Well-log parameterized full waveform inversion 1: analysis of the single parameter case
Ninoska Amundaray, Scott Keating, Matthew Eaid, Kristopher A. Innanen
Applications of multiparameter elastic full waveform inversion (EFWI) deal with several challenges posed by the intrinsic nonlinearity of technique. Parameter cross-talk is one of most critical examples of the latter. Though, existing resources like scattering radiation patterns and the addition of prior information into FWI formulations might provide means to prevent some level of cross-talk. Ultimately, they cannot guarantee the removal of these unwanted effects. As an alternative mitigation strategy, we are studying reducing the numbers of modeled parameters in an isotropic medium, from three to one in this study. We argue that the single parameter inversion can be constructed in any application where elastic parameters demonstrate a strong correlation in readily available data, like well-logs, by fitting a trendline. An example of strong parameter relationship was observed and formulated for the Carbon Management Canada Newell County Facility. Though, the single parameter formulation was previously analyzed. We are expanding its use in a synthetic time-lapse application for carbon dioxide (CO2
) sequestration at the same site. Current results demonstrate that inverted models converged towards the true solution with various levels of success. While inverted P-wave velocity (VP) and S-wave velocity (VS) values fluctuate between a small parameter overestimation and underestimation. Density (ρ) values are mostly overestimated.