Lessons learned from the damage produced by the tornadoes of 3 May 1999



Charles A. Doswell III [1] and Harold E. Brooks

NOAA/National Severe Storms Laboratory

Norman, Oklahoma


A manuscript submitted as an Article


Weather and Forecasting

(Special Issue on the 3 May 1999 tornadoes)


February 2001


Corresponding author address: Dr. Charles A. Doswell III, Cooperative Institute for Mesoscale Meteorological Studies, The University of Oklahoma, Sarkeys Energy Center, 100 East Boyd Street, Room 1110, Norman, OK 73019-1011

[1] Current affiliation: Cooperative Institute for Mesoscale Meteorological Studies, Norman, OK

NOTICE: This manuscript has been submitted to Weather and Forecasting, published by the American Meteorological Society. See their policy statement regarding copyright. There may be differences between the paper as it appears here and the final version, owing to revisions suggested by the reviewers.


After the tornadoes of 3 May 1999, the Federal Emergency Management Agency formed a Building Performance Assessment Team (BPAT) to examine the main tornado paths during the outbreak and to make recommendations based on the damage they saw. Some aspects of the BPAT final report are reviewed and considered in the context of tornado preparedness in Kansas and Oklahoma. Although the preparedness efforts of many public and private institutions apparently played a large role in reducing casualties from the storm, a number of deficiencies were found during the BPAT's evaluation. Especially in public facilities, there are several aspects of tornado preparedness that could be improved. Moreover, there is clear evidence that a non-negligible fraction of the damage associated with these storms could have been mitigated with some relatively simple and inexpensive construction enhancements.


1. Introduction

As events unfolded in Oklahoma and Kansas on 3 May 1999, it became clear that this day was going to test severely the tornado preparedness of communities in these two states. The largest outbreak of tornadoes ever to hit the state of Oklahoma left scores of paths of destruction, including an F-5 tornado with a 68 mile long path that crossed several Interstate highways and devastated several suburban areas of Oklahoma City. In the days following the event, various groups conducted numerous surveys in an attempt to assess the damage to a residential, public, and commercial buildings, and to evaluate the performance of the tornado preparedness efforts. This included the first-ever survey of a tornado event by a Building Performance Assessment Team (BPAT), created by the Federal Emergency Management Agency (FEMA). The intensity and extent of the damage was high enough to warrant the creation of FEMA's first ever tornado BPAT; as noted in the final report (BPAT 1999),

The number of tornadoes that occurred on May 3, 1999, in Oklahoma and Kansas, their severity, and the level of devastation they caused have not been seen in a generation within the United States.

BPATs have been used by FEMA to review how well structures have performed in various disasters (hurricanes, earthquakes, etc.), with the goal being to reduce the damage created by similar events in the future. The team created to review building performance in the paths of the 3 May 1999 tornadoes was charged specifically to evaluate the performance of:

In this paper, we are not going to review all the details of the BPAT's findings and recommendations; those are provided in the final report of the team (BPAT 1999) and interested readers are advised to consult that report for a comprehensive treatment of the findings. Rather, herein, we wish to consider how the findings of that report relate to the overall process of tornado preparedness that has been evolving in the tornado-prone parts of the United States. Although that process includes the meteorological tasks of forecasting and warning for tornadoes, we are not going to review the forecasts and warnings, either. They are discussed elsewhere (Andra et al. 2001; Edwards et al. 2001). It is becoming increasingly apparent within the meteorological community that even if their forecast products are done perfectly, if they do not lead to the saving of lives and property, those efforts will have been in vain. On this fateful day, it appears that preparedness in Oklahoma and Kansas paid off, in terms of lives spared.

The notion of an integrated warning system (IWS) has been discussed in Doswell et al. (1999), wherein it was suggested that "The IWS process actually can be said to begin well before any severe weather has even begun to loom on the horizon." That is, tornado preparedness involves considerable effort in the months and years preceding an eventful day, such as 3 May 1999. Since tornado and severe thunderstorm forecasting began in the United States, weather forecasters have assumed considerable responsibility for creating enough public awareness that their forecasts and warnings will have the desired effects. For example, the spotter program requires training of local storm spotters, usually by the nearest office of the National Weather Service (NWS). Although not specifically charged with considering all aspects of tornado preparedness, during the course of the BPAT's investigation some issues associated with tornado preparedness were discovered.

