Mitigation Assessment Team Report
Hurricane Charley in Florida
Observations, Recommendations, and Technical Guidance
FEMA 488
April 2005

Chapter 5. Building Envelope Performance

The ability of the structural system to perform without failure is critical to 
avoiding injury to occupants and minimizing damage to a building and its 
contents. It does not, however, ensure occupant or building protection. Good 
performance of the building envelope is also necessary. The envelope includes 
exterior doors, non-load-bearing walls, wall coverings, soffits, roof coverings, 
windows, shutters, skylights, and exterior-mounted mechanical and electrical 
equipment. Historically, poor building envelope performance has been the leading 
cause of damage to buildings and their contents in weak to moderate intensity 
hurricanes, with damage to roof coverings and rooftop equipment being the 
predominant envelope problem. Building structural capacities have improved 
because of stronger building codes and better enforcement, resulting in less 
structural damage overall from intense hurricanes such as Hurricane Charley. 
Consequently, the performance of the building envelope is becoming increasingly 
important. The following sections describe envelope performance as observed for 
residential, commercial, and critical/essential facilities.


5.1	Doors

Failure of an exterior door has two important effects. First, failure can cause 
an increase in internal pressure, which may lead to exterior wall, roof, 
interior partition, ceiling, or structural damage (as discussed in Chapter 4). 
Second, wind can drive water through the opening, causing damage to interior 
contents and finishes, and leading to development of mold. Essentials to 
effective high-wind door performance include product testing to ensure 
sufficient factored strength to resist design wind loads; suitable anchoring of 
the door frame to the building; proper flashing, sealants, tracks, and drainage 
to minimize water intrusion into wall cavities or into occupied space; and, for 
glazed openings, the use of laminated glass or shutters to protect openings 
against windborne missile damage as discussed in Section 5.2.

5.1.1	Personnel Door Damage
There were only a limited number of buildings where personnel door damage was 
observed. Observed damage included broken window panes (typically caused by 
missiles) and doors disengaged from their frames (likely caused by over-
pressurization). Sliding glass door damage is shown in Figure 5-1, where several 
doors disengaged from their tracks; this damage was caused by over-
pressurization. Water infiltrated the interior of a residence and caused damage 
because of a lack of weatherstripping between a pair of doors and their 
threshold. A 3/8-inch gap occurred between the door bottoms and the threshold, 
apparently allowing a substantial amount of wind-driven water to enter the 
residence. Double-entry doors were also observed to be damaged even when 
homeowners tried to support the doors by pushing heavy furniture against them. 
An example of double-entry door failures was shown previously in Figure 3-20.

Figure 5-1. Photograph. Sliding glass doors blown out of their tracks (Punta 
Gorda Isles)

A limited number of buildings with personnel door damage were observed in 
commercial and critical/essential facilities. Observed damages included broken 
window panes (caused by missiles as shown in Figure 5-2) and disengagement of 
doors from their frames (likely caused by over-pressurization). At one school 
being used as a shelter, a pair of exterior gym doors reportedly blew open. 
People pulled the doors shut and held on to the horizontal exit hardware bars 
for the duration of the hurricane. The right leaf had top and bottom vertical 
rods, and the left leaf had a horizontal bolt. Therefore, at the latch edge, the 
door on the right was attached to the frame at the top and bottom of the door. 
However, the door on the left was attached only at mid-height, where it bolted 
into the right door (Figure 5-3). If the left door had also been equipped with 
top and bottom vertical rods, it may not have blown open. 

Figure 5-2. Photograph. Tempered glass in office building entry door and side 
windows broken by missiles (Punta Gorda)

Figure 5-3. Illustration. Improper attachment of doors. This illustration shows 
a double door in which the left-hand door was secured to the right-hand door 
only at mid-height, with a bolt into the right-hand door.

5.1.2	Garage Door Damage 
Damaged garage doors were observed throughout the Port Charlotte and Punta Gorda 
areas. In some instances, the doors buckled and were pulled outward (suction 
failures), as shown in Figure 5-4. In other instances, the doors were pushed 
inward (positive pressure failures), as shown in Figure 5-5. The home in the 
center of Figure 5-5 had a 5V-Crimp metal panel roof that performed well. Many 
of the other houses in this area (which typically had asphalt shingle or tile 
roofs) had roof covering damage. Many of the failures occurred because the doors 
had inadequate wind resistance. In these cases, the doors buckled inward or 
outward, and the rollers were often pulled out of the tracks. Other failures 
were caused by use of weak tracks or inadequate attachment of door tracks to the 
buildings. It was clear that most of the double car garage doors in older homes 
were not high-wind or debris-impact rated. In a number of the newer homes, the 
doors had improved bracing, but the metal gauge was much thinner than that used 
in Miami-Dade County approved impact-resistant garage doors. In addition, where 
door failures were observed, the tracks were not of the heavier gauge or braced 
according to high-wind recommendations. The garage doors approved by Miami-Dade 
County are constructed of thicker gauge material because they must meet 
different performance criteria for debris impact than is required by the 2001 
FBC in other counties in Florida. 

