The Science Behind the Storm Report

A storm report is only as good as the reasoning under it. This explains, in plain terms, what each part of the Azimuth Stormglass Storm Report measures, why it matters, and where it stops. Every method here is published, and every source is named at the end.

1. Where the record comes from

Record	What it tells you	What it cannot
NEXRAD Doppler radar (about 160 National Weather Service radars)	Where precipitation is, how strong the echo is (dBZ), which way it moves, and the shape of what is falling	What reached the ground: the beam reads the storm aloft
NOAA multi-radar hail estimate (MRMS MESH)	An estimate of the largest hail over each square about 0.6 mile across, every two minutes	A measured stone; it is an algorithm estimate
Radar hail signatures and cell positions (NEXRAD Level III, NOAA NCEI)	For each storm cell on each scan: the probability of hail, the probability of severe hail (Section 8) and a maximum expected size; and where the cell was, scan by scan (Section 9)	Where the hail fell; the point is the cell’s centre
Warnings (NWS)	That forecasters expected severe weather over the property, with hail and wind tags	That it happened: a warning is a forecast
Storm reports (NWS) and Storm Events (NCEI)	That someone saw hail of a size, wind damage or a tornado at a time and place; Storm Events is the final, checked record, published months later (Section 11)	Anywhere nobody was looking
Hail pads (CoCoRaHS)	The size of the stones that struck a foam pad at a volunteer’s station, measured	Anywhere without a pad
mPING (NOAA NSSL and the University of Oklahoma)	Hail sizes the public reports from a phone app	Whether a report is right: they are not checked
Airport stations (ASOS and AWOS)	The measured wind and gust at 33 ft (10 m), minute by minute	The wind at the roof, or a downburst that missed the station
Weather balloons	The freezing level and the −20 °C level that day	The profile inside the storm hours later or miles away
NOAA surface analysis (URMA)	An hourly estimate of the gust on a 2.5 km (1.6 mi) grid, blending every station with a weather model; the hours it reached 50 mph (Section 14)	A gust between hours or under one cell
Multi-radar rain (NOAA MRMS) and smoke (NOAA HMS)	Rain at the property each day, estimated; smoke analysts saw over it on satellite images (Section 12)	A gauge reading at the roof; smoke at ground level
Building outlines (FEMA / ORNL USA Structures)	Which building at the pin is the main one (Section 15)	Its height or roof shape
Google roof model and elevation (Google Solar and Elevation APIs)	Each roof plane’s pitch, facing, area and height; the ground around the building (Section 15)	The roof’s condition

2. Radar and dBZ

A radar sends a pulse and listens for the echo. Reflectivity, written in dBZ, is the strength of that echo on a logarithmic scale: every 10 dBZ is ten times the returned power. For drops and small ice that are much smaller than the radar’s 10 cm wavelength, the echo grows with the sixth power of the particle’s diameter, so a few large particles outshine a sky full of small ones. That is why storm cores stand out.

For hail it is not that simple. Once stones are larger than about a centimetre (0.4 in) they no longer scatter the radar in that simple way, and reflectivity climbs much more slowly than size. Wet, melting hail can also read as strong as heavy rain. So reflectivity alone cannot size hail: it says a core is intense; the height of the core and the shape of what is falling say whether it is hail.

Reflectivity	What it usually means
below 20 dBZ	Light precipitation, or nothing reaching the ground
30 to 40 dBZ	Moderate rain
45 to 50 dBZ	Heavy rain
50 to 55 dBZ	Very heavy rain; small hail possible in the core
55 to 60 dBZ	Hail likely in a tall storm’s core
60 dBZ and above	Large hail common in the core

General guidance. Section 6 explains why the same 55 dBZ means hail in one storm and heavy rain in another.

3. Dual polarization: the shape of what falls

Since 2013 every NEXRAD radar sends pulses in two orientations, horizontal and vertical, and compares the echoes.

• Differential reflectivity (ZDR). Falling raindrops flatten, so they return more horizontally: ZDR is positive, typically above 1 and up to about 5 dB. Hail returns about the same both ways: ZDR near zero, generally between −1 and +1 dB, higher when the stone is melting.
• Correlation coefficient (CC). Close to 1 in pure rain; it falls, to roughly 0.70–0.95, where rain and hail are mixed.
• The Weather Service’s own hail classes. The National Weather Service runs a hydrometeor classification on every radar, and its Hail Size Discrimination Algorithm sorts hail into small (under 1 in), large (1 to under 2 in) and giant (2 in or more). Stormglass reads these classes over the property.
• Hail differential reflectivity (HDR). A published combination of reflectivity and ZDR (Aydin, Seliga & Balaji, 1986): a positive value signals hail; later studies tied about 21 dB to hail of 0.75 in or more, and values above about 30 dB to larger stones.

4. How hail grows

The embryo. A thunderstorm’s updraft carries small frozen drops and soft ice pellets (graupel) high into the storm, above the freezing level, where the air holds supercooled water: liquid drops colder than 32 °F that freeze the moment they touch something. The stone grows by collecting them. Most growth happens between about −10 and −30 °C (14 to −22 °F).

The updraft holds it up. A stone stays aloft, and keeps growing, only while the updraft is stronger than its fall speed. The stronger the updraft, the larger the hail it can hold:

Hail	Size	Updraft needed to hold it	Speed it falls at
Pea	¼ in	about 24 mph	9 to 25 mph (under 1 in)
Quarter	1 in	about 49 mph	25 to 40 mph (1 to 1¾ in)
Golf ball	1¾ in	about 64 mph
Baseball	2¾ in	about 81 mph	44 to 72 mph (2 to 4 in)
Grapefruit	4 in	about 98 mph

Updraft speeds: NOAA JetStream. Fall speeds: NOAA National Severe Storms Laboratory. Time in the updraft matters as well as speed: a stone must spend minutes in the growth zone to become large.

