When the elementary school in Mulhall, Oklahoma, was destroyed during the tornado outbreak of 3 May 1999, a great loss of young life was prevented by chance alone. The tornado hit after school was out of session. The weather scientists and researchers who read this journal attempt to save lives by improving forecasting and public warning systems. The author has attempted to save lives for more than two decades by designing storm shelters to serve students who do not have the option of fleeing an approaching storm. The reasons why so few of Oklahoma's schools provide storm protection are defined as a lack of adequate construction and maintenance funding and a widespread low-risk perception. The estimated cost to provide all Oklahoma school children with protection from a direct hit by an F5 tornado is about $1.3 billion. A plain-language discussion of the aboveground shelter that was designed for the rebuilt school includes site planning principles, structural design options, the importance of including a brick veneer, shelter door design options and criteria, strategies to resolve fire code violations, and the importance of designing for a rescue after the storm has passed. A call is made for a national storm shelter rating system based upon differing occupant safety levels. A call is also made for a partnership among the professionals who are designing large-scale shelters, scientists and researchers studying tornados, and building code officials similar to the partnership that has vastly improved earthquake-resistant building design.
The destruction of the Mulhall–Orlando Elementary School in Mulhall, Oklahoma, was a major state news event in 1999. As the architect for the rebuilt school, which included a new storm shelter, I was invited to give an oral presentation at the National Symposium on the Great Plains Tornado Outbreak of 3 May 1999, held in Oklahoma City, Oklahoma, on the one-year anniversary of the outbreak. I agreed to participate in the symposium for three primary reasons: 1) to learn what the scientific and research community has learned about storm shelter design, 2) to share with the scientific and research community what my architectural firm has learned and developed independently during the last 27 years, and 3) to encourage the development of information flow channels between the scientific community and the architects who are designing and building storm shelters.
The scientists who research tornados to predict more accurately their occurrence and to improve public warning systems have saved countless lives through their work. The author has attempted to save lives through the design of storm shelters. Two choices are typically available when facing an approaching tornado: flee or hide. Through the simple logistics of the matter, if nothing else, school children are denied this choice and their only available option is to hide.
We are an architectural firm that specializes in the design of public schools. With most of our clients located in “Tornado Alley,” we have more than two decades of experience in the design of large storm protection facilities for school children. Working independently during this time, the author's firm has developed design guidelines for public school storm shelters. To our knowledge, little or no research on tornado shelters was available to architects in the 1970s. Therefore, because the results of a tornado hit often look like those of an explosion, we looked to the explosion control industry for research data and commercially available products. This industry's traces can be located throughout our work.
Our design guidelines naturally have evolved over time, and the latest iteration was applied to the design of the aboveground storm shelter for the Mulhall–Orlando Elementary School. The design was completed in the autumn of 1999, and the design principles that were employed are presented below without reliance on the specialized language and expressions of architecture, which may not be easily understood by many readers.
The design, for reasons described below, departs significantly from the recommendations for community shelters published by the Federal Emergency Management Agency (FEMA) in July of 2000 in that it provides protection from higher wind speeds over longer time periods and from larger, more powerful wind-blown missiles. Thus, the design presented below should be viewed as only one of a myriad of possible approaches to building a community shelter.
2. Mulhall, Oklahoma
On the evening of 3 May 1999, Mulhall, a small rural community located about one hour of driving time north of Oklahoma City, was hit by a large tornado, and the consolidated Mulhall–Orlando Elementary School was destroyed. The damage to this school was such that, if school had been in session when the tornado hit, the loss of life would have been horrific. The school building was not equipped to provide even rudimentary storm protection.
It is a surprising fact that most Oklahoma schools do not provide storm protection for their students for the following reasons:
The odds of any school actually being hit by a tornado are extremely small, and a low-risk perception is widespread among school administrators.
This low-risk perception extends to the conventional wisdom that most tornadoes seem to form after 1500 local time, when school is out of session and the buildings are generally not occupied. For example, Oologah Public Schools, of Oologah, Oklahoma, was hit by an evening tornado about 10 years ago. This district has not, and does not plan to, build a tornado shelter.
