California aftershock hazard forecasts.

1) quantile points at the selected times, S, corresponding to the expected probability and ± 1 SD (Table 1). The standard error in P rapidly decreases with increasing time after the mainshock due to the inclusion of current data. For example, at S = 1 day after the mainshock, the ± 1 SD range about the generic 1-day interval probability (0.052) is 0.034 to 0.075 (Table 1). Rydelek suggests estimating parameters from subsets ofthe a priori data corresponding to particular tectonic regions. While this approach has potential merit, it was not very successful for the California data. Parameter estimates for subsets of the data corresponding to the strike-slip regime of central Califomia, the compressional regime of southwestern California and the strike-slip and extensional regime of eastern California do not differ significantly from each other, with one exception. The a value for sequences in eastem California is significantly higher than in central or southwestern California, which indicates a higher productivity of aftershocks there. In future applications of our method to other areas, however, a search for regional or tectonic subsets of earthquake sequences that significantly differ in some parameter values could provide an improvement over the single generic model approach. PAUL A. REASENBERG MARK V. MATTHEWS U.S. Geological Survey, 345 Middlefield Road, Menlo Park, CA 94025