The Effects of Coulomb Stress Change on Southern California Earthquake Forecasting
- Author(s): Strader, Anne Elizabeth
- Advisor(s): Jackson, David D
- et al.
I investigate how inclusion of static Coulomb stress changes, caused by tectonic loading and previous seismicity, contributes to the effectiveness and reliability of prospective earthquake forecasts. Several studies have shown that positive static Coulomb stress changes are associated with increased seismicity, relative to stress shadows. However, it is difficult to avoid bias when the learning and testing intervals are chosen retrospectively. I hypothesize that earthquake forecasts based on static Coulomb stress fields may improve upon existing earthquake forecasts based on historical seismicity. Within southern California, I have confirmed the aforementioned relationship between earthquake location and Coulomb stress change, but found no identifiable triggering threshold based on static Coulomb stress history at individual earthquake locations. I have also converted static Coulomb stress changes into spatially-varying earthquake rates by optimizing an index function and calculating probabilities of cells containing at least one earthquake based on Coulomb stress ranges. Inclusion of Coulomb stress effects gives an improvement in earthquake forecasts that is significant with 95% confidence, compared to smoothed seismicity null forecasts. Because of large uncertainties in Coulomb stress calculations near faults (and aftershock distributions), I combine static Coulomb stress and smoothed seismicity into a hybrid earthquake forecast. Evaluating such forecasts against those in which only Coulomb stress or smoothed seismicity determines earthquake rates indicates that Coulomb stress is more effective in the far field, whereas statistical seismology outperforms Coulomb stress near faults. Additionally, I test effects of receiver plane orientation, stress type (normal and shear components), and declustering receiver earthquakes. While static Coulomb stress shows significant potential in a prospective earthquake forecast, simplifying assumptions compromise its effectiveness. For example, we assume that crustal material within the study region is isotropic and homogeneous and purely elastic, and that pore fluid pressure variations do not significantly affect the static Coulomb stress field. Such assumptions require further research in order to detect direct earthquake triggering mechanisms.