Interparameter tradeoffs quantification and reduction in isotropic-elastic FWI: synthetic experiments and real dataset application

Wenyong Pan, Yu Geng and Kris Innanen


This report aims at evaluating and reducing the interparameter tradeoffs in isotropic-elastic FWI with multiparameter Hessian matrix-vector products. It is revealed that prod-ucts of multiparameter Hessian off-diagonal blocks with model perturbation vectors, de-fined as interparameter contamination kernels, mainly account for interparameter tradeoff. The multiparameter Hessian is applied to various vectors designed to provide information regarding the strengths and characteristics of interparameter contaminations locally or within the whole volume. Based on these findings, a novel strategy is developed to mitigate the influence of interparameter tradeoffs with approximate contamination kernels. Furthermore, I propose to quantify resolution of the inverted models with approximate eigenvalue volume and extended multiparameter point spread functions (EMPSFs) by preconditioned conjugate-gradient algorithm. Finally, the proposed inversion strategies are applied to invert isotropic-elastic parameters with synthetic data and Hussar practical seismic dataset. Resolution of the inverted models are also evaluated.

View full article as PDF (25.01 Mb)