1. Introduction

As part of the notorious “tornado alley,” the state of Oklahoma has the second highest ranking in both tornado concentration (approximately 7.75 tornadoes per 10 000 square miles per year) and number of killer tornadoes (approximately 1.44 killer tornadoes per year; Grazulis 1997). On 3 May 1999, a severe tornado outbreak occurred in Oklahoma and other southwestern states. It spawned violent tornadoes, rated up to F5 on the Fujita scale (Fujita 1981), which affected a large area, including Oklahoma City, Oklahoma, and its suburbs.

Forty-six people from Oklahoma and Kansas died. Thousands of homes and businesses in Oklahoma were destroyed by some of the most powerful tornadoes ever recorded in the state. In Moore, Oklahoma, a suburb of Oklahoma City, 7500 homes, about one-half of the suburb's single-family residences, were damaged or destroyed. Nearly 2000 houses in Oklahoma City were destroyed. According to Insurance Services Office, Inc., the insured property losses were in excess of $1.5 billion (including $955 million in Oklahoma)—one of the worst tornado-related catastrophes in U.S. history (Brooks and Doswell 2001).

Because the tornado struck some highly populated areas, it provided a vivid exhibition of the impact of hazardous natural forces on the built environment. In section 2 of this paper, aerial images are correlated with field surveys for postevent damage assessment. The effects of internal pressurization are discussed in section 3.

2. Damage assessment using aerial photography

Immediately following the event, the authors visited some of the hardest-hit areas in Oklahoma. The timely visit provided a unique opportunity to examine the damage patterns from an engineering perspective. A commercial aerial photography crew was also retained to survey the damage from Chickasha to the northeast of Oklahoma City. About 100 high-resolution aerial photographs were taken, clearly showing the damage along the tornado paths. The high-resolution imagery enables detailed observations of property damage, which are complementary to field observations.

Damage was most extensive in south Oklahoma City, Moore, Del City, Oklahoma, and Midwest City, Oklahoma. Rows and rows of houses were brought down to their foundations. The path of the destruction was 19 mi long and measured between a 0.5. and 1 mi wide. The tornado damage in these areas was categorized as F4–F5 by the National Weather Service. Figure 1 shows the tornado path at south Oklahoma City. The tornado was generally traveling from the southwest and, in this particular segment, from 240°. From the aerial photograph, the damaging width of this particular tornado path was estimated at 0.25 mi. The common residential constructions in this area are single-story wood frames with brick veneer and composition shingled roofs. As indicated on Fig. 1, a block of homes was in the tornado path, with one end of the street block at the edge of the path and the other end around the center of the tornado. These houses exhibit various levels of damage, from no damage, minor damage, and major damage, to completely destroyed (Fig. 2). The houses in this community were built in about the same period with almost identical style. As a consequence, the different damage patterns observed for these houses tend to reflect the different wind speeds experienced across the tornado path. At one end of the street, houses show no sign of damage. As the wind speeds increase toward the center of the tornado path, the houses begin to suffer roof damage, then wall damage, and finally, toward the other end of the street near the tornado path center, the houses were totally destroyed. This observed damage gradient demonstrates the powerful yet concentrated tornado wind forces. Based on Figs. 1 and 2, the damage states are reported as a function of the distance from the tornado path center, shown in Fig. 3. It clearly shows the trend of diminishing wind speeds and damage away from the path center.

As an interesting note, one of the authors revisited the disaster areas one year following the event. Out of the 28 destroyed or damaged homes on the street block shown in Fig. 2, 15 homes were repaired or rebuilt while 13 lots were left empty (Fig. 4). The recovery rate is slightly more than 50%.

Midwest City is at the tail of the 19-mi-long tornado track that cut through Oklahoma City. The Development Services Department of Midwest City provided a damage survey map as illustrated in the left portion of Fig. 5. Homes are marked by an X, /, \, or – to indicate destroyed, major, minor, and slight damage, respectively. The damage survey map corresponds very well with the aerial photograph in the right portion of Fig. 5. In this particular area, as illustrated in Fig. 5, 84 homes were destroyed, 37 homes suffered major damage, 76 homes had minor damage, and 46 homes survived with slight or no damage. Typical aerial image and ground-level photograph of each damage state are compared in Table 1. Because most of the wind damage is done to the building envelope, especially the roof system, the damage state could be identified by the observations from high-resolution aerial images. This approach enables rapid and fairly reliable damage assessments for large areas, especially when road access is limited immediately following a tornado outbreak. Federal and local governments, as well as insurance companies, could use this tool to assess damage rapidly and to allocate limited resources efficiently. The approach is also suited well to other wind-related catastrophes such as hurricanes and nontornadic severe storms.

