We observed shear wave splitting (birefringence) for two temporary surface deployments of three-component, digital seismographs, that were in place before and after M=6+ earthquakes near the Long Valley caldera, California. In one of these deployments, the data sample precedes the large events of the May 1980 Mammoth Lakes earthquake sequence by 6 months and cover the two most active months of the May 1980 aftershock sequence; one of the stations (WIT) from this deployment was reoccupied with identical instrumentation in 1988. Another deployment preceded the 1986 Chalfant Valley mainshock by 2 days and recorded events for 6 days. The polarization of the faster shear wave changes from N30¿W at Mammoth to due north at Chalfant Valley and in both cases is parallel to the strike of nearby surface faults and to the mean direction of P axes determined from focal mechanism groupings. Observations from nearby stations yield fast directions nearly at 23¿ from each other, and time separations of fast and slow shear waves show considerable station-to-station variation, showing no correlation with earthquake-station distance or earthquake depth.

These observations suggest that the observed anisotropy results primarily from near-station (presumably shallow) effects rather than from widespread aligned microcracks. The records at Long Valley station WIT show a slight variation of average fast shear wave polarization from 1979 to 1980, but with the limited amount of data available, this difference is not statistically significant. Time separation measurements of up to 0.48 s at station WIT yield estimates of 15--30% velocity anisotropy, among the highest reported for local earthquakes. Secular variation in the time separation of fast and slow shear waves between events recorded at station WIT in 1979 and 1980 is swamped by the scatter in the data, if it is present. The five events recorded in 1988 showed similar S wave polarizations and time separations to those in 1980, but there are not enough events for a statistical comparison. Unlocated events at Long Valley station MRS show a slight increase in time separation after the May 1980 M=6.0+ earthquakes, but the increse is not statistically significant. ¿ American Geophysical Union 1990