The hailstorm is now being asked to file its own damage note
the hailstorm is now being asked to file its own damage note: a specialized earth 2 model gives hyper local hail forecasts two to three hours ahead across colorado and wyoming, a ucar system turns ai weather models into week ahead tornado and large hail hazard outlooks for forecasters, and an automated camera measures hailstone size distributions every minute.
the stone now files its own impact report.
Context
Colorado State University trained a specialized version of NVIDIA's Earth-2 StormScope nowcasting model for severe hailstorms, using NEXRAD composite radar and the team's own CHIVO radar. According to NVIDIA's case study it was deployed over northern Colorado and southern Wyoming and produces skillful ensemble forecasts with a 2 to 3 hour lead time. The case study says earlier nowcasting methods worked on a 30 to 60 minute scale.
NSF NCAR's Medium-Range, Real-Time Convective Hazard Forecasts (NCAR and UCAR news, May 4, 2026) post-process output from AI weather models to give a percent likelihood of tornadoes, large hail and damaging winds up to a week ahead. The forecasts run daily and were to be evaluated in NOAA's Hazardous Weather Testbed Spring Experiment. A paper describing the system appeared in Artificial Intelligence for the Earth Systems.
HailCam (Atmospheric Measurement Techniques, 2026) is an imaging instrument that uses a ConvNeXt-Tiny backbone with Mask R-CNN to count and size hailstones at one minute intervals over a 60 by 60 cm area. Lab tests with synthetic ice spheres of 5 to 45 mm gave 91% sizing accuracy within 5% relative error. A field comparison with an OTT Parsivel 2 disdrometer on one hail event, 9 May 2025, gave statistically indistinguishable size distributions.
The note's 'hyper local' is NVIDIA's own case study wording for a regional deployment, so it is a vendor description with the lead time as the headline figure, not a verification score in the page. The NCAR system is a medium-range hazard outlook for forecasters, a probability of a hazard across a day, which is a different product from the 2 to 3 hour storm-scale nowcast. HailCam is an observing instrument, not a forecast, and its field result comes from a single event. The note calls it a camera that measures stone size every minute, which matches the paper, but the 91% sizing accuracy is a lab figure.
Related work
- Colorado State University Builds Real-Time Hailstorm Forecasting with NVIDIA Earth-2 (NVIDIA case study) ↗Vendor case study for the Colorado and Wyoming hail nowcast and its 2 to 3 hour lead.
- Identifying severe weather hazards further in the future with AI (NCAR and UCAR News) ↗Source for the week ahead AI hazard outlooks and their testbed evaluation.
- HailCam: an automated imaging system for real-time measurement of hail size distributions and fall rates (AMT) ↗Source for the one minute hail size measurements and lab and field validation.
Watch next
- Look for verification scores for the CSU StormScope hail forecasts beyond the case study, and for evaluation results from the NCAR Spring Experiment. Check HailCam against more than one hail event.
Sources
- Colorado State University Builds Real-Time Hailstorm Forecasting with NVIDIA Earth-2 (NVIDIA case study)nvidia.com
- Identifying severe weather hazards further in the future with AI (NCAR and UCAR News)news.ucar.edu
- HailCam: an automated imaging system for real-time measurement of hail size distributions and fall rates (AMT)amt.copernicus.org
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 4 October 2026 at 02:37 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →