Improving spatial resolution of the moment rate function in regions of high slip determined from finite fault inversions

Abstract:

Finite fault inversions attempt to resolve the spatial distribution of the moment rate function per unit area using seismic observations. However, the observational and synthetic limitations that exist due to the station distribution and frequency content of signals that can be modeled imply that various simplifications to the source representation need to be made in order to stabilize the inversions. These simplifications inevitably affect the final results, though it is largely unknown how serious the regularization can be. Here, we explore the effects caused by a pre-assumed slip rate function using synthetic data from the Source Inversion Validation (SIV) produced by a crack-like spontaneous dynamic rupture embedded in a layered, elastic medium. Thus we have exact Green’s functions, which allow us to consider only the effects of varying the slip rate function. We study the effect of three different inverted slip rate functions: i) asymmetric cosine functions; ii) modified Yoffe functions and iii) no negative functions within given time windows. The first approach has been used routinely in finite fault inversions and the second one is characteristic of dynamic simulations. We perform the inversions using the first two cases with a simulated annealing algorithm and attempt to answer whether the ideal synthetic data can differentiate between them. The last one has the inverted slip rate function with the least a priori constraints but uses many more free parameters. It is inverted using the newly developed Projected Landweber (PL) method and we will explore its dependence on the choice of the initial model and the bandwidth of signals.