One pass, not many loops. The old picture of hail cycling up and down through a storm many times to build its layers is not how most hail grows. NOAA’s severe storms laboratory says the layers “do not occur simply due to the hailstone going through up and down cycles,” and recent research finds large hail usually grows on a single pass through the updraft. The layers come from the path: where the air is wetter and the stone cannot freeze the water at once, clear ice forms; where it is colder and drier and the water freezes on contact, trapping air, milky ice forms.

It falls when the updraft lets go. Hail falls when it is too heavy for the updraft, or when it is carried out of the updraft. Larger stones fall closest to the updraft; small ones can be blown well downwind.

Melting on the way down. Below the freezing level the stone melts as it falls. Small hail can melt to rain before it lands; large hail falls fast and survives. A high freezing level — common in Florida and the Gulf states — is why a storm can show hail aloft and drop little of it on the roofs below.

5. How fast hail falls, and how hard it hits

The fall speed. A falling stone speeds up until the drag of the air on it equals its weight. From then on it falls at a steady speed, its terminal velocity. Set the weight equal to the drag and solve for the speed:

Symbol	What it is	Value used
Vt	Terminal fall speed of the stone	metres per second; × 2.237 for mph
D	Diameter of the stone	metres; 1 in = 0.0254 m
g	Acceleration of gravity, a constant	9.81 m/s²
ρh	Density of the hail	900 kg/m³ for solid ice; less for spongy, porous or partly melted stones
ρa	Density of the air	about 1.2 kg/m³ at the ground; about 0.8 at 14,000 ft (600 mb)
CD	Drag coefficient: how hard the air grips the shape	about 0.6 for a smooth ice sphere, about 0.5 for hail in a wind tunnel; 0.8 to 1.0 for natural stones measured by equivalent diameter (Heymsfield et al., 2018)
π · D³ ÷ 6 and π · D² ÷ 4	The volume of a sphere, and the area it presents to the air	geometry; π = 3.1416

Worked: a 1¾ in stone (D = 0.0445 m) of solid ice at the ground, C_D = 0.6: V_t = √(4 × 9.81 × 0.0445 × 900 ÷ (3 × 0.6 × 1.2)) = √727 = 27 m/s = 60 mph. Buoyancy is left out; it changes the answer by less than a tenth of one percent.

Real stones fall slower than an ice ball. Natural hail is lumpy, tumbles, and is often less dense than solid ice. The fit to measured natural hailstones is V_t = 8.4 · D^0.67 for stones over 1.5 cm, and 7.6 · D^0.89 below (V_t in m/s, D in centimetres; Heymsfield et al., 2018, as corrected in 2020). NOAA’s severe storms laboratory gives lower ranges still and notes “much uncertainty”. This page uses the measured fit.

Stone	Mass, solid ice	Ice-sphere formula	Measured fit	NOAA NSSL	Energy, still air
1 in	7.7 g	46–50 mph	35 mph	25 to 40 mph	0.95 J
1¼ in	15 g	51–56 mph	41 mph		2.5 J
1¾ in	41 g	60–66 mph	51 mph		10.8 J
2 in	62 g	64–71 mph	56 mph	44 to 72 mph; over 4 in, possibly over 100	19 J
2¾ in	161 g	76–83 mph	69 mph		77 J
4 in	494 g	91–100 mph	89 mph		390 J

Ice-sphere formula at the ground with C_D 0.6 and 0.5. Energy uses solid-ice mass and the measured fit. 1 joule = 0.738 foot-pound.

The same number is the updraft. Weight equals drag whether the stone falls through still air or the air rises past a stone that hangs still. So V_t is also the updraft needed to hold the stone up, which is where the updraft column in Section 4 comes from (64 mph for a golf ball, against 60–66 here). Aloft the air is thinner, and the stone falls faster in proportion to √(ρ_ground ÷ ρ_aloft). At 14,000 ft that is √(1.2 ÷ 0.8), about 22 percent, so the updraft that grows the stone must be stronger than its speed at the ground suggests.

Impact energy

E is the energy in joules; m the mass in kilograms; V the speed at impact in m/s.

Mass: what radar cannot see. Radar estimates the size of hail; it cannot weigh it. Put the fall speed into the energy, for still air:

Energy grows with the fourth power of the diameter, so size counts most. Density counts with its square: a denser stone is heavier and falls faster. Radar cannot see density at all. A 1¾ in stone of solid ice carries about 10.8 J; the same size at the density of porous ice (0.5) carries about 3.3 J. Measured natural hailstones carry, “on average, lower” energy than ice spheres of the same diameter (Heymsfield, Giammanco & Wright, 2014). Small hail matters too: in tests by IBHS, an institute funded by insurers, weathered shingles struck by hail under 1 in were about ten times more susceptible to later damage from 2 in hail (Meisenzahl, Giammanco & Hedayati, 2025). Two storms with the same radar size can leave very different roofs, which is why the roof is inspected by a trained, licensed professional whatever size the record shows.

Wind adds speed, and tilts the path

V_w is the wind speed; φ is the tilt of the stone’s path from vertical. This takes the stone as moving sideways with the full wind, which is the upper bound: a large stone near the ground may lag the wind. The same resultant is used in the roofing literature (Fedler et al., 2019, after ASTM E822).

A 1¾ in stone falling 51 mph in a 50 mph wind travels at √(51² + 50²) = 71 mph, tilted 44°. Energy goes with the square of speed, so it rises from 10.8 J to 21.1 J: nearly double.

The angle of the roof

θ is the pitch angle (6/12 is 26.6°); i is the angle between the stone’s path and the line square to the roof; v_⊥ is the part of the speed driving into the roof, v_∥ the part sliding along it; E_⊥ is the energy driven into the roof. When φ + θ reaches 90°, the leeward slope is sheltered and not struck.

