Towards holistic river pollution source apportionment: Multi-Faceted framework across scenarios, media, pollutants, and sources.

Rivers harbor numerous compounds posing threats to ecological and human health, Identifying pollution sources and quantifying their transport processes are prerequisites for effective pollution prevention and control. This review systematically examines the evolution of river pollution source identification methodologies and critically assesses the applicability of key approaches-including source inventories, transport-diffusion models, receptor models, isotopic tracing, and machine learning (ML)-across diverse pollutants and scenarios, and provides the main criteria for identifying pollutants based on existing research. However, single-method applications face limitations in complex environments characterized by spatiotemporal water quality heterogeneity, pollutant additive effects, and nonlinear transport dynamics. To address this, 'customized' technical pathways paths for multiple media, multiple pollutants, and multiple pollutants have been established, featuring two types of composite frameworks: receptor-model-based and ML-based techniques. Integrating site-specific hydrogeographic characteristics, pollutant physicochemical properties, and transport processes enables quantitative tracking of dynamic source contributions and future predictions, enhancing results accuracy and interpretability. The review also outlines potential limitations of these customized approaches and proposes improvement strategies. The inherent complexity of river source apportionment stems from their dual roles as both sources and sinks, governed by hydrological features, hydraulic conditions, and ecological status. This synthesis provides theoretical and technical pathways for advancing river pollution source apportionment methodologies, offering valuable guidance for enhancing precision pollution control within watersheds.