The data from the BARD sites generally are of high quality and measure relative horizontal positions at the 3-5 mm level. The 24-hour RINEX data files are processed daily with an automated system using high-precision IGS orbits. Final IGS orbits, available with 7-10 days of the end of a GPS week, are used for final solutions. Preliminary solutions for network integrity checks and rapid fault monitoring are also estimated from Predicted IGS orbits (available on the same day) and from Rapid IGS orbits (available within 1 day). Data from 5 primary IGS fiducial sites located in North America and Hawaii are included in the solutions to help define a global reference frame. Average station coordinates are estimated from 24 hours of observations using the GAMIT software developed at MIT and SIO, and the solutions are output with weakly constrained station coordinates and satellite state vectors. Due to the large number of BARD and fiducial stations, the BARD network is split into a "San Francisco Bay" subnet and a "Northern California" subnet, with the 5 fiducial stations and CMBB and FARB in common to both to help tie the subnets together. The weakly constrained solutions are combined using Kalman filter techniques using the GLOBK software developed at MIT that can apply tight a posteriori constraints. This helps to ensure a self-consistent reference frame for the final combined solution. The subnet solution for each day are combined assuming a common orbit to estimate weakly constrained coordinate-only solutions. These daily coordinate-only solutions are then combined with tight coordinate constraints to estimate day-to-day coordinate repeatabilities, temporal variations, and site velocities.
The estimated relative baseline determinations typically have 3-5 mm RMS scatter about a linear fit to changes in north and east components and the 15-25 mm RMS scatter in the vertical component. Figure 3 shows several time series from the automated 24-hour solutions for 1997-1998. These examples show some of the more unusual weather and tectonic activity recorded by the BARD network. The El Niño winter of 1997-1998 produced heavy rainfall along the coast and snowfall in the Sierra Nevada. The hillslope at the NUNE site became unstable during one rainstorm, causing the box housing the receiver to slide downhill and damage the antenna, which was more stably anchored at greater depth. This damage to the antenna caused an apparent 2-cm shift in position to the north (Figure 3a). MUSB, located at high altitude on Musick Mountain in the southern Sierra Nevada, becomes snowbound and inaccessible at times during the winter. Two episodes of large vertical motions occurred during this period in 1998, each beginning gradually but ending abruptly (Figure 3b). We speculate that these episodes were caused by the antenna becoming covered with snow that then abruptly fell off during melting. The Long Valley caldera experienced an episode of volcanic unrest, with increased levels of seismicity and high deformation rates, that began around May 1997 and abated in early 1998. The baseline length between CASA and KRAK, which crosses the caldera, shows the high levels of deformation (nearly 10 cm increase) measured during this period (Figure 3c). A similar deformation episode was measured by two-color laser EDM in 1989.
Figure 3: Time series of BARD station positions and baseline components from January, 1997 to August, 1998. Data are 24-hour estimates about their mean, with one standard deviation uncertainties, relative to the ITRF96 reference frame as defined by five North America and Hawaii IGS fiducials stations. a) The north component of NUNE. Typical relative scatter of horizontal components is 3-5 mm. The anomalous motion in early 1998 is non-tectonic, and is due to the antenna being damaged when the box housing the receiver slid downslope due to near-surface instability follow a severe winter storm. b) The vertical component of MUSB. Typical relative scatter of horizontal components is 15-25 mm. The anomalous motions in early 1998 are non-tectonic and are mostly likely due to heavy snow either covering the antenna or inducing severe multipath at the site. c) The length of the baseline from CASA to KRAK, which crosses the Long Valley caldera. The Long Valley caldera experienced increased seismicity and deformation, consistent with inflation of one or more magmatic sources at depth within the center and along the south edge of the caldera, during the latter half of 1997. The caldera became quiescent again in 1998. A similar episode of rapid deformation was measured by two-color laser EDM in 1989.
Average velocities for the longest running BARD stations during 1997-1998 are shown in Figure 4, with 95% confidence regions. Anomalous site behavior, such as the NUNE and MUSB episodes described above, have been removed from this analysis. The velocities are relative to stable North America, as defined by the five IGS fiducial stations. The Farallon Island site (FARB) off the coast of San Francisco is moving at nearly the rate predicted by the NUVEL-1A Pacific-North America Euler pole. Most of the Sierra Nevada sites (CMBB, QUIN, and ORVB), as well as SUTB in the Great Valley, show little relative motion, indicating that the northern Sierra Nevada-Great Valley is tectonically stable. The motion of these sites relative to North America is consistent with spreading in the Basin and Range Province. The sites near Long Valley caldera (CASA, KRAK, MINS, and JNPR) show anomalously high radial motions away from the center of the caldera, consistent with inflation of magmatic sources at depth. Deformation along the coast is dominated by the active San Andreas fault system, which accommodates about 35 mm/yr of right-lateral shear. These BARD results are being combined with older VLBI and spatially dense Geodolite EDM and campaign GPS measurements currently being collected by the USGS, Stanford University, and UC Berkeley. The combined velocity map will provide significantly improved constraints on three-dimensional locking depth and deep-slip models of strain accumulation, which will be used for seismic hazard assessment along the San Andreas fault system.
Figure: Average velocities of BARD stations in northern California (top) and in the San Francisco Bay area (bottom). The projection is the same as in Figure 1. Velocities are relative to stable North America. The NUVEL-1A predicted Pacific-North America 46 mm/yr relative motion is shown for scale.
One of the goals of the BSL effort is to develop real-time analysis techniques that will enable rapid determinations ( minutes) of motion following major earthquakes to complement seismological information and aid determinations of earthquake location, magnitude, geometry, and strong motion. We currently process data available within 1 hour of measurement from the 15 continuous telemetry BSL stations, and 3 U.S. Coast Guard stations. The data are binned into 1 hour files and processed simultaneously. Figure 5 shows a 5-day interval of hourly position estimates for 12 of the continuous telemetry stations and the 3 U.S. Coast Guard (USCG) stations (CME1, PBL1, PPT1). The scatter of these hourly solutions is much higher than the 24-hour solutions, 10 mm in the horizontal and 30-50 mm in the vertical; however, displacements 3 times these levels should be reliably detected. Several stations exhibit higher scatter than others, particularly TIBB and the USCG stations. The cause of this higher scatter is not well understood yet, but may be due to the higher multipath environment at these sites, which would have a greater effect on short interval data sets. The 2-cm apparent displacement caused by the damage to the CMBB antenna radome is discernable in the east component.
Figure 5: Results of automatically processed hourly position solutions for 5 days in December 1997 at 12 of the continuously telemetered BARD stations and 3 USCG stations, which provide data hourly. Solutions are available within 15 minutes after the end of the hour. Shown are the scatter about the average for the north, east, and vertical (note scale change) components. Horizontal offsets at the several-cm level are detectable at most sites. Note the 2-cm offset at CMBB in the east component on Day 337 that was caused by damage to the antenna radome.