Abstract
Several retrospective analyses have proposed that significant increases in moment release occurred prior to many large earthquakes of recent times. However, the finding of accelerating moment release (AMR) strongly depends on the choice of three parameters: (1) magnitude range, (2) area being considered surrounding the events and (3) the time period prior to the large earthquakes. Consequently, the AMR analysis has been criticized as being a posteriori data-fitting exercise with no new predictive power. As AMR has been hypothesized to relate to changes in the state of stress around the eventual epicentre, we compare here AMR results to models of stress accumulation in California. Instead of assuming a complete stress drop on all surrounding fault segments implied by a back-slip stress lobe method, we consider that stress evolves dynamically, punctuated by the occurrence of earthquakes, and governed by the elastic and viscous properties of the lithosphere. We study the seismicity of southern California and extract events for AMR calculations following the systematic approach employed in previous studies. We present several sensitivity tests of the method, as well as grid-search analyses over the region between 1955 and 2005 using fixed magnitude range, radius of the search area and period of time. The results are compared to the occurrence of large events and to maps of Coulomb stress changes. The Coulomb stress maps are compiled using the coseismic stress from all M > 7.0 earthquakes since 1812, their subsequent post-seismic relaxation, and the interseismic strain accumulation. We find no convincing correlation of seismicity rate changes in recent decades with areas of high stress that would support the AMR hypothesis. Furthermore, this indicates limited utility for practical earthquake hazard analysis in southern California, and possibly other regions.