Published June 5, 2026 | Version v1

The ASCAT U-Net cold-pool dataset.

  • 1. ROR icon University of Alabama in Huntsville

Description

The ASCAT U-Net Cold-Pool Dataset

This dataset provides a global observational record of oceanic convective cold pools detected from Advanced Scatterometer (ASCAT) surface wind observations using a U-Net deep learning framework. Cold pools are density currents generated by precipitating convection that play a fundamental role in convective organization, mesoscale convective system evolution, precipitation propagation, and air–sea interactions. Despite their importance, continuous observational datasets of cold pools remain limited, particularly over the global oceans.

The dataset is derived from ASCAT Level-2 ocean surface wind retrievals and applies a U-Net convolutional neural network to identify spatially coherent cold-pool signatures within scatterometer wind fields. Detected cold pools are subsequently converted into georeferenced polygon objects and gridded analysis products suitable for climatological studies, model evaluation, and investigations of convective organization.

This Zenodo deposit contains two complementary products:

1. GeoPackage Object Dataset

Monthly GeoPackage files containing vector representations of individual cold-pool objects detected from ASCAT observations. Each feature includes geometric, morphological, and quality-control information such as:

  • Cold-pool polygon geometry (MultiPolygon)

  • Observation time

  • Feature area

  • Centroid location

  • Orbit and swath identifiers

  • Dateline-crossing flags

  • Swath-edge quality flags

  • Connected-component identifiers

The GeoPackage product preserves the object-based nature of cold pools and is intended for feature tracking, object statistics, and event-based analyses.

2. Gridded NetCDF Dataset

Daily global gridded cold-pool fields provided on a 0.1° × 0.1° latitude–longitude grid. The gridded product includes sampling-aware information that distinguishes:

  • No ASCAT observation

  • Observed but no detected cold pool

  • Detected cold-pool feature

Each detected feature is linked to the object-based dataset through a unique feature identifier and accompanying lookup tables.

Spatial and Temporal Coverage

Current release:

  • ASCAT-A: January 2019 – December 2020

  • ASCAT-B: December 2019

  • ASCAT-C: December 2019

Geographic coverage is global over the oceanic ASCAT observational domain.

Applications

Potential applications include:

  • Tropical convection and mesoscale convective system studies

  • Cold-pool climatologies

  • Air–sea interaction research

  • Convective organization analyses

  • Numerical weather prediction evaluation

  • Earth system model evaluation

  • Satellite-based precipitation and convection studies

Related Publication

Garg, P., Nesbitt, S. W., Lang, T. J., Priftis, G., Chronis, T., Thayer, J. D., & Hence, D. A. (2020). Identifying and characterizing tropical oceanic mesoscale cold pools using spaceborne scatterometer winds. Journal of Geophysical Research: Atmospheres, 125(5), e2019JD031812.

Citation

Priftis, G. (2026). The ASCAT U-Net cold-pool dataset. Zenodo. https://doi.org/10.5281/zenodo.20562502

Files

ascat_unet_coldpools_dataset.README.txt

Files (123.1 MB)

Name Size
md5:3f780a62d70ba4203841b1bc4568b1b0
2.5 MB Preview Download
md5:796c5f3443cf0534f2311ef8f98a1b38
120.6 MB Preview Download
md5:49ffafea61def128c842c986fceb8bf6
8.5 kB Preview Download
md5:11e297f4334caf78e67aae680bb2aad2
11.8 kB Preview Download

Additional details

Dates

Created
2025-12-08
Initial release of the ASCAT U-Net Cold-Pool Dataset.
Issued
2026-06-05
Public release of the ASCAT U-Net Cold-Pool Dataset through Zenodo.