In collaboration with geodesists at JPL, Scripps, UCLA, and USGS, we have used EDM (1971-92), VLBI (1984-91), and GPS (1986-1997) measurements to infer the recent inter-seismic, co-seismic, and post-seismic motion of sites in southern California. During the past year we have concentrated our analysis on the GPS and VLBI data which might best address the nature and significance of velocity field changes subsequent to the Landers earthquake and those bearing on the ability of continuous GPS observations to discern small gradients in strain over the Los Angeles basin.
In a comparison of pre- and post-Landers velocities of four sites currently occupied by continuous GPS stations, Bock et al. [1997] found changes at the three sites closest to Landers that were significant at 40-95% confidence using a crude error model for the pre-Landers estimates [Feigl et al., 1993] and a somewhat more rigorous model for the post-Landers estimates from continuous data. The size of the changes ranged from 6 mm/yr at Pinyon to 4 mm/yr at JPL and Goldstone. Expansion of this study to include more sites is limited primarily by the accuracy of pre-Landers surveys in the affected region. Figure 1 shows our current estimates for pre- and post-Landers velocities of 17 sites that experienced co-seismic offsets as large as those at JPL and Goldstone. Neither the pre- nor the post-Landers solution has been thoroughly edited for outliers, so the velocities should not be taken too seriously at this point. Region-wide, reference-frame errors due to a single poorly analyzed survey could be present. However, the consistency of the changes for SIO (SOLJ/SIO1/SIO3), Monument Peak, Ocotillo, and L589 are noteworthy since the pre-Landers velocities for two (Monument Peak and Ocotillo), but not all four of these sites have VLBI data as a strong component. The 95% confidence uncertainties for the best-determined pre-Landers velocities are currently about 5 mm/yr. These stations were observed in the June, 1986, NGS/JPL survey (TREX0). We now have additional 1986 data from NGS that may allow comparable uncertainties to be obtained for other sites in the region. Use of long arcs in the orbital estimation may also reduce the uncertainties. An important issue is how the reference frame is to be defined consistently for the pre- and post-Landers period. To avoid distortions that might mimic geophysical signals, we have defined the frames through generalized constraints, minimizing the average departure of the horizontal velocities of a pre-defined set of stations from an external frame (e.g., ITRF94) [Dong et al., 1997]. Applying these constraints separately to the pre- and post-Landers data, we assure that the constraints themselves allow only a translation, rotation, and scale of the two frames.
We have been examining in detail the velocity field from continuous observations in the Los Angeles region. Our primary aim in this study has been to assess the reliability with which the secular velocity field can be deduced. Our analysis has been carried out in two parts: (a) GLOBK combinations of the SIO daily h-files to directly estimate the velocity; and (b) time series analyses of the daily h-files to determine the character of site position estimates. In both types of analyses, we use unconstrained solutions and apply generalized constraints to rotate and translate the solutions into the desired reference frame. We have been allowing explicit translation of the coordinate system and stochastic height variations (equivalent to ~30 cm daily changes) in the GLOBK analysis. To assess the quality of the velocity solution, and to minimize the effects of annual signals that are seen in time series for some sites, we estimate velocities using only data from the same time each year (essentially campaign style processing of continuous data). By changing the time of the year when experiments are used we can evaluate the stability of the velocity field estimates obtained from analyses that use non-overlapping data. Seasonal dependence on the quality of the GPS data can also be assessed by this means. We show in Figure 2 an example of the results obtained with this approach where the data sets chosen were the first 9 days of each month thus yielding 12 separate estimates of the velocity field. The h-files used here cover the interval from June, 1992 to November, 1997. We allowed 60 days of stochastic position variations for sites within 500 km of the Landers earthquake and 150 km of the Northridge earthquake. The error ellipses on this plot are generated from the RMS scatter of the velocity estimates from the 12 solutions used in this analysis. They vary from ±0.8 mm/yr at MATH for both the north and east components of velocity to ±3.4 mm/yr and ±2.9 mm/yr for the north and components of LONG. The orientation of the error ellipses indicates if the north and east velocity variations are correlated. The instability of some of the sites is evident in Figure 2. For example, JPL has much more variability in its east velocity estimates than in its north component. The CalState Northridge site (CNS1) has velocity variations negatively correlated between north and east.
As an example of time series for the data that went into generating the velocity estimates we show the north components of the daily time series for selected sites in the Los Angeles region (Figure 3). The character of the data from these sites can be clearly seen in the figure with some sites showing clear seasonal type variations (JPL and PVEP), and others showing relatively small systematic variations (e.g. LEEP which has 0.8 mm of white noise and an exponentially correlated component with variance [1.6 mm]2 and correlation time of 1.7 days). Detailed examination of the time series shows spatially correlated (over distances of 50 km), and few-day to week period variations which we believe are due to atmospheric delay modeling errors. For impact on the velocity field determination, the longer period variations whose origins are not so clear seem more important at this time.
References
Bock, Y., W. Wdowinski, P. Fang, J. Zhang, S. Williams, H. Johnson, J. Behr, J. Genrich, J. Dean, M. van Domselaar, , D. Agnew, F. Wyatt, K. Stark, B. Oral, K. Hudnut, R. King, T. Herring, S. Dinardo, W. Young, D. Jackson, W. Gurtner, Southern California Permanent GPS Geodetic Array: Continuous measurements of regional crustal deformation, J. Geophys. Res., 102, 18,013-18,033, 1997.
Dong, D., T. A. Herring, and R. W. King, Estimating regional deformation from a combination of space and terrestrial geodetic data, in press, J. Geodesy, 1997
Feigl., K. L, D. C. Agnew, Y. Bock, D. Dong, A. Donnellan, B. H. Hager, T. A. Herring, D. D. Jackson, T. H. Jordan, R. W. King, S. Larsen, K. M. Larson, M. H. Murray, Z. Shen, and F. H. Webb, Measurement of the velocity field of central and southern California, 1984Ð1992, J. Geophys. Res., 98, 21677-21712, 1993.
Captions
Figure 1. Pre- (black) and post (red) Landers velocities for selected sites in the southern San Andreas region, shown relative to the Pacific plate. Ellipses show regions of 95% confidence after scaling the formal uncertainties by a factor of two.
Figure 2. Velocity field in the Los Angeles region relative to the Pacific plate determined using the method discussed in the text. The error ellipses are 95% confidence. The sigmas and correlations between north and east components are computed from the scatter of 12 solutions averaged to generate the velocity field. Differences in the size of the error ellipses reflect to a large extent the magnitudes of the non-linear variations in the time series of individual sites. A complete velocity field solution generates a nearly identical field but the error ellipses are much smaller.
Figure 3. Time series plots for the north component of position for selected sites in the Los Angeles region. The sites are arranged approximately north to south (see Figure 2 for site locations). No error bars are shown but they are typically 1 to 2 mm for each daily value. The rms scatter of estimates varies from 1.5 mm for LEEP to 2.4 mm for JPLM. The velocity of USC1 has been removed from all time series. The offset due to the antenna change at CHIL was accounted for in the velocity field determination by estimating an offset to the position at this time. Although the antenna types changed at each of these sites were the same, the offsets do not agree, even in sign.
