Non-linear inversion travel-time tomography: imaging high-contrast inhomogeneities

Abstract

An iterative and constrained non-linear travel-time tomographic scheme is developed for imaging high-contrast inhomogeneities. The technique has been successfully tested using synthetic data. We have investigated the influence upon the image quality of the random errors in travel times for a cross-hole survey configuration. In the forward problem, the ray path is traced by finding the minimal travel time of each shot-receiver pair in the updated model. For non-linear inversion, we approximate the velocity distribution with a specific first-order surface function so that the analytical derivatives with respect to each grid-point value of the velocity variation can be calculated. A constrained inversion from an arbitrary starting model is implemented in an iterative fashion. Our results indicate that with 180° symmetric angular coverage and a cell size comparable to the minimum dimension of the target, it is possible to get a resolvable image of a high-contrast target only in the direction parallel to the boreholes, even without a priori information. Constraints from knowledge about the velocity distribution along the boreholes or along the surface are very important in overcoming the poor directional resolution. The combination of cross-hole data with VSP data substantially improves travel-time tomography as a means of imaging high-contrast inhomogeneities. The results also show that if the relative travel-time random errors in data exceed an estimated expected value of about 2.5% the fine detail in the shape of a reconstructed target could be severely distorted.