Section 2 of this paper provides a few, selected highlights from the BPAT final report; the goal is not to be comprehensive but, rather, to note specific issues that will be examined herein. Section 3 presents the findings of an informal survey conducted during the BPAT on the character of home remnants left in the wake of the F-5 tornado in the suburbs of Oklahoma City. In section 4, the finding from the BPAT survey are discussed in the context of the tornado preparedness program, and section 5 presents conclusions we draw from the study of the tornado damage.


2. Selected findings from the BPAT report

a. Load path integrity

In engineering terminology, the so-called load path is the set of structures designed to carry the weight of a building. For homes, this is typically the framing. Under normal conditions, the weight of the building is carried from the roof along the framing and, ultimately, to the ground (Fig. 1). In Oklahoma, most of the residences do not have basements. Rather, most are on concrete slabs; a few are built on "crawl space" foundations. On the other hand, most homes in Kansas were on foundations that included basements, with some having crawl space foundations.

The issue of the attachments along the load path of a home was clearly a major factor in the damage viewed by the BPAT (see BPAT 1999; Marshall 2001). In most cases, these attachments might marginally meet the building codes, but they provide little in the way of resistance to forces creating uplift. In some cases, the attachments did not meet building codes. In either case, when structures were subjected to tornadic winds, they would fail first at the weak points along the load path. Once failure was initiated at a weak point on the load path, considerable structural breakdown would follow. When a structural failure occurs, it typically begins with a breach of the external "envelope" of that building. This allows wind to enter the breach and exert additional force on the walls and ceilings of breached rooms, leading to additional failures such as wall collapse and loss of roofs. Side loads from the winds could literally slide a home off its slab or foundation, resulting in additional structural failures that could lead to total destruction of the home.

In most cases of damage to homes, either in Oklahoma or Kansas, the attachments along the load path were inadequate to resist side loads and uplift generated by the wind. Generally speaking, the building codes (which typically are not in force in rural areas) currently in effect should prevent structural damage in winds up to about 80 mph (i.e., corresponding roughly to F-1 windspeeds. Beyond that, increasing winds should result in increasing amounts of structural damage. For cases where the codes are not met, of course, much weaker winds could initiate structural failures.


b. Projectiles

Once structural failures occur, the tornadic wind field becomes filled with debris, acting as projectiles (or, "missiles") that fly at high speed. When tornadoes interact with homes, a common projectile is broken framing timbers and pieces of masonry. These projectiles can breach the external envelope of other homes and thereby can initiate failures that might not otherwise have occurred. The BPAT found numerous instances where structural failures occurred on the far periphery of a tornado path, typically attributable to building code violations or marginal construction practices. These failures would generate projectiles of various sizes, perhaps up to the size of whole roofs, that would hit other buildings and initiate further structural failures. In effect, off to the side of the tornado's centerline, there would be cones of damage that would widen as they approached the path center, illustrated schematically in Fig. 2. Each of these damage cones would begin at some notably weak structure, surrounded on three sides by other structures suffering no structural damage. One weak structure, therefore, could create additional structural damage in nearby structures that would otherwise have suffered little or no damage.

Of course, it is well known that projectiles are a major cause of casualties in tornadoes (in addition to persons becoming airborne and being crushed by collapsing structures). Within those parts of a tornado path experiencing F-2 and greater windspeeds, some structural damage is nearly inevitable. However, projectiles of all sorts create the potential for structural damage outside of the part of the path with F-2 and greater windspeeds.

The BPAT observed that some very large projectiles are created in violent tornadoes. When a large, heavy projectile, such as an automobile or utility pole strikes a structure, severe damage is obviously the result. When a violent tornado strikes a populated area, the projectile load carried in the debris cloud of the tornado represents a substantial hazard compared to an equivalent tornado in an open, rural setting. The projectile load adds considerably to the damage potential, beyond that of the winds themselves, as well as representing a major hazard to human (and animal) life.


