Predicting river water quality under different environmental factors and its significance with Machine Learning approach

River Water Quality (RWQ) is significantly influenced by natural and anthropogenic activities such as land use and land cover changes. Urbanization has led to an increase in impervious surfaces, which alters hydrological flow pattern and delivers non-point pollutants to the stream more efficiently. In addition, intensification of agricultural activities can result in the increased nutrient loads due to alternations in surface soil properties. Hence, it is necessary to understand the impact of surrounding environment with specific emphasis on geographical factors (e.g. climate change, land use patterns and landscape metrics) on the RWQ in order to develop sustainable water quality management strategies effectively. We collected pollutant concentration Biochemical Oxygen Demand (BOD), Total Phosphorus (T-P), and  Total Organic Carbon(TOC) from monitoring stations in the Nakdong River watershed. To utilize field monitoring data, we developed a Machine Learning (ML) models (DNN, XGBoost and Random Forest) to predict RWQ in accordance with different environmental factors. SHapley Additive exPlanations (SHAP) was used to illustrate the significance of land uses and landscape patterns on RWQ in Nakdong River. The results of this study can (1) demonstrate the relationship of water quality variables with land uses and landscape patterns, (2) identify pollution sources and factors that affect Nakdong River, and (3) support catchment managers and stakeholders in evaluating the benefits and risks of water management strategies in priority areas.