Lake water-level fluctuation forecasting using machine learning models: a systematic review.

Area of Science:

• Hydrology
• Environmental Modeling
• Data Science

Background:

• Lake water-level fluctuation is inherently complex, stochastic, and nonlinear, posing challenges for traditional modeling and forecasting.
• Recent advancements in machine learning (ML) offer promising solutions for accurately predicting these dynamic hydrological processes.

Purpose of the Study:

• To provide a comprehensive review of machine learning applications in modeling and forecasting lake water-level dynamics.
• To critically assess the strengths, limitations, and inter-comparisons of various ML models used in this field.

Main Methods:

• Review of seven popular machine learning model types: artificial neural network (ANN), support vector machine (SVM), artificial neuro-fuzzy inference system (ANFIS), hybrid models (e.g., WA-ANN, WA-ANFIS, WA-SVM), evolutionary models (GEP, GP), extreme learning machine (ELM), and deep learning (DL).
• Discussion of model inputs, data splitting strategies, performance evaluation criteria, and inter-model comparisons.

Main Results:

• Machine learning models have demonstrated substantial progress in forecasting lake water-level fluctuations.
• Detailed analysis of the advantages and limitations of each reviewed ML model type is presented.

Conclusions:

• This review offers a consolidated perspective on ML applications for lake water-level dynamics, aiding researchers and practitioners.
• Provides future research directions to enhance the sustainable management of lake resources through improved hydrological forecasting.