Methodology & sources

Every number on National Storms comes from public US government data. This page lists each source, how we turn it into storm-day and place pages, and where the data falls short.

Sources

Storm reports, 1955–2025: NOAA Storm Prediction Center severe weather database, SVRGIS: hail (inches), damaging wind (knots, converted to mph for display) and tornadoes (one row per state segment, with start and end points, path length and EF/F rating).
Storm reports, 2026: the Storm Prediction Center's daily filtered storm reports. These are preliminary; the counts change when NOAA finalizes the year.
Radar hail estimates, October 2019 onward: NOAA Multi-Radar Multi-Sensor (MRMS) Maximum Estimated Size of Hail (MESH), 24-hour maximum on a 0.01° grid (about 1 km), from the Iowa Environmental Mesonet MRMS archive. We keep cells of 0.75″ or larger over US land and draw them as the colored swath on storm-day maps. A few days are missing from the archive; for five days the nearest file is up to 2.5 hours earlier than 12:00 UTC.
Places and population: US Census Bureau 2024 Gazetteer files (counties and incorporated places), 2024 population estimates, and 2020 block-group centers of population for “people in the swath”.
Homes: total housing units (table H1) for each of the country's 242,000 block groups from the US Census Bureau's 2020 Census Redistricting Data (P.L. 94-171) Summary File, 140.5 million homes in all. A housing unit is a house, apartment, mobile home or other separate living quarters, occupied or vacant.
Hail size names (quarter, golf ball, baseball) follow the Storm Prediction Center's hail size table.

The radar swaths and counts are our own processing of NOAA data. NOAA has not reviewed or endorsed them.

How the pages are built

Storm-day pages list every report on that storm day, the states and cities with the largest hail, and (from October 2019) the radar-estimated hail swath, its area and the number of homes and people inside it.
City pages count storm days with a hail report within 5 miles of the city's Census center point. They also show the largest radar estimate near the city center.
County and state pages count storm days with at least one hail report inside the county or state boundary.
Thin pages are kept out of search engines: storm days with fewer than five hail reports and no tornado or sizable radar swath, and places with fewer than five hail days, are marked noindex.

Definitions

Storm day: 12:00 UTC to 12:00 UTC the next day (7 a.m. to 7 a.m. Eastern Standard Time), the Storm Prediction Center's convention, so an overnight storm stays on one day.
Hail day for a city: a storm day with at least one hail report within 5 miles of the city's Census center point. For a county or state: at least one hail report inside it.
Radar max over a city: the largest MESH value within about 2 miles (a 7 × 9 cell window) of the city center.
Radar quality control: MESH sometimes comes through corrupted, with patches of 7–55″ “hail” far larger than the 8″ US record. We drop a whole radar day when its typical (median) cell is 2″ or more, which happened on 26 days since October 2019 (among them July 15, 2024 across the Northeast and several winter days), and on other days we drop every cell over 6″ along with cells within about 2 miles of one. This removes about 0.5% of 1″+ cells. Storm reports for those days are unaffected.
Radar clutter: a few spots near radar sites, wind farms and coastlines show 1″ MESH on many days with no storm (near Albany, NY it was 185 days from 2020 on). We drop on every day the cells that had 1″ MESH on at least 15 days from 2020 on and at least three times as often as the land within about 25 miles, plus a 2-cell edge: about 2,600 of the 8.2 million radar cells over US land. The busiest real hail spots in the Plains run about twice their surroundings, so they are kept.
Data desk studies (/data/) also screen radar by reports: a county's 1″ radar on a storm day counts only if its state logged at least one hail report (any size) that day. That removes about 11% of 1″+ radar area, mostly leftover clutter and sea echo, and almost none in the Hail Belt. Each study's methodology box gives its own rules.
Area, homes and people in the swath: grid cells with MESH of 1″ or more, and the 2020 housing units and population of block groups whose center of population falls in those cells. Each block group counts whole or not at all, so the totals are estimates: a block group whose center is just outside the swath counts zero even if part of it was inside. Shown as “about 41,000 homes”.
Report locations: 2026 reports carry the place the reporter gave (“3 miles WSW of Aydelotte, OK”). Reports from the historical database carry a county, so they read “in Montague County, TX”.
Counties: today's counties and county-equivalents (2024 Census): Connecticut's nine planning regions, Alaska's boroughs and census areas, Louisiana's parishes. Older reports filed under counties that no longer exist (Connecticut's eight former counties, Shannon County SD, Dade County FL) are placed in today's county by their location.
Ranks compare hail days from 2016 to 2025 among places in the same state; city pages also give the rank among all US cities.
ZIP code pages count hail reports within 5 miles of the Census center point of the ZIP Code Tabulation Area, the same radius as city pages. They are listed in search engines only when the ZIP code has at least 5 hail days and its hail days differ from its city page's by at least a quarter (or it has no city page); the rest stay out of search results.
Hail season pages count hail days (storm days with at least one hail report in the state) by month for 1955–2025 and 2016–2025, find the 7-day stretches of the year with hail on the largest share of days (2006–2025), and give the median first and last hail report of the year. States with fewer than 250 hail days since 1955 or 50 from 2016 to 2025 get a page but are kept out of search engines.
Downloads: every report is available as CSV by state, year and type at storm report downloads. City and ZIP code pages offer embeddable hail record and hail season cards for other sites.
How often it hails near a city: distinct storm days with a hail report of 1″ or larger within 5 miles of the city center from 2006 to 2025, divided by 20 years, written as “about once every 1.4 years”.

Limits

Reports depend on someone seeing and reporting the storm, so rural areas, nights and earlier decades are undercounted. Report counts rose sharply after the 1990s as spotter networks grew.
Report locations are rounded by the reporter, often to the nearest town, so a report “near” a city may be several miles from any given street.
Radar MESH estimates the largest hailstone aloft. It can be off by an inch or more in either direction and does not confirm hail reached the ground at a given spot.
Radar hail estimates cover the lower 48 states and DC only. Alaska and Hawaii pages have storm reports but no radar swath, homes or people counts.
Neither source is a damage assessment of a specific property. See the disclaimer.

Guides

For reporters: data for journalists.

Updates

Data through November 6, 2026. New storm days are added daily from the Storm Prediction Center and MRMS feeds; the previous year's reports are replaced when NOAA publishes the final database.