Bayesian sparse wideband source reconstruction of Japanese 2011 earthquake

We consider the sparse inversion of seismic recordings from a Bayesian perspective. We have a prior belief that the spatially distributed seismic source should be sparse in the spatial domain. In a Bayesian framework, we assume a Laplace-like prior for a distributed wideband source and derive the corresponding objective function for minimization. We solve a sequence of convex minimization problems for finding a sparse seismic source representation from an underdetermined system of linear measurement equations using teleseismic P waves recorded by an array of sensors. The root mean square reconstruction error for the source distribution is evaluated through numerical simulations.