The Bay Area Regional Deformation (BARD) network is a collection of permanent, continuously operating GPS receivers that monitors crustal deformation in the San Francisco Bay Area (SFBA) and Northern California. Started in 1991 with two stations spanning the Hayward Fault, BARD has been a collaborative effort of the Berkeley Seismological Laboratory (BSL), the USGS at Menlo Park (USGS/MP), and several other academic, commercial, and governmental institutions. The BARD network is designed to study the distribution of deformation in Northern California across the Pacific-North America plate boundary and interseismic strain accumulation along the San Andreas fault system in the Bay Area for seismic hazard assessment, and to monitor hazardous faults and volcanoes for emergency response management. The BSL maintains and/or has direct continuous telemetry from 26 stations comprising the BARD Backbone (Table 3.11), while additional stations operated by the USGS, US Coast Guard and others fill out the extended BARD network. Twelve BARD Backbone sites are collocated with broadband seismic stations of the BDSN, with which they share continuous telemetry to UC Berkeley (Table 3.11).
With the completion of major construction on the Plate Boundary Observatory (PBO) portion of EarthScope, the number of GPS stations in Northern California has expanded to over 250 (Figure 3.21) and a number of BARD stations were folded into the PBO network. Together, PBO and BARD stations provide valuable information on the spatial complexity of deformation in the SFBA and Northern California, while the BARD network has the infrastructure and flexibility to additionally provide information on its temporal complexity over a wide range of time scales and in real-time. Many of the GPS stations in the BARD network are collocated with BDSN seismic instrumentation or are close to active faults where reliable access to real-time information could be critical following an earthquake.
The majority (24 of 26) of BARD Backbone stations now collect data at 1 Hz sampling frequency (Table 3.11). The data are collected continuously, as opposed to on a triggered basis, and transmitted to the BSL. The effort to expand the high-rate data collection was helped by upgrades over the past several years at 12 stations to Trimble NetRS receivers. The NetRS receivers feature a compact data stream, which has allowed us to collect high-rate data from locations with limited bandwidth telemetry. Furthermore, IP connectivity on the NetRS facilitates streaming of data over a Ntrip server to other agencies and the general public. Data streams from NetRS equipped BARD stations are currently available (http://seismo.berkeley.edu/bard/realtime).
The BSL has received funding through the American Reinvestment and Recovery Act (ARRA) to upgrade the remaining BARD sites with Topcon Net-G3A receivers that will provide BINEX streaming of data at 1Hz sampling over TCP/IP. The new receivers will also be capable of recording L5 data in addition to L1 and L2; L5 is a third frequency that will be added to GPS satellites in the coming years. In addition to upgrading existing sites, we have also received ARRA funding to install seven new stations at existing BDSN stations (Table 3.11), thereby taking advantage of shared telemetry. Three of these stations will be mounted on the existing seismic vaults, while the remaining four will be new, short-braced monuments.
Twelve BARD stations are currently equipped with high performance Trimble NetRS receivers, which have sufficient internal buffering to allow robust real-time telemetry at 1Hz. Recent upgrades include stations YBHB and SAOB in April, 2009, MODB in August, 2009, and SUTB in March, 2010. At MODB, we are able to telemeter 1Hz data using the USGS VSAT system that collects seismic broadband data as part of the National Seismic Network (NSN). Other stations are still equipped with older Ashtech Z-12 (A-Z12) and Ashtech MicroZ-CGRS (A-UZ) receivers. At these sites, the data are collected using direct serial connections and are susceptible to data loss during telemetry outages.
All BARD stations are continuously telemetered to the BSL. Many use frame relay technology, either alone or in combination with radio telemetry. Other methods include a direct radio link to Berkeley (TIBB, EBMD) and VSAT satellite telemetry (MODB). We also changed our data strategy by allowing some data to be transferred by web-based telemetry (ADSL lines). This will reduce our communication operational costs and, we hope, will not affect our ability to react in a large event. Twelve GPS stations are collocated with broadband seismometers and Quanterra data loggers. With the support of Integrated Research Institutions for Seismology (IRIS), the BSL developed software that converts continuous GPS data to MiniSEED opaque blockettes that are stored and retrieved from the Quanterra data loggers (Perin et al., 1998), providing more robust data recovery from onsite disks following telemetry outages.
