Planning-driven model diagnostics to determine the effects of observation and parameter uncertainty on groundwater management in Mexico City

Abstract:

Human population growth and urbanization in groundwater-dependent cities leads to increased pumping demand while simultaneously altering infiltration. In the Mexico City Metropolitan Area, land and water use changes have caused massive overdraft and subsidence, threatening water supply, particularly in marginalized communities. Regional-scale, spatially distributed groundwater models have been developed in prior work to test potential future scenarios and policy interventions. However, there is little understanding of how parameter and observational uncertainty affect decisions based on model output. This study develops a planning-driven diagnostic approach to support groundwater management. We use global sensitivity analysis to understand how changes in model parameters alter the decision between aquifer management alternatives, including demand management, targeted infiltration, and wastewater reuse. Observational uncertainty is represented by randomly selecting subsets of hydraulic head observations used to calibrate the model. The spatial alternatives are then compared across the range of uncertain model parameters and observation subsets to test their robustness. In general, the parameters governing total water use in the basin cause the greatest effects on the planning decision. However, geologic parameters such as hydraulic conductivity and specific yield also play a significant role, particularly in areas with concentrated groundwater pumping. Additionally, the optimal parameter values vary based on the spatial subset of well observations used to select them. This study highlights the importance of understanding how individual uncertain parameters and their interactions with the observations used to determine model error can affect water supply decision-making processes in densely populated urban areas.