[Research Progress on River and Lake Water Quality Assessment Based on Machine Learning]

Area of Science:

• Environmental Science
• Data Science
• Water Resource Management

Context:

• Machine learning (ML) is increasingly vital for pollution early warning and water quality assessment in aquatic ecosystems.
• A comprehensive review of 309 studies from Web of Science and CNKI databases highlights the growing trend in ML applications for water quality.
• The research focuses on rivers and lakes, analyzing trends and methodologies in ML-driven water quality assessment.

Purpose:

• To investigate the application scenarios, methodological focus, impact factors, bottlenecks, and future directions of ML in water quality assessment.
• To analyze the predictive goals and directions of ML in water ecosystems, specifically time-specific and time-series water quality prediction.
• To examine the influence of input factors and ML methods on the prediction accuracy of nutrients, chlorophyll-a (Chla), and organic matter concentrations.

Summary:

• Water quality prediction, encompassing specific-time and time-series forecasting, is the primary application of ML in water ecosystems.
• Dissolved oxygen (DO), water temperature (WT), and pH are the most frequent inputs for ML models, alongside internal/external sources and hydraulic conditions.
• Mechanistic model-ML coupling and interpretable machine learning (XML) are emerging as key research focuses to address limitations like data missing, overfitting, and interpretability.

Impact:

• ML models demonstrate powerful fitting capabilities for predicting contaminant concentrations, even without complete physical and chemical mechanisms.
• Identifying high-frequency input factors and core driving factors provides insights for improving ML model prediction accuracy.
• Addressing limitations such as data missing, overfitting, and insufficient interpretability through advanced methods will enhance the reliability and applicability of ML in water quality assessment.