<p>Accurate prediction of lake water levels is essential for addressing ecological, economic, and social challenges. However, the complex nonlinear dynamics of lake levels—shaped by diverse meteorological factors—pose significant challenges to traditional modeling approaches. Recent advancements in deep learning and big data analytics have transformed water level forecasting by enabling the modeling of intricate relationships among environmental variables such as precipitation, temperature, and historical trends. This study focuses on forecasting monthly water levels for three major lakes: Beyşehir in Türkiye, Balkhash in Kazakhstan, and Baikal in Russia. Key meteorological parameters—mean temperature, mean relative humidity, and precipitation—were employed as inputs, with lake water level as the output. The modeling process involved two independent approaches: artificial neural networks (ANN) and gated recurrent units (GRU). Additionally, a hybrid model integrating convolutional neural networks (CNN), long short-term memory (LSTM), and GRU architectures (CNN-GRU-LSTM) was developed to enhance predictive accuracy. The dataset, covering the period from 2002 to 2024, was divided into training (2002–2018) and testing (2019–2024) phases across three distinct scenarios. Model performance was evaluated using root mean square error (RMSE), Nash–Sutcliffe efficiency (NSE), coefficient of determination (R²), and graphical analyses. The results demonstrated that the CNN-GRU-LSTM model outperformed the GRU and ANN models, achieving an average RMSE of 0.578 m, NSE of 0.770, and R² of 0.93. The exceptional performance of the hybrid CNN-GRU-LSTM model underscores its potential for broader applications in lake water level prediction. This approach can be extended to other lakes, supporting the development of advanced monitoring systems and contributing to the prevention of large-scale flood events.</p> Graphical Abstract <p></p>