The Berkeley Digital Seismic Network is a regional network of very broadband and strong motion seismic stations which spans northern California, and is linked to UC Berkeley through continuous telemetry using wideband frame-relay telephone connections. This network is designed to monitor regional seismic activity at the magnitude M 3 level as well as to provide high quality data for various research projects in regional and global broadband seismology. The network upgrade and expansion, commenced in 1991, is continuing. Growing from the original 3 broadband stations installed in 1986-87 (BKS,SAO,MHC), the BDSN currently counts 19 stations and plans for the installation of 4 additional stations are under way. We take particular pride in high quality installations, which involves often lengthy searches for appropriate sites, away from sources of low frequency noise, as well as continuous improvements in installation procedures and careful monitoring of noise conditions at individual existing stations.
Future expansion of our network is contingent on the availability of funding and coordination with other northern California institutions, for the development of a denser state-of-the-art strong motion/broadband seismic network and joint earthquake notification system in this seismically hazardous region.
All BDSN stations are equipped with three component broadband sensors and strong motion accelerometers. A downhole broadband sensor package (Guralp CMG3T) is deployed at the BRIB station. A Guralp CMG40T is deployed at the FARB station. Other BDSN stations have Streckeisen broadband sensors (Wielandt and Streckeisen, 1982; Wielandt and Steim, 1986), of either STS-1 of STS-2 type. FBA-23 strong motion accelerometers with 2g dynamic range are deployed at all stations. Figure 2.1 shows the distribution of operational and planned BDSN stations.The equipment configuration and telemetry type are listed in Figure 2.2
During 1997-98 an eight channel Quanterra Model Q4120 recording system was deployed at BDSN station CVS. Eight channel Q4120s were earlier deployed at FARB, and WENL. The recording systems at all other stations are Quanterra Model Q935s which allows for data acquisition from all six channels of both the broad- band and strong motion accelerometers through a 24-bit A/D conversion at a variety of sampling rates.
Continuously sampled and telemetered channels are the following: ULP (0.01 samples/sec), VLP (0.1 samples/sec), LP (1.0 samples/sec) and VBB (20 samples/sec). Triggered channels (VSP, 80 samples/sec from the STS seismometers, and LG, 80 samples/sec from the FBA23s) use the Murdock, Hutt and Halbert event detection algorithm for triggering Individual data streams which are currently extracted through digital FIR filtering as shown in Figure 2.3.
In addition, at all BDSN stations, data from the FBA-23 strong motion sensors are continuously recorded at one sample per second and stored on the digitizer hard disk to provide low frequency data, retrievable by redundant dial-up modem in the event of saturation of the STS seismometers during a large local earthquake. Figure 2.3 illustrates the data filters through the telemetry subsystems to the Berkeley host. Figure 2.4 illustrates data flow from the sensor to the digitizer.
BDSN stations have continuous data telemetry links between the remote station and Berkeley data acquisition computer, in addition to dial-up telephone/modem access. During 1997-98, the dataloggers were upgraded at six stations to enable ethernet connectivity, and allow multiple remote recipients of the data.
During the 1997-98 fiscal year, a new broadband station was added at the Carmenet Vineyards, north of the city of Sonoma. The seismic instruments were placed within a wine aging and storage tunnel drilled 50 meters into the volcanic tuft which is characteristic of the local geology.
The overburden is sufficient to limit thermal variations. The signal variation on the temperature channel (LKS), over the last week of June 1998, is 1613 counts P-P or approximately 83 millidegrees Celsius peak to peak. The power spectral density plot (PSD) for station CVS is shown in Figure 2.5.
Although background noise at the Carmenet Winery location periodically increases due to the bottling, shipping, and refrigeration activity, data are typically suitable for observation of both local and teleseismic activity. Installation within the existing tunnel reduced station construction and operation expenses.
