From physical modeling to seismic precursors

In classical models, an earthquake is a sudden slip event taking place along a smmoth interface as a consequence of stress accumulation. The strain is accommodated as long as the failure stress is reached with a complex spatial and temporal evolution mediated by fault rheology and pore-fluid pressure which can now be reconstructed in detail thanks to current computational and observational power. However, while this approach is ale to provide more and more advanced knowledge of the different physical contributions to fault dynamics, it cannot be useful for predicting future seismic activity, except for peculiar cases such as induced events. In fact, the prediction time horizon of the coseismic dynamic is noise-embedded because of the chaotic behavior of the system and poorly constrained physical parameters. Since the number of degrees of freedom is extremely large in fault systems, it is unlikely data will ever be enough for the prediction of the incoming evolution of seismicity; therefore, we need to set up models to capture the dependence on observables with effective prediction power. A possible answer consists in considering collective properties of seismicity within extended crustal volumes associated with clustering and memory effects.