Iowa produces 2.5 billion bushels of corn a year. To support this kind of production, the soil requires huge amounts of nitrogen fertilizer. The excess nitrogen leaches into waterways as nitrate, a carcinogen linked to Iowa's rising cancer rate. Drinking water often exceeds the federal 10 mg/L limit, but real-time sensors are sparse, expensive, and shrinking due to budget cuts, leaving utilities, regulators, and farmers without visibility at most locations.

We built an XGBoost model that predicts nitrate concentration and violation risk anywhere in Iowa's waterways using only public weather and land-use data, no sensor required. Validated with leakage-safe cross-validation, it flags violations with a PR-AUC 2.4x better than chance and explains about a third of daily nitrate variance on unseen basins, turning a sparse sensor network into statewide coverage.