Northridge Aftershock Data

• Summary
• Introduction
• Station Map
• Station & Channel List
• Selected References

Summary

• Network Code: XA
• Experiment duration: 1994-01 to 1994-03
• Number of events: 4736
• Prinicpal investigators: Adam Edelman, Frank Vernon
• Dataless SEED volume currently unavailable. Please visit the IRIS DMC

Introduction

A collaborative effort to record aftershocks of the January 17, 1994 12:30:55 Mw 6.7 Northridge earthquake was undertaken by several institutions associated with the Southern California Earthquake Center (SCEC), including the California Institute of Technology (Caltech), Kinemetrics Inc., Lawrence Livermore National Laboratories (LLNL), San Diego State University (SDSU), the United States Geological Survey (USGS), the University of California at San Diego (UCSD), the University of California at Santa Barbara (UCSB), and the University of Southern California (USC). In the seven weeks following the main shock, over 100 portable instruments were deployed throughout the Los Angeles county area, with coverage extending from southern Ventura county, through the San Fernando Valley, and south into the Los Angeles basin. Instrumentation consisted of both short period and broadband velocity sensors, along with strong motion accelerometers each recorded on either 12, 16 or 24 bit dataloggers.

This aftershock deployment was conducted with several scientific objectives in mind from the beginning. Instruments were deployed directly above the aftershock zone to help constrain aftershock hypocenters. Many sites were chosen to investigate the near surface site effects which produced extensive damage during the earthquake. Strong-motion sites that recorded the main shock were instrumented to collect empirical Green's functions for separating source and propagation effects of the main shock. Broadband instruments were deployed to study the long period response of the Los Angeles basin, and a dense array using broadband and short period sensors was deployed to study the response of the San Fernando Valley basin.

The completed set of aftershock data consists of almost 3 GB of data comprising 4817 events, of which 428 were Ml = 3.0 and greater with locations provided by the Southern California Seismic Network (SCSN). In addition to the waveforms, both P and S-phase arrival picks are included and available from the SCEC data center, and complete instrument response functions are available in the SEED format data from the Incorporated Research Institutes for Seismology (IRIS) Data Management Center at the University of Washington.

Field Deployment

The initial response to the earthquake involved the mobilization and deployment of all available equipment to the main shock area. The Pasadena office of the USGS became the staging area due to its proximity to the aftershock zone, and access to network data. USGS and Caltech personnel were instrumental in providing preliminary aftershock locations as well as permanent network station locations, both of which were essential to the efficient deployment of the portable instruments. AirTouch Cellular also provided cellular telephones to all field teams which was extremely helpful for communication with the staging office, especially during the first few days following the main shock when damage to the infrastructure made travel and communication difficult. The portable instruments were provided by the SCEC instrument center at UCSB, Caltrans, LLNL, SDSU, UCSD, and the USGS offices in Pasadena, Menlo Park, and Golden, Colorado. Additional instruments were provided by the RAMP project at the IRIS PASSCAL Instrument Center at Lamont Doherty Earth Observatory (PIC).

Several field teams were established and the work was divided by geographic region. By the end of the first 48 hours there were 21 instruments deployed and recording data, with a total of 50 by the end of the first week. Figure 1 shows all portable station locations color coded by instrumentation type. Halfway through the experiment, several instruments were re-deployed in a dense array throughout the San Fernando Valley basin. Locations of these stations is shown in Figure 2. The time line in Figure 3a and Figure 3b shows the data availability from each site, with color indicating the sensor type. The time line gives a qualitative view of the time periods for which the stations were acquiring data which, due to variable trigger sensitivities, does not imply that all events during those time periods were recorded by those stations. Both the TERRAscope stations which are shown in green, and the GEOS stations, which are not included on the time line, contributed data for the 46 aftershocks of Ml = 4.0 and greater.

