A staged genetic algorithm for tomographic inversion of seismic refraction data

Abstract

A genetic algorithm has been developed to tomographically construct the sub-surface velocity distribution using first-arrival seismic refraction data. The performance of the genetic algorithm is improved by the inclusion of a 'zooming' technique able to perform a staged subspace search in the higher dimensional solution space. This allows the complexity of the model defining the solution to be progressively increased during the inversion and results in an increase in the speed and accuracy with which a solution can be obtained. The procedure has been tested with both synthetic and model data. The method proved able to correctly recover the geometry and velocity structure of a non-planar refractor and the overlying lower velocity layer without the need of any 'a priori' information.