Quantifying uncertainty in earthquake rupture models

Using dynamic and kinematic models, we analyze the ability of GPS and strong-motion data to recover the rupture history of earthquakes. By analyzing the near-source ground-motion generated by earthquake ruptures through barriers and asperities, we determine that both the prestress and yield stress of a frictional inhomogeneity can be recovered. In addition, we find that models with constraints on rupture velocity have less ground motion than constraint-free, spontaneous dynamic models with equivalent stress drops. This suggests that kinematic models with such constraints overestimate the actual stress heterogeneity of earthquakes.; We use GPS data from the well-recorded 2004 Mw6.0 Parkfield Earthquake to further probe uncertainties in kinematic models. We find that the inversion for this data set is poorly resolved at depth and near the edges of the fault. In such an underdetermined inversion, it is possible to obtain spurious structure in poorly resolved areas. We demonstrate that a nonuniform grid with grid spacing matching the local resolution length on the fault outperforms small uniform grids, which generate spurious structure in poorly resolved regions, and large uniform grids, which lose recoverable information in well-resolved areas of the fault. The nonuniform grid correctly averages out large-scale structure in poorly resolved areas while recovering small-scale structure near the surface.; In addition to probing model uncertainties in earthquake source models, we also examine the effect of model uncertainty in Probabilistic Seismic Hazard Analysis (PSHA). While methods for incorporating parameter uncertainty of a particular model in PSHA are well-understood, methods for incorporating model uncertainty are more difficult to implement due to the high degree of dependence between different earthquake-recurrence models. We show that the method used by the 2002 Working Group on California Earthquake Probabilities (WGCEP-2002) to combine the probability distributions given by multiple earthquake recurrence models has several adverse effects on their result. In particular, WGCEP-2002 uses a linear combination of the models which ignores model dependence and leads to large uncertainty in the final hazard estimate. In addition to analyzing current statistical problems, we present alternative methods for rigorously incorporating model uncertainty into PSHA.