A non-linear optimization technique for the inversion of long range refraction profiles

Abstract

The problem of interpreting long range crustal refraction profiles is usually addressed by a conventional estimation of the velocity-depth model from the travel time data followed by iterative forward modelling. In principle, the objective of this technique is to find the crustal model which gives the best agreement between the calculated and observed travel time data. Recent advances in the design of ocean bottom seismometers (OBS) have facilitated the use of a large number of closely spaced OBS along such profiles. This has increased the resolution of the technique and the constraints put on the calculated models but at the same time increased considerably the complexity of the interpretation process and the computation time. In order to reduce the tedious iterative interpretation process to manageable proportions, we propose to apply a development of the coherency inversion method, suggested for use in exploration reflection interpretation by Landa et al. (1987) and Koren et al. (1987), to the evaluation of long range crustal profiling data. This development is designed for the estimation of the velocity sequences and the geometry of the interfaces using selected record sections for a given profile. We are concerned with two-dimensional heterogeneous media and data which contain arrivals of reflected and refracted waves. The method is based on an iterative algorithm producing a model which maximises some measure of coherency; the latter is computed within a time gate for a record section along travel times generated by tracing rays through the model. This algorithm has the advantage that event picking on the record sections is avoided and it is not based on curve fitting of hyperbolic approximations of the arrival times.