Smart and Sustainable Water Quality Monitoring System Using Explainable AI

Reliable monitoring of water quality is essential for protecting aquatic ecosystems, ensuring sustainable aquaculture production, and safeguarding public health. Conventional monitoring approaches rely on periodic sampling and laboratory analysis, which often result in delayed detection of unsafe water conditions and limited monitoring coverage. This paper presents an interpretable machine learning framework for intelligent water quality monitoring that integrates predictive modeling with Explainable Artificial Intelligence (XAI). Fourteen physicochemical and biological parameters are used to train an XGBoost-based classification model that categorizes water conditions into SAFE and UNSAFE classes. To enhance transparency, SHAP-based explanations are applied to quantify the influence of individual parameters on the prediction outcome and translate model decisions into human-understandable reasoning. In addition, similarity-based evidence retrieval using a k-nearest neighbor approach provides supporting examples from historical observations to strengthen decision reliability. The system is implemented through an interactive monitoring dashboard that performs range-constrained validation, prediction visualization, and explanation delivery. Experimental evaluation on an aquaculture water quality dataset demonstrates strong classification performance while maintaining interpretability and decision transparency. The proposed framework provides a practical decision-support tool that can assist environmental monitoring authorities and aquaculture practitioners in identifying critical water quality parameters and enabling timely corrective actions in intelligent monitoring environments.