c. Garages

Since many of the structures involved in the tornadoes were typical suburban single-family residences, most of them had attached garages. Garage door construction is relatively flimsy, and many of the garage doors failed. As with other initial failure modes, this constituted a breach of the envelope, permitting winds to enter the garage and create additional side loads and uplift. It is not uncommon for garages to be built with framing and attachments that are below code requirements (since garages are not considered living space). The failure of the garage structure, being attached to the rest of the home, often initiated structural failures in the rest of the home. Depending on the orientation of the garage to the tornadic winds, the failure of the garage door often resulted in major structural damage to the entire home.


d. Manufactured homes

In several examples considered by the BPAT, it was evident that manufactured homes generally were considerably more vulnerable to tornadic winds than the typical site-built frame home. In some cases within the violent tornadoes, even when the manufactured homes were anchored to the ground, the anchor straps were either broken or pulled out of the ground. Anchoring varied considerably from place to place during the survey.

One manufactured home park in Kansas had a community shelter. It was found that a number of problems existed regarding that shelter, including:

The shelter was partially underground, and the flat roof was covered with a loose stone aggregate that likely would become a source of airborne projectiles in strong winds.


e. Tornado preparedness

Surprisingly, in many public facilities (some of the schools, many of the workplace buildings, etc.) there was only minimal preparation for dealing with tornadoes. Most of the facilities had little or no idea where the safest portions of the structure might be. In some examples, safe areas had been designated that did not appear to be consistent with the actual structural integrity of those areas. Most public facilities did not have a NOAA Weather Radio and had only a rudimentary plan in place for tornado safety. One school used what amounted to manufactured homes on their campus for additional classroom space, and in the event of a warning, the students in those buildings would have to evacuate outdoors into the main building to seek shelter. That school did not have a NOAA Weather Radio on-site. Moreover, the hallway in the main building designated as a shelter area did not appear to be structurally safe, owing to the presence of "clerestory" windows, which are vulnerable to projectile penetration and reduce the resistance of the corridor walls to lateral wind forces.

Some facilities, like one manufacturing plant in Kansas, had done considerable planning for tornadoes. They used NOAA Weather Radios, had well-conceived plans for getting occupants to adequate shelters, and actually practiced executing their plans during drills twice per year. The plant in Kansas was damaged by the tornado, but their tornado action plan was executed, such that there were no casualties. Interestingly, the person responsible for safety in the plant called for movement to shelter as if it was a drill, apparently to reduce the potential for panic.

Most of the public seems to have been using the information that the NWS and other organizations have been disseminating, regarding what to do in the event of a tornado. Given the excellent NWS performance in this event (see Andra et al. 2001; Edward et al. 2001), and the widespread media attention prior to the arrival of most tornadoes, it seems obvious that the death toll was much less than what might have been expected. Given that many thousands of structures were heavily damaged or destroyed, a community that was less prepared might have suffered many more casualties (Doswell et al. 1999). Note that although the violent tornado that struck in the Oklahoma City area was the first tornado ever that produced $1 billion in assessed damage, this figure may be misleading, owing to a steady increase of wealth in the United States (see Pielke and Landsea 1999; Brooks and Doswell 2001a).


f. Shelters

In spite of being within what might reasonably be called "tornado alley", there were relatively few tornado shelters found within the paths of the tornadoes in Oklahoma and Kansas. A number of below-ground shelters were found; many of them had a number of problems (notably, poor ventilation and water infiltration), but the biggest issue with most of them was the door and its attachments. Many of the doors were little more than plywood sheathed in metal and some had deteriorated since their installation. Most troublesome were the absence of solid attachments, with flimsy hinges and a simple sliding bolt to hold the door. The general recommendation for shelter doors is six attachment points, including three sturdy hinges and three deadbolts, with all six attachments spaced roughly equally along the door sides. Such a thorough attachment of a door was not found during the BPAT survey. A relatively new in-ground shelter design performed well in the F-4 damage associated with the tornado near Wichita, Kansas, in spite of having some minor deficiencies in ventilation and door attachment.

A handful of FEMA-recommended "safe rooms" had been constructed in and near tornado paths. Apart from some questions regarding door attachments, these safe rooms performed as intended. One was on the outskirts of F-4 damage in Midwest City, Oklahoma. Another was on the periphery of F-5 damage near Bridge Creek, Oklahoma.


