What can we learn from 20 years of interseismic GPS measurements across strike-slip faults?

• I compare GPS velocities across 13 segments of 8 major strike-slip faults.
• Fault parameters derived from 2D elastic half-space models are similar.
• No significant strain variations are detected through the interseismic period.
• Dislocation at depth tends to be located on the concave side of the fault trace.
• GPS derived fault slip rates are consistent with geomorphologically derived ones.

I use GPS interseismic velocities and classic 2D elastic half-space models with a screw dislocation to estimate the long-term fault slip rate, locking depth, and the offset between the surface fault trace and the location of the dislocation below the seismogenic zone for 13 segments along 8 major strike-slip faults. Using deduced strike-slip rates and the position of the dislocation to normalize the interseismic velocities to facilitate comparison of spatial patterns of deformation, I show that no substantial differences can be detected, ruling out a large asymmetry in interseismic velocities across the 8 faults used in this study. Only the Carrizo Plain segment of the San Andreas Fault shows a significant asymmetry that cannot be explained by shifting the position of the dislocation at depth relative to the fault trace. However, the resulting perturbation is less than 10% of total strike-slip rate. Fault traces are usually curved, defining a concave side. When the dislocation at depth is significantly offset from the fault trace, the shift is always toward the block on the concave side of the fault trace. This suggests that the fault zone in the lower crust may develop a simpler geometry more consistent with relative motion across the fault than its upper seismogenic part constrained by the structural complexity of the brittle crust. Since the faults used in this study are at different times in their interseismic period, comparing the interseismic velocity fields across them allows identification of possible variations of the interseismic velocities with time. When normalized by slip rate and dislocation location, all the faults show the same interseismic strain with no significant differences between deduced locking depths. These comparisons suggest that if temporal variations occur as suggested by some dynamic earthquake cycle models, they are small and below the accuracy of the available geodetic measurements.