Only the part square to the roof drives into it; “hail KE must be resolved into direct and shear energy at impact” (Fedler et al., 2019), and “perpendicular impacts impart the most severe impact energy” (Cope & Giammanco, 2022). The sliding part carries the stone off down the slope.

1¾ in stone: joules driven into the roof	Flat	4/12	6/12	8/12	12/12
Still air: 51 mph, 10.8 J, falling straight	10.8	9.7	8.6	7.5	5.4
30 mph wind: 59 mph, 14.5 J, φ 30° (windward / leeward)	10.8	13.9 / 6.3	14.4 / 4.3	14.5 / 2.8	13.6 / 0.9
50 mph wind: 71 mph, 21.1 J, φ 44°	10.8	17.1 / 4.4	19.1 / 2.2	20.4 / 0.9	21.1 / 0
70 mph wind: 87 mph, 31.0 J, φ 54°	10.8	20.6 / 2.9	24.5 / 0.9	27.3 / 0.1	30.3 / sheltered

Three things follow. In still air a steeper roof turns part of the energy aside: a 12/12 slope takes half. In wind, the windward slope takes nearly all of a much larger total, and the leeward slope little. A flat roof takes the same in wind as in still air, because wind adds only sideways speed. That is why hail damage gathers on the slopes that faced the storm, and why the report marks them.

Where the energy goes

Energy is neither created nor destroyed at impact; it is shared. What the stone or its pieces carry away — a bounce, or a spray of fragments — is energy that did not go into the roof. At hail speeds a stone striking a hard surface square on mostly crushes rather than bounces: ice balls fired at a rigid plate at 13 to 105 m/s handed almost all their momentum to it (Chen et al., 2022). Breaking the ice itself takes very little energy, about 0.04 J of a 60 J impact (Combescure et al., 2011, as cited by Pernas-Sánchez et al., 2015). So the visible bounce on a roof is mostly the sliding part, leaving down the slope.

Hardness: how sharply the blow lands. Energy is not the whole story. A hard stone delivers its energy as a sharp blow on a small spot; a soft one crushes as it lands, spreading the blow over a larger patch and a longer moment. Hardness is measured as compressive strength, in megapascals (1 MPa = 145 psi). In IBHS testing, ice spheres stronger than about 1.5 MPa struck with a “hard bounce” or “hard shatter”; spheres under about 0.8 MPa struck in “soft” modes (Brown-Giammanco & Giammanco, 2018). IBHS crushed natural hailstones to measure this (Giammanco et al., 2015), and the comparison with the ice balls used in roof testing found that “generally, iceballs are stronger than natural hail” (Fedler et al., 2019). So mass and strength are the two properties of a stone that decide what it does to a roof, and radar measures neither.

What decides how much the roof takes is what is under it. “The solidly supported areas of roofing tend to be the most resistant to hail damage,” and “any soft layer within the system permits the back of the layer above it to be in tension and fail more easily” (Greenfeld, National Bureau of Standards, 1969). A shingle that bridges a gap or rides on a soft layer bends at the strike, and the energy goes into it there. Ridge and hip caps over a ridge-vent opening, and shingle tab edges, are unsupported in the same way. Temperature works the other way from what many expect: the same study found shingles “more resistant to hail damage at higher temperatures”; cold, stiff shingles are the more fragile, and a fiberglass mat fractured under 1¾ in ice spheres at 40 °F in later testing (Crenshaw & Koontz, 2001). Age and weathering lower resistance further.

Indicators of damaging hail

Indicator	What it does to the roof	Does the record show it?
Size	Energy rises with the fourth power of the diameter: twice the size, 16 times the energy	Estimated: radar and ground reports
Wind	Adds speed, and energy rises with its square; tilts the path onto the slopes that face the storm	Yes: stations, Doppler radar, NOAA surface analysis
Slope and facing	E⊥ = E · cos² i: steep slopes turn energy aside; slopes facing the wind take the most	Yes: the roof planes, measured
Mass (density)	Energy rises with the square of density: solid ice carries about three times the energy of porous ice the same size	No: only stones collected on the ground show it
Hardness (MPa)	A hard stone lands as a sharp blow; a soft one spreads it	No
Support under the roofing	Shingles over a gap or a soft layer, caps over a vent opening and tab edges bend and crack	No: found on the roof
Temperature, age, earlier hail	Cold, stiff and weathered shingles break more easily; repeated small hail weakens them	Partly: the air temperature; the roof’s condition only on inspection

The record supplies size, wind and angle. Mass, hardness, support and condition are found on the roof, by a trained, licensed professional.

6. Altitude: why height decides hail

Because hail grows above the freezing level, how high the strong echo reaches above the freezing level is one of the best radar signs of hail. A 55 dBZ core that stays below the freezing level is usually rain; the same core lofted thousands of feet into the growth zone is the signature of an updraft strong enough to grow hail.

• The echo-top test. When the 45 dBZ echo reaches about 1.4 km (4,600 ft) or more above the freezing level, hail is likely (Waldvogel, Federer & Grimm, 1979). Stormglass reads the 45 and 50 dBZ echo tops from the raw radar over the property and sets them against the freezing level from the nearest weather balloon that day.
• The freezing level. Read from the balloon launch closest in time and place: the 0 °C and −20 °C heights.
• The beam. The radar beam goes out straight while the earth curves away, starting half a degree up, so the farther the property is from the radar, the higher the beam passes over it: about 500 ft at 10 miles, 3,600 ft at 50 miles, and about 9,600 ft at 100 miles (above the radar). Stormglass prints the beam height over the property, because a beam at 9,000 ft sees the hail aloft, not what fell.

