PARAMETERIZATION STUDY FOR ACOUSTIC AND ELASTIC RAY PLUS BORN INVERSION

We present a linearized 2-D elastic and acoustic multi-parameter inver sion of surface seismic data. We solve the forward problem by a combin ation of ray theory and Born approximation. The reference model is 2-D , and a 2.5-D approximation is made in order to model a point source p ropagation. The multi-parameter inversion method is based on the minim ization of a weighted misfit-function. The weighting function is estim ated from parameters associated with ray and paraxial ray theory, and is introduced in order to diagonalize approximately the Hessian matrix of the inversion. This form of the Hessian allows analytical study of its eigenvalues and eigenvectors (thus of inversion conditioning). We discuss the choice of the parameterization from the point of view of inversion. We conclude that the best parameterization with respect to inversion is not necessarily a good one for a geophysical interpretati on. We propose to use the conditioning number of the inversion as a Qu ality Control Factor to guide a posteriori the choice of parameterizat ion of the perturbation model.