2025-05-29 バージニア工科大学（Virginia Tech）
＜関連情報＞
- https://news.vt.edu/articles/2025/05/eng-cee-smarter-storm-predictions.html
- https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2024WR039054
長短期記憶ステーションに基づく近似モデルと転移学習技法による極端水位変動の予測 Predicting the Evolution of Extreme Water Levels With Long Short-Term Memory Station-Based Approximated Models and Transfer Learning Techniques
Samuel Daramola, David F. Muñoz, Paul Muñoz, Siddharth Saksena, Jennifer Irish
Water Resources Research Published: 14 March 2025
DOI:https://doi.org/10.1029/2024WR039054
Abstract
Extreme water levels (EWLs) resulting from cyclones pose significant flood hazards and risks to coastal communities and interconnected ecosystems. To date, physically based models have enabled accurate prediction of EWLs despite their inherent high computational cost. However, the applicability of these models is limited to data-rich sites with diverse characteristics. The dependence on high quality spatiotemporal data, which is often computationally expensive, hinders the applicability of these models to regions of either limited or data-scarce conditions. To address this challenge, we present a Long Short-Term Memory (LSTM) network framework to predict the evolution of EWLs beyond site-specific training stations. The framework, named LSTM-Station Approximated Models (LSTM-SAM), consists of a collection of bidirectional LSTM models enhanced with a custom attention mechanism layer embedded in the architecture. LSTM-SAM incorporates a transfer learning approach applicable to target (tide-gage) stations along the U.S. Atlantic Coast. Importantly, LSTM-SAM helps analyze: (a) the underlying limitations associated with transfer learning, (b) evaluate EWL predictions beyond training domains, and (c) capture the evolution of EWL caused by tropical and extratropical cyclones. The framework demonstrates satisfactory performance with “transferable” models achieving Kling-Gupta Efficiency (KGE), Nash-Sutcliffe Efficiency (NSE), and Root-Mean Square Error (RMSE) ranging from 0.78 to 0.92, 0.90 to 0.97, and 0.09–0.18 m at the target stations, respectively. We show that LSTM-SAM can accurately predict not only EWLs but also their evolution over time, that is, onset, peak, and dissipation, which could assist in operational flood forecasting in regions with limited resources to set up physically based models.
Key Points
- We present a deep learning framework that accurately predicts the evolution of cyclone-induced water levels across multiple domains
- An attention mechanism enhances the framework’s recognition of extreme water level patterns within and beyond training locations
- It effectively identifies unseen water level patterns, different from those in training; thus enhancing model’s transfer learning capability
Plain Language Summary
Water levels in rivers, estuaries, and bays rise significantly during hurricanes, leading to severe flood risks and hazards in low-lying areas and interconnected ecosystems. With climate change increasing the frequency of extreme events, it has become crucial to develop models that can accurately simulate extreme water levels in a short time frame and support emergency management for future events. Conventional modeling approaches that help us predict extreme water levels rely on either physically based or data-driven models. Unlike state-of-the-art data-driven models such as deep learning, the former models are site-specific and cannot be applied or transferred to other regions. In this study, we propose a framework that leverages a model trained on extreme water levels from one region to accurately predict those of neighboring regions through a technique known as “transfer learning”. We address the limitations associated with this technique, including the inability of transferable models to accurately generalize new input data from those neighboring regions and examine how changes in model parameters influence the development of efficient transferable models. We show that these models can effectively capture the magnitude and timing of extreme water levels, making this framework suitable for early and operational warning systems.