Extremely limited construction and maintenance funds are available. For example, the Norman Public Schools, in Norman, Oklahoma has recently removed the earth berm protection from its facilities. During the demolition of the old elementary school at Wetumka, Oklahoma, a serviceable underground storm shelter was found that had been walled off sometime in the past and forgotten.
Unfunded mandates such as Americans with Disabilities Act (ADA) compliance, fire safety upgrade compliance, and Title IX athletic compliance must be implemented. School districts are subject to lawsuit for failure to comply with ADA or Title IX and to involuntary closure for failure to perform fire safety upgrades. Without a mandate, tornado shelters fall to the bottom of the priority list.
These reasons establish why so few of Oklahoma's schools provide storm protection. We have estimated on a bid-today basis, without a contingency, the cost to provide all Oklahoma school children with shelters similar to that designed for the Mulhall–Orlando Elementary School. Oklahoma has 1824 public school sites. Site visits to approximately 80% of these sites over the years, including some very small rural sites where a single shelter could serve multiple schools, indicate that 1750 shelters would be required at an average cost of $750 000 each, totaling approximately $1 312 500 000. This estimated cost is an impossible amount for Oklahoma, or for any other state for that matter, to achieve. On the tenuous assumption that shelters meeting the FEMA community shelter recommendations could be built for only one-half of this cost, we are left with an equally impossible construction cost of $656 250 000. Only the federal government has the financial resources to provide the necessary grant.
Therefore, we were pleased when the Mulhall–Orlando school board mandated the inclusion of a storm shelter into the rebuilt elementary school that would, to the greatest extent possible, guarantee the safety of the shelter's occupants and would address the fears of the students and the community. For Mulhall, the residents had faced nature at its most violent and had survived. They did not believe that their good fortune would hold a second time—their low-risk perception vanished and their fear increased. The financial problems of building the shelter were resolved when FEMA agreed to provide mitigation funds for the construction project.
3. Occupant safety levels
Our experience profile includes aboveground storm shelters of vastly different degrees of storm resistance and, therefore, occupant safety levels. Beginning in the 1970s, our designs were subdivided into three categories of shelter as follows:
Class A Class A is a storm-proof facility that will provide occupants with a reasonable degree of protection from a direct hit by a 100-yr tornado with winds in excess of 318 m h–1. A 100-yr tornado is defined as a storm that can be expected to occur in a given area one time in a 100-yr time period (similar to the standard definition of a 100-yr flood). Note that, as stated in the introduction, a class-A shelter design exceeds the recent recommendations set forth by FEMA.
Class B Class B is a storm-resistant facility that will provide occupants with a reasonable degree of protection from a near miss by the largest of tornadoes or a direct hit from a weaker storm. Strategies such as earth berming of exterior walls and strengthened corridors are included in class-B shelters. This design generally meets the recommendations set forth by FEMA for community shelters.
Class C Class C is an enhanced protection facility that will remain basically intact and provide occupants with a minimum level of protection from such hazards as flying glass and other light debris.
To date, a standardized rating system for storm shelters providing differing levels of occupant safety has not been established. Therefore, a high level of confusion exists as to what constitutes a “storm shelter.” Many people in Oklahoma rely on so-called shelters, usually public buildings, that would only qualify as class-C facilities under our rating system. For the most part, these facilities were once designated as fallout shelters. This designation provided protection from a very different threat.
Reference has been made above to FEMA community shelter recommendations. These recommendations were published in July of 2000 in FEMA (2000). This publication is promoted as a “standard” but we do not agree with its one-size-fits-all idealogy. This author believes that certain facilities, such as schools, require more protection (class A) while other facilities may, for cost, low-risk perception, or other reasons, choose less protection (class C).
Some very serious liability issues are apparent. First, people trust that a shelter will protect them from all possible tornados and go to the shelter for safety because of that trust. Second, parents trust that a public school shelter will protect their children from all possible tornadoes. They do not rush to the school to retrieve their children and flee when a warning is issued because of that trust. Last, it is a tragedy when people are killed during a tornado. The litigation over people killed or injured in a shelter that was not correctly promoted for its actual level of occupant safety could be truly impressive. We have the means to provide the people with the truth and we should do so.
