همایش آروین البرز ETSJR 2 1 2026 05 07 Machine Learning-Based Prediction of Water Quality in European River Basins: Advancing Sustainable Management under Climate Change Machine Learning-Based Prediction of Water Quality in European River Basins: Advancing Sustainable Management under Climate Change 1 13 10.22051/jera.2021.31891.2698 FA اسرین دانشگاه تبریز 2026 04 08 European river basins face increasing pressures from climate change, agricultural intensification, and urbanisation, threatening compliance with the EU Water Framework Directive (WFD) and long-term ecosystem health. This study proposes a machine learning framework to predict the Water Quality Index (WQI) and key physicochemical parameters (e.g., dissolved oxygen, BOD, COD, nitrate, pH, turbidity) using readily available monitoring data from selected European watersheds. Advanced models, including ensemble methods (e.g., XGBoost, Random Forest) and recurrent neural networks (e.g., LSTM), were developed, optimised, and validated against historical datasets, achieving high predictive performance (R&sup2; > 0.93, normalised RMSE < 0.08 in cross-validation). Feature importance analysis via SHAP values highlights dominant drivers such as temperature, nutrient loads, and hydrological variability under climate scenarios. The framework supports proactive decision-making by enabling early detection of degradation trends, scenario testing for pollution mitigation, and adaptive management strategies. By minimising the need for resource-intensive sampling and facilitating real-time insights, this approach strengthens evidence-based governance, aligns with Sustainable Development Goal 6 (clean water and sanitation), and contributes to resilient, sustainable water resource management across Europe amid escalating climatic uncertainties. The methodology is scalable and transferable to other EU catchments, promoting harmonised monitoring and policy implementation. European river basins face increasing pressures from climate change, agricultural intensification, and urbanisation, threatening compliance with the EU Water Framework Directive (WFD) and long-term ecosystem health. This study proposes a machine learning framework to predict the Water Quality Index (WQI) and key physicochemical parameters (e.g., dissolved oxygen, BOD, COD, nitrate, pH, turbidity) using readily available monitoring data from selected European watersheds. Advanced models, including ensemble methods (e.g., XGBoost, Random Forest) and recurrent neural networks (e.g., LSTM), were developed, optimised, and validated against historical datasets, achieving high predictive performance (R&sup2; > 0.93, normalised RMSE < 0.08 in cross-validation). Feature importance analysis via SHAP values highlights dominant drivers such as temperature, nutrient loads, and hydrological variability under climate scenarios. The framework supports proactive decision-making by enabling early detection of degradation trends, scenario testing for pollution mitigation, and adaptive management strategies. By minimising the need for resource-intensive sampling and facilitating real-time insights, this approach strengthens evidence-based governance, aligns with Sustainable Development Goal 6 (clean water and sanitation), and contributes to resilient, sustainable water resource management across Europe amid escalating climatic uncertainties. The methodology is scalable and transferable to other EU catchments, promoting harmonised monitoring and policy implementation. /downloadfilepdf/309926