Browse individual events
Filter by hazard, year, morphology, state, or intensity, then inspect images as a grid, table, or map.
Open explorer →SeVere Reflectivity IMaGe Dataset
Storm-centered radar imagery for research and machine learning.
SVRCLIP pairs 136 × 136 radar reflectivity sequences with nine frames spanning −60 to +60 minutes around each report with severe-weather reports across the contiguous United States.
Use the dataset
One portal, three primary workflows.
The redesigned interface separates visual discovery, bulk access, and scientific documentation instead of mixing them into year-specific pages.
Get the underlying files
Browse the existing tarballs and CSV products directly from the data symlink, with file size and modification metadata.
Browse downloads →Use labels and Python tools
Review morphology classes, documented dataset splits, and links to the Python interface and example workflows.
Labels & ML →Dataset at a glance
Designed for repeatable analysis.
Method
Severe reports linked to gridded radar analyses.
SVRIMG centers radar imagery on severe-weather report locations and uses the Iowa State IEM 15-minute base reflectivity nearest the report time. Base reflectivity is regridded by nearest neighbor to the WRF-BCC 3.75-km Lambert conformal grid and extracted into consistent 136 × 136 images.
Read documentationFor developers
Bulk files and a Python interface remain first-class access paths.
The web explorer is intended for discovery and inspection. Reproducible analyses should use the downloadable metadata and imagery or the SVRIMG Python package.
View GitHub repository ↗