New Asphalt Shingle Hail Impact Performance Test Protocol and Damage Assessment

Building components, particularly roofing systems, are frequently damaged in hailstorms (Marshall et al. 2002), leading to large insured losses. Annual insured losses attributed to hail commonly exceed $10 billion (US) (Gunturi and Tippett 2017). This is a sharp increase compared to the $850 million estimated in 2009 (Changnon et al. 2009). Unfortunately, hail is a natural hazard that, in the United States, is currently not well accounted for in common building practices. Existing testing methods for building materials have shown little real-world predictive value and struggle to identify performance relativities between products, making it difficult to discern which materials can be used to mitigate hail damage. There is a need for a new testing protocol that replicates the damage modes associated with typical natural hailstones and provides a ratings assessment that moves beyond the binary outcomes afforded by the existing standardized tests.

The Insurance Institute for Business & Home Safety (IBHS) hail research program, launched in 2010 (Brown et al. 2012, 2014; Giammanco et al. 2014; Brown-Giammanco and Giammanco 2018), has focused on understanding the damaging potential of hail on roof covers with the objective of developing a new hail impact test method that can distinguish relative performance between products and account for real-world damage modes. The critical questions that needed to be answered were:

Laboratory experimental testing and field research have been coupled to answer these questions to produce and demonstrate a new protocol to test the hail impact resistance of roof cover products, specifically asphalt shingles. The work presented here applies recent advances in the understanding of hailstone properties (i.e., diameter-to-mass relationships, aerodynamics, strength), research and development of a laboratory hail manufacturing system, and the applications of machine vision techniques. The IBHS Impact Resistance Test Protocol for Asphalt Shingles (IBHS 2019) is comprised of three main components. The hailstone characteristics provide the basis for manufacturing of laboratory ice spheres used for impact testing, which drive the impact mode. The impact configuration describes the construction of test panels and provides requirements for impact locations on the test panels. The damage and functionality assessment provide a quantitative method for determining a performance rating based on damage in five categories.

Roofing products have varying resistances to hail based on the material properties of both the hailstones and the shingles. There are impact-resistant (herein referred to as IR) asphalt shingle roofing products marketed to consumers to perform better in hailstorms. These products have material and design differences relative to their non-IR counterparts, but generally all have the same performance ratings from standard test methods. IR products have been marketed in hail-prone regions of the United States as a possible means to mitigate hail damage. However, there are few laboratory, postevent-assessment, or closed-claims studies to evaluate differences between building product performance in the field between non-IR and IR products (Crenshaw and Koontz 2001; RICOWI 2012, 2017; Brown et al. 2015).

Steep-slope IR roofing materials are currently rated according to the standard test methods UL 2218 (Underwriters Laboratory 2012) or FM 4473 (FM Approvals 2005). Products that pass these tests often carry a visible stamp, usually on the packaging and/or on the material itself. These test standards assume that damage characteristics and severity will scale with a projectile’s impact kinetic energy. These tests use a projectile to match a theoretical impact kinetic energy for a similarly sized, spherical hailstone at sea level, with a hailstone density equal to that of pure ice ($0.9\mathrm{g}{\mathrm{cm}}^{-3}$). These standard test methods are based on the work of Laurie (1960), which derived its diameter-to-kinetic energy relationship from Bilham and Relf (1937). The UL 2218 test method uses steel ball bearings of a given mass and diameter dropped from specific heights to achieve the prescribed kinetic energy, while the FM 4473 method utilizes distilled water ice spheres made in molds, which are propelled at velocities needed to achieve kinetic energies near the Laurie (1960) values. Both the UL 2218 and FM 4473 methods use four classifications of increasing severity to assess impact performance (Table 1).

The performance assessment for these two tests has a binary outcome: pass or fail. After being impacted at several different locations, the roof covering is removed and inspected (in the case of the UL 2218 method, this includes being turned over to expose the back of the shingle, then bent over a curved mandrel under $5\times$ magnification, as shown in Fig. 1). If a crack or tear is visible for any of the impacts, the product fails the test. If no crack is identified for all impacts, then the product passes.