Distance from the radar	Lowest beam (0.5°) above the radar	Beam width
10 miles	about 500 ft	0.2 mile
25 miles	about 1,500 ft	0.4 mile
50 miles	about 3,600 ft	0.8 mile
75 miles	about 6,300 ft	1.2 miles
100 miles	about 9,600 ft	1.7 miles

Standard four-thirds-earth refraction and a 0.95° beam. A few radars now scan lower than 0.5°.

7. Reading hail from radar

NOAA’s hail estimate (MESH). The Maximum Expected Size of Hail weighs the reflectivity of 40 to 50 dBZ and above by its height above the freezing level, rising to full weight at the −20 °C level, and turns the total into a size (Witt et al., 1998). It was built to sit on the high side: about three in four observed stones should be smaller than it. NOAA publishes it every two minutes on a grid; Stormglass reads the day’s maximum at the property’s square and within 1, 3, 5 and 10 miles.

Three independent checks on the raw radar. The Weather Service’s hail classes, the echo-top test against the freezing level, and hail differential reflectivity are computed separately. Agreement among them is strong; disagreement is shown, not hidden. Stormglass does not use a single ZDR or CC cut-off for hail, because in testing that rule missed known hailstorms.

8. The probability of severe hail

The first figure on the report’s Summary is the probability of severe hail, POSH. It comes from the same NEXRAD hail detection algorithm as the size estimate in Section 7 (Witt et al., 1998), which the National Weather Service runs on every storm cell its radars identify, on every volume scan. The algorithm turns the strong reflectivity above the freezing level into a severe hail index (SHI), the same sum MESH is built from, and sets it against a warning threshold that rises with the height of the freezing level:

SHI and WT are in joules per metre per second; H₀ is the height of the freezing level above the radar, in kilometres (Witt et al., 1998). A higher freezing level raises the threshold, so the same storm aloft is given a lower probability where hail has farther to fall through warm air. POSH is a probability, from 0 to 100 percent.

Severe here means hail 3/4 in (19 mm) or larger: POSH is the probability that the cell is producing hail of at least that size. That was the National Weather Service’s severe-hail size when the algorithm was built. Since January 2010 the Weather Service’s criterion has been 1 in (NWS Directive 10-511), so POSH speaks to hail a quarter inch smaller than today’s warning criterion. The report labels it “3/4 in or larger.”

Stormglass reads the value NOAA archives for each signature, the probability-of-severe-hail field of the NEXRAD Level III hail index in NCEI’s Severe Weather Data Inventory, and takes these for the local calendar day of the date of loss:

On the Summary	How it is taken
Within 1 mile	The highest POSH of any signature whose cell centre was within 1 mi of the property
Within 3 miles	The highest within 3 mi, shown beside it when it differs; it leads when there is no signature within 1 mi
The closest signature	Its distance, time, radar, and its own POSH
None within 3 miles	The headline says so; it does not reach farther

POSH is a probability for the cell, at the cell’s centre aloft, not for the roof: hail can fall miles from the plotted point (Section 1). It is computed from radar alone, on every scan, so it does not depend on whether anyone was there to report hail (Section 10). Like MESH it reads the storm aloft and cannot see what melted on the way down. The storm history in Section 16 gives the highest POSH within 5 mi for every storm date.

9. The storm’s path over the property

Two public records give the storm’s motion. The report shows both, side by side, on its own page after the storm aloft.

The radar’s track of the cell

The National Weather Service’s storm cell identification and tracking algorithm (SCIT) finds each storm cell on each volume scan, gives it an identifier and follows it from scan to scan (Johnson et al., 1998). NCEI archives each cell’s position with the NEXRAD Level III storm structure and hail index products. Stormglass gathers every position on the date of loss that carries the same radar and cell identifier, and then:

• keeps the cells that came within 10 mi of the property, and for each takes the positions within 20 minutes either side of its closest one: a 40-minute window;
• uses a cell only if that window holds at least three positions spanning at least 8 minutes, and the cell moved between 2 and 90 mph; anything else is left out rather than guessed at;
• takes the direction and speed as a straight line from the first position in the window to the last;
• measures the closest pass as the distance from the property to that line, and the time the centre reached that point from the speed;
• and, of the cells that qualify, reports the one whose line passed closest.

d is the closest pass in miles, the speed in mph; 2 · √(1 − d²) is the length of the chord the track cuts through a 1-mile circle around the property. Worked: a cell centre passing 0.6 mi away at 30 mph cuts a chord of 2 × √(1 − 0.36) = 1.6 mi and is within 1 mi for 1.6 ÷ 30 × 60 = 3.2 minutes. A centre that passes more than 1 mi away counts 0 minutes.

The Weather Service warning

Every storm-based warning carries a line the forecaster writes at issue, TIME...MOT...LOC: the time, the direction the storm is moving from in degrees, its speed in knots, and its position (NWS Directive 10-511). A warning reading 320DEG 23KT describes a storm moving from the northwest at 23 knots, which is 26 mph (1 knot = 1.151 mph). Stormglass reads the warnings over the property on the date of loss from the Iowa Environmental Mesonet archive and uses the one issued closest in time to the radar cell’s closest pass; with no radar track, the one closest to 20:00 UTC on the date of loss.

The report leads with the radar track when there is one, since it follows the cell that actually passed; the warning’s motion is the forecaster’s estimate at the moment of issue. When both exist, the report prints how far apart they are in direction and in speed. When neither exists, the page says so.

From the path to the slopes

Section 5 shows that wind-driven hail strikes hardest on the slopes that face the wind. The storm’s motion gives a second direction to check against. On the report’s hail-energy page, a roof plane faced the storm’s approach when the direction it slopes down is within 60° of the direction the storm came from, the same 60° used for the wind in Section 15. The speed driven into each plane is worked in three dimensions from the plane’s own pitch and facing:

θ is the plane’s pitch; facing is the direction it slopes down and wind from the direction the wind blows from, both in degrees clockwise from north; V_t is the measured-fit fall speed and m the solid-ice mass of Section 5; V_w is the station’s peak gust, taken as the stone’s full sideways speed, the upper bound. For a plane facing straight into the wind this is the V · cos i of Section 5, and for one facing straight away it is the leeward case. Where v_⊥ is zero or less the plane is sheltered.

