This dataset provides a collection of atmospheric fields associated with intense extratropical cyclones (ETCs) over the North Atlantic, prepared for machine learning applications. It was created to address the limited availability of high-resolution data on extreme ETC events, which hampers the development of robust data-driven models in climate and atmospheric sciences. The dataset covers the North Atlantic basin, from 70°N to 20°N and from 20°E to 100°W. Using the Northern Hemisphere Extratropical Cyclone Tracks dataset (Crawford et al., 2021), we identified a subset of intense ETCs based on their normalized deepening rate (NDR ≥ 1), following established thresholds for rapid cyclone intensification. This filtering process resulted in 4,070 cyclone tracks. For each time step along these tracks, we retrieved the corresponding hourly ERA5 reanalysis fields at a spatial resolution of 0.25°. The extracted variables include mean sea level pressure, the u and v components of wind at 10 meters (from which wind speed was derived), and total precipitation (used to compute rainfall rate). The final dataset consists of 129,271 individual “snapshots” of ETCs. Each snapshot is a 200×480 grid containing the three atmospheric variables, offering a physically consistent and spatially coherent view of cyclone dynamics. These fields were specifically prepared to train a Progressive Growing Generative Adversarial Network (PG-GAN), aimed at learning to reproduce the structure and variability of ETCs. Although the dataset does not include the synthetic fields generated by the GAN, it provides the original, high-quality ERA5 inputs used in training. It can support a variety of tasks in machine learning and atmospheric science, including storm classification, intensity estimation, and the study of ETC-related impacts.
Crawford, A. D., D. Barber, D. Dahl-Jensen, J. C. Stroeve, M. C. Serreze, and N. Sommer, 2021: Northern hemisphere extratropical cyclone tracks. Dataset Version 12.2, Centre for Earth Observation Science - University of Manitoba. https://doi.org/10.34992/ebnw-s681.