A Generative Super‐Resolution Model for Enhancing Tropical Cyclone Wind Field Intensity and Resolution

• Published in: Journal of geophysical research. Machine learning and computation Vol. 1; no. 4
• Main Authors: Lockwood, Joseph W., Gori, Avantika, Gentine, Pierre
• Format: Journal Article
• Language: English
• Published: Wiley 01.12.2024
• ISSN: 2993-5210; 2993-5210
• DOI: 10.1029/2024JH000375

Extreme winds associated with tropical cyclones (TCs) can cause significant loss of life and economic damage globally, highlighting the need for accurate, high‐resolution modeling and forecasting for wind. However, due to their coarse horizontal resolution, most global climate and weather models suffer from chronic underprediction of TC wind speeds, limiting their use for impact analysis and energy modeling. In this study, we introduce a cascading deep learning framework designed to downscale high‐resolution TC wind fields given low‐resolution data. Our approach maps 85 TC events from ERA5 data (0.25° resolution) to high‐resolution (0.05° resolution) observations at 6‐hr intervals. The initial component is a debiasing neural network designed to model accurate wind speed observations using ERA5 data. The second component employs a generative super‐resolution strategy based on a conditional denoising diffusion probabilistic model (DDPM) to enhance the spatial resolution and to produce ensemble estimates. The model is able to accurately model intensity and produce realistic radial profiles and fine‐scale spatial structures of wind fields, with a percentage mean bias of −3.74% compared to the high‐resolution observations. Our downscaling framework enables the prediction of high‐resolution wind fields using widely available low‐resolution and intensity wind data, allowing for the modeling of past events and the assessment of future TC risks. Plain Language Summary Extreme winds associated with tropical cyclones (TCs) can cause significant loss of life and economic damage globally, highlighting the need for accurate, high‐resolution modeling and forecasting for wind. Traditional global climate models and weather forecasting models often lack the horizontal resolution needed to accurately model these extreme winds. We introduce a two‐part deep learning method that enhances the quality of wind speed predictions from coarse data. The first model corrects inaccuracies in initial low‐resolution data from 85 TC events. The second part uses an advanced generative deep learning model to refine these corrections, producing high‐resolution wind field images. This process allows us to generate more detailed and accurate maps of TC wind speeds, capturing finer spatial details that are crucial for effective impact analysis and energy modeling. Our downscaling framework enables the prediction of high‐resolution wind fields using widely available low‐resolution and intensity wind data, allowing for the reconstruction of past events, the assessment of future TC risks, and improved understanding of historical TC impacts even in the absence of satellite data. Key Points We demonstrate that a cascading deep learning generative model can effectively and efficiently model tropical cyclone wind maps The framework accurately models high‐intensity TC wind speeds from low‐resolution climate data The super‐resolution model enhances resolution, producing realistic radial profiles and fine‐scale spatial structures of wind fields