Covariance Matrix Adaptation Evolution Strategy Algorithm in the Context of Seismic Inversion: Proposal, Application, and Analysis Compared to Ensemble Smoother with Multiple Data Assimilation

Seismic reservoir characterization faces non-uniqueness challenges, requiring robust uncertainty quantification. This study evaluates the UH-CMA-ES algorithm against the ESMDA method using synthetic seismic data. Both methods effectively minimize seismic misfit. However, UH-CMA-ES provides significantly more focused uncertainty envelopes. Statistical analysis via the Kolmogorov-Smirnov test demonstrates that UH-CMA-ES can generate property distributions indistinguishable from the ground truth. While computationally more expensive and exhibiting a reduction in global coverage, UH-CMA-ES proves to be a high-fidelity alternative for precise local exploitation in ill-posed inversion problems.