Caig: prediction of groundwater uranium contamination using physics-informed and casual AI methods

The National Science Foundation Division of Undergraduate Education awarded Navajo Technical University $299,900 on August 1, 2026, under the STEM Education program (CFDA 47.076) to develop physics-informed and causal artificial intelligence methods for predicting uranium groundwater contamination across the Navajo Nation.

The project integrates physics-informed neural networks and structural causal models to predict uranium migration from more than 500 abandoned hard-rock mines that have contaminated groundwater across Navajo Nation territories. The AI framework combines governing groundwater physics equations with causal inference to produce physically consistent, interpretable, and scientifically explainable models. Neural operators, generative models, transfer learning, and uncertainty quantification enhance computational efficiency and prediction under limited monitoring data. Model performance will be benchmarked against established MODFLOW-based groundwater simulators and validated using field observations collected through the Navajo Technical University–Florida Atlantic University partnership.

The award funds research capacity building at Navajo Technical University through AI and geoscience education, undergraduate research experiences, field investigations, and mentoring that prepare students for STEM careers in artificial intelligence, geoscience, and environmental stewardship. Work is performed in Crownpoint, New Mexico, with a period of performance through July 31, 2029.

Abstract

Hard-rock mining has occurred across the navajo nation for more than 100 years, leaving more than 500 abandoned mines that have contaminated groundwater with uranium. This contamination can pose risks to the health of families who rely on the affected groundwater and represent a broader public health concern. This project integrates artificial intelligence, hydrogeology, environmental engineering, and geochemistry to enhance uranium migration prediction and assist groundwater management in the navajo nation. The project will also strengthen research capacity at a tribal university by providing artificial intelligence (AI) and geoscience education, undergraduate research experiences, field investigations, and mentoring that prepare students for stem careers in artificial intelligence, geoscience, and environmental stewardship. The project integrates physics-informed neural networks (pinns) with structural causal models (scms) to combine governing groundwater physics with causal inference, enabling models that are physically consistent, interpretable, and scientifically explainable. Pinns enforce groundwater flow and advection-dispersion-reaction equations through partial differential equation residuals, boundary conditions, and initial conditions, while scms identify the causal interactions among fracture connectivity, groundwater flow, and geochemical reactions and support counterfactual analyses of remediation scenarios. Neural operators, generative models, transfer learning, and uncertainty quantification improve computational efficiency, data-efficient inversion, and prediction under limited monitoring data. Model performance will be benchmarked against established modflow-based groundwater simulators and validated using field observations collected through the navajo technical university-florida atlantic university partnership, producing next-generation AI methods and scientifically credible decision-support tools for groundwater management in complex subsurface environments. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the foundation's intellectual merit and broader impacts review criteria.- subawards are planned for this award.