We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Research on ecological risk attribution and source apportionment of new pollutants in watershed water driven by explainable AI
Summary
Researchers applied an integrated SHAP-LIME explainable AI framework to attribute sources of microplastics and other emerging pollutants across two contrasting Chinese river basins, finding that urban sewage, aquaculture, and agricultural runoff contributed 38%, 28%, and 21% of contamination respectively, with microplastic concentrations averaging 1,831 items per liter — above background thresholds.
In response to the weak interpretability and poor accuracy of traditional methods for source apportionment of new pollutants in river basins, this study selected the Niyang River Basin on the Qinghai-Tibet Plateau (agricultural dominant type) and The Upper Reaches of the Taohe River Basin on the Loess Plateau (urban-rural composite type) as research areas, and constructed an ecological risk attribution and source apportionment system that integrates SHAP-LIME interpretable AI. According to the Technical Guidelines for Accuracy Assessment of New Pollutant Screening - Gas Chromatography Mass Spectrometry (Trial) and the Technical Guidelines for Accuracy Assessment of New Pollutant Screening - Liquid Chromatography Mass Spectrometry (Trial) (China National Environmental Monitoring Center, 2023), an active-passive joint sampling method was adopted to set up 20 monitoring points in two major watersheds, collect water samples during the wet and dry seasons of 2022-2023, and detect 33 antibiotics, 4 parabens, and 6 types of microplastics in the laboratory. Integrating landscape pattern data released by the National Geographical Condition Monitoring Cloud Platform with real-time climate data measured by hydrological stations in the basin to quantify the contribution of pollution sources and risk driving mechanisms. The results show that the fusion model achieves an analytical accuracy with R² = 0.65; Urban domestic sewage, large-scale aquaculture wastewater, and agricultural runoff are the core sources of pollution, contributing 38.2%, 27.5%, and 21.3% respectively; The average concentrations of microplastics and antibiotics were 1831 items/L and 55.33 ng/L, respectively, both exceeding the watershed background threshold. This study confirms that explainable AI can break through the black box bottleneck of traditional models, clarify the pathways of factor action, and provide technical support for precise control of new pollutants in watersheds. The full text data is sourced from field monitoring of the watershed and core journals such as Environmental Science and China Environmental Monitoring, and is authentic and traceable.