Tornado Damage Survey at Moore, Oklahoma

Abstract

On 4 May 1999, the Wind Science and Engineering Research Center at Texas Tech University dispatched three survey teams to the Oklahoma City area to conduct a tornado damage survey. The author was the leader of one of the teams whose purpose was to survey tornado damage in and around the suburb of Moore, Oklahoma. The survey team was given five tasks: 1) to map out the damage path and assign F-scale numbers to damaged buildings, 2) to document the performance of housing, 3) to interview witnesses, 4) to document projectiles, and 5) to assess the performance of any above- or belowground shelters within the damage path. This paper will present the methodology utilized for conducting the tornado damage survey and will summarize the observations and findings of the survey team. Wind speeds necessary to cause the observed damage to residences were found to be significantly lower than the established F-scale wind speeds. The author returned to the disaster area three months later and discovered that, in general, the quality of new home construction had not improved.

Full text

582 V
OLUME
17WEATHER AND FORECASTING
q2002 American Meteorological Society
Tornado Damage Survey at Moore, Oklahoma
T
IMOTHY
P. M
ARSHALL
Haag Engineering Company, Dallas, Texas
(Manuscript received 5 February 2001, in final form 28 June 2001)
ABSTRACT
On 4 May 1999, the Wind Science and Engineering Research Center at Texas Tech University dispatched
three survey teams to the Oklahoma City area to conduct a tornado damage survey. The author was the leader
of one of the teams whose purpose was to survey tornado damage in and around the suburb of Moore, Oklahoma.
The survey team was given five tasks: 1) to map out the damage path and assign F-scale numbers to damaged
buildings, 2) to document the performance of housing, 3) to interview witnesses, 4) to document projectiles,
and 5) to assess the performance of any above- or belowground shelters within the damage path. This paper
will present the methodology utilized for conducting the tornado damage survey and will summarize the ob-
servations and findings of the survey team. Wind speeds necessary to cause the observed damage to residences
were found to be significantly lower than the established F-scale wind speeds. The author returned to the disaster
area three months later and discovered that, in general, the quality of new home construction had not improved.
1. Introduction
A tornado outbreak occurred in central Oklahoma
during the evening of 3 May 1999. One tornado struck
densely populated suburbs of Oklahoma City. Burgess
and Magsig (2000) indicated this violent tornado began
near the rural community of Amber at 2327 UTC and
traveled northeast through mostly rural areas paralleling
Interstate Highway 44 before reaching southern sections
of Oklahoma City around 0020 UTC. The tornado con-
tinued through the city of Moore and crossed Interstate
Highway 35 (I-35) around 0030 UTC before turning
more northerly and striking Del City and Midwest City
before dissipating around 0047 UTC (Fig. 1). The Fed-
eral Emergency Management Agency (FEMA 1999a)
reported that this single tornado killed 42 people, in-
flicted 800 injuries, and caused over $1 billion in prop-
erty damage. The tornado lasted approximately 80 min
and left a continuous damage path 61 km (38 mi) long,
giving an average translational speed of 33 kt (38 mi
h
21
). The damage path averaged 400 m (1300 ft) wide
through suburban areas.
Within 24 h after the tornado, threesurvey teams were
dispatched to the disaster area by the Wind Science and
Engineering Research Center at Texas Tech University.
The survey teams focused their attention on the densely
populated communities where building damage was
most severe. The author’s team was given the task of
conducting a damage survey from rural Newcastle
Corresponding author address: Timothy P. Marshall, Haag Engi-
neering Co., P.O. Box 814245, Dallas, TX 75381-4285.
E-mail: timpmarshall@cs.com
through the city of Moore to Interstate Highway 240 (I-
240) on the east side of Oklahoma City. Our team was
asked 1) to map out the damage path and assign F-scale
numbers to damaged buildings, 2) to document the per-
formance of housing, 3) to interview witnesses, 4) to
document projectiles, and 5) to assess the performance
of any above- or belowground shelters within the dam-
age path.
Most of the structures damaged by the tornado were
one- and two-story, wooden-framed houses with one-
or two-car attached garages. The vast majority of houses
were constructed on concrete slab foundations (Fig. 2).
Such a large number of houses of similar construction
allowed a large sample size by which to assess and to
compare building damage. A report summarizing the
findings of all three teams was published by Gardner et
al. (2000).
2. Damage survey methodology
a. Logistics
A meeting was held in Oklahoma City the day after
the tornado to establish the procedures to be followed
during the damage survey. Radar images and newspaper
accounts were gathered to define better the locations of
the damage path. Team members realized it was essen-
tial to begin the damage survey as quickly as possible
to determine the extent of building damage. Cleanup
operations already had begun and had accelerated as
fair weather continued for the next several days. Work
crews had cleared all primary roads almost immediately
and opened secondary roads within a few days after the

J
UNE
2002 583MARSHALL
F
IG
. 1. Damage track of the tornado that struck Oklahoma City and suburbs on 3 May 1999.
The tornado had a continuous damage path of 61 km (38 mi) and lasted approximately 80 min.
Only a portion of the tornado path actually struck densely populated areas (hatched area). Our
damage survey covered the path from A to B. Map adapted from Burgess and Magsig (2000).
tornado. However, police and U.S. Army National
Guard personnel had cordoned off the damaged areas,
and permission was required to enter. Our team pro-
ceeded to walk much of the damage path to talk to
witnesses and to examine failed building components
closely.
An aerial survey by helicopter also was conducted of
the tornado damage path. We quickly determined the
overall extent of the tornado damage path and noted
interesting areas for later study on the ground. Similar
methodologies for conducting damage surveys have
been described by McDonald and Marshall (1984) and
Bunting and Smith (1990).
b. Equipment
It was important to have proper equipment for con-
ducting the damage survey. Detailed road maps were
obtained before the damage survey began. Still cameras
with both print and slide film were employed to pho-
tograph the damage. A wide-angle lens on one camera
captured the overall damage scene, whereas a zoom lens
on another camera captured specific details. A second
camera also served as a backup in case one of the cam-
eras malfunctioned. A notepad and writing pens were
brought along for documentation purposes. A tape re-
corder was utilized to record the locations of the pho-
tographs and to record pertinent observations. A tape
measure was helpful to determine the distances between
objects and to obtain dimensions of building compo-
nents and projectiles. Business cards and magnetic signs
mounted on the survey vehicle provided identification.
Hard hats were worn in the disaster areas. A hand-held
global positioning system receiver was available to pin-
point ground locations accurately.
c. Use of the F scale
Fujita (1971, 1973, 1981) developed the F scale for
rating the degree of wind damage to buildings. The F
scale is a subjective, visual interpretation of wind dam-
age ranging from F0 to F5 based on the increasing se-
verity of damage, primarily to a ‘‘well-constructed’’ or
‘‘strong’’ wooden-framed house. Team members as-
signed an F-scale rating to each damaged house within
our study area based on the damage descriptions pre-
sented by Fujita. However, as Grazulis (1993) noted,
the single-paragraph descriptions of damage given by
Fujita are vague and limited in scope. Thus, the author
developed additional damage descriptions to aid in as-
signing F-scale numbers. A house was rated F0 damage
if it had lost a few roof shingles, had a downed television
antenna, had broken windows, or had a damaged garage
door. A house was rated F1 damage if large areas of