IBHS has previously applied both test methods to evaluate non-IR and IR asphalt shingle products purchased through the normal supply chain as a consumer or contractor would purchase them. In the operational testing of UL 2218 or FM 4473 for product ratings, products are obtained and tested directly from the manufacturing plant. The IBHS test results for UL 2218 for 22 total non-IR and IR products were striking—none of the products that were rated to UL Class 4, which tests against a 5.08-cm (2-in.) steel ball (Table 1), passed during the testing at IBHS (Fig. 2) (IBHS 2014). Furthermore, the damage produced by this test method (such as crushed granules) is not representative of real-world hail damage. Similar results were found using the FM 4473 ice sphere impact test method (not shown), although the damage modes were different.

The testing results indicated that the methodology of each standard test warranted further scrutiny. The inability of either method to distinguish high-performing materials (if they exist) was concerning for consumers and their insurers, who often offer credits and incentives based on these test ratings. Furthermore, consumers deserve to have confidence that shingles labeled as IR (in this case through the UL or FM ratings) live up to expectations.

Some of the noted issues are a result of the reliance on theoretical kinetic energies with a lesser focus on the material properties of the projectile. The UL 2218 method uses a steel ball bearing as its projectile, which has a high mass traveling at a relatively low velocity to achieve the target kinetic energy. With such a hard object and low strain rates at impact, the steel ball does not deform at all, and the momentum and energy transfer between the steel ball and shingle is concentrated across a very small surface area, over a short time duration. In addition, the steel ball is hard enough and impacts the shingle with enough kinetic energy that granules on the shingle surface are crushed (Fig. 3). This type of damage is not seen in the field, as natural hailstones are not hard enough to pulverize the granules. The FM 4473 test method attempts to create more representative damage modes using ice as the projectile, but the damage severity is often too high (too damaging) for the representative hail size. Additionally, a pure ice sphere (assuming a density of $0.9\mathrm{g}{\mathrm{cm}}^{-3}$) will have a higher mass and density than its natural hailstone counterpart, because the hailstone has entrapped air and an irregular shape that does not have the same volume of matter as a sphere of the same maximum diameter (Giammanco et al. 2017). This requires equivalencies to be used. In addition, Giammanco et al. (2015a) (herein G2015) and Phelps et al. (2019) have shown that the FM 4473 ice sphere manufacturing process (using molds) produces a high variance in strength under compression. This is a cause for concern when applied in a standardized testing environment within a framework that uses a pass/fail approach, especially considering that any one impact can result in a failing performance rating.

The material properties of natural hail were not accounted for in the UL or FM methods. However, these material properties were likely important, as Knight et al. (2008) speculated that little property damage was likely to occur for an observed hail event which produced soft hailstones of sizes larger than 4 cm, which exceeds the 2.54-cm (1-in.) severe hail criteria. Historical studies often qualitatively describe hailstones as hard, soft, or slushy, with no quantitative means to describe them (Carte and Kidder 1966; Knight and Knight 1973, 2001; Knight et al. 2008). There were no material property data beyond mass, diameter, and density to guide the manufacturing of laboratory ice for use in impact testing to validate the Knight et al. (2008) hypothesis. Within the aerospace industry, attempts were made to account for this property in impact testing. The ASTM F320 test (ASTM 2010) uses cotton fibers within ice spheres to reduce the bulk density while increasing the strength (Swift 2013).

The G2015 study was the first to test natural hailstones in the field to provide a quantitative assessment of the strength of hail. The findings from the original study, and additional field campaigns conducted since, have filled the long-running knowledge gap regarding hailstone strength, and the data have been leveraged to determine target manufacturing requirements for laboratory ice spheres, which have then been used to determine the influence of hailstone strength on damage. G2015 and Phelps et al. (2019) showed that natural hail was slightly stronger (when using compressive stress as a proxy) on average than FM 4473 ice spheres. However, natural hailstones can exhibit a wide range of compressive stress values.

