Two-dimensional inversion of refraction traveltimes by progressive model development

Abstract

We develop a systematic approach for the analysis of 2-D refraction experiments using traveltimes that allows progressive improvement of velocity structure through a sequence from 1-D models to pseudo-2-D models, and then 2-D models. The approach consists of three steps. First, 1-D velocity models are constructed for each segment of the profile using a genetic algorithm inversion, and then pseudo-2-D models are constructed using a turning point approximation. The purpose of this step is to provide an approximate image of 2-D velocity structure, and to infer the number and general location of layer boundaries. The second step uses 1-D layered modelling, again with a pseudo-2-D conversion, to generate a rough 2-D layered structure. The third step consists of smoothing the pseudo-2-D model in order to create initial models for use in 2-D inversion and the construction of a 2-D model using a Bayesian formulation of non-linear iterative inversion. All the steps exploit traveltime data for the first arrival and do not use any trial-and-error forward modelling. The progressive approach is efficient because the results of each step are employed as the initial model for the next step. The method is applied to real data from an along-strike experiment and also to a related synthetic example so that the quality of the solution can be judged. The results indicate that the method is robust, and this is confirmed by a further synthetic example that represents a survey across a trench and dipping subduction zone. The systematic approach to the inversion of refraction data enables a complex inversion to be undertaken in 1 or 2 days. The sequential approach allows the incorporation of additional information if desired.