Regularized Bi-Directional Deconvolution for Wavelet Estimation and Its Field Application in CO₂ Storage Characterization

The conventional blind deconvolution methods for wavelet estimation, which involve solving for both the unknown wavelet and reflectivity, are hindered by nonlinear optimization leading to local minima and amplitude ambiguity between the reflectivity and wavelet. We propose a regularized bi-directional deconvolution method that employs a hybrid norm to enforce sparsity and spikiness in the estimated reflectivity. By incorporating prior knowledge as regularization terms and applying preconditioning to the unknown filters, our method aims to avoid local minima, recover amplitudes of the wavelets, and expedite convergence. In addition, we linearize the objective function and introduce a customized optimization schedule specifically designed for the regularized scheme problem with a hybrid norm. We investigate the impact of input parameters on the estimation accuracy, providing guidance for selecting the appropriate input parameters. We validate our method by conducting experiments on three synthetic examples, demonstrating its ability to accurately retrieve reflectivity and estimate mixed-phase wavelets. Our method proves effective in handling complex structures, such as wedge models with tuning effects and interfered reflectors. Furthermore, we apply our method to the Sleipner CO2 storage project in the North Sea, successfully capturing attenuation effects caused by the CO2 plume. The estimated wavelet, when combined with attenuation measurements, enables quantitative interpretation (QI) and enhances the characterization of the CO₂ plume.