Historical Severe Weather Events

Guidance on interpreting these data: Storm Events records reports, not observations from a fixed network. National counts rose roughly thirtyfold between 1955 and the late 1990s as population, spotter networks, radar and NWS verification practice changed, so trends spanning that era partly measure the growth of reporting. Trends from about 2000 on are far more defensible, and the higher thresholds are much less sensitive to reporting practice than “any report”. Derechoes are single storms, taken from the archive of Squitieri, Wade and Jirak (2026), the first comprehensive US derecho archive: 184 events from 1956 to 2025. The biggest-days lists are Storm Events throughout. June and July 1993 are missing from the Storm Events database entirely — the NWS offices' records for those two months were lost when NCEI migrated the archive — so 1993 is dropped from maps and annual plots rather than counted as a quiet year.

Loading…

drag · scroll to zoom

Methods

Source. NOAA/NCEI Storm Events Database, county-type records (CZ_TYPE = C), plus the SPC derecho archive of Squitieri, Wade and Jirak (2026) for the derecho swaths.
What ranks a day. The number of distinct counties in the chosen place that reported the hazard at or above the selected threshold on that calendar date. Counties rather than report counts on purpose: national report totals rose roughly thirtyfold between 1955 and the late 1990s as population, spotter networks and radar changed, so ranking by reports would largely rank by era and by population. A county either saw it or it did not, which travels far better across seventy years.
How many. The top 50 days for each state and the top 100 nationally, computed separately for every threshold — the biggest ≥ 2″ hail days are a very different list from the biggest “any hail” days. States with fewer qualifying days show all they have.
The day view. Counties shades each county by how many qualifying reports it logged. Reports draws every individual report as a point, coloured by magnitude; that detail is a separate download and is fetched the first time you ask for it.
Any date. The ranked lists are a way in, not a limit: the Date box takes any YYYYMMDD and maps that day, whether or not it ever made a top 50. Every date in the record is carried — well over ten thousand per hazard — against roughly 3,000 ranked ones. A date inside the record with nothing to report says so rather than drawing a blank map, and so does one outside it.
Tornado rows are county segments, not tornadoes. Storm Events splits a tornado at every county line, so a three-county tornado is three rows: 3 April 1974 holds 239 segments for 148 tornadoes, and 27 April 2011 holds 329 for about 200. The Counties column and the ranking are unaffected — each of those counties genuinely had a tornado, which is precisely why counties are the ranking unit — but the Segments column counts rows, not storms. They cannot be collapsed back into tornadoes across the record: TOR_OTHER_CZ_FIPS, which links continuation segments, is empty for every row before about 1990 and populated on only 13% of rows since 2010. Hail and wind rows are individual reports and carry no equivalent problem.
Ties are broken by date, earliest first.
Reporting practice dominates long comparisons. A 1960s day and a 2020s day are not measured the same way. Within the last two decades or so the lists are much more comparable. The higher thresholds are far less sensitive to this than “any report”.
June and July 1993 are missing from Storm Events entirely — NCEI lost those records — so no day in those two months can appear in any list.