584 V
OLUME
17WEATHER AND FORECASTING
F
IG
. 2. Typical cross section of exterior wall in a Moore, OK, house, showing building
nomenclature and fastener type.
the roof covering had been removed, if there had been
loss of any roof decking, if the gable end had been blown
in or out, or if the garage door had failed, causing uplift
of the garage roof or collapse of the garage walls. A
house was rated F2 damage if most of the roof structure
had been removed but perimeter walls remained intact.
A house was rated F3 damage if the roof structure and
most perimeter walls had been removed, leaving interior
walls standing. A house was rated F4 damage if the
house structure had collapsed, leaving a pile of debris
on the foundation. A house was rated F5 damage if the
majority of the house structure and contents had been
displaced downwind from the foundation.
Fujita (1971, 1973, 1981) also assigned wind speed
ranges to the numerical values in his F scale. Wind speed
ranges were derived empirically by dividing the gap
between Beaufort 12 (33 m s
21
,74mih
21
) and Mach
1 (about 330 m s
21
, 738 mi h
21
) into 12 nonlinear
increments. The F-scale wind speeds were defined as
the ‘‘fastest 1/4-mile speed’’ being longer in duration
than a gust, usually in the 5–10-s range for most tor-
nadoes. The F scale was deemed ‘‘experimental’’ by
Fujita, and he awaited engineering assessments of tor-
nado damage to help to calibrate the wind speed ranges.
Engineering assessments of tornado damage by Minor
et al. (1977) and Minor (1977) questioned the accuracy
of the F-scale wind speeds, especially when they ex-
ceeded 56 m s
21
(125 mi h
21
). Based on their engi-
neering damage assessments, they revised the F-scale
wind speed ranges downward. Marshall (1983) utilized
load and resistance statistics to demonstrate how un-
certainties in assessing building damage can lead to
large errors in assigning F-scale ratings, especially in
the upper ranges of the F scale.
There are a number of potential problems in deter-
mining tornado intensity based on assessing building
damage. Most of the tornado damage paths on 3 May
1999 occurred in rural areas in which buildings were
spaced relatively far apart. Thus, it was unknown how
long the tornadoes remained at a certain intensity level
or what the maximum intensity levels might have been
in rural areas. As Doswell and Burgess (1988) pointed
out, building damage and tornado intensity are related
but are not perfectly correlated. A destroyed building

J
UNE
2002 585MARSHALL
may have been built poorly, leading to an overestimate
of tornado intensity. Because tornadoes are rated by the
worst damage caused, there would be a tendency to
overrate tornado intensities unless the relative strengths
of the damaged buildings are known. Additional diffi-
culties in rating tornadoes occur when they are in open
country and do not cause building damage. Schaefer
and Galway (1982) found that tornadoes that strike pop-
ulated areas are more likely to be assigned a higher F-
scale rating than those tornadoes that remain in open
country.
Fujita (1992) realized residences were not constructed
homogeneously, and he devised corrections to compen-
sate for those differences in assigning F-scale numbers.
For example, Fujita indicated that a strong wooden-
framed house may sustain F2 damage whereas the same
wind might cause F0 damage to a concrete building or
F5 damage to a poorly constructed outbuilding. Thus,
Fujita believed that relative strengths of buildings must
be considered when assigning F-scale numbers. How-
ever, some people who utilize the F scale have yet to
consider the relative differences in building resistance.
This may be due to a lack of understanding in how
buildings are constructed and/or confusion in how to
apply corrections to the F scale. The large number of
similarly built residential structures within our study
area fortunately provided more uniformity in which to
assign or compare F-scale ratings. Thus, corrections to
the F-scale ratings were not employed given that all but
a few homes were constructed on concrete slab foun-
dations.
Reynolds (1971) indicated that flying debris is an
important factor in the destruction of buildings and that
the part played is not always obvious. Our team mem-
bers found many homes along the edge of the tornado
damage path that had been compromised by debris im-
pact. However, a portion of the building had to remain
to allow determination of the extent of debris impact.
Doswell and Burgess (1988) indicated that complete
failure of a building would yield only a lower-bound
estimate of tornado intensity. Therefore, our teammem-
bers were interested especially in less damaged or un-
damaged buildings in the tornado path where an upper
bound of tornado wind speed could be determined.
Phan and Simiu (1998) determined that wind speeds
of longer duration resulted in greater damage to resi-
dences in the Jarrell, Texas, tornado. Residences near
the center of the Jarrell tornado were subjected to tor-
nadic winds for about 3 min. By comparison, we cal-
culated that houses near the tornado center at Moore
were subjected to tornadic winds for about 30 s. Last,
there is the human factor in determining the tornado
intensity based on analyzing damage. A person with
knowledge of how buildings fail will tend to rate a
building differently than a person who does not possess
such knowledge.
d. Aerial survey
Fujita and Smith (1993) have demonstrated that near-
ground wind fields can be inferred better by viewing
damage patterns from the air. Our team employed a
helicopter to conduct the aerial survey. The aircraft was
flown between 300 m (1000 ft) and 900 m (3000 ft)
above the ground in overlapping circles parallel to the
tornado damage path. Flight clearance of air space had
to be obtained from air traffic control because portions
of the disaster area were restricted. Numerous photo-
graphs were taken of the damage path. The best per-
spective was obtained when photographing from almost
directly above the damaged buildings. Clear skies pro-
vided the best contrast and sharpest images for photog-
raphy, as opposed to cloudy skies. The aerial survey
was conducted in the morning when the air was least
turbulent and the sky was not hazy. In many cases,
specific buildings such as churches and schools were
identified and served as landmarks. In most instances,
it was not possible to determine from the air how well
a structure was built. This problem demonstrated the
importance of conducting a comprehensive ground sur-
vey.
3. The damage path
a. General observations
Our team began the damage survey in a rural area 3
mi west of Newcastle, Oklahoma. Rural houses were
built on concrete pier and wooden beam foundations or
concrete masonry foundations. Houses on pier and beam
foundations usually were secured to the concrete beams
with anchor bolts whereas houses built on masonry
foundations were not anchored. Some unanchored hous-
es along the edge of the tornado damage path were
moved intact from their foundations by as much as 90
m (295 ft; Fig. 3). Movement of unanchored homes off
their foundations illustrated how F5 damage could occur
at F1 or F2 wind speeds.
As the tornado entered southern sections of Oklahoma
City, it inflicted F4 and F5 damage to a number of
houses in the Country Place subdivision located just
south of 134th Street (Fig. 4). These wooden-framed
houses had been built recently on concrete slab foun-
dations that averaged 150 m
2
(1615 ft
2
). Houses had
wooden bottom plates attached to their concrete slab
foundations with tapered cut nails. Little was left of the
homes near the center of the damage path other than a
swath of concentrated debris. This swath of concentrated
debris continued through an open field extending about
1000 m (3280 ft) beyond the subdivision. Most of the
debris was composed of wooden boards from house
framing along with furniture and a number of auto-
mobiles. One vehicle traveled 1000 m (3280 ft) and
ended up inside a bridge culvert. It was apparent to team
members that houses were a major debris source that
contributed to their own destruction. As homes broke