The energies on that page use the gust measured at the nearest airport station, from the direction it was measured. When that direction is more than 45° from the storm’s approach, or the station gave no direction, the page adds a column: the same gust, applied from the side the storm came from. The hail size is the one used throughout the page: the NOAA hail estimate at the property, else the largest within 1 mi, else the largest hail reported or signature within 1 mi. The two columns give the answer under each direction; the record alone cannot say which way the wind blew at the roof when the hail fell.

Storm motion is the movement of the storm as a whole. The hail and the gust at the ground blow with the storm’s outflow, which can come from another side, and the roof’s exposure follows both. The minutes within 1 mile are for the centre of the radar cell: the hail core is wider or narrower than that, and the radar sees the cell every 4 to 6 minutes.

10. Gaps, scans and who reports

• Between scans. A radar completes a volume every 4 to 6 minutes in storm mode. At 30 mph a storm moves 2.5 miles in five minutes, so a swath on a map is snapshots joined together, and a short burst can fall between looks. Stormglass never relies on one scan, one radar or one product.
• Who reports. A hail report needs someone to see the hail. Reports cluster where people are: the standard study of U.S. hail reports finds “a spatial bias towards road networks” (Allen & Tippett, 2015). That is one reason hail maps built from reports look densest along highways and in towns. Radar estimates do not depend on who was there, which is why Stormglass leads with them and uses reports to corroborate. Section 11 sets out the reports themselves.
• Cities and storms. Cities can change storms themselves: the St. Louis study of the 1970s (METROMEX) found rain, thunderstorms and hail peaked 10 to 15 miles downwind of the city.
• Stations and outages. An airport anemometer measures one point; a microburst can be under 2.5 miles across. Strong wind can knock a station out, and radars go down for maintenance.

11. Ground evidence: who saw what

Radar reads the storm aloft. The ground record is what people and instruments saw where they stood. The report gives it two pages, after the hail pages, and every entry carries its distance from the property.

Storm reports, by distance and by who made them

The first page lists every hail, wind and tornado report the National Weather Service logged within the report’s search radius (10 mi unless it is set to 5, 15 or 25) and window (the date of loss and one day either side unless it is set otherwise), closest first: its distance and direction from the property, the time, the size or speed, the place and the remark. It counts them by who made them, using the Weather Service’s own source field: a trained spotter, an emergency manager, law enforcement, the public, a co-op observer, an airport station and so on. Reports on the date of loss are shaded. The sixteen closest are printed; the rest are in the storm record spreadsheet.

A local storm report is filed within hours and is preliminary. It records what was seen at that spot at that time; the absence of a report is not evidence that nothing happened (Section 10). Source: NWS local storm reports, Iowa Environmental Mesonet archive.

The NCEI Storm Events Database: the final record

Storm Data, published as NOAA’s Storm Events Database, is the Weather Service’s checked account of significant weather, entered by each forecast office’s warning coordination meteorologist (NCEI) under the Weather Service’s Storm Data instructions (NWS Directive 10-1605). Each entry has a begin and end point, a magnitude (hail in inches; wind in knots, marked estimated or measured), the office’s narrative of the event and of the whole episode, and the property and crop damage the office estimated in dollars. Stormglass reads NCEI’s bulk files for the state and lists the entries whose begin or end point lies within 10 mi of the property, from the day before the date of loss to the day after. An entry for a whole forecast zone, with no point, is matched on the county. Wind is converted at 1 knot = 1.151 mph.

It is the final record, and it is slow. NCEI receives each month’s data “approximately 75 days after the end of a data month” and updates the database “within 75-90 days.” When the month of the date of loss has not been published yet, the report says so, names the month NCEI’s file runs through, and asks for the report to be run again later. The damage figures are the office’s estimates, which NCEI says “should be considered as a broad estimate”; they are not adjusted losses.

CoCoRaHS hail pads

Volunteers of the Community Collaborative Rain, Hail and Snow Network (CoCoRaHS; Cifelli et al., 2005) set out a foam hail pad and, after a storm, measure the dents: the largest, average and smallest stone, and how many struck the pad. The second page lists every CoCoRaHS hail report within 10 mi of the property from the day before the date of loss to the day after, closest first. A pad is a measurement, not a sighting, but only at one point and only where a volunteer keeps one. Data: CoCoRaHS, CC BY 3.0.

mPING

mPING is a phone app from NOAA’s National Severe Storms Laboratory and the University of Oklahoma through which anyone can report what is falling where they are, hail size included (Elmore et al., 2014). The reports are the public’s own and are not checked. Access to mPING’s data needs a key issued by the mPING team. Stormglass keeps that key on its backend, never in the browser, and until a key is in place the page says mPING is not connected instead of showing reports. With a key, the report lists mPING hail reports within 10 mi from the day before the date of loss to the day after, by UTC day.

12. Rain and smoke around the date of loss

One page of the report records two more conditions at the property, day by day, from a week before the date of loss to a week after it (or to today).

Record	What the report shows	What it cannot
Rain: the NOAA Multi-Radar Multi-Sensor estimate (MRMS)	The rain each day at the property’s point, as a bar chart and a table, the date of loss marked, and the total. MRMS builds its rain estimate from radar and, where they report, rain gauges (Zhang et al., 2016); the daily values come from the Iowa Environmental Mesonet’s service.	A gauge reading at the roof: it is an estimate for a grid square
Smoke: the NOAA/NESDIS Hazard Mapping System (HMS)	Whether smoke lay over the property each day: analysts outline the smoke they see on satellite images as light, medium or heavy, with the times it was seen (UTC). Each day reads as the heaviest over the point, none, or no analysis.	Smoke at ground level: it is smoke seen from above, in the air column

Neither is evidence of hail or wind damage. Rain says whether water fell on the property around the date; smoke, whether wildfire smoke lay over it. The Verify tab reads the same two records for any range of dates (rain up to two years at a time, smoke up to 92 days) as a spreadsheet.

