science

What a Hail Swath Actually Is

A warning polygon, a radar footprint, a modeled swath, and the addresses that actually took damage are four different objects, not one.

Four different objects get called "the swath" in a storm conversation, and people argue past each other for an hour without noticing which one they are each holding. The four are the National Weather Service warning polygon, a radar-derived hail footprint, a vendor's modeled swath, and the set of addresses that actually took damage. They are drawn by different parties, at different times, for different purposes. Only one of them pays claims, and it is the one nobody publishes.

The warning polygon is a safety product

The polygon came from a deliberate change in how warnings were issued. The National Weather Service upgraded warning capabilities on October 1, 2007. Before that, the NWS issued county-based warnings, which encompassed the entire county regardless of what portion of the county the storm or threat was located in. After it, under storm-based warnings, warning polygons only cover the portion of the county actually threatened by the storm.

Notice the tense in that description. The polygon covers the portion that is threatened. It is drawn before and during the storm by a forecaster deciding who needs to take cover in the next several minutes. It is a statement about expected threat, made under time pressure, for the purpose of keeping people alive. It is not a record of what happened.

A watch is yet another object, and it is coarser still. Watches typically last 6 to 8 hours, and a watch is not a warning. Nobody should be reasoning about a specific roof from a watch area.

Whole counties warned, three cities outside the polygon

The NWS Birmingham severe weather handout carries a worked example that settles this faster than any explanation. In the case shown, tornado warnings were in effect for Hale, Bibb, Perry, Tuscaloosa, Jefferson and Shelby counties. At the same time, the cities of Tuscaloosa, Birmingham and Calera were not in the polygons and were therefore not under a tornado warning.

Six counties were warned. Three of the largest population centers inside those counties were not. The example is drawn with tornado warnings, but the mechanic being illustrated belongs to the storm-based warning system rather than to the tornado product: the polygon follows the storm's threatened corridor, and county lines are irrelevant to it.

That geometry is the reason "the whole county was under a severe thunderstorm warning" tells you almost nothing about a specific address, and the reason "this address was not in the polygon" tells you almost nothing about whether hail fell on it.

Why the radio and the app disagree on purpose

The same page explains a disagreement people run into constantly. NOAA Weather Radio All Hazards continues to alert the entire county, while several vendors alert only if you are within the storm-based warning.

So two homeowners in one county, in one storm, get two different answers to "were you in it." The one with a weather radio heard an alert. The one with a particular app heard nothing. Neither was misinformed. One system answers at county resolution and the other answers at polygon resolution, by design.

This matters when you are taking a statement from a homeowner. "We were under a warning" and "we were in the warning polygon" are different claims, and the homeowner's memory usually comes from whichever device alerted them. The distinction is not a gotcha, it is a translation problem, and it is worth resolving before anyone writes it down.

A radar footprint is a threshold applied to a continuous field

The second object people call a swath is radar-derived. Radar hail products such as MESH work by estimating hail size from reflectivity properties above the environmental 0 degree C level, from the Multi-Radar Multi-Sensor system. The underlying output is a continuous grid of estimated sizes, not a shape. The shape appears only when somebody picks a size and draws the line where the grid crosses it.

Change the threshold and the footprint changes. That is not a flaw in anyone's software, it is what a contour is. And the estimates themselves carry known error: the same research that compared radar estimates to the official record found severe hail hours from MESH running 2 to 4 times greater than those estimated from Storm Data even in plains areas where population density is relatively high, and discarded MESH values above 127 mm, 5 inches, as likely spurious. Coverage is uneven too: relative minima in the difference between radar estimates and reports show up across much of the Intermountain West, the West Coast, the Appalachians and the Northeast, and lack of radar coverage from beam blockage or widely spaced radars accounts for part of that.

A vendor's modeled swath is the third object. It typically starts from a radar field like this one, applies its own threshold, its own smoothing, and its own size relationship, and hands you a shape. The shape inherits every uncertainty above, and then presents it as a boundary.

What a swath is as a geometric object

Strip away the imagery and a swath you can query is a very simple thing. In GeoJSON, the geometry types are Point, MultiPoint, LineString, MultiLineString, Polygon, MultiPolygon and GeometryCollection, and a swath is a Polygon or a MultiPolygon.

A Polygon is built from linear rings. Each ring is a closed LineString holding at least four positions, and it has to return to where it began: the first and last positions carry identical values. The first ring is the exterior ring, and any additional rings are interior rings, which is to say holes. Exterior rings run counterclockwise and holes clockwise, following the right-hand rule, although parsers are told not to reject polygons that ignore this, for backward compatibility.

Each position is an array whose first two elements are longitude and then latitude, with an optional third element for altitude in meters. Longitude first is the single most common source of swapped-coordinate bugs in this work. The default coordinate reference system is the WGS 84 datum in decimal degrees, equivalent to urn:ogc:def:crs:OGC::CRS84, and alternative coordinate reference systems were removed in this version of the specification.

So the object is a closed ring of coordinate pairs. A point is either inside that ring or outside it. The answer is binary, immediate, and complete. Hail damage is none of those things.

The edge is the least reliable part of any swath

Everything that makes a swath useful in its interior makes it weakest at its boundary. The interior is where multiple lines of evidence agree. The edge is where a continuous estimate was cut by a chosen threshold, where radar resolution and beam geometry matter most, and where a model's smoothing has the most influence on the result. A point just inside the line and a point just outside it are, physically, the same place. The polygon insists they are different.

The fourth object, the set of addresses that actually took damage, has no public source. Damage is a function of hail size, of fall angle and wind, of roof age, of material, and of what was already failing before the storm. The official report record does not fill the gap either: if an event is missing from both Storm Data and the Storm Events Database, it was not reported to the National Weather Service, which is a statement about reporting and not about weather.

Three practical rules follow. Name the object before you argue about it, because a warning polygon, a radar footprint and a modeled swath will disagree and all three can be correct. Treat any address sitting close to a swath edge as unresolved rather than included or excluded. And never present a swath boundary as a finding about a specific roof, because the boundary is the part of the geometry carrying the most uncertainty and the least evidence. Storm Stream is an API over these same public feeds: it builds swath geometry, serves it as GeoJSON, and answers which of a list of addresses fall inside a given swath.