A Geo AI-Based Environmental Risk Assessment Framework For Chemical Exposure And Public Health Vulnerability In U.S. Industrial Corridors

Abstract

The United States industrial corridors are still subjected to unequal and disproportionate chemical exposures and community health hazards, but the current environmental justice instruments have low predictive power and poor heterogeneous datasets. This paper describes GeoAI RiskLab, a reproducible environmental risk assessment framework which combines EPA Toxics Release Inventory emissions, ambient and satellite monitoring and sociodemographic vulnerability indicators into one GeoAI architecture.

The model uses spatial deep learning models, uncertainty quantification, and explainable AI diagnostics in generating high-resolution estimates of cumulative chemical exposure risk. Outputs are projected to decision relevant spatial units e.g. census tracts and school zones, which has been used to target regulatory purposes and urban resilience planning. The usage in the selected industrial corridors has shown that the toolkit is useful in establishing hotspots of exposure, assessing model plausibility, and making decisions on environmental health. GeoAI RiskLab is a policy-friendly and transferable environmental governance platform based on data.