We propose a new method for defining segmentation of plate boundaries and faults, based on the directions of the stress tensor. Estimates for these directions are obtained by minimizing the average misfit between the theoretical and observed slip directions on fault planes of earthquake focal mechanisms. The misfit, f, for an individual earthquake is the parameter we use for defining the segmentation of plate boundaries. We hypothesize that the stress directions along plate boundaries, and faults, are uniform within segments, but different from other segments. If this is true, a cumulative plot of f as a function of space along strike will show constant, but different, slopes for each segment. The significance of the difference between segment-slopes can be estimated by the standard deviate z-test. Applying this method to the San Andreas fault from the Carrizo plains to its southern end, we identify quantitatively the same four boundaries between segments as proposed based on non-quantitative tectonic considerations, plus one additional segment boundary.

We interpret the relatively uniform, but segmented, distribution of stress directions as due to the changing strike, and possibly changing fault surface properties. Whether great earthquake ruptures, or their major asperities, may terminate at segment-boundaries, should be determined along faults that recently generated large earthquakes. This method of defining fault segmentation also allows identification of volumes with uniform stress directions, suitable for inversion for stress orientations, with a minimum of computing time. And finally the method affords an alternative estimate of the significance of differences in stress directions. ¿ American Geophysical Union 1995