Station CVS features a Q4120 data logger, STS-2 broadband seismometers, FBA-23's and 56 kbit/sec continuous telemetry to Berkeley. A GPS clock provides reference timing. Low loss co-axial cable (<2 db per 100 feet) was used to minimize the attenuation of the signal, and corresponding loss of external clock source.
During the past year, floods, lightning and weather related power outages affected data acquisition and telemetry at BDSN stations (FARB, WENL, KCC, and SAO), HFN station YBIB, and BARD station SUTB.
The BDSN station WENL was installed in 1996-97. During the past winter, groundwater in the tunnel rose one half meter over the seismic pier, submersing both the strong motion FBA-23 and the broadband STS-2 instrument. The water affected connectors at the seismometer interface, contaminating and effectively shorting the power to the instrument as well as the output signal lines.
At Berkeley, routine analysis of the data from WENL indicated a problem. The station was found flooded and we had to remove both seismometers. After returning to the lab, both were inspected for water damage. Fortunately, the instruments were sufficiently sealed so that no internal water was found, but the connecting cables were ruined.
When the ground water subsided, the instruments were replaced at WENL, and data acquisition resumed. It may be necessary to remove the instruments during next winter's rains to prevent another submersion, and we are exploring the possibility of moving this station to another location.
At station FARB, power for the system is supplied by diesel generators. The island facilities are operated by the US Fish and Wildlife Service. During Christmas week, the island generator failed. Dr. Uhrhammer and Dave Rapkin made a trip to the island to repair the generator before the back up batteries were drained. However, the generator failed again in late January, and data acquisition was interupted until early March.
In March, engineers from the BSL visited the island twice, once to repair the diesel generators again, and a second time to upgrade the data acquisition system and re-initiate data acquisition and continuous telemetry. For the second consecutive year, personnel from BSL received letters of commendation from the US Fish and Wildlife Service for technical efforts and critical support of the facilities at the Farallon Islands.
Lightning and surge protection are used on power and telemetry connections at all BDSN stations. After a lightning strike, the device must be replaced -requiring a station visit. Also, these devices do not protect against the magnetic component of a lightning strike. In the event of close lightning strikes, equipment beyond the protection can be damaged.
Lightning strikes during heavy winter storms affected data acquisition and telemetry at BDSN stations KCC and SAO. In both cases, primary surge protection was destroyed and telemetry equipment damaged. The lighting strike was particularly close to the instruments.
At KCC, BSL engineers made a hurried trip to the station to replace damaged equipment in early January, before heavy winter storms made the site inaccessible. A second lightning strike occurred in spring, which damaged the overhead communication (telephone) wire as well as BSL hardware. BSL hardware was replaced and continuous telemetry via wireless modems was re-established. Additional lightning and surge protection was added to both the power and telemetry interfaces.
At HFN station YBIB, a winter lightning strike destroyed the wireless telemetry modem, and one of the three accelerometers, damaged the signal preamplifier, and damaged the Q4120 data logger. Integrated circuits were blown off on the printed circuit board within the data logger. The station is located beneath the San Francisco-Oakland Bay Bridge. The bridge itself was likely hit by the lightning rather than the seismic equipment.
The BARD station at Sutter Buttes (SUTB) likewise experienced a lightning-related failure of the wireless telemetry link. Again, BSL engineers made a hurried trip to replace the damaged equipment and restore telemetry.
During the 1997/98 year, upgrades to BDSN stations were initiated, consistent with the broader goals of improving signal-to-noise ratios, and building redundant telemetry and data logging as part of REDI.
In order to improve signal-to-noise ratios, warpless baseplates were purchased and installed under the STS-1 seismometers, at BKS, SAO, YBH, CMB, ORV, KCC, and HOPS. The improved baseplate design developed by Streckeisen for the Albuquerque Seismological Lab enables the horizontal STS-1 components to be evacuated.