The national adoption and promotion of a shelter rating system is strongly encouraged. With the new elementary school at Mulhall incorporating our definition of a class-A shelter based on occupant safety, the discussions that follow will be limited to this type of facility.
4. Why build a class-A shelter?
A class-A shelter is an extremely strong facility and is designed to resist our definition of a tornado with a once in a 100-yr frequency. Therefore, such a shelter is not appropriate for every situation because a class-B, or even a class-C, facility will provide near-perfect protection from all but the most violent storms. The Mulhall shelter, as all of our class-A shelters have been, was designed to protect elementary school children. The level of risk that is acceptable for the lives of these children must be carefully evaluated. With a population of 200–600 children gathered in a single facility, can any failure be tolerated for any reason?
Oklahoma schools build shelters because of fear, and fear alone. Without fear, most schools are unwilling to spend their limited financial resources on even class-C protection. As the memory of 3 May 1999 has faded into the background, we are not being asked about shelter design by our current school clients. After the next significant tornado event, the questions will begin again.
With fear as the motivation, we typically will be asked “can you guarantee our childrens' safety?” during a public school board meeting where shelter design is being considered. This question cannot be answered “yes” with a FEMA community shelter (our class B)–based design. Public school shelter design must address the real fear that parents have. Parents need to hear that the shelter is designed to resist the strongest tornado ever recorded, they need to see the concrete walls under construction, and they need to feel the weight and strength of the shelter doors. Without this reinforcement, parents may not trust the lives of their children to the school, thus negating the very reason for the existence of the shelter.
5. Why build aboveground shelters?
Both practical experience and common sense dictate that shelters should be located underground. We agree strongly. However, schools have a specific problem that discourages underground shelter construction. Simply put, the problem is money. Lack of adequate funding impacts shelter design in the following ways:
Public school shelters must serve a dual use to be cost effective. Our class-A facilities have employed the dual uses of cafeterias, libraries, and, as at Mulhall, band/vocal music facilities.
Most Oklahoma schools are single-story facilities. The location of one essential dual-use function on a separate level would be highly disruptive to normal school operations.
The Americans with Disabilities Act (ADA) requires that an underground shelter in a government-owned facility be serviced by an elevator. Cost efficiency then demands that additional school functions also be located on the elevator-serviced basement level, creating an underground school. Educators in Oklahoma might state that such a facility would not be conductive to a good educational program.
Oklahoma schools do not have adequate maintenance funding. This funding inadequacy would impact an underground shelter in terms of elevator maintenance cost. Because many areas in Oklahoma have a high water table, an additional maintenance cost would be incurred in preventing water intrusion into the facility. The critical importance of the maintenance factor can be seen in the disappearance of most of the earth-bermed class-B facilities that were built throughout Oklahoma in the 1960s and 1970s.
These factors tend to force storm shelters to the surface and into harm's way.
6. Design of the Mulhall–Orlando Elementary School
a. Applicable design standards
As noted in the introduction, we developed design guidelines for public school shelters by working independently without access to scientific and research community input. The first realistic design guidelines (FEMA 2000) were published in July of 2000, 2 months after the symposium and almost 10 months after the design of the Mulhall shelter was completed and construction begun. This publication states that use of lower wind speeds than we used is acceptable but does not seem to have been written with an elementary school in mind. Design and construction of the shelter were also controlled by the BOCA National Building Code published by the Building Officials and Code Administrators International (1996), and the Life Safety Code published by the National Fire Protection Association (1997).
With data unavailable from the scientific and research community, we utilized the resources of the explosion control industry. Of particular utility over the years were knowledge and expertise provided to us by a number of explosion control door manufacturers.