586 V
OLUME
17WEATHER AND FORECASTING
F
IG
. 3. Unanchored rural house (in background) near Newcastle was moved approximately 90 m (295 ft) to the east. The
unanchored house was supported on concrete masonry units.
apart, their debris impacted neighboring homes, com-
promising them as well. Damage to houses on the edge
of the tornado path varied greatly depending on their
orientations. Houses with attached garage doors facing
the wind typically sustained more severe damage than
houses for which garage doors opposed the wind. Sim-
ilar observations were made by Marshall and McDonald
(1982) in their survey of the Grand Island, Nebraska,
tornadoes.
The tornado continued northeast through the Briar-
hollow subdivision where the damage path remained
about 400 m (1312 ft) wide. These were older tract
houses with attached two-car garages. Wooden-framed
houses had been built on concrete slab foundations that
averaged 150 m
2
(1615 ft
2
). Houses were attached to
their foundations with either tapered cut nails, shot pins,
metal straps, or anchor bolts. Without regard to method
of attachment, none of the houses survived within the
center of the damage path.
The Westmoore High School was located on the north
side of the damage path and sustained damage to the
roof covering along with removal of some metal roof
decking and metal cladding. An awards ceremony was
being held at the time the tornado struck the school, and
hundreds of people were in attendance. People were
evacuated to interior portions of the school, and all sur-
vived; however, many vehicles in the parking lot were
displaced. Some vehicles tumbled southward and
crashed into houses adjacent to the school.
As the tornado crossed Western Avenue, it struck the
Emerald Springs Apartments, reducing some of thetwo-
story wooden structures to one story or less. A pair of
traffic signals in front of the school remained intact
although the signal lights had been removed by the
wind. A nearby metal church building collapsed. The
tornado continued through the Greenleaf subdivision,
reducing many of the wooden-framed townhouses to
rubble. Numerous townhouses sustained F4 damage,
and one house sustained F5 damage. The tornado com-
pletely destroyed a two-story metal church building,
then proceeded across Santa Fe Avenue and entered the
heart of the city of Moore. The damage path extended
diagonally from NW 12th to NW 19th Streets through
a densely populated area of smaller wooden houses that
averaged 100 m
2
(1076 ft
2
) and that were typically one
story with one-car garages. An occasional interior closet
or hallway remained; however, most houses near the
center of the damage path had sustained F4 or F5 dam-
age.
The tornado struck Kelly Elementary School directly
and severely damaged the building. The school was a
steel-framed structure with load-bearing masonry walls
constructed on a concrete slab foundation. A concrete
bond beam had been constructed atop the masonry walls

J
UNE
2002 587MARSHALL
F
IG
. 4. Aerial view of Country Place subdivision in southern Oklahoma City where F5 damage was found. Wooden-framed
houses were attached to concrete slab foundations with tapered cut nails. The tornado continued northeast through the suburbs
of Moore, Del City, and Midwest City.
to secure the ends of open-web steel roof joists. Many
of the steel roof joists and pieces of the decking were
removed, along with some of the concrete bond beams,
and were transported downwind and deposited in a field.
The school was the only engineered building in the cen-
ter of the damage path within our study area. The survey
team unfortunately was not allowed inside the school
to conduct a more detailed assessment.
On the north edge of the damage path was the Re-
gency Park Baptist Church. The church sanctuary had
a high-pitched wooden-framed roof structure supported
by glued–laminated (glulam) wooden arches. Tornadic
winds removed the roof structure but left the arches
intact.
The tornado continued through another subdivision,
inflicting up to F4 damage before crossing I-35 at
Shields Boulevard. One fatality and several injuries oc-
curred when people tried to seek shelter under the over-
pass. However, the overpass had a concrete slab deck
with sloping abutments and offered no protection. The
tornado crossed I-35 and struck a two-story apartment
complex and the two-story Best Western Hotel. Roofs
of these wooden-framed buildings were removed, and
some second-story walls failed. The tornado then trav-
eled through the Ridgewood subdivision, inflicting up
to F4 damage before heading out of town over rural
areas. The tornado turned more northerly and then
crossed I-240, continuing through portions of Del City
and Midwest City before dissipating.
About 100 survivors were interviewed during our
damage survey. All of the people interviewed had
known the tornado was approaching. Most people said
they received the tornado warning via local television,
and others received the warning from relatives or friends
via telephone. Hammer and Schmidlin (2000) conducted
detailed interviews of the survivors and found that sev-
eral people actually drove away from the tornado. Al-
though driving away from the tornado was a successful
strategy, fleeing a tornado in lieu of seeking shelter in
one’s home is generally not recommended. We found
that most survivors who remained at home sought shel-
ter in a bathtub, closet, or interior hallway because none
of the houses had basements. We also found that the
majority of people who sought shelter inside their homes
survived the tornado without serious injury. In all, our
team rated F1 damage to 284 houses, F2 damage to 405
houses, F3 damage to 558 houses, F4 damage to 317
houses, and F5 damage to 17 houses.
b. Wind-borne projectiles
A tremendous amount of debris was generated by the
tornado as it traveled through the residential areas. De-
bris ingested into the tornado became high-speed pro-

