Integrating Explainable Artificial Intelligence and Machine Learning for Understanding Water Pollution and Its Management

Water qualityWater quality assessment is vital for water resource management and environmental sustainabilityEnvironmental sustainability. This study estimated the water qualityWater quality index (WQI)Water Quality Index (WQI) for Loktak Lake, Manipur, India, evaluated predictive modelsPredictive models for WQIWater Quality Index (WQI), and used explainable artificial intelligence (XAI)Explainable Artificial Intelligence (XAI) techniques for water pollution managementPollution management. Entropy theory was employed to determine WQI, with weights assigned based on parameter importance. WQIWater Quality Index (WQI) values ranged from 80.59 to 100.00, indicating varying water qualityWater quality levels. The study classified samples into good (24%), moderately polluted (25%), and severely polluted (26%) categories. Predictive modelsPredictive models were implemented using the AdaBoost and random forestRandom forest algorithms, with grid search-based hyper-parameter tuning. The random forestRandom forest model outperformed AdaBoost, showing high accuracy in predicting water qualityWater quality classes. Explainable AI techniques, SHAP, and LIME were used to gain insights into parameter impacts on WQIWater Quality Index (WQI) prediction. BOD, COD, and turbidityTurbidity negatively influenced WQI, while temperature and dissolved oxygen had positive effects. Site-specific analysis using LIME revealed parameter behavior at specific locations, guiding targeted management strategiesManagement strategies. The study recommended strategies to improve water qualityWater quality, including controlling and reducing BOD, COD, and turbidityTurbidity, enhancing dissolved oxygen levels, monitoring temperature and pHPH, addressing nitrates and TDSTDS, and implementing site-specific measures. Policy implications involve adopting the identified strategies by water resource managers and policymakers. Regular monitoring, adaptive management, and interdisciplinary collaboration are crucial for sustainable water resource management and improved water qualityWater quality. Considering local regulations and stakeholder involvement will contribute to successful implementation and positive environmental outcomes. Therefore, this study provided valuable insights into water qualityWater quality assessment and management. The predictive modelsPredictive models exhibited strong performance, and XAIExplainable Artificial Intelligence (XAI) techniques facilitated a better understanding of influential parameters. Implementing the recommended strategies can lead to effective water pollutionPollution control and resource conservation.