There were several different equipment configurations used during the deployment; many 16-bit stations consisted of three components of short-period velocity sensors and three components of strong motion accelerometers, some recorded six channels of velocity or acceleration data in a high/low gain configuration, and some recorded 3 channels of broadband velocity data on 24-bit streams. In addition, there were also seven 12-bit accelerographs deployed in close proximity to the main shock epicenter. Hardware configurations, station locations, data sampling rates, and site conditions are tabulated in Table 1. Timekeeping at most stations was supplied by either Omega or GPS time clocks, except for stations KMAR, KMCH, KMNH, KMNL, KMSF, KMVN, and WVES which all ran on internal datalogger oscillators only. Information on the time clocks used at the GEOS stations are available from the USGS, and the TERRAscope information is available from Caltech.

The total system response for each of these configurations was calculated and can be used to remove the instrumentation characteristics from the data. Figures 4a-d, 4e-h, 4i-l show the normalized velocity and acceleration phase and amplitude response for several of the portable station instruments.

Due to the urban location, environmental noise was a concern but resulted in minimal problems. At most sites, the sensors were sufficiently shielded from local noise sources. All broadband STS-2 sensors were plastered on concrete slabs and insulated with fiberglass, and were installed in private garages, fire stations, and other enclosed buildings. The short period velocity sensors and accelerometers were installed in a variety of conditions: buried in soil, on hard rock, and on asphalt.

Each station was serviced at regular intervals and the preliminary data processing was done at the field teams home institution. Preliminary processing consisted of reading the raw field tapes and converting to SEGY format. Station timing and data quality were assessed in order to determine if all hardware was functioning properly. The SEGY data was then archived on either 4 mm DAT or 8 mm exabyte tape.

Data Processing

After the initial raw data processing, all SEGY data was compiled into a single set. A large amount of disc space was required to handle the volume of raw data (over 5 GB) most efficiently. The majority of processing was shared between a Sun Sparc station LX and Sparc station 10 with 9 GB of SCSI disc space.

The first step was to organize all data by station and time, and convert it to CSS 3.0 format. At this point, a time based association with the the SCSN catalogue was done for all aftershocks of Ml = 4.0 and larger. This preliminary data set was made available in May 1994 via the internet in both SAC and CSS 3.0 format. The final data set processing continued as the data was run through an automatic phase picker provided by Roger Hansen at the University of Colorado, Boulder which uses an auto regressive algorithm to determine the P phase arrival. The data was then reviewed by analysts who checked the P picks, determined the S-phase arrival, and added error ranges.

In the meantime, all state of health logs were reviewed and any necessary timing corrections were determined utilizing software from the IRIS PIC. Corrections for improperly set Omega time clock parameters and periods of poor Omega reception were necessary at many stations. The program REFRA TE from the PIC was used which reads the state of health logs and calculates the offset and drift rate for time periods when the clock is unlocked. Table 2 shows the timing corrections required for each portable station, and describes the types of shifts that were applied. In some instances it is known that there are problems with the timing, but it was not possible to determine the exact shift necessary to confidently correct the error. These problem stations are also noted in Table 2, along with the approximate time offset that has been determined through a qualitative review of travel time residuals. Corrections were not made for those offsets since the exact error is unknown.

Figures 5a, 5b, 5c and 5d show the travel time residuals for both the P and S-phase for the portable stations described in Table 2. These were calculated when the association with the SCSN catalogue was done. The figures show that the majority of stations have residuals clustered primarily around 0, with a small static offset present due to the differences between the IASPEI travel time model used for association and the actual crustal velocities. Several stations show a large scatter in their residuals in the early stages of the deployment, which can be attributable to the high number of triggers being incorrectly associated with larger events. The association was done within a window of several seconds for both P and S-phases, and due to the high frequency of events many of these phases were associated with other events that occurred within the allowable time window.

