Jing Yang

College of Water Resoures and Architectural Engineering, Northwest A&F University, Yangling 712100, Shaanxi, China

Conventional variance-based sensitivity analysis method only considers parameter uncertainty but ignores the model uncertainty. To overcome this weakness and extend the scope of variance-based sensitivity analysis method, the work of Dai et al. (2017) and Yang et al. (2022) integrated the model averaging approach with variance-based method and developed the first-order PS_K and total-effect PS_TK process sensitivity indices to identify the sensitivity of the processes with explicitly considering multiple plausible process models, which also contains uncertainty parameters. A key concern of their methods is the computational cost, as the groundwater models usually take a long processing time. By expanding the study of Dai et al. (2022) for estimating PS_K with a computational efficient method, we propose an efficient method to estimate PS_K with a simple and intuitive implementation which could simultaneously estimate PS_K and PS_TK. We demonstrated the accuracy efficiency, and robustness of the proposed design by using a hypothetical one-dimensional (1-D) groundwater flow model and then applied the proposed design to estimate the process sensitivities of a two-dimensional (2-D) Arsenic (As) sorption and reactive transport model adopted from Duan et al. (2020), of which three processes were conceptualized and a total number of 12 = 2 × 2 × 3 system models were populated. The results show that: (1) the proposed design provides process sensitivities consistent with the those obtained by using brute force Monte Carlo (MC) method with excessive model runs; (2) the proposed design generally converges ~1000 times faster than the brute force MC method and achieves better robustness; (3) the proposed design applied to the As sorption and reactive transport model analyzing the time-dependent process sensitivities is a helpful tool for understanding model internal dynamics behaved by different processes. The proposed design is mathematically general and straightforward to use and can be applied to a wide range of sensitivity analysis problems in groundwater modeling.