3. Core remnants study

As an unofficial part of the study, two of us (CAD and PRM) conducted a non-systematic examination of those homes that had lost roofs and exterior walls but still had some interior walls standing (and so would be rated as F-3 damage). In effect, during the BPAT survey, we considered "targets of opportunity" as we walked along the damage path in residential housing areas. There was no attempt to find and assess all such remaining interior walls (hereafter called "core remnants"). The goal was to determine the type of rooms left standing in F-3 damage areas. According to the Fujita scale (see, e.g., Fujita 1981), at damage levels of F-4 and F-5, no interior walls are left standing, so occupant survivability is a matter of luck, if the occupants can't reach adequate shelter.

After the pioneering engineering studies done by the engineers at Texas Tech. University in the years since the Lubbock, Texas tornado of 11 April 1970 (Fujita 1970), the NWS and other agencies charged with tornado safety have recommended that occupants of residences who cannot gain access to proper shelter (either below-ground or within an in-house safe room) should seek shelter in interior rooms (without windows). The idea is to put as many walls between the occupants and the exterior as possible. Bathrooms, closets, and under stairways have been recommended, owing to the additional structural elements that might resist the tornadic winds. It seems that most residents within the path of the 3 May 1999 tornadoes followed these instructions. We were interested to confirm the validity of this advice.

Results of our informal study are shown in Table 1. During the first part of the work (data points 1-38), we only determined which type of room was left among the core remnants. As our study proceeded, however, we found that several rooms could remain within the core at the same time, so data points 39-94 do not necessarily represent independent core remnants. The finding is clearly in favor of seeking shelter in bathrooms or closets, if adequate shelter is not available. Kitchens often were reduced to one standing wall that often included cabinets, on the other side of the wall from a hallway leading to a bathroom or closet. Whereas the bathrooms or closets often provided what clearly would have been adequate shelter in most such cases, the kitchens on the other side of the wall were often open to the outside on three sides and so would be very hazardous. Occasionally, projectiles penetrated the walls of otherwise relatively intact core remnant rooms, rendering them unlikely places to avoid serious injury or death.

Table 1. Results of informal "Core Remnants" survey. Subjective determination was made concerning the likelihood of avoiding serious injury or death within the remnant rooms, corresponding to the "yes/no" columns under each room type. A "yes" means a high likelihood of avoiding serious injury or death, whereas a "no" means a lowlikelihood of avoiding serious injury or death. The first 38 data points were all independent core remnants, and each core remnant was determined to be one of the set (kitchen, closet, or bathroom). Subsequent data points (38-91) are not necessarily independent core remnants, because it was determined that core remnants might include any or all of the set (kitchen, closet, or bathroom).




Data point





























4. Discussion of findings

It is very clear from the outcome of this devastating outbreak of tornadoes that, on the whole, the public in Oklahoma and Kansas is reasonably well-prepared to deal with tornado disasters. The limited toll of fatalities, given the magnitude of the event, is a tribute to the many long hours of effort, spread out over many years, by the public and private institutions responsible for public safety in this part of "tornado alley." We believe it is unlikely that this sort of performance would be possible in parts of the country where the perceived hazard of tornadoes is much lower than in Oklahoma and Kansas.

However, the story regarding damage to structures from the storm is much less optimistic. It is hard to understand how and why construction practice is so marginal in a part of the nation that shows the highest likelihood of violent tornadoes (Fig. 3). It seems obvious that no affordable, practical home could survive intact from the impact of F-4 or F-5 tornadic windspeeds. However, several facts are relevant to this. First of all, violent (F-4 and F-5) tornadoes are unlikely events, even when given that a tornado has occurred. Violent tornadoes only represent a few percent of the total number of tornadoes. Moreover, in a violent tornado, the strongest winds (i.e., those of F-4 or F-5 intensity) occupy only a small percentage of the path. From what the BPAT report shows, some relatively modest enhancements to the construction of a house (e.g., "hurricane clips" that tie together the framing members along the load path, and anchor bolts attaching the bottom plate of a wall to the foundation or slab) could reduce the damage, and especially the structural damage, done to homes in the parts of a tornado path experiencing winds of F-3 or lower intensity. The installation costs for these enhancements are of order $100-200 for new construction, even including labor. Retrofitting these construction enhancements into existing construction is not very practical and would be much more expensive.


