Using resolution-constrained adaptive meshes for traveltime tomography

Abstract

We present a traveltime tomography method which exploits a greedy mesh refinement algorithm to construct adaptive parametrizations with guarantees on model resolution. The refinement strategy is implemented on an unstructured trigonal mesh of cells, each with a constant gradient of slowness. We extend previous work on this theme by adding a new class of refinement operations which produce higher quality meshes with fewer poorly formed triangles. Zones of refinement are chosen based on the properties of the diagonal of the model resolution matrix. The resulting mesh guarantees a well-posed inversion problem by construction. Our algorithm is tested on a synthetic crosswell tomography problem with spatially variable angular aperture and a region of negligible ray coverage; the results demonstrate the advantages of adaptive parametrization in comparison to inversion using a regular grid and a single regularization parameter. We also process a large multiwell seismic dataset with irregular ray coverage collected at a shallow site with groundwater contamination. © 2006 Geophysical Press Ltd.