Beyond the material properties of the projectile used for impact testing, recent studies (Heymsfield and Wright 2014; Heymsfield et al. 2014, 2018) have questioned the kinetic energies of hail previously published by Bilham and Relf (1937), which were used by Laurie (1960). Unfortunately, current test standards have relied on this outdated work ever since. More recently, Heymsfield et al. (2014) (herein HGW2014) revisited hail diameter-to-terminal velocity and kinetic energy relationships using improved aerodynamic assumption and accounted for the shapes of natural hailstones. Heymsfield et al. (2018) went further to include experimental testing of these assumptions, using a larger natural hail database and testing three-dimensional (3D) printed hailstones of varying shapes and densities in a vertical wind tunnel. The result was improved functions to calculate hail kinetic energy, using a maximum diameter at varying pressure levels, and the ability to express the uncertainty. The resulting kinetic energies for natural hailstones were less than those shown in Laurie (1960) and well below that used in the UL 2218 and FM 4473 standards (Table 1). The kinetic energies listed using the HGW2014 results represent the target for the new testing protocol presented here.

Hailstones are grown through complex and alternating microphysical processes, which are difficult to recreate to produce laboratory hailstones in large numbers for testing. While laboratory hailstones have been grown in a vertical wind tunnel, the process is very time consuming and inefficient (Gokhale and Spengler 1977; Pflaum et al. 1978). Therefore, a process was required to engineer manufactured ice to (1) meet a set of specifications tied to natural hailstones and (2) produce damage modes that were representative of natural hailstones. The initial ice design specifications were:

The breakpoints for the compressive stress target values were determined from qualitative observations of impact mode on asphalt shingles and the statistical distribution of the compressive stress from the IBHS hailstone database. The soft category breakpoint represents the 25th percentile of the distribution, and the breakpoint for the hard category is near the median of the data set and the 40th percentile.

Diffusing gas into water was identified as a feasible method to lower the ice density to that of natural hailstones from historical literature ($0.2–0.9\mathrm{g}{\mathrm{cm}}^{-3}$) (Macklin 1962; Browning et al. 1963). Initial experiments were conducted using seltzer water in injection molds. In later experiments, a sealed tank was used to control the process. Approximately 75% of the tank volume was filled with filtered water and the remaining volume filled with ${\mathrm{CO}}_2$. The tank was held under varying pressures at room temperature, approximately 20°C–22°C. The proof-of-concept testing showed that ice spheres could be produced at densities as low as $0.65\mathrm{g}{\mathrm{cm}}^{-3}$, which was well within the range of natural hailstones. However, the injection mold technique led to high variances in both density and compressive stress, even within a single batch. The extraction of the spheres from the molds also produced expansion cracks, likely due to thermal shock. The observations fostered research work to develop a stable automated process to produce consistent manufactured hailstones in quantities suitable for laboratory impact testing.

The laboratory hail production system was designed by Accudyne Systems (Newark, Delaware) and IBHS to consistently produce manufactured hailstones matching the properties previously described to accommodate the needs of repetitive testing that provided a better match in damage mode and severity. A photograph and process schematic of the system is shown in Fig. 4 (US Patent No. 20170122636). The mechanical system diffuses ${\mathrm{CO}}_2$ gas into water, then freezes 2-m-long cylinders of ice of varying diameters quickly (in less than 2 h). The ice is produced through a heat exchanger process, where the aerated water is pumped into a stainless steel cylinder. Valves close off the heat exchanger, and coolant is then circulated in an outer jacket surrounding the inner cylinder. The process increases the transfer of heat from the relatively warm liquid to the coolant in the outer jacket. During this process, ${\mathrm{CO}}_2$ comes out of the aqueous solution (e.g., effervesces), forming bubbles. The phase change of the liquid to a solid occurs over a sufficiently short period of time so that the bubbles are trapped within the ice. After this process, the ice cylinders are then extruded from the system and shaped into spheres using heated aluminum molds, at the diameters desired for testing. The result is an opaque, bubble-filled ice with a density below that of pure water, and manufacturing settings are used to obtain different ice sphere compressive stresses, masses, and densities. The appearance of the manufactured hail is very similar to the structure that results from dry growth of natural hailstones. For the purposes of the impact test protocol, spheres are used to maintain consistency in production and for their ability to be consistently propelled at the target speeds/kinetic energies. It is understood that natural hailstones are not spheres but more spheroidal and oblate in their shape, with large protuberances at times. However, beyond aerodynamics, there is no evidence to suggest that these features alter damage modes and/or severity. The system, though, can accommodate any shape that can be milled into aluminum molds using computer aided design (CAD). In the future, detailed experiments could be conducted using a variety of hailstone shapes.

