Joint Inversion of Receiver Function and Surface Wave Dispersion by Hamiltonian Monte Carlo Sampling

Abstract

Abstract We have proposed a new probabilistic inversion method to perform the joint inversion of receiver function and surface wave dispersion data. In this method, we apply the Hamiltonian dynamics in the Bayesian framework to efficiently sample the posterior probability distribution of this joint inverse problem. This method will lead to nearly 100% acceptance of each sample in theory. Semianalytical derivatives of both the datasets to the model parameters (including elastic parameters, density, and the thickness of each layer) are used to speed up this algorithm. Finally, we apply our method to both synthetic data and real data. The result shows that the velocity model can be recovered well within a much smaller number of samplings than the traditional Markov chain Monte Carlo method.