Dual‐Transformer Deep Learning Framework for Seasonal Forecasting of Great Lakes Water Levels

• Published in: Journal of geophysical research. Machine learning and computation Vol. 2; no. 2
• Main Authors: Chen, Yi, Xue, Pengfei
• Format: Journal Article
• Language: English
• Published: United States American Geophysical Union (AGU) 01.06.2025; Wiley
• Subjects: deep learning; Great Lakes; lake level; machine learning; transformer
• ISSN: 2993-5210; 2993-5210
• DOI: 10.1029/2024JH000519

The Great Lakes of North America form one of the largest freshwater systems on Earth, and their lake‐wide average water levels (lake levels) can fluctuate by more than 0.5 m on a seasonal scale. These fluctuations pose substantial challenges for coastal resilience, flood risk management, and navigation planning. Accurate seasonal forecasting of lake levels using traditional mechanistic models is challenging due to the complex physical mechanisms and coupled hydroclimatic processes involved. Recently, deep learning has gained prominence in geoscience applications for its ability to recognize intricate patterns within multiphysical data sets. Here, we introduce a novel Dual‐Transformer deep learning framework, tested on the Great Lakes. This architecture integrates two modified Transformer models: the Prophet, which predicts underlying trends, and the Critic, which refines the Prophet's predictions. The final lake level prediction is derived by weighting the outputs of both models through a multi‐layer perceptron, jointly trained with the Prophet and Critic to enhance overall accuracy. Our results demonstrate that the innovative learning framework achieves the highest prediction accuracy compared to established deep learning models when using identical input features. It attains a root mean square error of 4–7 cm in predicting lake levels up to 6 months in advance across the lakes. Additionally, the Dual‐Transformer model runs six orders of magnitude faster than conventional mechanistic models, producing results in less than one second on a typical personal computer. These findings suggest that our deep learning framework has strong potential to advance lake level prediction and carries important implications for water management and disaster mitigation, thereby enhancing the quality of life in coastal regions. Plain Language Summary The Great Lakes, one of the world's largest freshwater systems, experience natural seasonal water level fluctuations exceeding half a meter. These fluctuations pose challenges for coastal communities in water management and coastal resilience. Traditional forecasting methods struggle to capture these changes accurately due to the complex interactions between water and atmospheric conditions. To address this, we developed a new deep learning model, the Dual‐Transformer, which we tested on the Great Lakes. This model comprises two main components: the Prophet, which forecasts general trends, and the Critic, which adjusts these forecasts based on recent weather data. A third component combines the two predictions to improve accuracy. Our results show that the Dual‐Transformer, using seven key weather and lake variables, can predict lake levels with very high accuracy up to 6 months into the future. The model is also computationally efficient, running up to a million times faster than traditional methods and producing results in less than a second on a standard computer. This deep learning model has significant potential to enhance water management, safety, and community planning along the coasts of the Great Lakes. Key Points An innovative Dual‐Transformer deep learning framework is developed for seasonal forecasting of water levels in the Great Lakes The input features are widely accessible as standard outputs from most atmospheric and climate models, enhancing model applicability The model achieves unprecedented accuracy at very low computational cost in predicting lake levels with a lead time of up to 6 months