Tomographic velocity model inversion using kinematic wavefield attributes

Abstract

Summary A tomographic inversion scheme is introduced, that makes use of seismic kinematic waveeld attributes (NIP-wave curvatures and normal ray emergence angles) at given measurement surface positions and zero-oset traveltimes to construct a velocity macro-model. These kinematic waveeld attributes can be extracted from multi-coverage prestack data in a data-driven way by application of the Common-ReectionSurface stack. The model to be inverted for is made up of two parts: the smooth velocity model itself, given by two-dimensional B-splines on a rectangular grid, and the starting positions and starting directions at depth (i.e. local reector dips) of the normal rays corresponding to the input data. The non-linear inversion problem is solved by iterative leastsquares solutions to the linearized problem. Forward modeling and calculation of Fr echet derivatives for the tomographic matrix is done by dynamic ray tracing and ray perturbation theory. Results of a rst 2D numerical test of the inversion algorithm are presented, where the input data have been produced directly by dynamic ray tracing.