Linearized versus non-linear inverse methods for seismic localization of underground sources

The problem of localization of underground sources from seismic measurements detected by several geophones located on the ground surface is addressed. Two main approaches to the solution of the problem are considered: a beamforming approach that is derived from the linearized inversion problem, and the Bayes nonlinear inversion method. The travel times used in the beamformer are derived from solving the Eikonal equation. In the linearized inversion method, we assume that the elastic waves are predominantly acoustic waves, and the acoustic approximation is applied. For the nonlinear inverse method, we apply the Bayesian framework where the misfit function is the posterior probability distribution of the model space. The model parameters are the location of the seismic source that we are interested in estimating. The forward problem solver applied for the nonlinear inverse method is a Finite Difference elastic wave-field numerical method. In this paper, the accuracy and performance of the linear beamformer and nonlinear inverse methods to localize a underground seismic source are checked and compared using computer generated synthetic experimental data. © 2013 Acoustical Society of America.