Figure 3. Frequency of days per millenium with one or more violent (F4/F5) tornado touchdowns within a grid box 80 km on a side, based on tornado data from 1921-1995.

Limiting the damage to residences is not so simple as having one's home built properly. As the BPAT findings make clear, having poorly-constructed homes in one's neighborhood increases your chances for serious damage from F-3 or lower windspeeds even if one's own home is capable of resisting structural damage on its own. The only way to reduce the damage potential from surrounding structures is for all the structures to be enhanced over existing codes. That is, it would be necessarily to make entire communities more resistant to tornadic winds to gain most of the benefit from the construction enhancements.

It is clear, as well, that efforts must be undertaken to reduce the likelihood of serious injuries or fatalities associated with manufactured homes. Curiously, in the Oklahoma City area tornadoes, no manufactured home parks were hit. Only in the unincorporated community of Bridge Creek was there a concentration of manufactured homes hit by the tornado, resulting in 11 of the 36 fatalities produced by the F-5 tornado. The chances of having another violent tornado sweep through a populated area without hitting manufactured home parks is probably not very high (see Brooks and Doswell 2001b).

In some parts of the country, there are "parks" where recreational vehicles of those on extended vacations stay for months at a time, representing another highly vulnerable residence. Local authorities often don't know how many people are in such temporary housing. In locations dominated by the presence of manufactured homes or recreational vehicles, the inherent vulnerability of such residences suggests that construction enhancements are unlikely to do much to reduce the potential casualties. Therefore, the way to protect the residents (even if they are only temporary) must be to provide adequate shelters. There are many possible options for this that might be practical.

It is also difficult to understand how and why the public in Oklahoma and Kansas generally has lost interest in building tornado shelters, either below ground or as in-home safe rooms. Naturally, with FEMA's financial assistance, some fraction of the rebuilt homes in the paths of the 3 May 1999 storms will now include an in-home safe room. The challenge clearly is to convince the public that it is worth an investment of several thousand dollars to construct some form of tornado shelter, even though they've not yet experienced a tornado and the event's return frequency is on the order of 1000 years for any particular square mile.

Although the 3 May 1999 tornadoes were devastating, they are by no means indicative of the worst possible events. In addition to not hitting very many manufactured home parks, the tornadoes didn't hit any public facilities with a large vulnerable population. Such events have happened in the past, but it seems that many public facilities (including workplaces) are not very well prepared to deal with tornado disasters. The United States has not had a large school in session hit by a significant tornado for since the 1950s. Given the findings of a considerable shortfall in tornado preparedness by some schools and other public facilities, this represents a considerable challenge to those institutions that have accepted responsibility for public safety.

Recently, the North Texas Council of Governments has used the data regarding the 3 May 1999 event to simulate what might happen if a similar outbreak of tornadoes hit in North Texas. The results of this study are staggering, including the potential for several billion dollars in damage and thousands of casualties. The outcomes of these experiments depend strongly on the specific circumstances of the outbreak: time of day, day of the week, the details of just what lies in the path, etc. Such hypothetical scenarios represent what we propose to be a sobering dose of reality. As noted in Doswell et al. (1998), public apathy as a result of the long interval between major disasters continues to be a major factor in tornado preparedness.


5. Conclusions

The events of 3 May 1999 contain several lessons regarding tornado preparedness. The first is that, at least within the Plains region of the United States where tornadoes are most common, the level of preparedness is sufficient to result in a substantial reduction in casualties. Most people were aware of what to do in the event of a tornado and the actions they took resulted in a remarkably low death toll for this event, given the intensity and number of tornadoes. The forecast and warning system in place is working well, especially for major events. The concept of the integrated warning system, that includes a collaboration between public and private sector institutions provided excellent warnings and information to the public. Although we have not discussed it, the emergency response after the event appears to have been excellent and no doubt was responsible for amelioration of considerable post-event suffering.

However, another important lesson is that the preparedness program still has some aspects that need improvement. This is particularly noticeable in the area of public facilities, including workplaces. Even within the nation's center for weather awareness, many schools, sporting facilities, workplaces, shopping areas, and other public facilities do not have adequate tornado preparedness plans. The current minimal use of the NOAA Weather Radio as a means for getting NWS warnings as soon as possible, especially for public-use facilities (schools, factories, sporting events, etc.) is a disturbing finding and we conclude that this sorely needs to be addressed. Many public facilities need to work with someone trained to recognize the presence (or absence) of truly safe shelter areas within their structures. If no safe shelter exists within the existing buildings, then shelters clearly need to be constructed. Practical plans for getting occupants to shelter in the event of a tornado need to be developed in many places, and need to be practiced at least twice per year.

