Abstract A new inversion methodology is developed to determine recurrence rates of earthquake rupture scenarios from observed fault-slip rates in a multisegment, multifault rupture system. The inversion requires specification of the probabilities of ruptures jumping segment and fault boundaries. It solves for nucleation rates on each fault segment, given the slip rates on all segments, such that the number of unknowns equals the number of data points. The inversion, with smoothing constraints on the solution, is overdetermined. Multiple inversions are run to find jumping probabilities that better fit the data. High probability of jumping segment boundaries (≥0.9) results in a characteristic earthquake model, for which the rate of earthquakes is highest for ruptures that span the entire fault length. Lower probability of segment jumping produces higher rates of smaller earthquakes and rates that decay steeply with magnitude. The inversion methodology is applied to the southern San Andreas (SAF)–San Jacinto (SJF)–Garlock fault system to determine the rates of 1047 rupture scenarios (M 6.7–8.2). Predicted rates of ruptures from the inversions are compared to observations from paleoseismology and historic rates of M ≥ 6.6 earthquakes. High probability of jumping segment boundaries (0.9) on most of the SAF results in earthquake rates consistent with observations. The observed paleoevent rate on the SJF requires a lower probability of segment jumping of 0.5. The inversion with these segment-jumping probabilities predicts higher rates of M ≤ 7.2 earthquakes on the SJF than the SAF, consistent with historic observations. The inversions indicate that the lower slip rate observed on the Mill Creek section of the SAF requires stronger segment boundaries at the northern and southern ends of this section, reflecting a restraining bend on the SAF. This inversion methodology incorporates geologic and other information on segmentation and fault connectivity to calculate earthquake recurrence rates for probabilistic seismic hazard analysis.
Arthur D. Frankel (Fri,) studied this question.