AIR Currents

July 25, 2014

Editor's Note: AIR has released a major update to its U.S. severe thunderstorm model. This article—the third in a series highlighting model enhancements—describes the data and research that went into updating the vulnerability module.

Severe thunderstorms—and their accompanying tornadoes, hailstorms, and straight-line winds—cause property damage ranging from the relatively mundane (displaced roof shingles, dented car exteriors) to the catastrophic (entire blocks of leveled houses).

A better understanding of the damage mechanisms associated with this complex risk was a primary goal when AIR undertook a comprehensive update of the U.S. severe thunderstorm model.

The updated vulnerability component leverages state-of-the-art wind and structural engineering research, damage survey data, and analyses of nearly 40 billion dollars of loss data from AIR's sister company Xactware® and 3 billion dollars of location-specific insurance company claims data. We'll discuss some of our most significant findings in this article.

Collaborating with the Engineering Community to Enhance Hail Damage Estimation

Modeling hail damage is a challenging task because of the limited amount of published research in the engineering community. AIR engineers supplemented existing literature with post-disaster damage survey data from the 2011 Dallas-Fort Worth hailstorm, in-house claims data analyses, and experimental and claims data studies from the Insurance Institute for Business & Home Safety (IBHS).

Agerneh DagnewBy: Agerneh Dagnew, Ph.D

Kathryn FobertBy: Kathryn Fobert

Edited by Nan Ma

The 2011 hailstorm damage survey, on which AIR engineers collaborated with the Roofing Industry Committee on Weather Issues (RICOWI), provided a wealth of information on the vulnerability of different roofing materials. More than 100 roofs were inspected and their relative vulnerability was rated from 0 (no visible damage) to 5 (severe damage resulting in potential water leakage). Survey findings were used to develop the relative vulnerability of roof covers for the model's new secondary risk module, which allows the user to input the impact resistance class of various roofing materials.

While damage surveys present ideal opportunities to observe actual damage inflicted by actual events, physical simulations such as those conducted at IBHS's full-scale testing facility in Chester County, South Carolina, are also an invaluable resource. The vulnerability component of the AIR model was informed by results from the IBHS's first-ever indoor hailstorm, a test conducted on February 20, 2013 (and observed by AIR engineers). Results were used to better quantify the exacerbating effects of hail on the building envelope when hailstones are accompanied by strong wind. In addition, this realistic test demonstrated how key construction features such as roof covers (non-impact vs. impact resistance asphalt shingles), wall sidings (fiber-cement vs. vinyl), and windows (vinyl vs. aluminum) perform under a hailstorm with varying hailstone sizes. This experiment, coupled with other data sources (Underwriters Laboratory (UL), FM Global standards, and literature) informed how AIR engineers set the relative importance of the secondary risk features for hail. AIR also used claims analysis studies conducted by IBHS following the 2011 Dallas-Fort Worth1 and the 2003 North Texas2 hailstorms to inform the new model and further validate the relative vulnerability of different roofing materials.

Blending Post-Disaster and Location Level Claims Data to Create Realistic Tornado Wind Fields

An integral part of estimating losses from tornadoes is a realistic representation of the wind field—that is, how wind speeds degrade as a function of distance from the tornado core. While hurricane wind fields have long been a focus of research, the literature on tornado wind fields is sparser.

Thus a primary goal of AIR's damage survey after the 2013 Moore, Oklahoma tornado was to examine the variation in damage within the footprint. AIR visited areas along the tornado track where EF-4 and EF-5 winds were reported. Figure 1 shows the progression of damage from the core of the tornado (photo A) to the outer edge (perpendicular to the track) of the damage footprint (photos B-D) in residential areas.

Cat Bond Figure 1 Event 1
Figure 1. These photos from AIR's damage survey of the 2013 Moore, Oklahoma, EF-5 tornado show complete devastation along the centerline of the tornado in (A) and progressively lesser damage as the distance from the centerline increases, such as unseated roof decking, roof covering torn loose, and damage to windows and walls in (B), moderate damage on roof covering, roof decking, and siding in (C), and finally minor damage from flying debris on the upper part of the buildings in (D). (Source: AIR)

AIR also undertook a collaborative study of the 2011 Tuscaloosa, AL and Joplin, MO tornadoes, this time with the National Wind Institute (NWI) of Texas Tech University (TTU). The AIR and TTU teams together analyzed damage survey data on more than 11,000 geocoded buildings within the affected areas, including high-definition video recordings, interviews with local residents, and detailed photos. Ratings were assigned according to the EF scale "Damage Indicators" along with the severity of damage for each building surveyed (Figure 2). Conditional mean damage ratios were also calculated for all the insurance claims within the damage footprints and were used to validate the damage ratings assigned to the field data. Both claims and survey data analyses confirm that risks at the tornado core experienced complete damage, while the severity of damage decreased as distance from the tornado core increased.