Installation of the warpless baseplates involves removing the seismometer from the old baseplate, removing the baseplate, placing the warpless baseplate, replacing, re-orientating, releveling, and evacuating the seismometer within the bell jar housing. BSL engineers designed and built special laser fixtures to assure that the reinstalled instruments were aligned as originally installed. Replacing of a single baseplate takes up to eight hours.
Concurrent with the baseplate upgrades, the dataloggers were upgraded to enable full ethernet connectivity (see subsection on Quanterra upgrades). In addition to the STS-1 stations above, the ethernet upgrade was performed at stations with STS-2, and Guralp CMG-40T seismometers (WDC, BRIB). Station KCC was not upgraded to ethernet connectivity because the telemetry link currently will not support the necessary protocol and data rate.
The ethernet upgrade required modification to both the hardware and software within the Q935 data loggers. Because the system would be stopped and powered down, the data logger modifications were co-ordinated with the seismometer baseplate upgrades. In one instance, a planned upgrade was delayed due to the occurrence of a large teleseismic event (Mw>7.0), in order to preserve undisturbed 24-hour records for normal mode data analysis.
Also, as part of the overall upgrade, a device to turn the entire station off when the back-up battery voltage falls below a minimum threshold was installed. Although such a circuit exists within the Quanterra data loggers, the seismometers, telemetry equipment and reference clocks have traditionally continued to be powered, after the data logger turned off due to low battery power. With the charging circuit off, the equipments continue to draw constant power from the batteries. (Electrically, power = voltage x current). As the battery voltage gets lower, the current increases to the point of failure (ie -fuse blowing or circuit fails at the component level). Either case requires a station visit for repair. The low voltage cut-off device therefore protects equipment, and eliminates the need to visit a station after an extended power outage.
Thirteen stations of BDSN currently have continuously recorded and telemetered barometric and temperature data channels.
With continued support from Pacific Bell's CalREN (California Research and Educational Network) to Caltech and UC Berkeley, the Berkeley Seismological Laboratory has maintained and expanded its frame relay network for real-time data acquisition from its seismic and GPS network.
The frame relay network uses DS0 (56 KBit/second) and DS1 (or T1, 1.5 MBit/second) digital phone circuits. Frame relay is a packet-switched network, which allows a site to use a single physical circuit to communicate with multiple remote sites through the use of permanent virtual circuits (PVCs). Frame Relay Access Devices (FRADS), which replace modems in frame relay network, can simultaneously support multiple interfaces such as RS-232 async ports, synchronous V.35 ports, and ethernet interfaces.
Our CalREN grant provided for a 56 KBit/second line at each of the remote seismic stations and GPS stations, and a single 1.5 MBit/second T1 circuit at UC Berkeley to receive the data from all of the stations that use frame relay. In addition, we were granted an additional 1.5 MBit/second at UC Berkeley and a corresponding T1 circuit at USGS Menlo Park to be used for the real-time exchange of seismic data between the BSL and Menlo Park.
Frame relay was installed at the seismic sites CVS, BDM, at the GPS sites MONB and DIAB, and at the CA Office of Emergency Services (OES) in Sacramento. The site CVS (Carmenet Vinyards, Sonoma) is currently operational, but BDM still awaits additional cabling to be installed at the site between the Pacific Bell demarcation and the seismic site. The GPS sites MONB and DIAB are both in operation, and provide real-time GPS data over async RS-232 interfaces.
The installation of the OES circuit was completed last year, but the PVCs to six seismic stations (YBH, ORV, HOPS, CMB, BKS, and SAO) were just recently completed by Pacific Bell and CalNET. The OES installation will serve as a secondary central telemetry and REDI processing site that is outside the Bay Area. The REDI computer system has been installed, and should be operational by the end of Fall 1998.
Ethernet interfaces and TCP/IP software were purchased for all of the Quanterra Q935 dataloggers and installed together with the latest version of Quanterra UltraSHEAR software. The software corrected several problems associated with real-time control of data channel telemetry, and provides a uniform software base for our central site data collection software. All of the software upgrades were performed with when the new disks drives were installed, or by downloading and installing the software over the network from UC Berkeley.
