Developments and Trends in Water Level Forecasting Using Machine Learning Models—A Review

Water level forecasting in rivers, lakes, and reservoirs is crucial for effective water resource management, flood control, and environmental planning. This review examines the latest developments and trends in water level forecasting research from 2011-2024. A wide range of methods are explored, including traditional statistical models (ARIMA, regression) and advanced techniques like artificial neural networks (ANN), fuzzy logic, support vector machines (SVM), and deep learning models (LSTM). The study assesses the performance and accuracy of these applied models, analyzing their strengths and limitations in capturing water system dynamics and uncertainties. It investigates how data sources (hydrological, meteorological, historical) and variables (rainfall, evaporation, inflow) impact forecast accuracy. The significance of different variables for improving model predictive capabilities is determined. Spatiotemporal aspects are explored, examining model applicability across local, regional, and global scales. Approaches to quantifying and communicating uncertainties associated with probabilistic forecasting for decision-making are evaluated. Detailed analysis identifies proven model efficiencies, potential challenges, and suggests future research directions. By comprehensively reviewing recent water level forecasting literature, this study provides state-of-the-art knowledge on applying machine learning models for reservoir water level prediction. It guides water resource strategies, flood mitigation measures, and decision-making for sustainable water systems management. This review is a valuable resource for researchers and practitioners in hydrology and related fields.