BARD station monumentations broadly fall into three types. Most are anchored into bedrock, either directly or via a steel-reinforced concrete cylinder. The five ``mini-PBO'' stations that are still operated by the BSL are collocated with USGS strainmeters and the GPS antennas are bolted onto the borehole casing using an experimental mount developed at the BSL, which has since been adopted by PBO for their strainmeter sites. Four sites (UCD1, SRB1, UCSF, SBRB) are located on the roofs of buildings. Most of the last type have been installed in the past three years, and their stability over long periods of time is yet to be evaluated.
Most BARD stations use a radome-equipped, low-multipath choke-ring antenna, designed to provide security and protection from weather and other natural phenomena, and to minimize differential radio propagation delays. Four stations are equipped with Trimble Zephyr Geodetic antennas, though these are scheduled to be upgraded to choke-rings under ARRA funding. A low-loss antenna cable is used to minimize signal degradation on the longer cable setups that normally would require signal amplification. Low-voltage cutoff devices are installed to improve receiver performance following power outages.
In September 2009, the BSL established the telemetry of high-rate data from 13 PBO stations in the Parkfield region. These stations were installed as part of the collaborative NSF/MRI program between the BSL, UC San Diego and Carnegie Institution of Washington nicknamed ``mini-PBO.'' Since September 2009, 1 Hz GPS data from these 13 stations flow through the T1 line from Parkfield to Menlo Park and then on to Berkeley. From here it is sent back to UCSD via a NTRIP server. We plan to participate in a state-wide real-time geodetic network that will eventually be integrated with the CISN for earthquake notification purposes. The acquisition of real-time data from the Parkfield subnetwork is the first step towards linking Southern and Northern California real-time GPS networks.
Raw and RINEX data files from the 26 BARD Backbone stations and several other stations run by BARD collaborators are archived at the Northern California Earthquake Data Center (NCEDC). The data are checked to verify their integrity, quality, completeness, and conformance to the RINEX standard, and are then made accessible, usually within 2 hours of the end of the UTC day, to all BARD participants and other members of the GPS community through the Internet, both by anonymous FTP and through the World Wide Web (http://ncedc.org/). BARD data are also available to the community through GPS Seamless Archive Centers (GSAC), such as that hosted by the Scripps Orbit and Permanent Array Center (SOPAC, http://gsac.ucsd.edu). High-rate raw data are also decimated to create 15 s RINEX data files. 1 Hz RINEX files are available for all BARD Backbone sites after May 2010; high-rate RINEX files for earlier dates will be backfilled in the coming months.
As part of the activities funded by the USGS through the BARD network, the NCEDC has established an archive of the 10,000+ survey-mode occupations collected by the USGS since 1992. The NCEDC continues to archive non-continuous survey GPS data. The initial dataset archived is the survey GPS data collected by the USGS Menlo Park for Northern California and other locations. The NCEDC is the principal archive for this dataset. Quality control efforts were implemented by the NCEDC to ensure that raw data, scanned site log sheets, and RINEX data are archived for each survey. All of the USGS/MP GPS data has been transferred to the NCEDC, and virtually all of the data from 1992 to the present has been archived and is available for distribution. These survey-mode data are used together with data from BARD and PBO stations to produce BAVU (Bay Area Velocity Unification), a united set of continuous and survey data from the wider San Francisco Bay Area, processed under identical conditions using GAMIT (d'Alessio et al., 2005).
Data from five of our sites (HOPB, MHCB, CMBB, OHLN, and YBHB) are sent to the National Geodetic Survey (NGS) in the framework of the CORS (Continuous Operating Reference Stations) project (http://www.ngs.noaa.gov/CORS/). The data from these five sites are also distributed to the public through the CORS FTP site.
Average station coordinates are estimated from 24 hours of observations for BARD stations and other nearby continuous GPS sites using the GAMIT/GLOBK software developed at MIT and SIO (King and Bock, 1999, Herring, 2005). GAMIT uses double-difference phase observations to determine baseline distances and orientations between ground-based GPS receivers. Ambiguities are fixed using the widelane combination followed by the narrowlane, with the final position based on the ionospheric free linear combination (LC or L3). Baseline solutions are loosely constrained until they are combined together. GAMIT produces solutions as “H-files”, which include the covariance parameters describing the geometry of the network for a given day and summarize information about the sites.
