Hail on Date

Where each number in a Hail on Date report comes from

A report is built at the moment you order it, from NOAA's own records, for one address and one local date. This page explains every step, what each record can and cannot show, and how to check a row yourself.

The address

The address is matched with the US Census Bureau geocoder, which returns a point and the county. If Census cannot match it, the report uses OpenStreetMap's Photon geocoder instead. The report prints the matched address and its coordinates so you can confirm the point is your property. Every distance is measured from that point.

The date and the time zone

NOAA stores every record in UTC. Your date is a local calendar day, so the report converts it using the time zone the National Weather Service assigns to your county (from the NWS county file effective April 16, 2026), with the daylight saving rules that applied on the date. Every row shows local time and UTC. A local day in the continental US runs across two UTC days: May 27, 2024 in Texas is 05:00 UTC on May 27 to 05:00 UTC on May 28, which is why evening storms carry the next day's UTC date.

The report covers your date and the three days either side, seven local days in all, so a storm dated a day off in a claim still shows up.

Radar hail signatures

NOAA's network of NEXRAD (WSR-88D) radars runs the Hail Detection Algorithm on every volume scan, about every 4 to 6 minutes when storms are around. For each storm cell it writes three estimates, which NOAA's National Centers for Environmental Information publish in the Severe Weather Data Inventory:

  • Maximum expected hail size (field MAXSIZE), in inches.
  • Probability of hail 3/4 inch or larger (field SEVPROB, "probability of severe hail"). The algorithm was designed when 3/4 inch was the severe threshold; the National Weather Service now uses 1 inch, and NOAA's Radar Operations Center treats a 70% value as roughly a 50% chance of 1-inch hail.
  • Probability of hail of any size (field PROB), from 10% to 100%.

NCEI publishes two versions. nx3hail_all holds every cell the algorithm rated, from a 10% to a 100% probability of hail, plus rows marked -999 where it produced no estimate. nx3hail is a filtered subset that keeps only cells rated 100%. The report's week tables, map and counts, the free check and the radar figures on the storm-date pages use nx3hail_all, skipping the -999 rows, and every row shows both probabilities. For the sample address on May 27, 2024, that is 26 signatures within 5 miles, 9 of them rated 100%. The ten-year record uses the filtered set, so it lists the days a cell within 3 miles was rated 100%.

Most radar IDs are NEXRAD sites. A few, such as KDAL and KDFW near Dallas, are airport Terminal Doppler Weather Radars whose hail estimates NOAA's archive also carries.

Each row is placed at the storm cell's position at that scan. The report lists every row within 1, 3 and 5 miles of the address, counted by great-circle distance. Several radars often scan the same storm, and each writes its own rows with its own cell ID, so one storm usually appears as a run of rows from two or three radars.

What a signature cannot tell you: it is an estimate made from radar reflectivity, high in the storm, and the stones that reach the ground can be smaller or larger than the estimate. Hail also falls outside the plotted points, which mark only the cell's core at each scan. A signature shows where the algorithm placed a storm cell it rated as producing hail at that time; it is not a measurement at your roof.

Spotter reports

The Storm Prediction Center publishes a daily log of every hail report the National Weather Service receives from trained spotters, emergency managers, law enforcement, the public and apps such as mPING. Each daily file covers 12:00 UTC to 12:00 UTC the next day and gives the time, the size in hundredths of an inch, a place described as miles and direction from a town ("3 NE Arlington"), the county, coordinates to two decimal places (0.01 degree, about 0.7 mile) and the NWS comment. The report lists every one within 10 miles of the address during the seven days, with the comment exactly as published.

These logs are preliminary. NOAA later reviews them for its Storm Data publication and can correct or drop reports. A storm nobody saw or reported is not in them, which is common at night and in rural areas.

The ten-year record

The report lists every local day on which the algorithm rated a storm cell within 3 miles of the address a 100% probability of hail, from ten years before your date to the latest day NOAA has published, with the count within 1 and 3 miles, the largest expected size, the highest probability of 3/4 inch hail and the nearest distance. It answers a question adjusters ask: were there earlier storms that could explain the damage? It uses the 100%-only set because it marks the stronger storms and keeps the record to a size a person can read; the days are found with SWDI's daily tile summaries and then checked row by row.

Gaps in NOAA's radar archive

NOAA's radar hail archive has holes. For example, on September 28, 2026 it held no hail signatures anywhere in the country for April 19 and April 20, 2023 (UTC), although the Storm Prediction Center logged dozens of hail reports around Oklahoma City, Omaha and Chicago on those days. A report built on a day like that would wrongly show no radar hail, so before you pay we check NOAA's national count for the UTC days your date covers. If either is empty, we do not sell a report for that date and say why. Empty days elsewhere in the week are named in the report. Some holes are filled later, so the check is made live each time.

How anyone can check a row

A radar row's reference is the radar ID, the cell ID and the UTC scan time, for example KGRK X1 2024-05-28T00:31:28Z from the sample report. The report prints the SWDI query that returns every row near the address for your date; it looks like this, and it opens in a browser as plain text:

https://www.ncei.noaa.gov/swdiws/csv/nx3hail_all/2024052705:2024052805?bbox=-97.193,32.663,-97.018,32.810

A spotter row's reference is its daily file and UTC time, for example 240527_rpts_hail.csv 0038Z, from:

https://www.spc.noaa.gov/climo/reports/240527_rpts_hail.csv

What the report is not

It is a compilation of public NOAA records for a place and time. It is not a roof inspection, an engineering opinion, an adjuster's estimate or advice about an insurance claim, and it cannot say whether hail damaged a particular roof. It does not predict what an insurer will decide.

How the storm-date and metro pages are built

Each metro is a fixed set of whole counties, listed on its page. A spotter report counts for a metro when its coordinates fall inside one of those counties (county outlines from the NWS county file). Reports are grouped by the metro's local calendar date. A metro has a page when its counties logged at least 150 hail reports from January 1, 2016 onward; a storm date has a page when a metro logged 5 or more reports on one local date from January 1, 2021 onward. On each storm-date page, the radar figures count the signatures in nx3hail_all for that local day (every rated cell, 10% to 100%) whose positions fall inside the same counties, and say how many were rated 100%.

The pages were built from Storm Prediction Center logs through September 27, 2026 and radar data queried from NOAA the same week. They are rebuilt as new storms come in.

Sources

NOAA data are US Government works in the public domain. NOAA does not endorse this site.