Upscaling groundwater models for inclusion in management models - Case study of California's Sacramento Valley

This article describes and evaluates the upscaling of a calibrated groundwater model of California's Sacramento Valley for the purpose of incorporating it into general management (optimization) models. Upscaling is performed in two stages, by vertical integration, than by combining grid elements into isolated lumped subbasins and switching from a distributed to a simplified lumped subbasin model. Simulation results of the upscaled models are compared to those of the initial spatially distributed model. Because anthropogenic stresses (pumping and recharge) outweigh interbasin flows, the fully upscaled formulation estimates groundwater levels very close to those of the vertically integrated model. An optimization formulation based on the upscaled model efficiently retains almost all relevant management information contained in more detailed models without recourse to expensive distributed embedding methods.