588 V
OLUME
17WEATHER AND FORECASTING
F
IG
. 5. Leg of steel folding chair penetrated a solid wooden post. The post partially supported a second-story porch on a
house; the house sustained only F0 damage.
jectiles that penetrated building roofs and walls and ve-
hicles. Most projectiles were broken pieces of wood
from houses, furniture, and trees. Larger wooden pro-
jectiles included 2.4-m (8 ft) long, wooden, 5 cm 310
cm (2 in. 34 in.) or 5 cm 315 cm (2 in. 36 in.)
lumber from residences. One 5 cm 315 cm (2 in. 3
6 in.) board had entered through the roof of a house
and penetrated the refrigerator freezer.In another house,
we found a 2.4-m (8 ft) long, wooden fence post that
had gone through a window and lodged in an interior
wall. A number of wooden boards also were found pen-
etrating brick veneer. Projectiles were found on both the
north and south sides of the damage path, usually in
areas where F0–F2 damage had occurred to residences.
The largest projectile found was a 3.7-m (12 ft) di-
ameter, 4.3-m (14 ft) tall steel oil tank that had tumbled
end over end for 276 m (905 ft), leaving gouge marks
in the ground just west of the Newcastle overpass. The
tank had not been anchored. The longest projectile found
was an 11-m (36 ft) long, steel beam from a mobile
home that had traveled 300 m (984 ft). The beam was
twisted but remained on the ground surface. A number
of vehicles had been rolled or tossed up to 1000 m (3280
ft) and were barely recognizable. In some instances,
debris filled the passenger space of the vehicles or pen-
etrated body metal. A barbed wire fence was rolled up
into a ball measuring 1 m (3.28 ft) in diameter.
Our survey team found a few unusual projectiles. The
leg of a steel lawn chair had penetrated a solid wooden
post that measured 13 cm 313 cm (5 in. 35 in.; Fig.
5). The post had supported a second-story balcony on
a rural house that sustained only F0 damage. A 1.8-m
(6 ft) long section of steel sewer pipe, weighing about
23 kg (50 lb), went through the front door of a residence
and came to rest in an interior hallway.
4. Performance of housing
a. Foundation connections
There were four methods of anchoring stud walls to
the concrete slab foundations. Most houses had wall
bottom plates attached to their foundations with 5-cm
(2 in.) long, tapered cut nails. Tapered cut nails had
been driven through the top sides of wall bottom plates
at intervals ranging from 30 to 130 cm (12–51 in.). Wall
bottom plates were 5 cm 310 cm (2 in. 34 in.) lumber
with actual dimensions of 3.8 cm 38.9 cm (1.5 in. 3
3.5 in.). Thus, the tapered cut nails extended a maximum
of only 1.3 cm (0.5 in.) into the concrete slab foundation.
In destroyed structures, scrape marks in the surfaces of
concrete slab foundations extended from the points
where the tapered cut nails had been installed, indicating
the walls moved laterally (Fig. 6). Bottom plates were
found in the debris with tapered cut nails still attached
to them (Fig. 7). Such foundation attachments were in-

J
UNE
2002 589MARSHALL
F
IG
. 6. Scrape mark and conical spalled area in the surface of
concrete slab foundation indicating where the wall bottom plate along
with the tapered cut nail connection were moved laterally. The entire
house was swept away, resulting in an F5 damage rating.
F
IG
. 7. Wall bottom plate found in debris still had tapered cut nail
attached.
F
IG
. 8. Tapered cut nail that remained in the foundation. The wall
bottom plate simply pulled through the fastener, leaving the nail.
herently weak, and, in many instances, conical spalled
areas or divots were found in the slab surface where the
tapered cut nails had been driven. These divots resulted
when the tapered cut nails were installed or when the
connection failed during the tornado. In either case,fail-
ure of the nailed connection at the foundation indicated
to us it was weaker than the nailed connection between
the wall stud and bottom plate. In other instances, the
bottom plates pulled through the fasteners, leaving only
nails attached to the concrete foundation (Fig. 8).
Some homes had perimeter walls attached toconcrete
foundations with metal shot pins. These pins were 7.6
cm (3 in.) long and had square metal washers. Pins were
driven vertically at 40–45-cm (16–18 in.) intervals
through the top sides of the bottom plates with a powder-
actuated tool. The pins extended up to 3.8 cm (1.5 in.)
into the concrete slab, or about 3 times the depth of
tapered cut nails. Failure of houses with pin attachments
occurred when the wooden bottom plates broke around
the fasteners or when the wall studs pulled away from
the bottom plates (Fig. 9).
FEMA (1999a) indicated that most residential con-
struction in Oklahoma, including the city of Moore, was
required to meet the design requirements of the one-
and two-family dwelling building code as published by
Council of American Building Officials (CABO 1995).
However, it should be noted that houses built prior to
1995 were governed by a less restrictive building code.
The CABO (1995) building code specified that floor
systems be anchored to concrete foundations with 1.3-
cm (0.5 in.) diameter, steel anchor bolts spaced a max-
imum of 1.8 m (6 ft) apart and not more than 30 cm (1
ft) from the wall corners. Anchor bolts also should ex-
tend a minimum of 18 cm (7 in.) into the concrete.
Nailing wooden bottom plates to concrete slab foun-
dations would not meet the requirements set forth in the
CABO (1995) building code. Local building officials
indicated that shot pins could be utilized to anchor bot-
tom plates around the perimeter of concrete slab foun-
dations whereas tapered cut nails were allowed to attach
bottom plates of interior walls within the city of Moore.
Such variances in the building code probably were al-
lowed because straps, cut nails, or shot pins would be
adequate to keep the bottom plates of walls in place for
normal gravity loads. However, these variances did not
adequately consider how these connections would work
when wind forces were applied.
The remaining houses in our survey had wall bottom
plates attached to the concrete slab foundation with met-
al straps or anchor bolts. These connections were stron-
ger than the tapered cut nails because they usually held
the wooden bottom plates in place. Failure of these
houses occurred when the wall studs pulled away from
the bottom plates. Wall studs were end-nailed with pairs
of 16-penny (8.9 cm or 3.5 in.) nails driven from the
bottom side of the wooden bottom plate, which is the
minimum requirement under the CABO (1995) building
code. The nailed connection frequently had pulled apart,
leaving pairs of nail shanks pointing skyward from the
strapped or bolted wooden bottom plates (Figs. 10 and
11). Straight-nailed connections were weak in tension.
Marshall (1983) tested 30 pairs of 16-penny straight-

590 V
OLUME
17WEATHER AND FORECASTING
F
IG
. 9. Bent shot pins remained in concrete foundation. The wall bottom plate had been
broken around the fasteners as the house was destroyed by the tornado.
F
IG
. 10. Wall stud pulled out, leaving only the strapped bottom
plate intact. The wall stud was straight nailed, a connection that is
inherently weak in uplift/tension.
F
IG
. 11. Wall studs pulled out, leaving only the anchor-bolted bot-
tom plate. Local building codes unfortunately still allow straight-
nailed connections between the wall plates and studs.
nailed connections and found an average pullout
strength was 980 N (220 lb), with a standard deviation
of 167 N (37 lb). In F4- and F5-damaged areas, some
of the wooden bottom plates had split around the anchor
bolts and had been pulled apart, leaving only theanchor
bolt attached to the foundation. In no instance did we
find anchor bolts that had failed on residences.
b. Wall connections
Stud walls were framed using 5 cm 310 cm (2 in.
34 in.) lumber and stiffened with 2.5 cm 310 cm (1
in. 34 in.) let-in bracing as called for in the CABO
(1995) building code. Let-in braces were installed di-
agonally at wall corners extending from the tops of the
wall plates to the bottom plates at angles between 458
and 608. Notches were cut into the wall studs to receive
the let-in braces. The purpose of the let-in braces was
to transfer lateral wind loads to the foundation applied
from the perpendicular wall. Nominal dimensions of the
let-in wall brace were 1.9 cm 39 cm (0.75 in. 33.5
in.). Team members noted that let-in wall bracing
worked adequately as long as the perpendicular wall
was subjected to positive external pressure. However,
such wall bracing was ineffective when walls were sub-
jected to internal pressures. The CABO (1995) building
code permits the use of wall sheathing as an alternative
to let-in braces. Such sheathing is usually 1.2 m 32.4
m(4ft38 ft) plywood or oriented strand board (OSB)
placed on its short side so that it can be nailed to the
top plates, wall studs, and bottom plates. Sheathing
nailed to the plates and studs help stiffen the wallframe
to resist racking (side to side) forces. However, a sheet
of plywood costs more than a 2.5 cm 310 cm (1 in.

