Storm Stream.

science

How NWS and SPC Hail Reports Get Made

Trace one hail report from a spotter's call to the NCEI archive, and read the biases the agencies that publish it document themselves.

Every hail size in the federal record started as somebody's comparison. Local National Weather Service offices build their storm reports from damage surveys, emergency managers, and SKYWARN spotters. There is no dense national network of hail gauges feeding the database. A person saw the stones, a person called it in, and a person at a forecast office decided what to write down. If you argue about hail data for a living, that chain is the thing worth knowing in detail.

The stone and the coin

Because the report comes from a person, the size is almost always a comparison rather than a measurement. NWS spotter guidance publishes the pairings forecasters expect to hear, and the pairings are ordinary objects:

Those are the pairings on the NWS hail and wind spotter page, which also sets the severe thunderstorm criteria: hail 1 inch in diameter, quarter size, or larger, and winds 58 mph or greater, or trees and power lines blown down. The Storm Prediction Center uses the same line for hail and adds a second one above it: severe hail is 1 inch or larger and significant severe hail is 2 inches or larger.

A quarter and a nickel are one eighth of an inch apart. One is severe and one is not. That is the resolution the ground truth is collected at, and no amount of downstream processing adds precision that the original call did not have.

The SPC day starts at 1200 UTC

The daily storm report pages at SPC are drawn from NWS Local Storm Reports, usually sent in near real time. The same page explains the clock, and the clock is where most date arguments come from. Reports are grouped on a 1200 UTC to 1159 UTC day. Today's page updates every 10 minutes, starting at 6 AM CST or 7 AM CDT, and a separate page covers just the last three hours. Yesterday's reports cover 6 AM to 6 AM local time.

Read that last detail twice. The local day boundary on the archive is 6 AM, not midnight. Hail that falls in the small hours of the morning is filed under the previous calendar date. A homeowner who says the storm was on a Tuesday and a report page that files it on Monday can both be correct. Archived reports can be pulled by date, and SPC offers KML files and a REST service for them.

Preliminary is a status, not a disclaimer

SPC labels the daily reports preliminary and presents them as is. The reports index says it in capital letters: All Reports Are Considered Preliminary. The severe weather database carries the same caveat for the current year's reports.

Two other limits on the real-time feed are worth holding onto. SPC does not decode non-thunderstorm Local Storm Reports, such as hurricane-related wind, so a tropical system can generate a pile of local reports that never appear in the severe weather pages. And for post-storm summaries, SPC points users away from its own daily pages and toward NWS Storm Data through the NCEI Storm Events Database.

Months later, the same event comes back

The second life of a hail report is the archive. The Storm Data publication and the reports in the Storm Events Database are available 90 to 120 days after the event. SPC's severe weather database is itself compiled from NWS Storm Data, with tornado reports going back to 1950 and hail and damaging wind reports back to 1955. Storm Events covers tornadoes since 1950, severe thunderstorms since 1955, and all other weather events since 1996.

The preliminary record and the archived record are not guaranteed to match. The first version is what came in while the storm was happening. The second is what survived review at the local office. If you built a position on a preliminary count, pull the archived version before the other side does, and know which one you are quoting.

One procedural point that matters when a file is going to be contested: certified, official copies must be requested from NCEI, not from local NWS offices. A screenshot of a report page is not the same artifact as a certified copy.

The biases are published by the agency, not discovered by critics

You do not have to infer the weaknesses in this data. SPC lists them on the severe weather database page. Reports cluster around population centers. Wind reports are mostly verified by falling trees rather than measured gusts, so they tend to cluster where trees are more common. The data are used for NWS warning verification and may not capture all storm events. And monetary loss figures in the database are described as highly suspect.

The wind entry is the most instructive, because it shows what verification often means in practice: a tree went over, so the gust must have met criteria. That is a reasonable operational inference and a poor measurement. Hail carries the same structure. The report exists because the storm met a human being.

The size distribution is also bottom-heavy. NWS Burlington notes that most local hail reports there are 3/4 inch or smaller. That is a local statement about one county warning area, not a national figure, but it fits the shape of the problem: small hail is common, and the reports that matter commercially sit in the tail.

Then there is the line that should end most arguments about empty map space. If an event is missing from both Storm Data and the Storm Events Database, it was not reported to the National Weather Service. The sources say nothing about whether it happened.

The 2010 threshold change

In 2010 the NWS raised the severe hail threshold from 0.75 inch to 1.00 inch. For legacy purposes, 0.75 inch reports remain in the annual hail files, so users may need to filter.

That one sentence invalidates a lot of casual analysis. Count severe hail reports straight out of the annual files across the full period of record and you are mixing two definitions of severe, with the older and looser one applying to the earlier decades. Any trend you compute without filtering is partly an artifact of the definition change rather than the weather. If somebody hands you a chart of rising hail frequency since the 1950s, the first question is whether they filtered.

What a report is actually good for

A hail report is evidence that someone was present, saw hail, and judged its size against a familiar object. It is strong evidence that a storm produced hail near that point at roughly that time. It is weak evidence about size, because the size came from a comparison. And it is not evidence at all about the storm's full footprint, because the map of reports is also a map of where people were.

Three habits follow. Check the date convention before you argue about the date, since the archive day starts at 6 AM local and groups on 1200 UTC. Treat the preliminary feed as a lead and the archived Storm Data record as the citation, and expect them to differ. And never treat a gap in the report map as a finding, in either direction: a cluster of reports does not prove a roof was hit, and an empty block does not prove it was spared.

Sources

More on this