Besides direct high winds, tornadoborne debris is another source of damage (Zhao and Martinez 2000). In one instance, some metal roof trusses of a church building were carried across the street and impacted an auto repair shop, causing substantial damage (Fig. 6). Doswell and Brooks (2002) illustrated the “cones-of-damage” pattern. The tips of the cones are usually associated with weak structures. The failure of these weak structures may generate windborne projectiles that can cause additional damage to nearby downwind structures. The damage area would tend to widen toward the path center in a cascading fashion. Figure 7 clearly demonstrates the observed cones-of-damage pattern. The picture inset in Fig. 7 identifies one of the weak structures. The impact of windborne debris may cause breaches to the envelopes of downwind buildings. In many instances, the sudden breach of a building envelope may lead to internal pressurization. This increase of internal pressure, combined with uplift forces acting on the roof, can lead to catastrophic roof failures.

3. Effects of internal pressurization

The Federal Emergency Management Agency (FEMA)–sponsored Building Performance Assessment Team (BPAT) surveyed the damage following the tornado outbreak. Data collected by one of the authors as a member of the BPAT indicate again that internal pressurization is a major contributor to poor building performance under severe wind loading conditions (BPAT 1999). Several of the figures shown in this section are reproduced from BPAT (1999).

The majority of the homes in the subdivision shown in Fig. 8 are single-story wood frames with gable or hip roof styles, or combinations thereof. Most house plan configurations are simple L, T, or rectangle shapes. Roof decking was observed to be mostly dimensional lumber with some oriented strand board (OSB) applications. Roof rafter and wall top-plate connections were typically toe nailed with three nails and no added straps or clips. Roof covering was typically composition shingles. Most homes had single-skin aluminum, noninsulated, and nonreinforced double garage doors.

The damage states of the two homes (at locations 1236 and 1237) shown in Fig. 9 are significantly different, even though they are located directly across the street, approximately 95 ft, from each other, and may have experienced relatively similar wind conditions. The home located at 1236 had seven broken windows from debris impact, one breached glass entry door, a permanently deformed garage door from negative pressure (see right picture inset in Fig. 9), and lost approximately 60% of its roof cover. The home at 1237 lost its entire roof and several exterior walls. The remaining homes in Fig. 8 show similar “across-the-street” damage gradients. The home located at 1217 did not lose its entire roof and the home located at 1213 did not lose any substantial roof covering (suggesting lower wind speeds). Both of them did sustain severe roof framing damage, however, owing to increased uplift loads following internal pressurization. Observed damage states listed in Table 2 quantify the differences between the odd- and even-numbered homes. It is clear that the damage level decreased as the distance from the tornado path center increased.

Several failed garage doors were observed lying to the back of the garage for many of the odd-numbered homes, suggesting that the garage doors failed in positive pressure. These failures of garage doors are believed to have initiated or contributed significantly to the catastrophic roof and exterior wall failures for the odd-numbered homes, a direct consequence of load increase due to a large breach in the building envelope. A partial roof failure is shown in Fig. 10. In this case, the garage door was found within the garage (see picture inset in Fig. 10, taken from a different angle). The observed location of the failed garage door and the localized roof damage suggest that the failed garage door may have initiated or played an important role in the roof failure. It is observed that many homes had a significant amount of structural damage to the garage area and immediate surrounding structure, but less structural damage at the opposite side of the building where no garage structure was located. Another example of internal pressurization and roof uplift is shown in Fig. 11 for the dwelling located at 1213. The garage door was found inside the garage, again suggesting the door failed in positive pressure. Figure 12 shows strong evidence of the early stages of roof uplift between the garage roof and exterior wall. The ceiling drywall connected to the garage ceiling rafters was pulled away from the exterior wall perimeter, indicating that the whole roof frame was in the process of uplifting. Tension cracks in the brick veneer and a large gap along the length of the right exterior wall between the roof and top plate were observed (Fig. 13), all evidence of roof uplift.

A potential mitigating factor to entire roof systems failing may be the overall roof geometry itself. Roof geometry will affect the overall strength of the roof system based on its framing configuration. For example, hip framing will have more wall-to–top plate connections than will gable framing. Because the total integrated uplift forces acting on both shapes of roof are approximately the same for most wind directions, the hip roof construction yields a stronger system (Surry 1990).

Field observations following the Oklahoma tornadoes provide strong evidence of roof and exterior wall failures initiated by breaches to the building envelope leading to internal pressurization and subsequent load increases. For residential buildings, a significant contributor to catastrophic failures due to internal pressurization appears to be the failure of single-skin, noninsulated, and non-reinforced double garage doors. Breaches of windows and entry doors may also cause significant damage through internal pressurization. However, if wind speed and direction are unfavorable for the development of high local loads, the effects may not be as dramatic as those associated with larger breaches such as garage doors.

4. Conclusions

The 3 May 1999 tornado outbreak produced extreme wind hazards. It provides a unique opportunity to evaluate some innovative damage assessment methods and to examine damage patterns from an engineering perspective. The two following findings summarize this paper:

1. The damage states of structures correlate well with the aerial image observations. High-resolution aerial photography is a rapid and fairly reliable tool for large-area damage assessment following wind-related catastrophes.

2. Increased internal pressurization due to the sudden breach of residential garage doors was again observed to contribute to progressive failures.