Analysis and Modeling of Earthquake Interaction

Earthquake occurrence is a random process well represented by stochastic point process models such as the Epidemic-Type Aftershock Sequence (ETAS) model rather than deterministic ones. The robustness of some scientific conclusions about earthquakes could be weakened by the randomness together with the incompleteness and inaccuracy of earthquake data.;Here I investigated how to model earthquake interactions accurately, analyze characteristics of earthquake triggering and evaluate earthquake potential in California. I address four fundamental questions: (1) How to estimate accurately standard errors of parameter estimates in the ETAS model? (2) How do incomplete data bias earthquake interaction modeling? (3) Are spontaneous earthquakes stationary in California? (4) How to estimate better earthquake potential in California based on different choices?;First, I constructed a unified earthquake catalog covering California by combining different previous earthquake catalogs. I converted all non-moment magnitudes to moment magnitudes using orthogonal regression, and I estimated magnitude uncertainties from pair comparison. Unknown focal mechanisms are inferred from existing ones. I also used extended sources rather than point sources to represent large earthquakes. Second, I estimated standard errors of parameter estimates in the ETAS model by simulating the earthquake process repeatedly, estimating the parameters corresponding to each simulation and comparing the estimates with the initial value. Third, I found that characteristics of earthquake triggering, especially the proportion of spontaneous earthquakes, are biased by ignoring earthquakes outside the target space-time-magnitude window. Moreover, I concluded that there is no strong evidence to support that spontaneous earthquakes are not stationary and remote large earthquakes have significantly influenced local seismicity in California. Finally, I found that using spontaneous earthquakes only and using extended sources to represent large earthquakes could significantly improve the estimation of 5-year and 10-year earthquake potential in California. However, neither the longer earthquake catalog nor the anisotropic smoothing kernel helps to get better forecasting results.