Strain and stress tensors along plate boundaries in Iran and southern California are oriented coaxially if they are estimated based on small to moderate earthquakes. If large earthquakes (M>7) are included in the analysis, the estimate of stress directions does not change, but the orientation of the released strain rotates significantly. Thus the direction of the total released strain is different from the stress by about 20¿ in both Iran and southern California, and by about 45¿ in central California. We interpret this observation to indicate that the smaller earthquakes occur on faults favorably oriented for failure in the prevailing stress field, in which they are possibly generated. The larger events rupture near-vertical strike-slip faults that are unfavorably oriented but weak. This weakness causes failure under low ratio of shear to normal stress. This view is supported by the fact that, in the southern California data set, the average angle between the fault planes and the greatest principal stress is about 40¿, whereas it forms an angle of 70¿ with the San Andreas fault (nearly 90¿ in central California).

For estimating the seismic strain release by Kostrov's method along about a 10-km-wide section of the San Andreas fault zone, we used the data set from which Jones determined the stress directions for earthquakes in the range of 2.6<M≤4.3. We estimated the directions of principal strain for Iran based on 111 focal mechanisms and the scalar moment for earthquakes in the range of 4.4<M<8. We inverted regional subsets of these data for volumes with dimensions of a few hundred kilometers for the stress directions using the method of Gephart and Forsyth. The average misfits for these inversions ranged from 2¿ to 5¿. Based on test inversions of synthetic data with known misfits, we found that these misfits can be attributed solely to inaccuracies of less than 10¿ in the fault plane solutions. Therefore the results for Iran are most likely obtained for volumes with uniform stress directions. We propose that stress inversions with misfits smaller than 6¿ can be accepted as meaningful, as an approximate general guideline, because our tests for Iran agree with those for Hawaiian data sets performed by Wyss. ¿ American Geophysical Union 1995