J
UNE
2002 591MARSHALL
F
IG
. 12. Typical failure of attached garage when exposed to tornadic winds. The garage door blew in, resulting in
displacement of the roof from combined loads of aerodynamic pressure and internal pressure.
34 in.) lumber; thus, contractors were more apt to
install a let-in brace. A problem with using let-inbraces
at wall corners occurred when the intersecting wall was
interrupted by a large opening such as a garage door or
window. Thus, walls containing large openings usually
were not braced. In such instances, sheathing would
have been a better alternative. Another problem arose
when notches were cut too deep into the wall studs,
leaving let-in braces elevated by their nail shanks. It is
important that let-in braces fit snugly into the notches
and be fastened correctly to transfer lateral loads.
Wooden-framed walls were capped by a doubled top
plate. The top plates were doubled in order to support
the roof framing and to connect intersecting wall cor-
ners. Wooden top plates were straight-nailed into the
tops of the wall studs using pairs of 16-penny nails.
Such connections were inherently weak when uplifted;
however, they would meet the minimum requirements
set forth by the CABO (1995) building code. Team
members did find a few instances in which top plates
had separated from the wall studs or where both boards
composing the top plates had pulled apart.
Many houses within the damage path had brick ma-
sonry exterior walls. Such walls were erected around
the perimeters of the concrete slab foundations and were
nonstructural; that is, they did not support the roof. Ac-
cording to the CABO (1995) building code, brick ma-
sonry walls must be secured to the wooden stud walls
by metal wire or ties about every 60 cm (2 ft) vertically
and horizontally. Team members found many instances
in which wall ties had not been installed or in which
ties had not been bent to engage the masonry. Such
masonry walls were free standing and could be pushed
back and forth by hand.
As mentioned previously, most houses had attached
garages. Team members identified numerous garage
door failures along the edge of the tornado path. Garage
doors in newer homes typically were aluminum panels.
Most two-car garages had continuous doors that were
4.9 m (16 ft) wide and 2.1 m (7 ft) tall. Garage door
failures initiated when the aluminum door panels buck-
led and the metal or plastic rollers pulled out from the
door cleats. Garage door failures resulted from external
pressures as well as internal pressures. Loss of the ga-
rage door on the windward side of the house allowed
wind to enter the garage, increasing internal pressures
on the roof and walls. If winds were sufficiently strong,
the garage lost its roof and/or walls (Fig. 12).
c. Roof connections
Team members found that roof structures frequently
pulled apart from the top plates. Rafters were toenailed
to the top plates, usually with two fasteners. In some
instances, the fasteners were pulled out along with the
rafters; in other instances, the rafters split around the

592 V
OLUME
17WEATHER AND FORECASTING
fasteners, leaving the nails in the top plates. Rafters
secured with pairs of 16-penny nails toenailed into the
top plates met the minimum requirements as called for
by the CABO (1995) building code. However,toenailed
connections were relatively weak in tension. Marshall
(1983) also conducted tests on 30 pairs of 16-penny
toenailed connections and found an average pullout
strength of about 1313 N (295 lb) with a standard de-
viation of 304 N (68 lb). Canfield et al. (1991) conducted
similar pull tests and found similar results. However,
they could increase the average pullout strength to 1842
N (414 lb) if the wood did not split when nailed. By
comparison, Canfield et al. (1991) also conducted pull
tests on metal ‘‘hurricane’’ clips and found an average
strength of 5341 N (1200 lb) depending on the type of
clip used. Thus, a metal clip can be about 3 times as
strong as a 16-penny toenailed connection.
Wood roof decking was attached to the rafters with
staples, usually at 15-cm (6 in.) intervals. However,
some houses had lost large portions of the roof decking,
leaving the roof structure exposed to the weather. Close
examination revealed that deck staples partially or to-
tally had missed the underlying rafters. Therefore, a
poorly installed roof deck was very vulnerable to being
uplifted and removed by the wind.
d. Sequence of house failure
Wind was forced to go over and around a house in
its path. As a result, aerodynamic pressures were applied
to the building exterior, which included (inward acting)
positive pressures on the windward walls, uplift pres-
sures on the roof, and (outward acting) negative pres-
sures on the leeward walls. Windward walls literally
were pushed inward when they could not adequately
transfer the wind load to the foundation or to inter-
secting walls. Uplift on the roof resulted in the removal
of the roof covering, roof decking, or roof structure. If
the building was breached on its windward side from a
broken window or door, internal pressures added to the
aerodynamic uplift pressures on the roof. This frequent-
ly resulted in forcing the roof upward and the exterior
walls outward. Once the roof structure had been re-
moved, it was relatively easy for the winds to topple or
push out the perimeter walls because these walls no
longer were braced at the top. Liu et al. (1989) indicated
that inadequate tie-down of roofs was the most serious
common problem with wooden-framed houses. In our
study, we found a fairly even split between the numbers
of houses that sustained wall–foundation connection
failures and those that sustained roof–wall connection
failures.
Buildings are designed with a concept of a continuous
load path by which applied loads are transferred from
the roof down through the walls and into the foundation.
The ‘‘dead weight’’ of a building normally keeps it in
place on its foundation. However, theload path direction
reverses when wind lifts the roof. Connections adequate
under gravity-load situations actually may pull apart
when the load-path direction is reversed. Thus, con-
nections must be designed to accept loads from opposite
directions. Each connection between structural members
can be thought of as a link in a chain, with a building
only as strong as its weakest link.
The CABO (1995) building code is not without its
problems. It still allows top and bottom plates to be
straight-nailed to the wall studs. Marshall (1983, 1992)
has shown that such nailed connections are weak in
tension and can be pulled apart easily. Fasteners must
be placed in shear, not tension, to provide greater re-
sistance to wind uplift. Solid plywood sheathing also
can provide more formidable protection and help to
stiffen the frame if it is fastened to the top plates, studs,
and bottom plates in lieu of let-in braces.
e. Building codes
Houses are generally nonengineered structures. Thus,
the details of house construction are left to the discretion
of the building contractor. Larger towns and cities have
adopted building codes, but most rural areas and smaller
towns do not have building codes or inspectors to en-
force the codes. However, building codes serve only as
aminimum design requirement. It has been the author’s
experience that builders whose goal it is to achieve a
minimum design invariably fall short of this goal be-
cause of variations in such things as workmanship and
material qualities. Thus, a ‘‘beyond-code’’ approach
should be made in an effort to surpass the minimum
design requirements. Although it is true there are no
provisions in building codes to construct houses to resist
tornadoes, building codes are still important in miti-
gating tornado damage. A better-built house wouldyield
less debris in a tornado, and occupants would have a
better chance of surviving in such a house. In addition,
structural improvements could reduce the level of dam-
age to a house that experiences weaker tornadoes or
straight-line thunderstorm winds that are just above
code-required design levels and hence could mitigate
economic loss and improve safety.
The CABO (1995) building code lists the design wind
speed for the Oklahoma City area at 31 m s
21
(70 mi
h
21
) for a fastest-mile wind at 10 m (33 ft) above the
ground in open, unobstructed terrain, which equates to
about a 40 m s
21
(90 mi h
21
) 3-s wind gust. McDonald
(2001) indicated that gust wind speeds are utilized to
calculate tornado wind pressures on structures. So, it
should not be surprising that houses would begin to
sustain some damage when winds approach the code
design wind speeds. Safety factors are built into the code
design wind speeds such that buildings have a reserve
strength. According to Gardner et al. (2000), wind gusts
in the range of 58–72 m s
21
(130–160 mi h
21
) could
overcome the reserve strength and completely destroy
a structure. Therefore, houses that sustain F5 damage
rating actually could fail in wind gusts in the original

