Study of crosstalk reduction in multiparameter acoustic FWI

Qi Hu, Kristopher A. Innanen

Crosstalk is the phenomenon in which data signatures of different physical properties are confused in full-waveform inversion (FWI). The challenges associated with crosstalk are a major obstacle to the effective implementation of multiparameter FWI. In this study, we focus on acoustic media with variable density, and discuss the reduction of parameter crosstalk from three aspects: optimization method, acquisition geometry, and model parameterization. Optimization methods such as steepest-descent, conjugate gradient and quasi-Newton methods are compared in terms of the inversion quality and computational cost. For geometry, we show how sub-surface sources and receivers provided by vertical seismic profile (VSP) and crosswell help mitigate the crosstalk from surface seismic. Scattering patterns of four parameterizations: velocity-density, impedance-density, impedance-velocity, and bulk modulus-velocity, are illustrated and employed to study their capabilities relative to crosstalk.