Calibration of complex subsurface reaction models using a surrogate-model approach

摘要: Abstract Automatic calibration of complex subsurface reaction models involves numerous difficulties, including the existence multiple plausible models, parameter non-uniqueness, and excessive computational burden. To overcome these this study investigated a novel procedure for performing simultaneous (SCMM). By combining hybrid global-plus-polishing search heuristic with biased-but-random adaptive model evaluation step, new SCMM method calibrates via efficient exploration multi-model space. Central algorithm components are an assignment preference weights, mapping functions relating uncertain parameters alternative shuffling step that efficiently exploits pseudo-optimal configurations models. The approach was applied to two nitrate contamination problems involving batch reactions one-dimensional reactive transport. For chosen problems, produced improved fits (i.e. up 35% reduction in objective function) at significantly reduced expense 40–90% evaluations), relative previously established benchmarks. Although effective test cases, relies on relatively ad-hoc assigning intermediate weights functions. Despite limitations, results numerical experiments empirically promising reasoning structure provide strong foundation further development.