Tornado-generated winds create structural loads of less than 3 lb in.–2. This loading level is very small in comparison with explosion-generated loading that the engineers employed by these manufacturers must contend with on a daily basis. The knowledge and resources held by these companies are significant and, after many discussions over the years, we have come to trust their expertise. In addition, we cannot find a reason to differentiate among missiles thrown by an explosion, a hurricane, or a tornado.
b. Site design
Because it is widely known that most Oklahoma tornadoes move with a significant eastward component of direction, the shelter is located on the eastern side of the new school (Fig. 1). This allows the bulk of the school building to block wind-blown missiles and, possibly, to dissipate wind strengths. The eastside location also tends to eliminate obstacles that may trap debris around the shelter making occupant exiting difficult after the storm has passed.
It was unfortunate for this shelter design that the Mulhall Elementary School was rebuilt on the original site of the destroyed facility. Only 2.4 acres in size, this original school site is significantly smaller that the 10-acre smallest recommended area for an elementary school. Therefore, teacher, administrative, and visitor parking had to be located on the eastern side of the site adjacent to the shelter to allow an adequate playground. If at all possible, this situation should be avoided and vehicle parking should be located remote from the shelter on the western side, where the bulk of the building will protect the shelter from wind-blown vehicles.
c. Structural Design
Structural design (Fig. 2) for an aboveground storm shelter must be capable of withstanding two primary loads: loading from the force of the wind and impact loading from wind-blown missiles. For wind loads, a 318 mi h–1 wind speed was used. We are aware that ground-level winds of more than 300 mi h–1 have not been observed. Buildings are, however, three-dimensional, and the gymnasium of the new elementary school rises to 30 ft above ground level and the shelter roof rises to about 16 ft above ground level. We do not have adequate assurance that wind levels on the full height of exterior walls will not approach or exceed the 300 mi h–1 level to risk the lives of the sheltered children. The foundation is designed to resist the overturning moment from the wind load.
The missiles likely to impact the shelter probably would come from the materials located uptrack. An analysis of these materials before design is futile because they are subject to vast and significant changes after construction of the shelter is complete. Therefore, the worst-case scenario must be addressed. The following standard missiles as developed by the explosion control industry were used with respect to impact loading:
4 in. × 12 in. wood plank weighing 200 lb traveling at 140.5 mi h–1,
3-in.-diameter schedule-40 pipe weighing 114 lb traveling at 84.5 mi h–1,
1-in.-diameter rod weighing 8 lb traveling at 136.4 mi h–1,
6-in.-diameter schedule-40 pipe weighing 285 lb traveling at 74.3 mi h–1,
12-in.-diameter schedule-40 pipe weighing 743 lb traveling at 67.5 mi h–1, and
13.5-in.-diameter wood pole weighing 1490 lb traveling at 52.5 mi h–1.
The impact load from the 13.5-in.-diameter pole equals almost 0.75 × 106 lb. Because of the large surface area and the soft, crumpling, energy-absorbing nature of modern vehicles, such impacts are not considered.
Three optional structural designs were evaluated for this shelter. We have never employed design option 3.
Exterior walls of 12-in.-thick concrete block supporting a 9.5-in.-thick reinforced concrete roof poured over steel-form decking are used. All of the concrete block cells are grout filled with vertical steel reinforcing in every cell and horizontal steel reinforcing installed at every fourth course vertically.
Exterior walls of 12-in.-thick concrete supporting a 9.5-in.-thick reinforced concrete roof poured over metal-form decking are used. Steel reinforcement is embedded in the walls in a staggered pattern to form a 4 in. × 4 in. reinforcement grid.
A 1-in.-minimum-thickness welded steel plate in the walls and roof is used with numerous available structural support options.
The concrete wall option was selected for the Mulhall–Orlando Elementary School because of the specific shape of the floor plan (Fig. 3) and the high level of fear held by both the students and parents who witnessed their destroyed school.
d. The importance of the brick veneer
A veneer of brick masonry is installed on the exterior side of the shelter walls (Fig. 2). This veneer is anchored to the concrete walls with ties that allow both horizontal and vertical movement between the two materials. In addition to the aesthetic role, the brick veneer provides a crumple zone that will absorb energy and lessen impact loads. The brick veneer will also distribute large missile impact point loads over a wider area of the concrete wall thus reducing the potential for interior spalling of the concrete that can occur under large impact loads. Interior concrete spalling can create secondary missiles inside the shelter that could kill or severely injure occupants.