13. Wind: from the station to the roof

Airport anemometers stand at 10 m (33 ft) over open ground — the World Meteorological Organization and National Weather Service standard. Wind is slowed by the ground, so a roof at a different height, over different ground, sees a different wind. ASCE 7-16, the standard the building codes adopt, gives the profile Stormglass uses:

Exposure (ASCE 7-16 §26.7)	Ground upwind	1/α	b̂
B	Towns, suburbs, woods, closely spaced buildings	1/7.0	0.84
C	Open country and grassland — an airport	1/9.5	1.00
D	Open water, mud flats, salt flats	1/11.5	1.07

Wind pushes on the wall it strikes and pulls up and out on the roof and the far wall; the pull is strongest where the flow breaks away at the eaves, rakes, ridge and corners.

The hill speed-up. Wind accelerates over the brow of a ridge or the edge of an escarpment. ASCE 7-16 §26.8 gives the factor from the hill’s height and length and the building’s place on it. Stormglass reads the bare ground along the wind direction from the USGS elevation service and applies the standard’s tests before using it.

Why not the plain power law? The power law alone assumes the station stood in the same terrain as the building, and can rank exposures backwards: a tall roof in town comes out windier than the same roof on the coast. The ASCE profile corrects for the slower wind near rough ground. The report shows both and leads with ASCE.

Two more views of the wind. The Doppler radar measures the part of the wind along its beam, aloft: a floor on the wind at that height, not the gust at the roof. NOAA’s surface analysis (URMA) estimates the gust hourly on a 1.6-mile grid between stations; Section 14 counts the hours it reached 50 mph. Both corroborate the station; neither replaces it. In a thunderstorm downburst the strongest wind can sit within a few hundred feet of the ground, so any profile is an engineering estimate there.

14. Hours of wind at 50 mph or more

The second headline on the Summary counts hours. NOAA’s Unrestricted Mesoscale Analysis (URMA) is the National Weather Service’s hourly analysis of the weather at the surface on a 2.5 km (1.6 mi) grid: every surface observation blended with a short-range weather model by the method of the Real-Time Mesoscale Analysis (De Pondeca et al., 2011), and run hours later so that late observations are included. Among its fields is the gust at 10 m (33 ft). For each hour of the date of loss, taken as a local calendar day at the property, Stormglass reads that gust from NOAA’s public archive at the grid square holding the property and at the eight squares around it.

On the Summary	How it is counted
At the property’s grid square	The hours whose analyzed gust at the property’s own 2.5 km square was 50 mph or more, with the times
Within about 2.4 mi	The hours in which any of the nine squares reached 50 mph: a block 7.5 km (4.7 mi) on a side, centred on the property’s square
Hours read	How many of the day’s hours NOAA’s archive answered for; an hour that could not be read is not counted, and is not taken as calm

50 mph is the line this report draws. It is not the National Weather Service’s severe-wind criterion, which is 58 mph (50 knots; NWS Directive 10-511). Each hour is the analysis valid at that hour, so a short gust between analyses can be missed, and one 2.5 km square smooths a downburst narrower than it. The analysis is an estimate between stations, not a measurement at the roof: it corroborates the station gust in Section 13 and does not replace it. NOAA’s public archive of it begins in 2015.

15. The roof: height, planes and zones

Height. The mean roof height is the height z_roof in the wind profile of Section 13. Stormglass measures it in four steps.

Step	What Stormglass does	Source
1. Find the main building	Of the building outlines within about 250 ft of the pin: the one the pin falls on (400 sq ft or more; the largest, if more than one), otherwise the nearest of 800 sq ft or more whose centre is within 200 ft. The roof model and the aerial are read there, so a detached garage beside the pin is not taken for the house.	FEMA / ORNL USA Structures
2. Read the roof	Google’s model of the building closest to that point, from the best imagery it has (high, then medium, then low): each plane’s pitch, facing, area, and height at its centre, which Google gives “in meters above sea level.” The mean roof height is the area-weighted mean of the plane-centre heights; a sloped plane’s centre sits halfway between its eave and ridge.	Google Solar API, building insights
3. Read the ground	Eight points 4 m (13 ft) outside the box around the building, at its corners and the middle of each side, from Google’s Elevation API; the median of the eight, so one point that lands in a ditch or against a wall does not move it. The mean roof height is the roof less this ground.	Google Elevation API; USGS 3DEP as the fallback
4. Check it	A mean roof height under 8 ft is refused; a roof under 900 sq ft is flagged (below).	—

The ground is read from Google because the roof heights are Google’s. Earlier versions subtracted the U.S. Geological Survey’s bare-earth height (3DEP, on the NAVD 88 datum) taken at the pin, which can be a hundred feet and a slope away from the building, and the two sources disagreed by several feet: enough to put a one-story roof at 5 ft. Google documents its roof-plane heights as above sea level and its Elevation API as relative to “local mean sea level”; it does not name a datum for either. USGS 3DEP is read at the same eight points as a check. When Google’s ground does not answer, 3DEP is used and named; when both answer and the roof height from 3DEP differs by 2 ft or more, the report prints both.

Two checks follow. A mean roof height under 8 ft is lower than any one-story roof, so it is not used: the report says why and asks for the number of stories (a one-story roof is about 10 to 12 ft at mid-roof) or a measurement from a photograph. A roof model under 900 sq ft is flagged as the size of a garage or shed, not a home, to be checked against the satellite view. A height already on file for the property, or typed in, is kept; without the roof model, the height comes from OpenStreetMap, the story count or a photograph, named on the report.