The hail production system is highly configurable and offers numerous controls. The dominant controls on density and strength are the amount of dissolved gas, the minimum target temperature of the heat exchanger, and the duration the ice cylinder is kept at the target temperature. The aeration system is controlled for pressure and duration of diffusion to control the amount of dissolved gas. The higher the pressure and longer the duration the liquid is kept under pressure, the more gas is aerated into the water, which will produce a lower density ice. The amount of gas can also be reduced, which produces clearer ice, representing the wet-growth process in natural hail. In addition, the manufactured hailstones can be stored in a laboratory freezer at low temperatures (a minimum of $-30°\mathrm{C}$) to help increase the strength beyond what the system can produce. Various combinations of the controllable variables were tested to arrive at those used to create the ice sphere characteristics defined here and in the test protocol (IBHS 2019) to mimic natural hailstones.

During the development of the test protocol, several different asphalt shingle products were impacted with manufactured hailstones of different density and strength combinations, within the values determined by G2015 and additional data collected since. The development work identified three distinct impact modes or ice sphere responses that occurred when impacting asphalt shingles. Manufactured hailstones that were produced with a lower total ${\mathrm{CO}}_2$ content and an ice density near $0.8\mathrm{g}{\mathrm{cm}}^{-3}$ ($>0.4\mathrm{mPa}$) produced impact modes qualitatively described as hard bounce or hard shatter (Figs. 5 and 6). The hard-bounce impact mode occurred when the laboratory hailstone bounced off the shingle, while remaining nearly completely intact. The hard-shatter mode occurred when the laboratory hailstone shattered into numerous small pieces, leaving no ice residue on the surface of the shingle. Despite their different impact behaviors, the configurations used on the laboratory hail production system are identical, as are the mean mass and strength data (Table 2), which indicates that the internal structure of the lab ice sphere and perhaps the radial distribution of density likely controls the final impact mode. Both hard-impact types generally caused deformations (i.e., dents) or breaches in the shingle surface and some amount of granule loss. Manufactured hailstones with a relatively lower strength ($<0.35\mathrm{mPa}$) and lower density from a higher ${\mathrm{CO}}_2$ content produced an impact mode described as soft (Fig. 5). The soft mode resulted in the manufactured hailstone liquifying and leaving a slushy/conical ice residue on the surface of the shingle. This impact mode generally produced a larger surface area of visible granule loss on the surface of the shingle, with less noticeable deformations to the shingle.

The differences in damage produced by the two primary modes (hard and soft), both of which have been observed on real roofs damaged by hail (RICOWI 2012, 2017), along with data showing that these hailstone characteristics occurred in real storms, warranted including each in the new test protocol. Dents and breaches could allow water penetration through the shingle and into the roof, causing damage to the roof structure, building interior, and contents. The loss of granules on the surface of an asphalt shingle is a key factor that affects how fast a shingle may degrade due to exposure to day-to-day weather. The underlying asphalt used in shingles will degrade when exposed to ultraviolet (UV) radiation, making them more brittle and prone to additional damage that could affect their water-shedding function and functional life (Terrenzio et al. 1997; Shiao et al. 2004; Wu et al. 2009). General weathering and heavy rainfall can also remove granules; however, products are often manufactured with excess granules because of this. Hail impacts, even at subsevere sizes, are more efficient than rain drops in removing granules, and repeated exposure to hail, even if it is small, has the potential to reduce the life span of a roof as a result of granule loss, even if no detectable dents or breaches were produced.