We combine daily, ambiguity-fixed, loosely constrained H-files using the Kalman filter approach implemented by GLOBK (Herring, 2005). They are combined with solutions from the IGS global network and PBO and stabilized under a North America-fixed reference frame. The estimated relative baseline determinations typically have 2-4 mm long-term scatter in the horizontal components and 10-20 mm scatter in the vertical. The most recent velocity solutions (Houlié and Romanowicz, in press, Figure 3.22) are in good agreement with previous work (e.g. d'Alessio et al., 2005).
Real-time streaming: We have begun the process of making all our 1Hz data available in real time; a step toward our goal of integrating GPS with the Northern California Seismic System (NCSS) for use in hazard assessment and emergency response. Stations with IP connectivity (currently those that are NetRS equipped) are the first to be streamed over our NTRIP server. Stations with serial connections will be phased in over the coming months. Similarly data are currently available in BINEX format, but RTCM streams and other raw formats will be added. Data are available to the general public, but an account must first be established; see http://seismo.berkeley.edu/bard/realtime for details.
Time-series analysis: Testing continues to re-establish automatic time-series generation of BARD Backbone data; an activity that was funded under the ARRA program. Daily processing ensures that bad data is caught quickly and any problems can be fixed in a timely manner. Several products used in time-series generation are available on different time scales. Final orbit files are generally available after 7-10 days, while IGS and PBO solutions have their own lag times. Figure 3.23 shows the detrended (residual) time series for station FARB on the Farallon Islands, for 5/1/2010 through 8/24/2010, as produced on 8/27/2010. The time series has been cleaned by removing common mode errors, which were determined using all California stations (though most are in the Bay Area). Overall scatter is very low, as would be expected for a time period with no major or moderate events, with root mean square (RMS) values of 0.9 mm, 0.6 mm, and 2.9 mm for the North, East and Up directions, respectively. The scatter in the data is not dramatically affected by being processed with ISG rapid orbit files (blue points) or when not combined with PBO solutions (green points), though the calculated error bars are affected. Motions above the several mm level should be detectable within 2-3 days, while smaller motions will be evident when final orbit files are available in 1-2 weeks. Once finalized, these plots will be posted on the BARD website and updated daily.
Metadata overhaul: Another major activity of the past year has been the updating, consolidation and presentation of site metadata and quality control information. Station log files (ftp://ncedc.org/pub/gps/site/) are now 100% up to date and conform to the IGS standard for metadata reporting. The BARD webpage (http://seismo.berkeley.edu/bard) has also been redesigned and upgraded to provide more information on individual stations. The web pages also include plots of data completeness (how many epochs are present in the data files) and estimated multipath for the L1 and L2 signals (Figure 3.24). These are updated daily and provide a measure of the antenna and telemetry performance and of the effect of the surroundings on the data quality. Changes to these values correspond to equipment changes, equipment failure and changes to the environment surrounding the site. The last is particularly important, as changes such as construction or tree removal can occur near a station without the BSL's knowledge.
The BARD program is overseen by Barbara Romanowicz and Ingrid Johanson (since February 2010). Rich Clymer, Bill Karavas, Rick Lellinger, John Friday, Doug Neuhauser, Mario Aranha and Jennifer Taggart contributed to the operation of the BARD network in 2009-10. Operation of the BARD network is partially supported by funding from the USGS/NEHRP program grants #07HQAG0031 and #G10AC00141 and infrastructure upgrades were made possible by funding from the ARRA grant #G10AC00079.
d'Alessio, M. A., I. A. Johanson, R. Bürgmann, D. A. Schmidt, and M. H. Murray, Slicing up the San Francisco Bay Area: Block kinematics from GPS-derived surface velocities, J. Geophys. Res., 110, B06403, doi:10.1029/2004JB003496, 2005.
Herring, T., GLOBK: Global Kalman Filter: VLBI and GPS Analysis Program, version 10.2, 2005.
Houlié, N. and Romanowicz, B., Asymmetric deformation across the San Francisco Bay Area faults from GPS observations in northern California, Phys. Earth Planet. In., in press.
King, R., and Y. Bock, Documentation of the GAMIT software, MIT/SIO, 1999.
Perin, B. J., C. M. Meertens, D. S. Neuhauser, D. R. Baxter, M. H. Murray, and R. Butler, Institutional collaborations for joint seismic and GPS measurements, Seismol. Res. Lett., 69, 159, 1998.
Berkeley Seismological Laboratory
215 McCone Hall, UC Berkeley, Berkeley, CA 94720-4760
Questions or comments? Send e-mail: firstname.lastname@example.org
© 2007, The Regents of the University of California