Spatial and temporal forecasting of large earthquakes in a spring-block model of a fault

We study a recently proposed statistical physics model of earthquake dynamics that includes stress relaxation in the plates as a fundamental ingredient. The model is known to reproduce many realistic features of seismic phenomena, such as: the Gutenberg-Richter law for the event size distribution, the Omori law for aftershocks and an overall velocity-weakening dependence of the average friction force. Here, we analyse the dynamics of the model in detail, in order to investigate to what extent the occurrence of large events in the model can be anticipated. We systematically find that large events occur in fault patches where strain accumulation has exceeded some threshold value. The spatial extent of these patches (which correlate with the magnitude of forthcoming events) can be calculated if the strain state of the system is supposed to be known. In addition, we find that some large events are preceded by well-defined precursor activity. This allows, in a fraction of cases, to complement the forecast of magnitude and spatial location, with a sensible prediction of time of occurrence. Although our work is exclusively limited to the numerical model analysed, we argue that it gives new breath to earthquake forecast techniques that combine the historical analysis of seismic activity with a search of appropriate precursor activity.