The steel-form decking on which the concrete roof is poured serves the same function in preventing spalled missiles from being emitted from the concrete roof. Additional wall spalling protection is provided by a layer of 5/8-in.-thick gypsum board anchored to metal furring strips on the interior side of the concrete wall.
e. The shelter floor plan
The storm shelter for the Mulhall–Orlando Elementary School also serves the dual use of a band and vocal music facility (Fig. 3). The size and shape reflect this use as well as the necessity to accommodate the students, teachers, and staff of the school in the event of an emergency situation. With about 1580 ft2, to include office and storage areas, the shelter provides about 8 ft2 per standing occupant. This is oversize for the shelter. However, because the shelter also serves the community at large, it will serve 450 standing occupants at 3.5 ft2 each. Although small for the entire community, this is all the space that the available construction funding allowed.
The shelter has an opening on the west side to provide an interior access connection to the remainder of the school. A second opening to the exterior is located on the probable lee side at the northeast corner. Both openings are provided with daily use doors that swing outward from the shelter and are protected by wind- and missile-resistant doors that swing to the interior of the facility.
f. Door design criteria
The doors are the weakest points of any storm shelter and require special design attention. The doors must resist the same wind-generated loads as the structure. Two design strategies are available. They are 1) design the doors to resist both wind and missile impacts and 2) design the doors to resist wind forces only and to allow missile penetration. Any penetrating missiles are caught and rendered harmless in a catchment vestibule located behind the doors. This strategy is less expensive. The dual use as a music facility with its acoustic requirements prevented the use of catchment vestibules. Therefore, missile penetration–resistant doors were used for the Mulhall–Orlando Elementary School.
Doors manufactured by the explosion contaminant industry were used. They were designed to meet the following criteria for a 100-yr tornado:
The doors must resist 3 lb in.–2 loads (400 mi h–1), which generate loads that equate to about 4.5 tons on a 3 ft wide by 7 ft high door.
The doors must resist the design wind load with the loading direction being either toward or away from the interior of the shelter for a duration of 5 min.
The doors must resist penetration or significant distortion by repeated impact loading from the standard missiles noted above for a duration of 5 min.
The door operating mechanism must be on the interior side of the door so that it is protected from missile damage.
The doors must open inward such that they are not obstructed by debris that prevents escape.
The installed doors provide the safety necessary during a tornado hit, but they are a major and very serious violation of the fire code as noted in section 7 below.
g. The late-arriver problem
All storm shelters, including an in-home safe room, have a late-arriver problem generated by the timing of the doors being closed and the willingness of the shelter inhabitants to reopen the doors if they hear someone frantically knocking. Public schools are a supervised environment well accustomed to fire drills during which the evacuation of all children is ensured. The same administrative solutions will apply to assure that all students, teachers, and staff are safely inside the shelter before the doors are closed. When used as a community shelter after school hours, a responsible person such as a member of the police or fire department, available on site to open the shelter, must be designated to determine when the doors will be closed. We are open to all suggestions for a design in lieu of an administrative solution to this problem.
7. Fire code violation
All construction is controlled by building and fire codes that are written by national associations and adopted into law by states, cities, towns, and other jurisdictions. Therefore, compliance with the requirements of these codes is mandatory. Construction of the Mulhall–Orlando Elementary School was controlled by the BOCA National Building Code (Building Officials and Code Administration International 1996), which requires that all doors shall swing in the direction of egress. The code is very clear when it states that doors must swing toward the outside of the structure to prevent entrapment. This dictum has been in place since the disastrous nineteenth-century ladies garment factory fire in New York City in which hundreds of trapped women died. It does little good to protect our children from a tornado only to see them perish in a fire.
A variance from the code must be granted to achieve a successful class-A shelter design (Fig. 4). We secured this required variance for the Mulhall Elementary School in the following way:
The entire school, to include the storm shelter, is protected by an automatic fire sprinkler system. The shelter is provided with a separate underground water supply to maintain operational sprinkler capacity in the event that the remainder of the building is destroyed.
