Ensemble data assimilation to forecast slow-slip events and earthquakes: a synthetic laboratory test

Abstract:

Our ability to forecast earthquakes and slow slip events is hampered by limited information of the absolute state of stress and strength of faults and their governing parameters. Ensemble data assimilation provides a means to estimate these variables by combining physics-based models and observations taking into account their uncertainties. This study aims to improve the estimates of the fault state with ensemble data assimilation to forecast both earthquakes lasting for seconds and slow slip events lasting for months. Our framework consists of an Ensemble Kalman Filter implemented in the Parallel Data Assimilation Framework, which is connected with a 1D forward model based on a C++ numerical library GARNET. The setup represents a meter-scale straight-fault governed by rate-and-state friction at a boundary of the model. We assimilate shear-stress and slip-rate observations and their uncertainties acquired at a small distance in the homogeneous elastic medium. A perfect-model test shows that ensemble data assimilation can estimate shear stresses and slip rates acting on the fault and can thus be used for forecasting events on time scales of seconds. The results suggest that the observations provide significant improvements in the timing of forecasted events over time. The results also illustrate the impact of assimilating observations during different phases of the seismic cycle. In future work, we will evaluate the performance of ensemble data assimilation in a laboratory setting with non-perfect physics and real observations of shear-strain gauges and piezoelectric transducers.