Cat Bond Figure 1 Event 1
Figure 2. Field survey data from the May 2011, Joplin, MO tornadoes (red represents a high severity of damage and blue represents minimal damage) (Source: AIR)

AIR engineers measured the relative distance of each damaged building from the centerline of the tornado damage footprint, then estimated the wind speed at each building's location based on the severity of the observed damage. The results were used to inform the new tornado wind speed profile implemented in the updated AIR model, which agrees well with experimental tests.

Distinguishing the Damage Mechanism of Straight-Line Winds and Tornadoes

AIR engineers updated and validated the straight-line wind damage functions using insurance company loss experience, a wealth of claims data from our sister company Xactware, damage survey data, and published engineering research. These were complemented by research using computational fluid dynamics (CFD) to better understand the aerodynamics of buildings under the wind fields of tornadoes, straight-line winds, and hurricanes.

CFD simulates fluid flows around buildings using numerical solutions (read more about CFD in this AIR Currents article). It provides valuable insights into the damage mechanisms and relative vulnerability of various building types and their features by evaluating the distribution of wind pressures on building envelopes.

While hurricane winds are cyclonic at the macro scale, experimental research and CFD analyses reveal that hurricanes and straight-line winds have similar vertical wind profiles at location level. Thus, the extensive amount of research undertaken at AIR on building vulnerability to hurricane winds was used to inform damage function development for the straight-line winds that accompany severe thunderstorms. Like hurricanes, the intensity of straight-line winds gradually increases with height—unlike tornadoes, which have very high wind speeds in the near-surface regions and lower wind speeds at higher elevations (Figure 3). The negative correlation of tornado winds with height causes overloading on low-rise structures, which can lead to their collapse. In addition, the high suction of tornadic wind introduces uplift of building roofs. Based on this and other numerical and experimental studies and the fact that tornadoes are always accompanied by wind-borne debris, straight-line wind damageability in the AIR model is less than that of tornadoes for the same intensity.

Cat Bond Figure 1 Event 1
Figure 3. Vertical tornado wind profile as compared to a hurricane wind profile (Source: AIR)

Peer Review

AIR's updated model, including the damage module, underwent an extensive peer review by Timothy Marshall, P.E., of Haag Engineering and Dr. Harold Brooks of the National Severe Storms Laboratory. In the video below, the reviewers share some of their thoughts.

Beyond Modeled Loss Validation: Incorporating Claims Data to Assess the Relative Vulnerability of Sub-Perils

AIR engineers analyzed nearly 300,000 location-specific insurance company claims and 4 million Xactware losses by sub-peril (see Figure 4) to validate the results of the newly updated model.

Cat Bond Figure 1 Event 1
Figure 4. Spatial distribution of residential Xactware claims for major events in 2010 and 2011 (Source: Xactware)

The observations from the claims data analyses were generally in line with expectations. They revealed, for example (and not surprisingly), that newer buildings are less vulnerable than older buildings to both wind and hail damage. Roof covers deteriorate over time, especially for shingle-roofs, which were common in the claims data. Newer buildings are generally better able to resist wind loads since they are better designed and conform to updated building codes and newer design requirements.

The location-specific claims data was useful for assessing the relative vulnerability of various building types and contents to each of the modeled sub-perils (tornadoes, hail, and straight-line winds) and for developing the model's new secondary risk module, as a significant portion of this data contains important building characteristics such as construction type, year-built, and number of stories, as well as key secondary building attributes such as roof cover, roof type, roof age, and siding material. In addition to the building damage functions, the claims data was also useful for accessing the relative vulnerability of contents to sub-perils (Figure 5).

Cat Bond Figure 1 Event 1
Figure 5. Claims data was useful for accessing relative content vulnerability (Source: AIR)


The vulnerability update to the AIR U.S. severe thunderstorm model was the result of the best sort of collaboration between AIR and the wider scientific community, as well as an unprecedented wealth of detailed claims data that enables risk differentiation at a highly granular level. Touchstone® users can now enter many secondary risk features based on their own exposure information—or in the absence of detailed building characteristics, use the vulnerability functions that AIR researchers developed for "model" buildings appropriate to the building's location and year-built. Companies can now analyze results for tornadoes, hailstorms, and straight-line winds individually, as well as for all three combined to gain further insight into how this complex peril impacts their risks.

1 Brown, T., and Pogorzelski. H. (2013). "Claims analysis study of May 24, 2011 Hailstorm in Dallas-Forth Worth," Insurance Institute for Business and Home Safety (IBHS).

2 IBHS (2003). "IBHS insurance claim hail study, Investigation into insured losses and damages to single-family homes resulting from the April 5, 2003 North Texas hailstorms," Institute for Business and Home Safety (IBHS), Tampa, FL.




You’re almost done.
We need to confirm your email address.
To complete the registration process, please click the link in the email we just sent you.

Unable to subscribe at this moment. Please try again after some time. Contact us if the issue persists.

The email address  is already subscribed.