Rate-and-State Southern California Earthquake Forecasts: Resolving Stress Singularities
Tuesday, 16 December 2014
In previous studies, we pseudo-prospectively evaluated time-dependent Coulomb stress earthquake forecasts, based on rate-and-state friction (Toda and Enescu, 2011 and Dieterich, 1996), against an ETAS null hypothesis (Zhuang et al., 2002). At the 95% confidence interval, we found that the stress-based forecast failed to outperform the ETAS forecast during the first eight weeks following the 10/16/1999 Hector Mine earthquake, in both earthquake number and spatial distribution. The rate-and-state forecast was most effective in forecasting far-field events (earthquakes occurring at least 50km away from modeled active faults). Near active faults, where most aftershocks occurred, stress singularities arising from modeled fault section boundaries obscured the Coulomb stress field. In addition to yielding physically unrealistic stress quantities, the stress singularities arising from the slip model often failed to indicate potential fault asperity locations inferred from aftershock distributions. Here, we test the effects of these stress singularities on the rate-and-state forecast’s effectiveness, as well as mitigate stress uncertainties near active faults. We decrease the area significantly impacted by stress singularities by increasing the number of fault patches and introducing tapered slip at fault section boundaries, representing displacement as a high-resolution step function. Using recent seismicity distributions to relocate fault asperities, we also invert seismicity for a fault displacement model with higher resolution than the original slip distribution, where areas of positive static Coulomb stress change coincide with earthquake locations.