An odd thing has been happening over the past several years: property insurance hail claims in many parts of the country have been increasing, but official National Oceanic and Atmospheric Administration (NOAA) hail reports have not. If no more hail has been falling, why are more people filing claims for hail damage?

One prominent reason for the increase in hail claims is the increased prevalence of automated hail track products. You've probably seen them or used them yourself: digital maps with hail swaths color-coded to indicate hail size, available online for instant download. These products are offered by a number of different companies and are heavily marketed to contractors as well as to insurance adjusters. It's an equal opportunity sales approach: pitch the roofer who will encourage the homeowner to file a claim, as well as the insurance adjuster who will investigate it.

Related: Hail-related insurance claims by the numbers

The marketing is not subtle. One prominent hail map provider advertises, "Learn how contractors are using [our product] for explosive sales growth!" As you can imagine, that explosive sales growth is not being funded out-of-pocket by property owners; rather, it's being funded by a parallel growth in insurance hail claims.

All modern, automated hail track products are built upon the same fundamental data: output from National Weather Service (NWS) dual-polarization weather radar processed through a hail detection algorithm. Since these products all rely upon dual-polarization radar's ability to distinguish different precipitation types, they are all subject to the same data-based uncertainty and limitations.

This is not to say that such hail track products aren't valuable. They are: they provide a convenient and generally accurate account of where hail is being produced within a storm. Meteorologists, both operational and forensic, often use hydrometeor classification algorithms (techno-speak for algorithms that offer a best-guess as to the type of precipitation occupying a given section of sky) as a first step in identifying likely regions of hail production, although we take the algorithm output with a healthy grain (or two) of salt when conditions warrant.

Dual-polarization radar (bottom) collects data about both the horizontal and vertical characteristics of precipitation. Since different precipitation types have characteristic shapes and movement, this two-dimensional information can be used to distinguish between different precipitation types. Source: National Weather Service

Dual-Polarization Radar: The basics

By collecting information from two dimensions instead of one (hence the "dual" in dual-polarization), dual-pol radars can provide sophisticated information about the shape, density and variety of precipitation particles within a storm.

Such detailed information wasn't always available. Before the National Weather Service upgraded its network of NEXRAD radars to dual-polarization operation during 2011-2013, those radars were single-polarization and collected only one-dimensional information that did not explicitly reveal hydrometeor properties. Radar algorithms had to infer precipitation type based on characteristics like storm structure and the atmospheric freezing level. 

Related: Severe storms bring rain, hail and questionable claims

Dual-pol changed that. The upgraded radars and their likewise upgraded hydrometeor classification algorithms don't have to rely on such inferences. Instead, they can directly discern and discriminate between areas of heavy rain, hail, sleet and snow based on collected data about hydrometeor shape, density and variety.

So dual-pol radar is a powerful tool, but like any tool, it can be abused when its limitations and inherent uncertainty are not well understood (or simply ignored). Which brings us to the dark side of dual-pol.

The relationship between distance from the radar site, height of the radar beam, and width of the radar beam (which determines spatial resolution). Source: National Weather Service (modified)

The dark side

What conditions warrant skepticism when it comes to automated (algorithm-based) hail detection?

Discrepancies between hail detection algorithm output and what's actually observed on the ground generally boil down to the following: hail melts as it falls out of a storm, no two storms are exactly alike, and hail track maps can indicate an artificial degree of certainty and detail.

Artificial certainty

As a radar beam propagates away from the dish, it spreads out and moves upward in the atmosphere, like a flashlight beam tilted slightly upward and shining across a vast room. For every 10 miles from the radar dish, the radar beam becomes about 1000 feet wider. Distance therefore limits both how low in the atmosphere the radar can "see" as well as the spatial resolution of the radar data. Farther away = higher radar beam and lower resolution data.

Related: LiveHailReports.com introduces free hail swath map alerting service

Yet most hail track products provide street-level detail regardless of distance from the radar site. The hail swath map will show a tidy, smooth dividing line between hail size zones — one house in the 1.5" hail swath and another a block away in the 0.5" hail swath.

But nature and the data aren't nearly so tidy. While those smooth lines and precise hail size zones look lovely on the screen, they're actually the result of smoothing and interpolation — the algorithm's attempt to "fill in the blanks." Don't get too attached to the exact location of the hail size zones or the estimated hail sizes — they are best guesses not absolutes.

Melting

Hail aloft does not always mean hail at the ground. Powerful hail storms require warm, moist air at the surface and cold air aloft to form and sustain themselves. Hail is produced in the cold, high regions of the storm, but once it falls beneath the freezing level (typically between 7,000′ and 15,000′ AGL, depending on location and season), it starts to melt. The amount of melting hail experiences once it falls beneath view of the radar depends on:

• Distance from the radar site — as previously mentioned, the radar beam gains in altitude with distance from the radar dish. At a range of 90 miles, the lowest radar beam is already at 10,000 ft altitude — the radar cannot "see" any lower than that. A lot can happen below 10,000 ft (like substantial melting of hail).

• The initial size of the hailstones — small hail melts proportionally faster than large hail, so the smaller the hail is initially, the more it will melt before reaching the ground. Hail that is initially 0.5" diameter or less often melts completely before reaching the ground.

National Weather Service radar coverage across the U.S. regions marked in gold, gray, or white do not have radar coverage below 6,000 ft AGL. In these regions, it is more difficult to determine whether hail is melting significantly before reaching the ground. Source: NOAA

No two storms are alike

Finally, no two storms are alike, yet the thresholds used for classification and size estimation within hail detection algorithms are fixed. They are designed to describe most, but not all, conditions. Dual-pol radar is a powerful tool, one that literally allows us to peer into the heart of storms. However, interpretation of what we see is by its nature complex, and no single algorithm can capture that complexity. An algorithm that provides an accurate estimate of hail size in one storm is not guaranteed to provide similar performance with all storms.

The bottom line

Hail detection and hail size algorithms are useful first-alerts — they reliably show where hail is found aloft within a storm and can generally distinguish between large vs. small hail. They are less reliable at estimating exact hail size on the ground or providing street-level detail of the location of hail swaths.

Claims adjusters who use such products for their own investigations should understand and consider the capabilities, limitations and inherent uncertainty. Uncertainty is highest when:

• Only small hail is indicated aloft — hail can melt significantly or completely before reaching the ground

• The location of interest is a long distance from the nearest radar site — the radar cannot "see" what's happening close to the ground

• The atmosphere is warm with a high freezing level — most common during the peak summer months, this increases the probability of significant melting

Take advantage of dual-pol radar data and the insight offered by algorithm-based products, but do so with a healthy degree of caution. And if you have any doubts about the reliability of the algorithm output, work with a forensic meteorologist who will investigate the data not with a fixed formula but with rigorous training, years of experience, and an understanding of local conditions and storm types.

Related: Where the hail is my estimate?

A forensic meteorologist will provide a more nuanced analysis of dual-polarization data, taking into account those factors that automated hail track products typically don't: distance to the radar, melting considerations, comparison with on-the-ground hail reports, data reliability and the like.

When the stakes are high (i.e., when litigation is involved or a claim is particularly costly), working with a certified meteorologist who will rigorously reconstruct and verify the weather conditions at the insured property — and who can provide expert testimony should litigation proceed — can ultimately save thousands to tens of thousands of dollars.

Don't get taken by the dark side of dual-pol — understand the capabilities and limitations of this valuable tool, and don't hesitate to seek an expert opinion when needed.

Megan D. Walker-Radtke, CCM, is the chief meteorologist at Gainesville, Florida-based Blue Skies Meteorological Services. Email her at [email protected].