J
UNE
2002 593MARSHALL
T
ABLE
1. Estimated failure wind velocities in comparison with F
scale, based on the degree of damage to wooden-framed residences
on concrete slab foundations.
F-scale
rating
Failure wind speed* after
Fujita (1971)
(mi h
21
)(ms
21
)
Failure wind speed**
estimated by author
(mi h
21
)(ms
21
)
F0
F1
F2
F3
F4
F5
40–72
73–112
113–157
158–206
207–260
261–318
18–32
33–50
51–70
71–92
93–116
117–142
,80
80–100
100–120
120–140
140–160
.160
,36
36–45
45–54
54–63
63–72
.72
* fastest 0.25-mi wind.
** 3-s wind gust.
F2 range. Phan and Simiu (1998) reached a similar con-
clusion in their analysis of damage in the Jarrell, Texas,
tornado.
Table 1 presents the author’s adjustments to wind
speed ranges of the F scale based on results of engi-
neering assessments of residential damage in Moore,
Oklahoma. For engineered buildings, the wind speed
ranges would be higher. Likewise, for more poorly built
houses, the wind speed ranges would be lower. It is
emphasized that failure wind speeds represent the lower
bound of tornado intensity. The upper bound of tornado
intensity could not be determined in this study because
no homes survived in the center of the tornado damage
path. Even if we had gained access to Kelly Elementary
School and determined the upper-bound wind speeds
for those building components that did not fail, this
would have represented only one sampled location with-
in the tornado path.
f. Shelter performance
Two belowground shelters were found within our
damage survey area. One shelter was located adjacent
to a rural house near Newcastle where the team began
the damage survey, and the other shelter was located in
east Moore. Both shelters were adjacent to houses that
sustained F3 damage. Shelters were constructed ofsteel-
reinforced concrete and had wooden doors lined with
sheet steel. The shelters were not damaged during the
tornado and occupants who used the shelters during the
storm were not injured.
Two aboveground shelters were located by other team
members in Bridge Creek and Del City, Oklahoma. Ac-
cording to Gardner et al. (2000), one aboveground shel-
ter was located in rural Bridge Creek estates southwest
of Oklahoma City and the other shelter was located in
Del City. The Bridge Creek house sustained F1 damage;
the Del City house sustained F3 damage. Both above-
ground shelters survived without damage.
5. Calculation of failure wind speeds
Mehta et al. (1981) showed there was a good cor-
relation between wind speed and building damage. Meh-
ta et al. (1976) and Minor et al. (1977) have shown how
to calculate failure wind speeds from buildings damaged
by tornadoes and hurricanes. Weights of various build-
ing components and strengths of critical connections
must be determined. The following wind speed–pressure
formula from ASCE (1988) has been utilized to cal-
culate failure wind speeds:
2
p5fC V ,
p
(1)
where pis the wind pressure, fis a density constant,
C
p
is the pressure coefficient, and Vis the wind speed.
Reynolds (1971) noted that one of the greatest uncer-
tainties in estimating a failure wind speed involves de-
termining the value of the pressure coefficients. This
variable depends on certain building characteristics and
its surroundings, including the size and shape of the
building and its proximity to neighboring buildings.
Such pressure coefficients are derived from wind-tunnel
studies under constant wind velocities and can only be
estimated for tornadoes. Mehta (1976) acknowledges
these uncertainties but states that pressure coefficient
values obtained from wind-tunnel studies appear to
work reasonably well in calculating failure wind speeds
on buildings in tornadoes.
As noted from our damage survey, numerous housing
failures initiated with destruction of attached garages.
In a typical case, garage doors blew in, allowing internal
pressures to act in combination with external aerody-
namic uplift pressures to remove the garage roof struc-
ture. Uplift on the roof structure caused toenailed con-
nections in the wooden top plates to pull apart. Destruc-
tion of the attached garage frequently led to damage or
the removal of the remaining roof structure on the res-
idence.
In an effort to determine a range of failure wind
speeds, calculations were made for the typical attached
garage that measured 6 m 36 m (20 ft 320 ft) and
had a gable roof covered with 1.3 cm (0.5 in.)-thick
plywood and asphalt composition shingles. Rafters were
assumed to be conventionally spaced 61 cm (24 in.)
apart. Using standard building data, the dead load of the
garage roof would be about 59 kg m
22
(12 lb ft
2
). Given
the size of this garage in this example, there would be
22 toenailed connections between the rafters and wall
top plates, resulting in about 78 kg m
22
(16 lb ft
2
)of
uplift resistance. Failure of the roof would occur when
the wind uplift force is equal to the weight of the garage
roof plus the resistance of the connections. Pressure
coefficients were utilized from ASCE (1988). If the
wind blew directly into the garage and caused the garage
door to fail, internal pressures would combine with ex-
ternal aerodynamic pressures, resulting in a failure wind
speed of only 38 m s
21
(85 mi h
21
). However, if there
was no internal pressure contribution, the resulting fail-
ure wind speed would be 56 m s
21
(125 mi h
21
). How-
ever, internal pressure is rarely nil, because houses usu-
ally have some natural ventilation. Thus, conventionally
constructed wood-framed roofs would likely fail at wind