The test protocol presented here includes two manufactured hailstone sizes, 3.8 and 5.1 cm (1.5 and 2 in. respectively). For each size category, impacts for the two strength categories and three impact modes previously described are required. The tolerances for the manufactured hailstones used in the test protocol are provided in Table 2, along with the specified impact kinetic energies based on HGW2014. The tolerances in strength are approximately one standard deviation from the mean using a sample size of 150 manufactured hailstones for each size and strength category. The target compressive stress for the soft impact mode is near the 25th percentile of the IBHS hailstone data set (over 2,000 strength measurements of natural hailstones), while the hard impact–mode target compressive stress falls between the mean and median of the field data set, as shown in Fig. 6. Natural hail, especially hail produced in supercells across the Great Plains, trends toward hard-impact modes. However, given the environmental dependencies noted by Sirico et al. (2018), storm-to-storm and regional variabilities are likely and could be large.

While standard test methods use shingles directly from the factory, the IBHS test protocol specifies that shingles must come through standard consumer distribution channels, and must be less than 2 years from the date of manufacture when they are tested. While a single batch of shingles is usually sourced for any given project, the shingles may ship from other parts of the country, particularly if they are IR shingles, which are not commonly sold in South Carolina where the IBHS Research Center is located.

The test specimen design outlined in detail in the IBHS test protocol (IBHS 2019) follows that used by the UL 2218 method. An approximate $1\text{-}{\mathrm{m}}^2$ (3 ft by 3 ft) $2\times 4$ wood frame is constructed with a middle structural member to simulate the presence of a roof truss. A high-grade plywood roof deck and underlayment is installed, and roof cover (i.e., shingles) is installed per the manufacturer’s instructions.

Impacts are focused on the main portion (or field) of the shingles, not along edges, joints, or corners, which is different from the UL 2218 method. Additionally, the test protocol avoids impacts along the outer frame and middle structural member of the test specimen. For 3-tab shingles, which are constructed as a single layer, the total number of impacts required for a product for a given ice sphere size is 20; for architectural shingles, which have portions with single and multiple layers, the total number of impacts required is 40, with 20 for each layer. The acceptable impact locations are outlined in Fig. 7, and impact quantities by mode are given in Table 3. The total number of hard impacts allows for some variation in quantity between hard shatter and hard bounce.

The UL 2218 method assesses damage to the back of a shingle under a microscope. In the field, this is simply not possible without damaging the shingle likely more than that caused by the hail itself. Contractors, inspectors, and claims adjusters assess damage on real roofs by looking at the top side with the naked eye. Additionally, the current test methods only evaluate performance based on the presence of a crack, but other damage modes may play a role in shingle longevity and the potential for water shedding capabilities to be degraded. As the work progressed to develop a new protocol, it became clear that three main categories of damage with underlying subcategories were required and their severities were needed to adequately assess product performance (Fig. 8). The categories are:

Deformations and granule loss are assessed using a damage evaluation tool (described in the next section), while the breach category relies on expert judgment, although development is underway to add this capability to the tool in future versions.

Objective tools capable of quantifying damage states are extremely important and necessary to improve upon the expert judgment-based pass/fail ratings of existing impact test methods (UL 2218, FM 4473). A previous, unpublished study conducted at IBHS utilized four or five investigators to assess two different classes of UL 2218 performance for three different shingle products, and five replicates of each product. For this study, the coefficient of variation for investigators $C{V}_{m,n,p}$ was defined aswhere $x_{m{n}_{1}p}-x_{m{n}_{5}p}$ refer to the mean passing ratings for individual investigators $n_1-n_5$ for panel $p$ of product $m$. The results showed coefficients of variation ranging from 0 to 1.73.

While damage categories could be quantified by conventional means (i.e., caliper, depth gauge, etc.) instead of through expert judgment, the measurements would be crude and the process cumbersome. For example, the use of a digital depth micrometer only provides a single depth dimension, typically to one significant digit, and is subject to human error in using the measurement device. The potential for error with a nonuniform impact crater is relatively high. To duplicate the surface area and volumetric measurements, it would also be highly impractical and time consuming to conduct a large sample of manual depth measurements with a depth gauge coupled with distance measurements. Thus, a better way to quantify damage was needed. Advances in the ability to build 3D mesh grids of objects using sets of photographs provided the means to capture the damage categories in a way that was not previously possible. In addition, this technology offers an efficient means to measure impacts systematically and output the associated data in a uniform fashion.

