Spatiotemporally adaptive optimization of groundwater remediation: Well-network system design and multi-stage operation.

Groundwater remediation under stringent resource and time constraints requires coordinated decisions on where extraction infrastructure is deployed and how pumping is operated over time. Conventional pump-and-treat optimization often addresses strategic well-network design and dynamic pumping operation separately, limiting the coordinated use of evolving hydraulic and concentration fields in heterogeneous aquifers. Here, we developed an optimization framework incorporating mixed-variable well-network design and multi-stage pumping optimization under concentration and hydraulic-head constraints. Mixed-variable NSGA-II identifies Pareto-optimal combinations of well number, location, pumping rate and remediation duration, after which the selected solution is refined by reallocating pumping between stages according to simulated plume evolution. The framework was evaluated using two-dimensional and three-dimensional heterogeneous aquifer models and applied to a chromium-contaminated field site initialized from monitoring data. In the three-dimensional case, multi-stage operation reduced estimated pumping energy cost by 35.3% relative to the selected single-stage solution. At the field site, multi-objective optimization reduced cumulative pumping by 27.3% using existing wells, and multi-stage operation achieved a further 2.8% reduction while maintaining modeled contaminant mass removal within 1.6% of the single-stage solution. This work shifts groundwater management from experience-led heuristic designs toward model-driven, precision decision-making, providing a resource-efficient approach for pump-and-treat remediation of groundwater contamination.