It seems inexplicable that the public has increasingly chosen not to build tornado shelters in the tornado-prone areas of the United States. The choice to build a shelter is clearly a personal one, since it requires a considerable investment, and should remain so for private homes. In tornado-prone areas, home builders should be encouraged to offer "safe rooms" as an option in new construction, where it is possible to amortize the cost of a safe room over the length of a mortgage. However, we conclude that serious consideration should be given to mandating the construction of shelters for mobile home and recreational vehicle parks, as well as in many other public facilities (especially in schools and day-care centers).

Damage mitigation has been given only minimal attention over the years. Reducing damage may well be of lower priority than reduction of casualties, and it is apparent that many believe there is nothing that can be done to resist violent tornadoes. It seems likely that reduction of damage could well be a factor in continuing the reduction of hazards to the public. That is, if we can minimize the projectile load in tornadoes by eliminating unnecessary structural damage outside of F-4 and F-5 windspeed areas (only a tiny fraction of the total area affected by tornadoes), then it is likely that this would result in reduced casualties. Home builders in tornado-prone areas should be encouraged to offer enhanced load path attachments as an option, preferably within whole developments, to reduce damage created by debris from weakly-constructed nearby homes. Perhaps the insurance industry could be encouraged to offer premium reductions for homeowners who live in disaster-resistant developments (including safe rooms). We are very much encouraged by FEMA's Project Impact, which seeks to support the development of communities that are resistant to a wide range of disasters, including tornadoes.

As our understanding of tornadoes has grown, it has become increasing apparent that damage mitigation from tornadoes is quite possible. In addition to the cost savings associated with taking steps to prevent unnecessary tornado damage, this may also cut down on casualties due to tornado-generated projectiles. As with other aspects of tornado preparedness, a process of damage mitigation necessarily begins well before a potential tornado day. If we are to have an impact on how decisions are made regarding damage mitigation and shelter construction, a considerable effort must be expended by those public and private institutions responsible for public welfare in disasters. Hopefully, we can develop a collaboration among all those institutions, rather than a disorganized collection of individual programs.

Acknowledgments: We are especially grateful to Peter Montpellier, for his contributions to the study of core remnants during the BPAT survey.



Andra, D. L., Jr., E. Quoetone and W. F. Bunting, 2001: Warning decision making: The relative roles of forecaster expertise and radar algorithms. Wea. Forecasting, 16,[this issue].

Brooks, H. E., and C. A. Doswell III, 2001a: Normalized damage from major tornadoes in the United States: 1890-1999. Wea. Forecasting, 16, [in press].

_____, and _____, 2001b: A brief history of deaths from tornadoes in the United States. Wea. Forecasting, 16, [this issue].

BPAT, 1999: Midwest Tornadoes of May 3, 1999: Observations, Recommendations, and Technical Guidance. FEMA 342, Federal Emergency Management Agency, Mitigation Directorate, Washington, D.C., 195 pp.

Doswell, C. A. III, A. R. Moller and H. E. Brooks, 1999: Storm spotting and public awareness since the first tornado forecasts of 1948. Wea. Forecasting, 14, 544-557.

Edwards, R., S. F. Corfidi, R. L. Thompson, J. S. Evans, J. P. Craven, J. P. Racy, D. W. McCarthy and M. D. Vescio., 2001: Storm Prediction Center forecasting issues related to the 3 May 1999 tornado outbreak. Wea. Forecasting, 16, [this issue].

Fujita, T. T., 1970: The Lubbock tornadoes: A study of suction spots. Weatherwise, 23, 160-173.

Marshall, T. P., 2001: Damage survey of the Moore, Oklahoma tornado. Wea. Forecasting, 16, [this issue].

Pielke, R. A., Jr., and C. W. Landsea, 1998: Normalized hurricane damages in the United States, 1925-1995. Wea. Forecasting, 13, 621-631.