The storm doors are held in the open position by 1500-lb electromagnetic locks until an emergency arises.
The emergency lock release switch that allows the storm doors to be closed is located securely in the school principal's office.
The release of the storm door locks sounds a buildingwide audible tornado alert alarm.
A second keyed emergency lock release switch for after-hours operation is located on the shelter exterior with the community police or fire department personnel in possession of the keys.
In this way, protection is provided against tornadoes and also for the more likely event of a fire.
8. After-the-storm rescue
A successful storm shelter design must provide safety from the storm and must also provide a safe exit from the shelter after the storm has passed. The shelter design must reduce the possibility of interior entrapment or help to reduce the length of any such entrapment to very short time periods. The possibility of panic among occupants, occupant injury requiring medical attention, and foul air are obvious reasons for this mandate.
First, a rescue plan must be established and implemented between school personnel and local civil defense agencies. In addition, the following strategies were employed:
Battery-powered emergency lighting that provides foot-candle levels far in excess of building code minimums, and ceiling tiles provided with hold-down clips to help to prevent them from falling in the event of a shelter depressurization during the storm, will reduce the possibility of panic.
Two remotely located exits, with one on the probable lee side, reduce the possibility of debris entrapment, as does the interior swing of the storm doors.
Out-swinging daily use doors are fully glazed. If they are blocked by debris, the glass will be broken by the debris or can be easily broken by the shelter occupants to allow escape or, at worst, to allow air and light flow into the interior.
Mechanically operated gravity air vents are provided, in the event that powered ventilation is lost, allowing the occupants to introduce fresh air into the shelter.
Although not provided at Mulhall–Orlando Elementary School, we also recommend the installation of an exterior underground communication system that is independent of building systems.
These design elements should provide the shelter occupants with an escape even if the new school is again destroyed by a massive storm.
Given the staggering cost of large-scale shelter construction, the very real concern architects have with liability issues, and the lack of research data presented in architectural journals and seminars, information flow to those working on tornado shelter design is critical for improved, cost-effective storm shelters.
A partnership between architects designing shelters, building code development officials, and the scientific researchers in the field is strongly suggested. This partnership should be similar to the partnership that has developed modern standards for earthquake-resistant building design. This partnership should be accomplished before a tragic loss of life occurs. This loss of life would have happened if the Mulhall tornado had occurred a few hours earlier.
For a starting point, data that we need from the scientific community to improve our shelter designs include the following:
a clear definition of a once-in-a-100-yr tornado similar to the definition of the well-known and accepted definition of a 100-yr flood;
measurements of actual tornado wind speeds at building level, from the ground to about 30 ft for single-story buildings and higher for multistory buildings;
analysis of building structural performance under rotational, in lieu of straight-line, wind loading of varying diameters and speeds;
analysis of building structural performance under shear stress from multiple-vortex tornadoes;
analysis of rotational wind effects among multiple buildings similar to the urban plaza wind studies conducted after the John Hancock Building glazing problem in Boston when swirling winds continuously blew the windows out of this high-rise office tower, and
analysis of glass and glazing performance under very high wind speeds.
New information on the design of tornado shelters has been released in the year since the symposium was held. As noted above, FEMA released its design recommendations for community storm shelters. Although we may disagree with some of this publication's recommendations as they apply to community shelters for elementary schools, the recommendations do provide architects designing shelters with the first available design liability standard and provide many useful ideas and strategies. This publication is an important first step. It also is extremely important to the architects who have put their careers on the line with their shelter designs and who have been working without a safety net. Second, the U.S. military has developed a design guide to protect its facilities against terrorist explosions. This currently classified publication has the potential to greatly improve the designs of all types of community shelters.
For the Mulhall–Orlando Elementary School shelter design we tried to err on the conservative side. New ideas and performance data will continue to allow shelter design to become safer and more cost effective. The people of Mulhall, Oklahoma, however, can take comfort in the fact that if their community is threatened in the future, their children will have a safe haven from the storm.
Corresponding author address: John W. Buckley, 4015 Morain Court, Norman, OK 73072. Email: email@example.com