The ethernet interfaces provide multiple benefits: faster data telemetry speed over TCP/IP socket connections than over RS-232 serial lines, support for data connections to multiple data collection systems, FTP access for retrieving older data and downloading software updates, and telnet services to the dataloggers. We simultaneously upgraded the disks drives in the datalogger to 4 Gbyte disks. The new disk drives store more data, are faster, and in some instances, are lower power than the original disk drives. The new disks provide significantly larger circular buffers for data storage, and increase our ability to retrieve data and provide a continuous data archive even after significant telemetry problems.
In preparation for the ethernet upgrades, YBH, BKS, and CMB stations were changed to run Serial Line IPO (SLIP) software using beta software from Quanterra, which provided TCP/IP connectivity over a serial line. This allowed us to test the TCP/IP capabilities of the dataloggers, and uncovered several bugs with the SLIP software which were corrected in later versions. We have installed SLIP on the station YBIB in order to provide TCP/IP connectivity, since that station uses an async spread spectrum radio connection instead of frame relay.
In conjunction with IRIS, the BSL has developed software for MultiSHEAR (the next generation Quanterra datalogger software) that will integrate GPS data into the Quanterra telemetry and data stream. This will allow both GPS and seismic data to be sent in real-time across a single comserv telemetry link and to be written to tape and disk files. The BSL is in the final stages of testing the software, and is waiting for the release of the MultiSHEAR software by Quanterra. We plan to upgrade all of our stations to use MultiSHEAR, and to integrate the GPS data into the Quanterra telemetry streams at that time. Transmitting GPS data through the Quanterra datalogger provides a local on-site system that will continue to collect and buffer data even during telemetry outages.
Two sites are in the final stages of construction in the San Francisco Bay Area: Black Diamond Mines (BDM) and Potrero Hills (POTR), south and north of San Pablo Bay respectively (figure 2.1.0), and will complement our coverage of Bay Area earthquakes on the east side. These sites will be completed before the end of 1998. The BSL is currently negociating for a new site in extreme northeast Califoria (Modoc County) and for a quieter site in Humbolt County, as a replacement for the current Arcata site. The new station in Modoc County will be housed within an underground adit, within 15 kilometers of the California/Oregon/Nevada boundary. The Modoc site is not located near any public telemetry facility, and arrangements have been made with the US National Seismic Network to install and operate a VSAT (Very Small Aperature Telemetry) satellite dish as part of the USNSN network, which will provide data telemetry to both the USNSN and the BSL. The new site near Arcata should be serviced by frame relay. Two additional sites further south are planned as a joint effort with the USGS, as part of the CREST program (NOAA funded tsunami warning network). The BSL will install frame relay and spread spectrum radio system to provide continuous telemetry from these sites to the BSL.
Discussions are still ongoing with Pacific Bell about the possibility of installing seismographic equipement in their facilities on a large scale basis.
This section covers some issues involved in the operation of the Berkeley Digital Seismic Network (BDSN); namely the installation, calibration, performance, and quality control issues involved in the operation of broadband seismic stations. In order to achieve the performance of which a broadband seismometer is capable, one must install the instruments in an appropriate seismic vault. At the request of the seismological community, we have written a guideline on installing broadband seismic instrumentation based on the 35+ years of experience that the staff at Berkeley has in installing broadband instruments. An important aspect of operating a broadband network is the maintenance of transfer function files which describe the responses of the various instrumentation at each BDSN station and the routing checking of the quality of the seismic data and the noise Power Spectral Density (PSD) at each station. Such routine checking is invaluable for expediting the identification of problems with the broadband instrumentation since a broadband instrument malfunction is not always obvious when merely viewing the broadband output from the sensors. Over the past few years we also have added thermistor and pressure sensor channels at many BDSN stations to record relative seismometer temperature and ambient pressure changes so that temperature and pressure generated noise in the seismic band can be reduced via cross-correlation techniques.
