Summary

Ray-Based Stochastic Inversion (RBSI) was introduced in Verdel et al. (2004) as a new seismic inversion method for reservoir parameter estimation claimed to be more accurate than conventional Stochastic Inversion (SI) in laterally heteroge-neous subsurfaces. In this paper, an RBSI-variant is presented which utilises a 1D convolutional forward-modeling kernel as found in common inversion software, offering great practical benefits. Synthetic datatests demonstrate the superiority of this variant, in its proper application regime, over SI in resolving reservoir-layer- P-velocities and thicknesses in strongly dipping subsurface structures. SI, under those circumstances, is severely affected by wavelet distortion due to migration