A Stochastic Simulation-Optimization Approach for Groundwater Flow Modeling and Management in an Over-Exploited Aquifer

Abstract. To guide the sustainable management of the over-exploited Western Baiyangdian Plain aquifer, this study developed an integrated simulation-optimization framework. A high-resolution stochastic aquifer model, constructed using Transition Probability/Markov Chain (TP-MC) geostatistics, was coupled with a groundwater flow model and a multi-objective optimization algorithm solved via a genetic algorithm with an adaptive penalty function (GA-APF). The stochastic model demonstrably outperformed its deterministic equivalent in replicating observed hydraulic heads. Scenario projections for 2035 indicate that combining ecological water replenishment with pumping reductions could raise groundwater levels by approximately 12, 10, and 6 m in wet, normal, and dry years, respectively. However, this recovery concurrently increases phreatic evaporation, reaching up to 0.97 × 10⁸ m³/yr in wet years, revealing a critical management trade-off. The optimization model effectively balanced competing objectives, reducing projected water shortage rates to 11.86–14.06% by 2035 while simultaneously mitigating evaporative losses. This framework provides a robust, transferable tool for restoring and optimally allocating water resources in stressed aquifer systems.