As the deployment of broadband seismometers becomes increasingly common practice, the need for specific installation guidelines has emerged. Indeed, in order to operate efficiently, broadband instruments require a much more controlled environment than standard short-period seismometers. In order to help those who are contemplating the installation of broadband instruments, we have written a guideline on the installation of broadband seismic stations (Uhrhammer et al., 1998). An online version of the guidelines is available on the BSL homepage at URL http://seismo.berkeley.edu/bdsn/ instrumentation/guidelines.html.
Rather than trying to cover all aspects of the installation procedures in detail, we concentrate on those aspects which have the largest impact on the overall performance of a seismic station housing a broadband seismometer and a high-resolution (24-bit integer) data logger. The two aspects of the installation process which most influence the overall performance of the broadband seismometer are the construction of the seismic pier and the application of thermal insulation around the sensor and pier.
When a potential site for a broadband seismic station has been found, it is necessary to deploy a portable broadband sensor recorded by a 24-bit resolution data logger at the site for a few days in order to determine the broadband power spectral density (PSD) characteristics of the background noise. This is an important step because broadband noise levels in general and low-frequency noise levels in specific can not be reliably predicted. Our experience is that a quiet short-period site will not necessarily be a low-noise broadband site and vice versa.
Typically, the broadband seismometer is capable and responsive to sub-nanoradian tilts. As a practical matter, a nanoradian tilt step will induce a signal on a 360 second free-period broadband seismometer (a Streckeisen STS-1, for example) which is generally about 20 times the RMS background noise level at quiet sites. Site selection on low porosity, hard rock is therefore critical. Any unconformity beneath or within the pier (such as rock aggregates or rebar) will add to the ambient noise. The primary concern is that the pier be stable and affect neither the response of the earth nor of the seismometer. Also, the concrete pier should be approximately level and simply be capable of holding the seismometer. The concrete mixture should be as homogeneous as possible and a rich mixture of 50 percent portland cement and 50 percent sieved sand will produce a very hard smooth surfaced pier that is ideal for supporting broadband sensors.
Of all the things that can be done when installing a broadband seismometer, thermal insulation has perhaps the biggest impact on the overall performance of the seismometer, and it has the added advantage of being both inexpensive and easy to install. The objective is to achieve a thermal time constant of sufficient duration to significantly attenuate at least the thermal noise in the seismic band (0.3 mHz and above) and preferably also the diurnal thermal signature. The insulation of choise is semi-rigid polyisocyanuratic foam board with aluminum foil facers on both sides. This insulation produced in 4 by 8 foot sheets at thicknesses of up to 4 inches. The insulation, see Figure 2.6 for example, is cut to shape with a hand saw and glued together with expanding polyurethane resin. Retrofitting an existing pier with thermal insulation is an inexpensive way of significantly reducing the broadband noise. For example, insulating the pier at the BDSN station at Mt. Hamilton (MHC) increased the thermal time constant of the broadband seismometers from 1000 seconds to 2+ days which significantly reduced the noise power (by 20+ dB) at diurnal frequencies as shown in Figure 2.7.
The new Ultrashear and software installed in the BDSN Quanterra dataloggers allows central site triggering and considerable flexibility in the signal processing, both under software control, at the expense of requiring significantly more complex file structures to keep track of the instrument transfer functions and software settable parameters. The complexity is necessary for the generation of dataless SEED records, as required for distribution of BDSN data from the NCEDC in the form of SEED volumes, where the entire signal path must be unambiguously documented.
The staff has spent considerable time in the construction and verification of the files which describe the sensor, preamp, digitizer, and filtration components for each channel and data stream a the BDSN stations. Dr. Uhrhammer set up the initial files and the engineering staff has been verifying the files. Once this process is completed, the data will be entered into the new relational database to keep track of the instruments at each station, to generate the dataless SEED records for export of NCEDC data, and to construct the BDSN transfer functions used by the BSL researchers.
