Statistical separation strategy to analyze velocity structure obtained by seismic tomography

General solutions of inverse problems often can be obtained by introducing probability distributions to sample the model space. We have developed a simple approach to define an a priori space in a tomographic study and retrieve the velocity/depth posterior distribution by a Monte Carlo method. Utilizing a fitting routine designed for very low statistics to set up and analyze the obtained tomography results, we can statistically separate the velocity/depth model space derived from inverting seismic refraction data. A profile acquired in the Lesser Antilles subduction zone reveals the effectiveness of this approach. Resolution analysis of the structural heterogeneity includes a divergence analysis that can dissect long, wide-angle profiles for deep crust and upper mantle studies. The complete information of any parameterized physical system is contained in the a posteriori distribution. Methods for analyzing and displaying key properties of the a posteriori distributions of highly nonlinear inverse problems are therefore essential in the scope of any interpretation. It is possible to map velocity variations in their extent and structure by measuring the total as well as relative divergence of the velocity structure in the a posteriori space. We have applied the divergence analysis to a part of the transect where a backstop structure has been identified, and the method resolves shallow features and returns information concerning the confidence level of results. Assuming a relationship between forearc and backstop, we can obtain a structural image in accordance with previous interpretations. ©2010 Society of Exploration Geophysicist