Methodology and limits

Source

Every number on this site is derived from the NOAA Storm Events Database, published by the National Centers for Environmental Information (NCEI). The database compiles reports from National Weather Service offices, trained spotters, emergency managers, law enforcement, storm chasers, media and the public. It is a US Government work and is in the public domain.

Source files: ncei.noaa.gov/pub/data/swdi/stormevents/csvfiles. This aggregate covers 2016-2025 and was retrieved on 2026-09-17.

How the aggregate is built

1. Each yearly file is downloaded and every row with an event type of Hail is kept.
2. Events are grouped by state and county FIPS code. 87,696 hail reports across 3,026 counties are represented.
3. For each county we count reports, record the largest reported size, count reports at or above 2 inches and at or above 4 inches, and note the first and most recent dates in the window.
4. Individual large reports (4 inches and above) are kept with their date and nearest place name, capped at 25 per county: the largest counties have far more than that.
5. The largest reported size is stored in hundredths of an inch as an integer, so values survive serialisation without floating-point drift.

The pipeline is a single script in the site repository and is re-run by hand when NOAA publishes new yearly files, usually in the first half of the following year. Re-running it is what updates this site; there is no automatic process, and no request to NOAA happens when you load a page.

Four ways a county hail count misleads

1. Reports need people to file them

A storm over a city generates dozens of reports; the same storm over ranch land may generate none. This is the single largest bias in the dataset and it is not correctable from the data alone.

2. Hail swaths are narrow

A severe hailstorm typically damages a swath a few miles wide and tens of miles long. A county can be 900 square miles, so "hail in this county" can mean hail 40 miles from your house.

3. Sizes are mostly estimates

Most reports compare a stone to a coin, ball or fruit. The database converts those to decimal inches using conventional values, which is why 1.75 inches appears so often and why the decimals imply more precision than exists.

4. Reporting practice has changed

Observation networks, spotter training and reporting technology have all changed since 1950, and they continue to change. That is why this site uses a ten-year window rather than the full record: all years in it are broadly comparable, and the trend line compares like with like.

What this data is good for

• Establishing whether large hail is plausible where you live, and how recently it happened.
• Comparing regions over the same period to see where hail is concentrated.
• Sizing a hail risk before choosing roofing material or buying a home.
• Public-interest research and journalism that cites the source and its limits.

What it is not good for

• Judging whether a specific property was damaged in a specific storm. That requires the storm date and an inspection.
• Underwriting, actuarial pricing, or any decision where a mis-stated probability has financial consequences.
• Comparing counties with very different populations on raw counts.

If you need storm-specific evidence, for an insurance claim or a dispute with a contractor, the useful sources are the National Weather Service's local storm reports and the Storm Prediction Center's archived severe-weather reports, which carry times and locations that this county aggregate deliberately discards.

Corrections

NOAA revises its yearly files, and a revision can change counts slightly. This site records the date the aggregate was built on every page so a discrepancy can be traced to a particular version. If you spot an error, check the source file for that year first; if the source disagrees with what is shown here, the aggregation is at fault and should be rebuilt.