An adjoint data assimilation method for optimizing frictional parameters on the afterslip area

Afterslip sometimes triggers subsequent earthquakes within a timescale of days to several years. Thus, it may be possible to predict the occurrence of such a triggered earthquake by simulating the spatio-temporal evolution of afterslip with estimated frictional parameters. To demonstrate the feasibility of this idea, we consider a plate interface model where afterslip propagates between two asperities following a rate-and-state friction law, and we adopt an adjoint data assimilation method to optimize frictional parameters. Synthetic observation data are sampled as the slip velocities on the plate interface during 20 days. It is found that: (1) all frictional parameters are optimized if the data sets consists not only of the early phase of afterslip or acceleration, but also of the decaying phase or deceleration; and (2) the prediction of the timing of the triggered earthquake is improved by using adjusted frictional parameters.