Velocity field imaging with an anomaly recovery algorithm, incorporating later arrivals

Abstract

We have developed a novel velocity‐field recovery algorithm incorporating later arrivals which is aimed at overcoming several drawbacks encountered with standard tomographic inversion schemes. The anomaly recovery algorithm (ARA) places a series of anomalies, each fully described by only a few parameters, into a given velocity‐field estimate, varying the parameters to minimize the residual errors in first‐arrival times and in reflection, refraction, and diffraction times, which are obtained by manual or automatic picks from seismograms. We successfully recover anomalies in synthetic variable velocity fields where standard conjugate gradient‐based tomographic inversion schemes fail, find the ARA to be very robust against errors in traveltimes of up to 8%, and obtain results that are economically more meaningful. A comparison of results of the ARA applied to a mineral field crosshole data set with that of a standard conjugate gradient least‐squares (CGLS) inversion scheme indicates the ARA is a viable option for real‐world applications. It recovers the major velocity features found in subsequent drilling. An examination of the solution space exhibits a generally smooth topography with few significant minima interspersed by a larger number of minor local minima, suggesting the applicability of several standard nonlinear inversion schemes. The central idea underlying the ARA, to represent the velocity‐field inversion problem as low dimensional rather than high dimensional, makes the method generic, highly flexible, and relatively easy to analyze in terms of stability and local minima.