IBHS partnered with Nemesis Inc. to develop the IBHS-Nemesis Impact Damage Evaluation Tool, a cloud-based computing tool to extract measurements of four of the damage categories listed in Table 4 (excludes breach). The web-based tool allows the user to enter information about an individual impact (i.e., panel name, shingle brand, impact number, etc.), upload a series of digital photographs, and process and quality control the data. The application can be run on a computer workstation with photos captured from another device (i.e., digital camera) or can be run from a mobile device using its integrated camera (phone or tablet). The photographs are taken at an oblique angle (approximately 45°), at an approximate distance (15 cm or 6 in.), but the application is flexible enough that it can accept a range of angles and distances. The images cover a full 360° rotation of the impacted area. Positioning targets (two 0.64-cm-diameter red circles) are placed on the shingle samples to provide orientation and scaling, and to provide boundaries for cropping the images. During processing, the images are stitched together, and a 3D mesh is created. The application takes approximately 10 min of computational time. Adjustments to crop, scale, and calculate granule loss take a fraction of the time once the 3D mesh has been generated.

The gridded 3D data are used to extract the volume of the deformed (dent) and raised (ridge of dent) regions associated with the impact. The granule loss detection algorithm uses the 3D point cloud model at a $0.1\times 0.1\text{-}\mathrm{mm}$ grid resolution. The 3D color image is transformed into a new grayscale image, as not all shingles are the same color. This allows an influence of shingle color to be eliminated from the analysis. With color extracted, tone is used. A thresholding operation is applied to the grayscale image by assigning a value cutoff, such that any pixel less than that value is considered one class, while pixels greater are considered a second class. This defines the background of the shingle (first class) relative to the granules (second class). It produces a binary, black-white image. Next, a mean filter is applied to the binary image to reduce the variation between one pixel and the next, resulting in a grayscale image that represents the local evaluation of the density of white pixels. A second thresholding operation is applied to distinguish normal conditions from those in which the density of white pixels is abnormally low. This step produces another binary image where the black color represents areas with a high probability of granule loss. The final step in the process calculates the surface area of the black areas in the final image. Areas that have a surface area less than $2.58{\mathrm{mm}}^2$ are considered individual granule loss, while those greater are classified as patch granule loss.

An example of the visualized output is shown in Fig. 9. The quantitative nature of the tool’s output allows for severities of each category of damage to be evaluated to determine performance, rather than treating all damage as equal, as is done in the existing test methods (e.g., binary pass/fail determination). It is noted that the tool is more accurate and precise than a human using conventional measurement tools (i.e., caliper, depth gauge).

The quantities determined by the IBHS-Nemesis tool for dent, ridge of dent, and patch and individual granule loss are utilized to assign a nondimensional severity level using integers from 0 to 3, where 0 represents no detectable or very minor damage, and 3 represents severe damage, as outlined in Table 4. The breach damage mode is determined outside of the tool using expert judgment based on visual assessment to determine the severity level, as image processing–based quantification of breaches is still under development. While breaches are often colocated with areas of enhanced granule loss and/or lie at the edge of a deformation, they do not interfere with the tool’s ability to assess the various damage modes.

The breakpoints between quantities for severity levels for dent, ridge of dent, patch, and individual granule loss were defined by the probability and cumulative density functions fit to data from the IBHS-Nemesis Impact Damage Evaluation Tool. A total of 280 5-cm laboratory hail impacts from three common IR shingle products were used to fit a distribution. The data for the dent/ridge categories were best represented through a log-normal distribution, and the granule loss categories were fit to an exponential probability distribution (Fig. 10). The breakpoints between severity levels 0–1, 1–2, and 2–3 were defined as the 25th, 50th, and 75th percentiles for the cumulative distribution fit as provided in Table 4. These values were in general agreement with expert judgment severity scores made by several hail damage experts during the development and validation process. Once the severity levels are obtained for a given impact, they are combined according to the schematic shown in Fig. 11 to produce an individual impact severity score. The individual impact severity scores are averaged for the total number of required impacts to calculate the final performance evaluation rating for the product, which is a rating on a continuous scale between 0 and 3, as given in Table 5.