594 V
OLUME
17WEATHER AND FORECASTING
F
IG
. 13. A new house being constructed after the tornado had a metal strap bent outside the
base of the exterior wall rather than being nailed to the wooden bottom plate. As a result, the
wall was not attached to the foundation. The exposed metal strap was likely hidden later when
the brick masonry was installed.
speeds around 50 m s
21
(112 mi h
21
) resulting in F2
damage. Once the roofs have been removed, it would
not take much additional wind to destroy the structure.
Mehta and Carter (1999) calculated failure wind speeds
in the Jefferson County, Alabama, tornado and con-
cluded that complete destruction of houses occurred in
the wind speed range of 51–67 m s
21
(114–150 mi h
21
).
6. New housing
The author revisited the disaster area three months
after the tornado to check the quality of new house
construction. A total of 40 houses were examined in
Moore and southern Oklahoma City on sites at which
houses previously had been destroyed. The author found
that the quality of new home construction generally was
no better than homes built prior to the tornado. Most
newly built homes were attached to their concretefoun-
dations with tapered cut nails or shot pins as had been
noted in homes destroyed by the tornado. Of the 40 new
houses inspected, 5 houses had bottom plates bolted to
their foundations, 6 houses had bottom plates strapped
with metal to their foundations, and 29 houses had bot-
tom plates attached to their foundations with tapered cut
nails and/or shot pins.
Homes anchor bolted to their foundations had bolts
properly spaced with nuts tightened over washers.
Homes strapped to their foundations had most straps
fastened to the bottom plates. However, a few straps
were found bent outside the walls and extending from
beneath the bottom plates (Fig. 13). Such straps were
not secured to the bottom plates and exterior insulation
board already had been installed on the perimeterwalls.
These straps evidently were going to be hidden behind
the brick veneer. Tapered cut nails utilized to secure
bottom plates to newly built homes were the same type
as in homes destroyed by the tornado (Fig. 14). These
nails only extended 1.3 cm (0.5 in.) into the concrete
foundation (Fig. 15). It remains unknown why four dif-
ferent types of foundation anchor systems were being
employed in the same general area, especially given that
it was apparent that anchor bolts were superior in
strength.
There was no change in house-framing techniques
from original construction except in one instance in
which a house had larger 5 cm 315 cm (2 in. 36 in.)
perimeter walls with additional fasteners. Some wall
corners were covered with OSB sheathing. However,
this better-built house was in the middle of a subdivision
of conventionally built houses. Building beyond the
code is commendable, but its effectiveness is limited
when a better-built home is situated in a neighborhood
of houses with less wind resistance. Weaker neighboring
houses could become debris sources that could damage
or destroy better-built houses. In essence, a house is
only as tornado resistant as its neighbors. However, this
finding should not discourage homeowners from in-
sisting on structural improvements to their home.
None of the houses inspected had hurricane clips or
other wind-resistant connections. In one house, cutouts
had been made in wall framing for let-in braces, but the
braces had not been installed (Fig. 16). Insulationboard
already had been fastened to the exterior wall, preclud-
ing installation of the let-in braces. The same house only
had one nail securing each rafter to the top plate of
walls. Nails also had been installed too close to the

J
UNE
2002 595MARSHALL
F
IG
. 14. Perimeter wall plate was fastened to the foundation with tapered cut nails. Nails were spaced between 30 and 130
cm (12 and 51 in.) apart.
F
IG
. 15. Tapered cut nails that securedwall bottom plates extended
only 1.3 cm (0.5 in.) into the foundation. Such nails were used in
lieu of anchor bolts or metal straps both in new construction and in
houses built prior to the tornado.
inside edge of the top plate, splitting the wood. (Fig.
17).
Six of the 40 new houses contained ‘‘safe rooms’’
(FEMA 1999b). These closet-sized rooms were con-
structed with steel-reinforced concrete walls and ceil-
ings. Entry into each safe room was through a heavy,
metal door.
7. Summary and conclusions
A violent tornado struck the Oklahoma City metro-
politan area on 3 May 1999, providing an opportunity
to assess building performance. Our survey team studied
building damage in and around the city of Moore. The
majority of buildings damaged by the tornado were one-
and two-story residences that were constructed conven-
tionally on concrete slab foundations. Many homes were
completely destroyed, and no building managed to sur-
vive in the center of the tornado path.
A methodology was developed as to how to conduct
the damage survey and how to assign F-scale numbers
to damaged houses. Additional damage descriptions
were defined by the author to aid in assigning F-scale
numbers. It was concluded that tornado damage to hous-
es occurred at significantly lower wind speeds than those
established by the original F scale. Houses with F4 or
F5 damage likely failed when wind gusts reached F2
on the original F scale. Thus, failure wind speed ranges
were adjusted by the author to match better the F-scale
number. It should be noted that such failure wind speeds
represent lower-bound estimates of this tornado.
Approximately one-half of the houses surveyed failed
where the walls were fastened to their foundations. Wall
bottom plates had been attached with either tapered cut
nails, shot pins, metal straps, or anchor bolts, with ta-

596 V
OLUME
17WEATHER AND FORECASTING
F
IG
. 16. Lack of let-in wall brace in new house being built after the tornado. Note that notches
were cut in the wall studs to receive the let-in brace; however, exterior insulation board already
had been installed.
F
IG
. 17. Rafters were fastened with a single nail to the wall top plate in this house under
construction after the tornado. The nails were installed too close to the edge of the top plate,
splitting the wood. Two nails should have been used at each connection.
pered cut nails being predominant. Tapered cut nails
usually pulled out of the foundations, leading to the
destruction of the houses. Shot pins remained in place;
however, some bottom plates broke around the pins.
Metal straps and anchor bolts tended to secure the wood-
en bottom plates; however, failure then occurred where
the wall studs were straight-nailed to the bottom plates.
The remaining one-half of the houses surveyed failed
where the roof structure was toenailed to the wall top
plates. These failures were found predominantly on the
edge of the tornado damage path and depended highly
on orientation of the attached garage to the wind. Houses
with attached garage doors facing the wind typically
sustained more severe damage than houses for which
garage doors opposed the wind.
A consistent damage sequence was found for resi-
dences. Houses typically failed as the attached garage
doors buckled inward and/or windows broke, allowing
the wind to enter and to pressurize the buildings. Roofs
then were uplifted by the combination of internal pres-

