Published January 2, 2026
| Version v1
Dataset
Open
Swiss Storm-Object Dataset for ML-Based Hail Detection: Polarimetric Radar Predictors and Crowdsourced Labels (2020–2024)
Contributors
Supervisor (3):
-
1.
University of Bern
-
2.
Federal Office of Meteorology and Climatology MeteoSwiss
Description
This dataset contains a set of object-based predictors derived from operational C-band polarimetric weather radar data in Switzerland, matched with ground-truth hail labels derived from crowdsourced reports. The predictors are calculated for storm cells identified and tracked by the Thunderstorms Radar Tracking (TRT) algorithm. This dataset was generated to support the development of machine learning models for hail detection and hail size estimation.
The dataset covers convective storm activity over Switzerland and a 50-km buffer zone around its borders for the years 2020–2024.
Files
radar_predictors.zip
Files
(3.7 GB)
| Name | Size | Download all |
|---|---|---|
|
md5:d461e6dd7a606a1cd1933da49f774d3a
|
3.7 GB | Preview Download |
Additional details
Funding
- Swiss National Science Foundation
- Seamless coupling of kilometer-resolution weather predictions and climate simulations with hail impact assessments for multiple sectors (scClim) 201792