The aerial of the roof is Google’s, centred on the building that was measured rather than on the pin, and sized to hold it: a radius of the larger of half the building’s diagonal plus 12 m (about 40 ft) and 1.3 times its longer side, kept between 30 and 60 m (about 100 to 200 ft). The roof planes are drawn on it from the same model.

Planes that faced the wind. A plane “faces” the way it slopes down. One facing within 60° of the wind took it head-on, and the report marks it. Section 9 applies the same test to the direction the storm came from.

Zones. Wind pulls hardest where it turns over an edge:

The edge strip a is 10% of the building’s least width or 0.4 times the roof height, whichever is smaller, but not less than 4% of the least width or 3 ft (roof heights to 60 ft). Above 60 ft it is 10% of the least width, not less than 3 ft (ASCE 7-16 Figure 30.5-1). On a gable over 7° the ridge has zones of its own; on a hip of 25° or less, ASCE 7-16 treats the corners as zone 2.

Ratings. ASTM D3161 fan tests shingles at 60, 90 and 110 mph (Classes A, D, F); ASTM D7158 uplift classes D, G and H correspond to 115, 150 and 190 mph; UL 2218 impact classes 1 to 4 use steel balls of 1¼ to 2 in. A rating is a test on new material; age and installation lower what a roof can take.

16. The date of loss in context

A roof can be struck by more than one storm. The page after the Summary places the date of loss among every storm date in the public record at the address.

A storm date is a local calendar day with radar hail of a chosen size or larger, or a hail report, within a chosen distance (0.75 in and 5 mi unless set otherwise), and, when asked for, a wind report or warning. The history covers the range run on the Verify tab. When no history on screen covers the date of loss, the report reads one itself, from five years before the date of loss to five years after it, or to today.

Row	What it is
Storm before	The last storm date before the date of loss
Date of loss	The date itself, with its rank among all the storm dates by hail at and around the property; or, if it does not meet the listing threshold, a statement that it does not
Storm after	The next storm date after it
Most severe in the period	The storm date with the largest hail at the property, then within 1, 3 and 5 mi, then the largest hail report, compared in that order

For each row the report gives the hail at the property and within 1 and 3 mi, the highest probability of severe hail within 5 mi (Section 8), the largest hail reported, and the wind reported or warned, with any tornado. Two bar charts follow: every storm date in the period, and the six months either side of the date of loss, with the date of loss, the storm before, the storm after and the most severe marked.

Hail at the property and within 1 and 3 mi is the NOAA multi-radar hail estimate (MESH, Section 7), which Stormglass reads for up to 80 dates with the largest radar hail, from October 14, 2020, where its archive begins; for other dates the largest NEXRAD Level III signature stands in and is marked. A storm after the date of loss can damage the same roof; a storm before it can explain older damage. The roof decides which (Section 18).

17. Starting a report from an open claim

The Reports tab starts a Storm Report from an open claim on the Claim Compass register. The date of loss is the one set on the claim: the report uses it and writes nothing back, and another date can be reported only from the Verify tab. The claim’s address (or its pin, when the claim is on a property in the book), the insured and the claim number fill the report, and the Summary adds the claim’s carrier and peril. A claim with no date of loss or no address cannot be pulled until the claim is filled in. The same box finds a property in the book or takes any address as typed; for those, the date of loss is entered by hand.

18. Confidence, and the roof decides

Confidence comes from independent records agreeing: a radar hail estimate, a hail class aloft, a warning, a spotter’s report, a hail pad and a station gust are made by different instruments and different people. One is a lead; three that agree on the same date and place are strong. Every figure in the report carries its source, its time and its distance, so it can be checked against NOAA line by line.

The record shows that a qualifying storm reached the property, and how hard. The roof decides whether that storm did the damage: fractures and bruising with the mat exposed, fresh granule loss and dented soft metal for hail; creased and torn tabs and broken seals gathered at the edges and corners that faced the wind, for wind. The report is built to sit beside the inspection, never to replace it.

Sources

† Funded by, or written with authors from, the Insurance Institute for Business & Home Safety (IBHS), which insurers fund; used here only for measurements of hail itself. Reports from engineering and weather firms retained by insurers are not used.