J
UNE
2002 597MARSHALL
sure and aerodynamic uplift. Last, perimeter walls fell
or were pushed outward because they no longer were
braced at their top ends by the roof structure.
Numerous projectiles were generated by the tornado.
Most of these projectiles were broken wooden pieces
from house structures, furniture, and trees. Many pro-
jectiles were found in areas in which residences had
sustained only F0–F2 damage. Only a few tornado shel-
ters were found in our study area. Shelters performed
well, and occupants who sought refuge in them were
not injured. About 100 survivors were interviewed, and
all had known the tornado was approaching. Most peo-
ple who stayed in their houses sought shelter in an in-
terior bathroom, closet, or hallway.
The author returned to the disaster area three months
later and inspected 40 houses that were under construc-
tion in replacement of destroyed houses. The author
found that, in general, the quality of the new construc-
tion was no better than the quality of construction of
destroyed homes. One exception was a house that had
thicker perimeter walls. Building contractors still were
constructing homes to reach the minimum design re-
quirements set forth by the CABO building code and
in several instances fell short. Building beyond code
was rarely done. In essence, the tornado that struck
Moore, Oklahoma, was too intense to have significant
impact on the way residential housing was rebuilt. If
anything, this event actually bolstered public perception
that no building could survive a tornado.
The author found that F-scale wind speeds were too
high in relation to the extent of damage to residences,
especially for houses that sustained F2 damage or great-
er. When wind speeds are overstated in the F scale, the
general public, designers, and builders may conclude
that it is economically unreasonable to design for these
higher wind speeds and associated loads on the building.
In fact, most residential housing can be made consid-
erably more resistant to damages from most tornadoes
by using construction techniques described in this paper.
When the author visited the post-tornado reconstruction
area, 6 of the 40 new houses contained tornado shelters
or safe rooms; however, these homesgenerally were not
built any better than prior to the tornado. Thus, hom-
eowners appeared to be making the decision to provide
for personal safety (i.e., building a safe room) instead
of to increase housing strength (to provide property pro-
tection and increased safety). This decision may, in part,
be based on the assumption that it is unreasonable or
uneconomical to try to construct houses for the high
wind speeds described on the F scale.
Acknowledgments. I thank Drs. Ernst Kiesling and
Kishor Mehta of Texas Tech University for inviting me
to be part of the damage survey team and my teammates,
Mr. Mark Conder and Dr. Zhongsan Zhou, for enduring
long hours of survey work. Doctor Charles Doswell,
Stoney Kirkpatrick, Scott Morrison, Jim Weithorn, and
Susie Meyer provided helpful comments and sugges-
tions to improve the manuscript.
REFERENCES
ASCE, 1988: Minimum Design Loads for Buildings and Other Struc-
tures. American Society of Civil Engineers, 94 pp.
Bunting, W. F., and B. E. Smith, 1990: A guide for conducting damage
surveys. NOAA Tech. Memo. NWS SR-146, 44 pp. [NTISPB93-
148427INZ.]
Burgess, D., and M. Magsig, 2000: Understanding WSR-88D sig-
natures for the 3 May 1999 Oklahoma City tornado. Preprints,
20th Conf. on Severe Local Storms, Orlando, FL, Amer. Meteor.
Soc., 33–36.
CABO, 1995: One and Two Family Dwelling Code. International
Code Council, Council of American Building Officials, Falls
Church, VA, 350 pp.
Canfield, L. R., S. Niu, and H. Liu, 1991: Uplift resistance of various
rafter-wall connections. For. Prod. J., 41, 27–34.
Doswell, C. A., III, and D. Burgess, 1988: On some issues of United
States climatology. Mon. Wea. Rev., 116, 495–501.
FEMA, 1999a: Midwest tornadoes of May 3, 1999. FEMA Publ. 342,
Item 9-1035, 216 pp. [Available online at http://www.fema.gov/
mit/bpat/fnlrpt.]
——, 1999b: Taking shelter from the storm: Building a safe room
inside your house. FEMA Publ. 320, 28 pp. [Available online
at http://www.fema.gov/library/ism2.pdf.]
Fujita, T. T., 1971: Proposed characterization of tornadoes and hur-
ricanes by area and intensity. SMRP Research Rep. 91, Uni-
versity of Chicago, Chicago, IL, 15 pp.
——, 1973: Experimental classification of tornadoes in FPP scale.
SMRP Research Rep. 98, University of Chicago, Chicago, IL,
15 pp.
——, 1981: Tornadoes and downbursts in the context of generalized
planetary scales. J. Atmos. Sci., 38, 1511–1534.
——, 1992: Mystery of Severe Storms. University of Chicago Press,
298 pp.
——, and B. E. Smith, 1993: Aerial survey and photography of tor-
nado and microburst damage. The Tornado: Its Structure, Pre-
diction, and Hazards, Geophys. Monogr., No. 79, Amer. Geo-
phys. Union, 479–493.
Gardner, A., K. C. Mehta, L. J. Tanner, Z. Zhou, M. Conder, R.
Howard, M. Martinez, and S. Weinbeck, 2000: The tornadoes
of Oklahoma City of May 3, 1999. Wind Science and Engi-
neering Research Center Rep., Texas Tech University, Lubbock,
TX, 38 pp.
Grazulis, T. P., 1993: Significant Tornadoes. Environmental Films,
Inc., 1326 pp.
Hammer, B., and T. Schmidlin, 2000: Use of vehicles to flee the 3
May 1999 Oklahoma City tornado: Reasons and relative injury
rates. Preprints, 20th Conf. on Severe Local Storms, Orlando,
FL, Amer. Meteor. Soc., 64–65.
Liu, H., H. S. Saffir, and P. R. Sparks, 1989: Wind damage to wood-
framed houses: Problems, solutions, and research needs. J.
Aerosp. Eng., 2, 57–70.
Marshall, T. P., 1983: Utilization of load and resistance statistics in
a wind speed assessment. M.S. thesis, Dept. of Civil Engineering,
Texas Tech University, 91 pp.
——, 1992: Lessons learned from analyzing tornado damage. The
Tornado: Its Structure, Dynamics, Prediction, and Hazards,
Geophys. Monogr., No. 79, Amer. Geophys. Union, 495–499.
——, and J. R. McDonald, 1982: An engineering analysis of the
Grand Island tornadoes. Preprints, 12th Conf. on Severe Local
Storms, San Antonio, TX, Amer. Meteor. Soc., 293–296.
McDonald, J. R., 2001: T. Theodore Fujita: His contribution to tor-
nado knowledge through damage documentation and the Fujita
scale. Bull. Amer. Meteor. Soc., 82, 63–72.
——, and T. P. Marshall, 1984: Tornado damage documentation. In-

598 V
OLUME
17WEATHER AND FORECASTING
stitute for Disaster Research Publ. 1984-20, Texas Tech Uni-
versity, 27 pp.
Mehta, K. C., 1976: Windspeed estimates: Engineering analysis.
Proc. Symp. on Tornadoes, Lubbock, TX, Texas Tech University,
89–103.
——, and R. Carter, 1999: Assessment of tornado wind speed from
damage in Jefferson County, Alabama. Proc. 10th Int. Conf. on
Wind Engineering, Copenhagen, Denmark, 265–271.
——, J. E. Minor, and J. R. McDonald, 1976: Windspeed analysis
of April 3–4, 1974 tornadoes. J. Struct. Div. Amer. Soc. Civ.
Eng., 102, 1709–1723.
——, ——, and T. A. Reinhold, 1981: Wind speed–damage corre-
lation in Hurricane Frederic. J. Struct. Div. Amer. Soc. Civ. Eng.,
107, 37–49.
Minor, J. E., 1977: Engineered-oriented examinations of the tornado
phenomenon. Preprints, 10th Conf. on Severe Local Storms,
Omaha, NE, Amer. Meteor. Soc., 438–445.
——, J. R. McDonald, and K. C. Mehta, 1977: The tornado: An
engineered-oriented perspective. NOAA Tech. Memo. ERL-
NSSL-82, 196 pp. [NTIS PB93-148435INZ.]
Phan, L., and E. Simiu, 1998: The Fujita tornado intensity scale: A
critique based on observations of the Jarrell tornado of May 27,
1997. NIST Tech. Note 1426, 20 pp. [NTIS PB98-175136INZ.]
Reynolds, G. W., 1971: Complication in estimating the magnitudes
of tornado forces from damage analysis. Preprints, Seventh Conf.
on Severe Local Storms, Kansas City, MO, Amer. Meteor. Soc.,
179–182.
Schaefer, J., and J. G. Galway, 1982: Population biasesin the tornado
climatology. Preprints, 12th Conf. on Severe Local Storms, San
Antonio, TX, Amer. Meteor. Soc., 51–54.