This grant provided funds to measure and model the response of the San Jacinto Fault to the stress increment from the 1992 Landers earthquake. Unfortunately, we cannot report as much progress as we would like to on this research. We felt it was more appropriate for us to concentrate on our part of an ``infrastructure'' task (the SCEC Geodetic Velocity Model), namely archiving GPS data. This turned out to require enough effort so that we were not able to pursue our non-infrastructure research fully. As explained in the accompanying proposal, we therefore seek to roll over the funds we received for this work so as to pursue it in the coming year--a strategy made possible (we hope) by the flexible nature of SCEC funding.
We can, however, report on work we did undertake that is relevant both to the Geodetic Velocity Model and to the response of faults to large earthquakes such as the Landers event. This is a comparison of strain-rates measured over different time spans, including this earthquake. Most of the data on deformation along the San Jacinto Fault (and elsewhere in Southern California) come from EDM measurements made by the Crustal Strain Group of the USGS from 1972 through 1987, roughly annually. In the late 1980s GPS began to overtake EDM as the measurement technique of choice, and the USGS stopped making these EDM measurements. The SCEC velocity model combines the EDM and GPS datasets (which barely overlap in time), implicitly assuming the strain rates to be constant.
We were funded by NSF to remeasure (with GPS) a part of the EDM network, namely that section across the San Jacinto Fault near Borrego Springs, with the goal of looking for possible temporal changes in strain-rate that might reflect stress relaxation after the 1968 Borrego Mountain earthquake. Remeasuring this part of the network requires helicopters; given the minimum charge for doing so, we felt that it made sense to extend these measurements to the northeast, across the Salton Sea and the San Andreas fault. Such a wider set of measurements would allow us to look for possible changes in the rate of deformation in this area (most notably, possible rate changes on this part of the San Andreas fault from the Landers sequence), at a relatively low marginal cost. The sites visited are shown in Figure 1; we stress that the cost of visiting the additional survey sites (beyond those funded by the NSF grant) was not paid for by SCEC, but used funds from other relevant grants. Our SCEC support was however needed to cover the cost of analyzing the GPS data (which have of course been archived at the SCEC Data Center).
Figure 1 also shows our estimates of the principal strain rates, found for each triangle of lines, assuming that strain has accumulated steadily with time. A better picture would of course come from combining this with the GPS data in the SCEC geodetic velocity model; we show these rates partly to illustrate how spatially variable the strains are--and also how unusual in the southeast corner, where we seem to be seeing significant dilatation, hardly the expected pattern for strain in a shear zone. Most of the rest of the observations imply pure shear along strike-slip faults, though with a rotation in the direction of shear from east to west, and higher strain rates over the San Jacinto fault than over the San Andreas.
Figure 2 shows the time histories of some of the line lengths. (Note that to compare the GPS and EDM distances we have included the correction of Savage et al. (1996).) On almost all the lines measured, the length change with time can be fit by a constant rate (with offsets at the times of significant earthquakes). For the lines shown we do not need to include any offset from the Landers event; none would be expected from the usual model of dislocations in a half-space, but this shows that this part of the San Andreas has not responded in any unusual way to this stress change. In a few cases the EDM data are not fit by such a line unless a particular EDM measurement is discarded (which is not unusual for this dataset); but once such points (shown by stars in Figure 2) are dropped from the fit, the combined EDM/GPS measurements are fit by a constant rate--with one exception.
This exception is the line between Alamo and Old Beach, at the southeastern edge of the network. Figure 2 shows this time series along with a linear fit, which in this case is to the EDM data only; this fit includes an offset at the time of the Westmorland earthquake of 1981. Looking at all the other series that include these two stations (all of which are shown in Figure 2) it is clear that the motion must be at Old Beach, as the Alamo-Soda line (at roughly the same azimuth) shows no anomaly. Once we found this discrepancy, we made an additional set of measurements of this line and of the ties from Old Beach to its two reference marks. We obtained the same value for the line length (included in Figure 2) and values for the distances to the reference marks that are within 5 mm of the values measured by NGS (probably in the 1940's). Thus, if there is local instability, it affects more than just the one mark, and requires all the marks to move uphill. As noted above, this is also where we see areal dilatation; we computed the dilatational results after excluding the GPS data from Old Beach.
At this point this result remains a puzzle, not least because the station involved is outside the deforming zone indicated by seismicity. Because the line involved is close to one edge of the EDM net (there are lines farther east, but they are difficult to resurvey) this result will not affect much of the SCEC geodetic velocity model; otherwise, it appears to be safe to combine EDM data from 1972-87 with GPS data from 1986-97. We should however note that the site Salton shows a significant displacement at the time of the the 1987 Superstition Hills event; we assume this to be a consequence of triggered shallow slip on the San Andreas fault nearby (which was observed to be 3 mm, from creepmeter measurements just to the North).
