How often the radar is right
Radar does not see hailstones on your roof. It estimates the largest stone in a storm from high above. So we checked it against 52,029 reports from storm spotters and weather offices since October 2020, and we publish the results here, good and bad.
Radar and the ground disagree on size
When radar shows big hail, it usually reads high: at 3 inches or more on radar, the typical ground report is 1.75 inches. At small sizes the reverse happens, because people report the biggest stone they find and rarely bother with small hail. That is why each storm shows a "likely on the ground" range next to the radar number.
| Radar estimate | Typical report | Usual range | Reports tested | Inside range |
|---|---|---|---|---|
| 0.5 to 0.75 in | 1 in | 0.75 to 1.5 in | 1,087 | 94% |
| 0.75 to 1 in | 1 in | 0.75 to 1.5 in | 1,889 | 91% |
| 1 to 1.25 in | 1 in | 0.75 to 1.75 in | 2,724 | 93% |
| 1.25 to 1.5 in | 1 in | 0.88 to 1.75 in | 2,312 | 88% |
| 1.5 to 1.75 in | 1 in | 0.88 to 2 in | 2,518 | 90% |
| 1.75 to 2 in | 1.25 in | 1 to 2.25 in | 1,852 | 85% |
| 2 to 2.5 in | 1.5 in | 1 to 2.5 in | 2,637 | 89% |
| 2.5 to 3 in | 1.75 in | 1 to 2.75 in | 1,081 | 92% |
| 3 to above in | 1.75 in | 1 to 2.75 in | 768 | 91% |
We set these ranges using reports up to the end of 2024, then tested them on 16,868 later reports they had never seen. 90% of those fell inside the range. Reports tend to describe the biggest stone someone found, so most stones in a storm are smaller.
Does the radar catch hail people report?
| Reported size | Reports | Any radar hail within 5 km | Within 10 km | Radar 1 inch or more within 5 km |
|---|---|---|---|---|
| 0.75 to under 1 inch | 6,994 | 82% | 88% | 50% |
| 1 to under 1.75 inches | 32,069 | 95% | 97% | 76% |
| 1.75 to under 2.75 inches | 10,722 | 99% | 100% | 92% |
| 2.75 inches or larger | 2,077 | 100% | 100% | 98% |
Radar rarely misses a big storm. It is weaker on small hail. Hail also drifts as it falls, and a report's location can be off by a mile or more, so a storm can land just beside the place the radar marks.
When radar shows hail, did anyone report it?
This is the harder test. We took each US ZIP code center on days radar showed hail there, then asked whether anyone reported hail of three-quarters of an inch or larger within 10 miles that day. These figures use 2025 onward only.
| Radar size at the ZIP center | Place-days | Ground report within 10 miles |
|---|---|---|
| 0.5 to under 0.75 in | 39,597 | 15% |
| 0.75 to under 1 in | 17,583 | 25% |
| 1 to under 1.25 in | 8,025 | 38% |
| 1.25 to under 1.5 in | 3,065 | 49% |
| 1.5 to under 1.75 in | 1,768 | 61% |
| 1.75 to under 2 in | 675 | 67% |
| 2 to under 2.5 in | 573 | 73% |
| 2.5 to under 3 in | 165 | 76% |
| 3 or larger in | 61 | 79% |
A missing report does not mean there was no hail, because storms over empty land go unreported. But it does mean you should treat a radar-only entry as a lead, not a fact. Hailstory marks which storms have ground support, so you can tell them apart.
One caution: some of the checks below use ground reports to remove bad radar patches, so the detection and report figures on this page flatter the radar a little. The size ranges were fitted and tested on separate years.
What we removed
Raw radar contains errors, so every day passes through the same automatic checks before it reaches this site. We removed 480 patches with values above the largest hailstone ever recorded in the US. The worst was a fault on July 15, 2024 that painted about 174,217 square km of false giant hail. We removed 110 patches where almost everything was extreme and no ground report came within 50 km, and 26,735 cold-season patches, October to March, with no ground report within 100 km. We also removed 4,657 patches in 44 small areas where radar keeps showing hail that nobody reports, such as parts of the south Texas coast and a spot near Albany, New York, unless a report came within 50 km. In total that is 6.5% of all radar area at 1 inch or larger. These rules can remove real hail, and some false hail can still get through. Storms with no ground report are marked so you can tell.
What this data cannot tell you
- Whether your roof was damaged. Only an inspection can show that.
- The exact spot. A storm counts if radar shows hail in your 1 km square or one touching it, because hail drifts as it falls.
- The exact time hail fell at your address. Dates are storm days, which run 12:00 to 12:00 UTC, so evening storms carry the date they began.
- Anything before October 14, 2020 from radar, or anywhere outside the lower 48 states.
Do not use Hailstory for insurance claims or legal decisions. Figures last rebuilt 2026-11-06.