Figure 5-4. Photograph. Door lacked sufficient strength to resist the suction 
load (Deep Creek)

Figure 5-5. Photograph. Garage door at the home in the center buckled and the 
rollers pulled out from their tracks; garage door at the home on the right also 
failed (Deep Creek).

Some garage doors observed were designed with removable stiffener bars. One 
garage door with this type of design at a post-2001 FBC residence did not have 
the stiffener bar in place at the time of the hurricane, and it was damaged by 
wind pressure (Figure 5-6). There were instances where owners had left their 
homes for the summer season and had neglected to put into place the stiffener 
posts required to make their garage doors resist winds as they were designed.

Figure 5-6. Photograph. Garage door failed because the removable stiffener bar 
was not in place at the time of the hurricane (Punta Gorda Isles).

5.1.3	Rolling and Sectional Door Damage 
Damage to rolling and sectional doors (e.g., service garage doors and loading 
dock doors) was observed. Newer doors generally performed well. However, in one 
instance, a new door failed (the drawings were dated 1997), resulting in the 
failure of an interior partition wall, as shown in Figures 5-7 and 5-8. The 
failed door shown in Figure 5-7 was attached with 1¾-inch long by 3/8-inch 
diameter expansion bolts into concrete that spalled at the bolt locations, 
likely due to the placement of the bolts too close to the edge. There were no 
ties between the wall itself and the end wall (Figure 5-8). The drawings 
indicated continuous angles on each side of the wall, but they were not 
installed. Another sectional door around the corner and perpendicular to the 
door shown in Figure 5-7 failed after the door in Figure 5-7 failed. The buildup 
of internal pressure exerted a positive load on the other door, which was also 
loaded in suction on its outer surface. One of the expansion bolts along the 
right track sheared off, and the concrete spalled at the other bolts. The bolts 
were typically 2 feet on center, but some were closer.

Figure 5-7. Photograph. New door that failed. Non-load bearing CMU wall at the 
left tilted (see Figure 5-8) (Punta Gorda).

Figure 5-8. Photograph. After the door shown in Figure 5-7 failed, buildup of 
internal pressure tilted the wall (Punta Gorda).

The rolling and sectional doors at the older fire stations lacked sufficient 
wind resistance and typically failed by suction or positive pressure. Common 
modes of failure observed included doors disengaging from the tracks shown 
previously in (Figure 3-57), or tracks or track blocking pulling away from the 
walls, or breakage of glazed or metal panel doors. In several cases, the tracks 
bent or bowed enough to allow the wheels to disengage from the tracks. Damage to 
older doors was not surprising, illustrating the need to replace weak doors on 
these important buildings. However, when doors are replaced, it is important to 
replace all of the doors and the track hardware as illustrated by Figure 5-9. 
There were six sectional doors at the fire station shown in Figure 5-9. Five of 
the doors were damaged. On the leeward side, one door blew out, one buckled, and 
one had minor outward bowing. On the windward side, two of the doors blew inward 
as shown in Figure 5-9. The door that did not fail was a newer door. The tracks 
on the newer door were attached with ¼-inch screws at 18 inches on center. There 
were two stiffener ribs per 24-inch high door section. It is notable that one of 
the sidewalls was pushed out when the attachment of the wall angle to the slab 
failed, likely due to an increase in internal pressure. The wall angle was 
attached with nail-ins at 3 feet 3 inches on center. The concrete spalled at the 
fasteners. Because of the door damage, this station was taken off-line after the 
hurricane.

Figure 5-9. Photograph. Windward side of a fire station; two doors blew inward, 
but the newer center door remained intact (Punta Gorda).

Surprisingly, sectional doors on two newer fire stations also failed. At the 
Aqui Esta Fire Station (east of Punta Gorda Isles), five of the six doors were 
not damaged, but the sixth door pulled out of the track. This station was first 
occupied in 2000. At a fire station in Deep Creek (Figure 5-10), all three 
windward doors were blown in. The tracks were attached with ¼-inch lag screws at 
24 inches on center. The leeward doors were not damaged, but the roof structure 
blew off the apparatus bay.

Figure 5-10. Photograph. At two of the windward doors, the doors were pushed out 
of the tracks; at the third door, one of the tracks was pushed from the wall 
(Deep Creek).

At several of the fire stations, the sectional doors blew inward on the 
apparatus (i.e., fire engines or ambulances). In some instances, the doors 
caused damage, such as a broken windshield, but there were no reports of door 
damage disabling a piece of apparatus. 
 

