AGU, 77, F451, 1996b.

Johnson, C., A. Bittenbinder, B. Bogaert, L. Dietz, and W. Kohler, Earthworm: A flexible approach to seismic network processing, IRIS Newsletter, XIV (2), 1-4, 1995.

Joyner, W., and D. Boore, Peak horizontal accelerations and velocity from strong-motion records including records from the 1979 Imperial Valley, California, earthquake, Bull. Seis. Soc. Am., 71, 2011-2038, 1981.

Murdock, J., and C. Hutt, A new event detector designed for the Seismic Research Observatories, USGS Open-File-Report 83-0785, 39 pp., 1983. Romanowicz, B., D. Dreger, M. Pasyanos, and R. Uhrhammer, Monitoring of strain release in central and northern California using broadband data, Geophys. Res. Lett., 20, 1643-1646, 1993.

Pasyanos, M., D. Dreger, and B. Romanowicz, Toward real-time estimation of regional moment tensors, Bull. Seis. Soc. Am., 86, 1255-1269, 1996.

Uhrhammer, R., S. Loper, and B. Romanowicz, Determination of local magnitude using BDSN broadband records, Bull. Seis. Soc. Am., 86, 1314-1330, 1996.

Uhrhammer, R., L. S. Gee, M. Murray, D. Dreger, and B. Romanowicz, The Mw 5.1 San Juan Bautista, California earthquake of 12 August 1998, Seismol. Res. Lett., 70, 10-18, 1999.

Wald, D., T. Heaton, and K. Hudnut, The slip history of the 1994 Northridge, CA, earthquake determined from strong motion, teleseismic, GPS and leveling data, Bull. Seis. Soc. Am., 86, s49-s70, 1996. Wald, D., V. Quitoriano, T. Heaton, H. Kanamori, C. Scrivner, and C. Worden, TriNet "ShakeMaps": Rapid generation of peak ground motion and intensity maps for earthquakes in southern California, Earthquake Spectra, 15, 537-556, 1999.

next up previous contents
Next: Data Archive and Distribution: Up: Operations Previous: Data Acquisition and Network

The Berkeley Seismological Laboratory, 202 McCone Hall, UC Berkeley, Berkeley CA 94720
Questions and comments to
Copyright 2000, The Regents of the University of California.