Machine learning approaches for ecological compensation in transboundary waters

Abstract

The Tongshun River, a vital waterway facing severe pollution, requires effective strategies for environmental management and sustainable resource use. This study utilizes an LSTM-XGBoost composite model to predict key water quality indicators—CODMn, NH3-N, TP, and TN—at the Gangzhou Village and Huangling Bridge monitoring sections. Using historical data from January 2021 to November 2023, the model demonstrates significant predictive accuracy, outperforming traditional CNN, LSTM, and XGBoost models by achieving lower MSE, RMSE, and MAE values, and higher R2 values. These results facilitate the implementation of ecological compensation mechanisms, enabling proactive interventions and efficient resource allocation. By integrating predictive modeling with ecological compensation frameworks, this approach enhances transboundary water pollution management, supporting the preservation of water quality and aquatic ecosystem health. The study highlights the potential of machine learning models in advancing water pollution control and informs effective policy formulation in environmental protection.