One procedure that is used on an ad hoc basis to check the instrument transfer functions as well as the response of the instruments is to deconvolve to ground velocity the signals from a large teleseismic P-wave. Figure 2.8 shows the vertical-component P-wave ground velocity in the 5 mHz to 8 Hz frequency band for a Mw 7.7 teleseism which occurred on 14 October 1997 in the Fiji Islands. The traces are scaled to absolute ground velocity and ordered by increasing distance from top to bottom. Note the similarity in the waveshapes and amplitudes which implies that all the BDSN vertical-component broadband seismometers are responding in a consistent manner. If one of the traces differed considerably from the other traces, it would imply that something is amiss with either the transfer function or the sensor.
One of the characteristics of broadband instruments is that the wide frequency bandwidth output by the sensor can render visual detection of a sensor malfunction difficult to detect at best. For this reason we have implemented automated algorithms which routinely process the broadband data in order to expedite the identification of a broadband sensor which has malfunctioned.
An automated algorithm, which calculates the mean signal level and standard error for each broadband sensor, runs daily and flags instruments where the mean signal level is more than 1 percent of full scale. This information is particularly useful for rapidly identifying instruments which have malfunctioned (eg, either drifted off center, tilted, or gone dead). The algorithm generates a daily email report.
Likewise, an automated BDSN PSD analysis algorithm is run weekly to characterize the seismic noise level recorded by each broadband seismometer. The algorithm sends the results via email to the engineering and some research staff members and it also generates a bargraph output, as shown in Figure 2.9, which compares all the BDSN broadband stations by components. Note that inland stations are generally quieter than the coastal stations. The quietest stations are mostly in old mines in remote locations and the noisiest stations are either near the ocean or on soft material or in the case of FARB, on the Farallon Islands off the coast west of San Francisco.
To show the performance of the BDSN broadband instrumentation over the past year (July 1997 to June 1998) we generate an annual synopsis of the weekly PSD values for each broadband instrument and produce a bar graph as shown in Figure 2.10.
As a demonstration of the capabilities of a modern very-broadband seismometer coupled to a high-resolution (24-bit) digital data logger, Dr. Uhrhammer examined in detail the performance of the BDSN station at Yreka (YBH). YBH is located in a remote hard-rock mining drift, in the Klamath National Forest (430 km N of Berkeley and 120 km from the coast), with no significant cultural noise sources within a radius of 10 km. Consequently, YBH has the lowest background noise level and also the most thermally stable vault (less than 60 millidegrees Celsius variation per year), owing to its location in a remote mine with stable air stratification, of any BDSN station. We also record the barometric pressure and the seismometer temperature along with the seismic signals to allow for estimation and compensation of the temperature and pressure correlatable noise.
Under the ideal conditions that exist in the YBH vault, the vertical component broadband sensor is capable of recording frequencies over a 8+ decade frequency range from 60 nHz to 8 Hz. Thus we can record such varied signals as the fortnightly gravitational tides, the free oscillations of the earth, low-frequency surface waves, and high frequency body waves with the same sensor. An example of the low-frequency capabilities is shown in the Figure 2.11. The signals are, from top to bottom, the original seismometer signal (RHZ), the pressure signal (RDS), the temperature signal (RKS), the compensated seismometer signal (RCZ), the theoretical gravitational tides (RGZ), and the residual signal (REZ) (REZ = RCZ - RGZ). The signal duration is 6 months and the data have been low-pass filtered using a sharp-cutoff FIR filter with a corner frequency of 0.4 mHz and decimated to a 1 mHz (1000 second) sampling rate. Note that the largest signal remaining in the bottom trace is due to a Mw 7.8 earthquake exciting the gravest free oscillations of the earth.