IBHS has published a list of product performances for the most widely sold IR architectural asphalt shingles. The publicly released ratings provide an overall performance evaluation rating for 5-cm (2-in.) hail, but also provide performance in the primary damage categories.

The test protocol presented here was clearly able to discern performance differences across the array of IR products tested. Current standardized testing methods have no ability to distinguish product performance relativities. Thus, contractors and consumers have previously had little guidance available to assist in selecting a product better able to protect against hail damage.

IBHS publicly released the first set of performance rating results in June of 2019. It was the first time a product performance test attempted to capture the true properties of hailstones and provide a quantitative method to evaluate performance. Over the coming years, evaluation through this protocol can be compared to the real-world performance of these products. The ability to move beyond a pass/fail framework will provide improved information for not only consumers, but manufacturers who wish to take steps to improve their products. The test itself, using manufactured hailstones, can easily be applied to other roof cover or wall siding materials or different hail sizes. However, the damage assessment tool is solely for use with asphalt shingles in a laboratory setting.

Beyond new product performance, work is ongoing to understand shingle performance for naturally weathered products. Five roof-aging farms have been constructed to address this issue, specifically for hail, wind, and fire performance (Giammanco et al. 2015b). The first set of products has been removed and tested after five years of natural weathering in South Carolina, Wisconsin, and Ohio, and analysis is ongoing. Specific to hail performance, the same damage modes (breach, deformations, and granule loss) will be evaluated to determine performance changes after impacting with 5.09-cm (2-in.) ice spheres. Another ongoing project evaluates the granule loss performance of shingles that have been subjected to high concentrations [$94–141\mathrm{impacts}/{\mathrm{m}}^2$ ($8–14\mathrm{impacts}/{\mathrm{ft}}^2$)] of very small hail [1.69–2.54 cm (0.67–1 in.)] for both new products and those naturally weathered for one year in South Carolina. After impacting, these same panels were reinstalled outdoors and will be assessed at future time periods to determine whether granule loss has accelerated as a result of the impacts.

Since 2008, insured losses from severe convective storms have become comparable to those of landfalling hurricanes. Insured losses now routinely exceed $10 billion (Gunturi and Tippett 2017). Hail damage, based on industry catastrophe modeling efforts, in any given year is estimated at 60%–80% of the insured losses from severe convective storms. Brown et al. (2015) was able to show that a large percentage of insurance claims, during a significant hail event in the Dallas-Forth Worth area, was driven by roof damage. As hail-related losses continue to rise, this test protocol and ability to more effectively determine which asphalt shingles may be more resilient to hail will help raise the level of performance and arm consumers with more information than was previously available to obtain quality and cost-effective roofing options for hail-prone regions.

The test method itself, the IBHS Impact Resistance Test Protocol for Asphalt Shingles, is available for download to allow others to conduct testing (https://ibhs.org/wp-content/uploads/2019/06/ibhs-impact-resistance-test-protocol-for-asphalt-shingles.pdf). Some data, models, or code that support the findings of this study are available from the corresponding author upon reasonable request. Data sets available include the IBHS Hail Study repository of hailstone characteristics used to set the parameters for laboratory ice spheres in the new test method. Some data, models, or code generated or used during the study are proprietary or confidential in nature and may only be provided with restrictions. This includes the IBHS-Nemesis Damage Evaluation Tool, which is proprietary.

The authors would like to acknowledge all the personnel who have participated in the IBHS Hail Field Research Program and contributed to the safe collection of the hailstone data used to form the backbone of this testing protocol. The authors would also like to recognize the contribution of Steven A. Cope from Accudyne Systems Inc., who passed away during the development of the laboratory hail manufacturing system. He was instrumental in helping develop the ice production controls, the design and process for the heat exchanger system, and the prototype unit. The advances made using this system would not have been possible without his contributions. The authors also wish to thank Jessie Brown and Tracy Dolan for their tireless work on the research, development, and fabrication of the hail manufacturing system. The Nemesis team, comprised of Jean Brassard, Andre Arsenault, and Kaven Theriault, developed the damage detection algorithm, and we are grateful for their contributions and support in filling this gap. Thanks are also extended to the two anonymous reviewers and the editor for valuable comments and suggestions in improving this manuscript.