A framework for tracing the sources of nitrate in surface water through remote sensing data coupled with machine learning

As an integral component of the global nitrogen cycle, nitrate are readily transferred from urban sewage discharge, agricultural activities, and atmospheric sedimentation to surface water. This paper introduces an innovative framework that combines multi-source remote sensing technology with stable nitrate nitrogen (δ¹⁵N-NO₃⁻) and oxygen (δ¹⁸O-NO₃⁻) isotopes mixing model, to identify nitrate sources quantitatively in surface water for the first time (R² range from 0.50 to 0.99, RMSE range from 0.05‰ to 2.31‰, MAE range from 0.03‰ to 1.35‰). By reconstructing the historical nitrate isotopes from 2006 to 2023, we found that manure and sewage were the main contributing sources, followed by soil nitrogen, fertilizer and atmospheric deposition (contribution ratio of 3.5:2.5:2.5:1.5), wastewater discharge and fertilizer application in Xijiang river had a significant impact on this. This framework fills a gap in the research pertaining to remote sensing technology’s identification of surface nitrate sources, facilitating straightforward and user-friendly forecasting of nitrate source spatio-temporal sequences.