Storm Report science
Where every line of a Hail Donkey™ Storm Report comes from.
A Storm Report lists the hail and the 55 mph or stronger wind on record for one address, with the date, the time, the size or speed, and the source that confirmed it. This page explains the physics and the data behind those lines, and where each can be wrong.
The short version
- Primary sources only. Every row comes from the National Weather Service, NOAA's National Centers for Environmental Information, the Storm Prediction Center or the NEXRAD radar network. Nothing is filled in from a weather-model guess.
- Two different kinds of evidence, never blurred. A person or instrument on the ground reported it, or radar estimated it. The report says which, and radar sizes are always marked as estimates.
- Units and clocks are corrected, not copied. Wind is converted to miles per hour from the knots the federal sources use; every time is stored in UTC and shown in the property's local time, daylight saving included.
- One storm, one row, every source named. When several sources describe the same storm they are reconciled into a single entry that names all of them.
- It says what it does not know. A time that no source recorded is shown as not recorded. A report is a record of weather, not a statement about damage.
1. Two kinds of evidence
Hail and damaging wind are short, narrow and local. No single instrument sees all of it, so a defensible record uses two independent kinds of evidence and keeps them apart.
| Ground reports | Radar estimates | |
|---|---|---|
| What it is | A trained spotter, emergency manager, law enforcement officer, member of the public or weather station reports what happened at a place and time. | The WSR-88D (NEXRAD) radar network scans the storm every few minutes and algorithms infer hail from what the beam sees. |
| Strength | It is an observation of the thing itself: a stone measured against a coin, a gust read from an anemometer. | Continuous in time and space, including places nobody was looking. |
| Weakness | It exists only where someone was present and chose to report. Estimated wind speeds run high. | It is an inference. Beam height, storm structure and the freezing level all shape the answer. |
Ground reports reach us through three federal channels with different review levels: National Weather Service Local Storm Reports, issued by the forecast office as the storm happens; the Storm Prediction Center's national report archive; and the NCEI Storm Events Database, which is reviewed and published after the fact. A storm that appears in more than one is stronger evidence, and the report shows every source that carried it.
2. How radar tells hail size
Hail is large, wet and bounces back far more energy than rain, so intense reflectivity high in a storm, above the level where the air drops below freezing, is the signature. The published method that the National Weather Service uses turns that signature into a number in two steps (Witt et al., 1998).
First, reflectivity measured above the freezing level is weighted by how much hail kinetic energy it implies and by how cold the air is at that height, then summed up the column. The result is the Severe Hail Index (SHI). Second, SHI maps to a size and a probability:
MESH (mm) = 2.54 × SHI0.5
POSH (%) = 29 × ln(SHI / WT) + 50, with WT = 57.5 × H0 − 121 (WT is never less than 20)
MESH is the Maximum Estimated Size of Hail. POSH is the Probability of Severe Hail. H0 is the height of the freezing level in kilometres, which is why a hail estimate is only as good as the temperature profile it was computed with. The same physics runs in two places in a Storm Report: the radar sites' own Level III hail detection, scanned volume by volume with a time stamp to the second, and NOAA's Multi-Radar Multi-Sensor (MRMS) system, which merges overlapping radars into one three-dimensional grid of about 1 km resolution and tracks the largest estimated hail over each 24 hours.
Because MRMS gives the day's maximum rather than a moment, a hail day found only there is listed with its date and the time marked as not recorded. When a time-stamped scan or report covers the same storm, the time-stamped entry is the one shown.
The swath map in a report draws the radar hail estimates of every storm in the period you chose on one map, keeping the largest size seen at each spot and coloring it by size band, over a topographic basemap from the U.S. Geological Survey. The wind reports marked on it are the ones listed in the report. Radar swath data begins in October 2020, so a longer period shows the storms since then.
3. How wind speed is determined
Federal wind reports are in knots. A Storm Report converts them to miles per hour with the exact factor (1 knot = 1.15078 mph) rather than rounding by eye, and lists the gust that was recorded. A “measured” gust came from an instrument. Where a source did not say, the report says nothing more than the number. The National Weather Service calls a thunderstorm severe at a gust of 50 knots, which is 58 mph; a Storm Report starts a little lower, at 55 mph, so a gust that fell just under the warning line is still on the record.
Estimated gusts are shown as recorded and not adjusted. The research literature documents that human estimates tend to run high and cluster on round numbers, which is the reason the measured flag is kept and the reason a measured gust and an estimated one are not treated as equals.
4. Time and place
Sources disagree about what time means: the start of a storm, the moment a spotter called it in, or the radar scan that detected it. Every time is converted to UTC at the source, kept in UTC, then shown in the local time zone of the property with daylight saving applied for that date. Radar scans carry their own time to the second.
Place is just as uncertain. A ground report is a point that a person described, a storm cell moves while it is scanned, and a hail swath is a band, not a dot. The report therefore never relies on a single point on a track: a cell whose strongest reading was a few miles away may still have had an earlier or later scan directly overhead, and the record reflects that.
5. Reconciling sources into one record
The same storm is routinely described by a spotter, a forecast-office report, the national archive and two or three radars. Left alone that produces five rows for one event. Hail Donkey reconciles them with a proprietary matching process built on the physics above, keeps the strongest credible value, and carries every contributing source forward so you can see which agreed. The matching rules are ours and are not published; the inputs and the outputs are, and you can check any row against the named source.
6. What the report does not claim
- It is not a damage report. It records weather at and around an address. Whether a roof was damaged is a question for an inspection.
- Silence is not proof of calm. Ground reports depend on people being present, so quiet areas and quiet hours are under-reported. Radar fills part of that gap; it does not close it. A comparison of radar-derived hail estimates with national storm reports found the radar estimates can supply useful hail information where population is thin (Wendt & Jirak, 2021).
- Radar has known biases. MESH tends to under-estimate hail from storms that are strongly tilted by wind shear and from left-moving supercells. That is why radar sizes are labelled estimates.
- Recent storms are preliminary. National review of ground reports lags the weather by months, so the newest entries may be refined later. A report is a snapshot of the record on the day it was generated, and carries its own ID and date.
7. Why you can rely on it
- It is built on the same federal sources that forecasters, insurers and researchers cite, and every row names its source so it can be checked.
- It separates observation from inference on every line.
- Conversions, clocks and time zones are applied by rule, not by hand, so the same address and window on the same data always give the same report.
- It declines to guess: a missing time stays missing, an estimate stays an estimate, and a model never stands in for a measurement.
References
- Witt, A., Eilts, M. D., Stumpf, G. J., Johnson, J. T., Mitchell, E. D., & Thomas, K. W. (1998). An enhanced hail detection algorithm for the WSR-88D. Weather and Forecasting, 13, 286–303. Summary: NOAA Warning Decision Training Division, Maximum Estimated Size of Hail; COMET, MESH and POSH algorithms.
- Wendt, N. A., & Jirak, I. L. (2021). An hourly climatology of operational MRMS MESH-diagnosed severe and significant hail with comparisons to Storm Data hail reports. Weather and Forecasting. Storm Prediction Center (PDF).
- NOAA National Centers for Environmental Information, Storm Events Database.
- NOAA Storm Prediction Center, Storm reports.
- National Weather Service, Thunderstorm basics and severe criteria.