1. NOAA National Severe Storms Laboratory, Severe Weather 101: Hail (formation, layers, fall speeds). nssl.noaa.gov/education/svrwx101/hail
2. NOAA JetStream, Hail (updraft speed and hail size). noaa.gov/jetstream
3. Kumjian, M. R., and K. Lombardo, 2020: A hail growth trajectory model for exploring the environmental controls on hail size. J. Atmos. Sci.
4. Nixon, C. J., J. T. Allen and M. Taszarek, 2023, Weather and Forecasting (the −10 to −30 °C hail growth zone).
5. Waldvogel, A., B. Federer and P. Grimm, 1979: Criteria for the detection of hail cells. J. Appl. Meteor. 18.
6. Witt, A., et al., 1998: An enhanced hail detection algorithm for the WSR-88D. Wea. Forecasting 13; Wendt, N. A., and I. L. Jirak, 2021: An hourly climatology of operational MRMS MESH-diagnosed severe and significant hail. Wea. Forecasting.
7. Aydin, K., T. A. Seliga and V. Balaji, 1986: Remote sensing of hail with a dual linear polarization radar. J. Climate Appl. Meteor. 25; Depue, T. K., et al., 2007, J. Appl. Meteor. Climatol.; CSU-CHILL radar notes.
8. NOAA NWS Warning Decision Training Division and NWS offices: dual-polarization products, Hydrometeor Classification and Hail Size Discrimination Algorithms.
9. Allen, J. T., and M. K. Tippett, 2015: The characteristics of United States hail reports: 1955–2014. Electronic J. Severe Storms Meteor. 10(3).
10. Changnon, S. A., et al., METROMEX: Illinois State Water Survey reports on urban effects on precipitation and severe weather at St. Louis (1970s).
11. American Meteorological Society, Glossary of Meteorology: Rayleigh scattering.
12. NOAA Radar Operations Center (WSR-88D scan strategies); NOAA MRMS; NOAA NCEI Severe Weather Data Inventory and Storm Events Database; NOAA/NWS URMA; Iowa Environmental Mesonet (archives).
13. ASCE 7-16, Minimum Design Loads and Associated Criteria for Buildings and Other Structures, §§26.7–26.8, Table 26.11-1, Chapter 30; World Meteorological Organization anemometer standard; U.S. Geological Survey 3D Elevation Program; Google Solar API.
14. ASTM D3161, ASTM D7158, UL 2218, as summarized by the National Roofing Contractors Association.
15. Foster and Bates, 1956: A hail size forecasting technique. Bull. Amer. Meteor. Soc. 37(4) (weight equals drag; the updraft that holds the stone).
16. † Heymsfield, A., M. Szakáll, A. Jost, I. Giammanco and R. Wright, 2018: A comprehensive observational study of graupel and hail terminal velocity, mass flux, and kinetic energy. J. Atmos. Sci. 75(11), with the corrigendum in J. Atmos. Sci. 77(1), 2020.
17. † Heymsfield, A. J., I. M. Giammanco and R. Wright, 2014: Terminal velocities and kinetic energies of natural hailstones. Geophys. Res. Lett. 41(23).
18. † Meisenzahl, B., I. Giammanco and F. Hedayati, 2025: Sub-severe hail: the missing piece in assessing asphalt shingle risk in North America. Frontiers in Materials 12, 1603074 (Insurance Institute for Business & Home Safety).
19. † Fedler, C. B., D. J. Spradling, M. Phelps, I. M. Giammanco and M. Smith, 2019: Evaluating the compressive strength of hail and its relationship to freezer iceballs. RCI Interface, January (also hail kinetic energy in wind, after ASTM E822); Cope, A., and I. Giammanco, 2022: Hail no! Professional Roofing, March.
20. Greenfeld, S. H., 1969: Hail Resistance of Roofing Products. National Bureau of Standards Building Science Series 23; Crenshaw and Koontz, 2001: Simulated hail damage and impact resistance test procedures. RCI Interface, May.
21. Chen et al., 2022: Experiments on an ice ball impacting onto a rigid target. Int. J. Impact Eng.; Pernas-Sánchez et al., 2015, Experimental Mechanics 55(9) (fracture energy of ice).
22. † Giammanco, I. M., et al., 2015: Evaluating the hardness characteristics of hail through compressive strength measurements. J. Atmos. Oceanic Technol. 32(11); Brown-Giammanco, T. M., and I. M. Giammanco, 2018: An overview of the IBHS hail research program. 29th Conference on Severe Local Storms, American Meteorological Society.
23. NOAA National Weather Service, Directive 10-511, WFO Severe Weather Products Specification (severe thunderstorm criteria: hail 1 in, wind 58 mph or 50 kt; the TIME...MOT...LOC line of a storm-based warning); Directive 10-1605, Storm Data Preparation.
24. Graham, R. A., and J. A. Nelson, 1998: An examination into the utility of the Build 9 Hail Detection Algorithm (HDA) across northern Utah. NOAA NWS Western Region Technical Attachment 98-27 (POSH and its warning threshold in Weather Service use).
25. Johnson, J. T., P. L. MacKeen, A. Witt, E. D. Mitchell, G. J. Stumpf, M. D. Eilts and K. W. Thomas, 1998: The Storm Cell Identification and Tracking algorithm: an enhanced WSR-88D algorithm. Wea. Forecasting 13 (storm cells and their positions; NEXRAD Level III storm structure and hail index, NCEI Severe Weather Data Inventory).
26. De Pondeca, M. S. F. V., et al., 2011: The Real-Time Mesoscale Analysis at NOAA’s National Centers for Environmental Prediction: current status and development. Wea. Forecasting 26; NOAA URMA 2.5 km surface analysis, NOAA Open Data Dissemination archive (noaa-urma-pds).
27. NOAA National Centers for Environmental Information, Storm Events Database, Frequently Asked Questions (when data are published, who enters them, damage estimates). ncei.noaa.gov/stormevents/faq.jsp
28. Cifelli, R., N. Doesken, P. Kennedy, L. D. Carey, S. A. Rutledge, C. Gimmestad and T. Depue, 2005: The Community Collaborative Rain, Hail, and Snow Network: informal education for scientists and citizens. Bull. Amer. Meteor. Soc. 86(8); CoCoRaHS hail reports, cocorahs.org, CC BY 3.0.
29. Elmore, K. L., Z. L. Flamig, V. Lakshmanan, B. T. Kaney, V. Farmer, H. D. Reeves and L. P. Rothfusz, 2014: mPING: crowd-sourcing weather reports for research. Bull. Amer. Meteor. Soc. 95(9).
30. Zhang, J., et al., 2016: Multi-Radar Multi-Sensor (MRMS) quantitative precipitation estimation: initial operating capabilities. Bull. Amer. Meteor. Soc. 97(4); Iowa Environmental Mesonet, Iowa State University, IEM Reanalysis (IEMRE) daily service.
31. NOAA NESDIS Office of Satellite and Product Operations, Hazard Mapping System Fire and Smoke Product.
32. Oak Ridge National Laboratory and the Federal Emergency Management Agency, USA Structures (building outlines with occupancy class).
33. Google Maps Platform documentation: Solar API, building insights and data layers (roof segment pitch, azimuth, area, and “the height of the roof segment plane, in meters above sea level”); Elevation API (elevation relative to “local mean sea level”). U.S. Geological Survey, Elevation Point Query Service (3D Elevation Program).