A Review of Water Quality Forecasting Models for Freshwater Lentic Ecosystems

Water quality (WQ) monitoring is critical for Mexico and the world due to water pollution and scarcity problems in recent years. In this article, a systematic review was conducted considering only forecasting models focused on lentic freshwater bodies (to specialize the analysis of variables, problems, considerations, etc.) from 2019 to 2025 (to ensure the inclusion of the most relevant and new studies). This review analyzes 52 articles focused on the monitoring place, predictors, forecasted variables, configuration of each forecasting model, results with or without multiple forecast horizons, monitoring conditions, forecasting horizon, data availability, and model replicability. Our review shows that the main models documented used to predict WQ are based on machine learning (where RFs are the most used), AI (where ANNs are the most used and LSTM-based architectures are the most implemented), and statistical methods (where MLR is the most used). The principal forecasted WQ variables are Chl-α, DO, and TP. In comparison, the most used predictors are TP, temperature, and Chl-α. Furthermore, only 10 articles have made their databases available, and nine articles share the configuration of their models. Future research should investigate the real impact of data (quantity and inputs) variation in forecasting values for multiple forecast horizons.