Journal Article

Source Tracing for Pollutants in River Channels Based on a Physics‐Informed Neural Network

Yin, Xinan, Yi, Yujun, Zhao, Xu, Liu, Haifei, Leng, Fei, Yang, Wei

2026

Overview

The river pollutant traceability problem represents a critical challenge in environmental monitoring and water resource management. In this work, we propose an approach based on a physics‐informed neural network (PINN) for identifying key parameters of pollutant sources, including the release intensity and location, on the basis of cross‐sectional observations. The accuracy of the proposed method was validated through experiments conducted on steady, unsteady, and noisy unsteady flows, with numerical simulations of real‐world river systems as test cases. The results demonstrate that the method can not only accurately identify source parameters beyond the gauging river reach but also effectively capture the spatial distribution of pollutants across the entire computational domain. This approach provides a novel solution for addressing the challenges of tracing pollutant sources in river channels.

• Publisher: John Wiley & Sons, Inc
• Subject: Algorithms / Artificial intelligence / Boundary conditions / Channels / Environmental monitoring / Inverse problems / Neural networks / Normal distribution / Numerical simulations / Parameter identification / Parameters / Partial differential equations / Physics / Pollutants / Pollution / Resource management / River channels / River systems / Rivers / Simulation / Spatial distribution / Statistical methods / Stream flow / Tracing / Unsteady flow / Water pollution / Water resources / Water resources management