The temporal behavior of 39 aftershock sequences in southern California, 1933-1988, was modeled by the modified Omori relation. Minimum magnitudes for completeness of each sequence catalog were determined, and the maximum likelihood estimates of the parameters K, p, and c, with the standard errors on each, were determined by the Ogata algorithm. The b value of each sequence was also calculated. Many of the active faults in the region, both strike slip and thrust, were sampled. The p values were graded in terms of the size of the standard error relative to the p value itself. Most of the sequences were modeled well by the Omori relation. Many of the sequences had p values close to the mean of the whole data set, 1.11¿0.25, but values significantly different from the mean, as low as 0.7 and as high as 1.8, exist. No correlation of p with either the b value of the sequence or the mainshock magnitude was found. The results suggest a direct correlation of p values is with surface heat flow, with high values in the Salton Trough (high heatflow) and one low value in the San Bernardino Mountains and on the edge of the Ventura basin (both low heat flow).

The large fraction of the sequences with p values near the mean are at locations where the heat flow is near the regional mean, 74 mW/m2. If the hypothesis that aftershock decay rate is controlled by temperature at depth is valid, the effects of other factors such as heterogeneity of the fault zone properties are superimposed on the background rate detemined by temperature. Surface heat flow is taken as an indicator of crustal temperature at hypocentral depths, but the effects on heat flow of convective heat transport and variations in near-surface thermal conductivity invalidate any simple association of local variations in heat flow with details of the subsurface temperature distribution. The interpretation is that higher temperatures in the aftershock source volume caused shortened stress relaxation times in the fault zone materials, leading to a faster decay rate (higher p value). In 1967, Mogi proposed such a relation on the basis of Japanese data. Seismic velocity distributions published for southern California generally support the hypothesis, with low velocities (higher temperatures) corresponding to high p values, and vice-versa. ¿American Geophysical Union 1991