Unveiling the Hidden Connections: Using Explainable Artificial Intelligence to Assess Water Quality Criteria in Nine Giant Rivers

Abstract

The degradation of water quality constitutes a significant global environmental issue with serious consequences for ecosystems, human health, and sustainable development. Notwithstanding comprehensive study, considerable knowledge gaps persist in comprehending the complex interrelations among water quality measures and in determining effective evaluation methodologies. This study utilizes explainable artificial intelligence (XAI) to thoroughly examine the connections among eight essential water quality measures and to determine the most dependable predictions for efficient monitoring and management. A thorough comparison examination of four machine learning models was performed, employing a unique scoring mechanism created exclusively for this research. The Random Forest model exhibited superior performance, as indicated by its attainment of the lowest Root Mean Square Error (RMSE) values, a generally recognized measure of model correctness. Dissolved Oxygen was identified as the primary predictor, exhibiting a minimum RMSE of 0.127589 across several river datasets. This discovery highlights the significance of Dissolved Oxygen as a crucial metric for evaluating water quality. The study underscores the significant impact of river-specific attributes, highlighting the necessity for customized assessment methodologies for various aquatic ecosystems. This study enhances the scientific comprehension of water quality evaluation by incorporating explainable AI to elucidate the intricate interrelationships among metrics. The results confirm the efficacy of AI-driven methods in environmental monitoring and offer practical insights to inform the creation of precision water management strategies. This work addresses essential information deficiencies, aiding in developing more effective and sustainable strategies for reducing water quality deterioration and protecting aquatic ecosystems. The results indicate that the suggested methodology provides a dependable and comprehensible technique for predicting water quality, which can greatly assist water experts and policymakers.