In this study, we developed a new type of water quality model that uses artificial intelligence (AI) and deep learning to better capture how nitrogen pollution varies across space. Our model, called dPL-SPARROW, learns how pollution sources like fertilizer, manure, and atmospheric deposition contribute to nitrogen levels based on local land use and soil conditions. We trained the model using 20 years of monitoring data from the Upper Mississippi River Basin.

Compared to existing methods, our dPL-SPARROW model made more accurate predictions and provided insights into why pollution levels differ across regions. By identifying which local features most influence pollution, our approach can help decision-makers design more targeted strategies to reduce nutrient runoff and protect water quality. This represents a step forward in combining AI with environmental modeling for practical water management.