Stations BRCY, CWHP, KM:AR, KMCH, KMNH, KMNL, KMSF, KMVN, LA03, and WVES are flagged in red in Figure 5 and Table 2 and all show deviations due to errors in timing. Prior to 94029:10:50, station BRCY had a problem with the Omega time clock hardware which resulted in the offset apparent in Figure 5. At 94029:10:50 the difference between the data time and actual time was -1.968 seconds which can be applied to the data to correct for the error at that time, but the choice was made not to apply that to the data released do to uncertainties in the time clock behavior prior to 94029:10:50. After that time, the timing quality is reliable. Station CWHP has a timing discrepancy of approximately 2.00 seconds prior to 94030:00:00:40 which is apparent in Figure 5, though there are no clock messages that allow for any calculations of the actual error in order to correct the data. The Kinemetrics stations, KMAR, KMCH, KMNH, KMNL, KMSF, and KMVN were left to free run on internal oscillators and were never synchronized to a known time source during installation, leaving the time quality unknown. Station LA03 had time clock hardware problems and never locked on to a signal, and station WVES was installed without a time clock and left to run on the datalogger's internal oscillator.

After all phase picking was completed, the timing corrections were applied. An association was then done with the SCSN catalogue utilizing software from the JSPC at the University of Colorado, Boulder which associates predicted arrival times from the SCSN origin table with the actual P and S-phase picks. The data was then converted to the SCEC database format and shipped to the SCEC data center at Caltech, and to SEED format and shipped to the IRIS Data Management Center at the University of Washington.

Data

The completed data set consists of almost 3 GB of data comprising 4817 events, all of which are associated with an SCSN location. The aftershock distribution is shown in Figure 6 with portable station locations included for reference. The aftershock magnitude distribution is shown in the upper half of Figure 7 indicating that the majority of aftershocks were MI = 3.0 and smaller. The lower half of Figure 7 is a histogram showing the number of events recorded versus the number of stations recording each event. Almost 200 events were recorded by more than 20 portable stations, with some having data from up to 50 different sites. Since the majority of aftershocks were in the MI1.0 - 3.0 range and not large enough to trigger many of the instruments, most events were recorded at 20 or less stations. The 46 events of MI = 4.0 and larger are listed in Table 3.

Figure 8 is a good example of typical station coverage, showing the vertical time series from 43 stations of an MI = 4.1 event. The hypocentral distances range from less than 15 km to over 230 km. Another interesting event example is shown in Figure 9. This figure shows a 900 second time series from 6 broadband stations, and the vertical component S coda spectra from a Los Angeles basin and a San Fernando Basin site. The broadband recordings show large amplitude, low frequency shaking which can be attributed to basin reverberation.

Discussion

The Northridge portable instrument deployment was not the first aftershock recording experience for many of the SCEC personnel involved. There were many lessons learned in previous experiments in Landers and Joshua Tree that were successfully employed during the Northridge work. Issues of organization, communication, and logistics were handled extremely well in the early stages with daily meetings at the Pasadena office of the USGS. Communication with field personnel, which was almost non-existent in previous deployments, was made possible through the use of cellular telephones. Jennifer Haase of Caltech became a single point of contact for field teams, providing updated station location information continuously. This level of communication resulted in the need for minimal site re-Iocation, and played an essential role in the success of this deployment.

Since this was the first major aftershock deployment in a populated urban environment, there were several new issues to contend with. Damage to transportation routes sustained in the main shock made travel to sites difficult, and sometimes impossible in the early stages. Sites selection was complicated since most land is privately owned, though field teams received much cooperation from local fire stations and relatives in the area. Also, much consideration was put into site selection to minimize the affects of urban noise sources as well.

We learned the benefits of installing broadband seismometers in conjunction with short period sensors from the Landers deployment, and we found the same benefits during this deployment as evident in Figure 9. The low frequency response of the broadband STS-2 sensors captured the basin reverberation, which is not apparent in the short period records.

Acknowledgments

Several people from many different institutions played important roles in all aspects of preparing this data set, and we would undoubtedly leave someone out if we attempted to name every individual. Instead, we will list the contributing institutions and organizations: Caltrans, Kinemetrics Inc., LLNL, AirTouch Cellular, the IRIS P ASSCAL Instrument Center at Lamont Doherty Earth Observatory, the Southern California Earthquake Center, including Caltech, SDSU, UCSB, UCSD, USC, and the USGS; and the University of Colorado at Boulder. Funding from SCEC fund #78023A helped support the data processing work.

Station Map

Station List

(plain text version)

Total: 115

Channel List

(plain text version)