5.2	Windows, Shutters, and Skylights 

Exterior windows are very susceptible to missile breakage during hurricanes 
unless they are protected against windborne debris (via use of laminated glass 
or shutters). Although the probability that any one window will be struck by 
windborne debris is typically small (except for manufactured housing in parks), 
when it does occur, the consequences can be significant. The probability of 
impact depends upon local wind characteristics and the amount of natural and 
manmade windborne debris in the vicinity. The greater the wind speed, the 
greater the amount of windborne debris that is likely to become airborne. 
Windows can also be broken by over-pressurization, but this damage is not as 
common as debris-induced damage. 

The 2001 FBC defines windborne debris regions (see Figure 2-1) as those 
specified in ASCE 7-02, except in the Florida Panhandle, where the 2001 FBC has 
different requirements than ASCE 7. This difference in windborne debris regions 
is discussed in Section 2.2. In windborne debris regions, the 2001 FBC requires 
glazing to be impact-resistant or protected by shutters (glazing above 60 feet 
from grade is exempt). The Port Charlotte and Punta Gorda areas are in the 
windborne debris region, but inland areas along Hurricane Charley’s track (such 
as Arcadia) are not.

One of the notable successes observed was the greatly increased use of shutters 
on both residential and commercial buildings, in both the windborne debris 
region as well as inland areas. Although some windows were shuttered during 
Hurricane Andrew in 1992 (FEMA FIA-22, Building Performance: Hurricane Andrew in 
Florida, 1992), it was apparent that many residents of Florida now have a 
greater appreciation of the benefits of protected glazing. The increased glazing 
protection is likely due to code requirements, development and increased 
availability of protection products, and the public’s awareness of the 
vulnerability of unprotected glazing.

5.2.1	Residential Buildings
In a manufactured housing park in Zolfo Springs, an area not in the defined 
windborne debris region, windborne debris broke windows in several homes. The 
winds (estimated at 100 to 115 mph in Exposure B) generated a large amount of 
windborne debris. The majority of the windborne debris was from accessory 
structures and attachments as discussed in Section 4.6. In a manufactured 
housing park east of Port Charlotte, windows were broken in most of the homes 
and, in some cases, nearly all of the windows on the windward wall were broken 
(Figure 5-11). This park was in a windborne debris region, but the windows were 
not protected. Figure 5-12 illustrates broken windows in a new home that was 
still under construction. Because this house was still under construction, the 
contractor may have intended to install shutters in order to meet the windborne 
debris requirement; however, this was not done before the hurricane arrived. 
Window breakage was also caused by the failure of attached structures and pool 
cages. Figures 4-25 and 4-27 illustrated this type of damage.

Figure 5-11. Photograph. Most of the windows on this side of a manufactured home 
were broken by windborne debris (east of Port Charlotte).

Figure 5-12. Photograph. Three of four panes broken by windborne debris; other 
windows in this house also broke (Deep Creek).

Many windows in the windborne debris region were equipped with shutters, 
although most were not. Shutters were made of wood sheathing, metal panels, or 
plastic panels of various designs. A common shutter design used metal panels 
that were held by top and bottom tracks permanently mounted to the wall. Figure 
5-13 shows a house with roll-up shutters at the windows and metal panel shutters 
at the garage (garage door shuttering was rare).

Figure 5-13. Photograph. This house, which appeared undamaged from windborne 
debris, had roll-up shutters at the windows and metal panel shutters at the 
garage (Deep Creek).

Figure 5-14 shows a common metal awning shutter. These types of shutters provide 
very limited protection and should not be considered impact-resistant; they have 
not been tested to 2001 FBC requirements. A plastic roll-up shutter was observed 
that had been broken by windborne debris. It is doubtful that this shutter met 
the impact-resistance requirements specified in the 2001 FBC. In one case, 
windborne debris (a roof tile) was observed to have penetrated a Miami-Dade 
approved shutter (shown previously in Figure 3-32).

Figure 5-14. Photograph. Metal awning shutter penetrated by a missile (Zolfo 
Springs)

Some of the shutters did not have the strength to withstand the forces of the 
wind or the impacts of windborne debris. Others may have had sufficient 
strength, but were improperly installed. Figure 5-15 shows a house that used 
plastic shutters that blew off.

Figure 5-15. Photograph. All of the windows on this house were covered by 
plastic shutters, many of which were blown off during the hurricane, resulting 
in several broken windows (North Captiva Island).

The MAT observed some laminated glass windows, but none of them had been 
impacted by windborne debris, or if they had been impacted, they did not break. 
Some broken tempered glass windows were observed. Although tempered glass is 
more resistant to windborne debris than common glazing, when tempered glass 
breaks, it shatters into small pieces and falls out of the frame. Wind-driven 
rain could then be driven into the residence and substantially increase the 
internal pressure. When laminated glass breaks, the glass remains bonded to the 
plastic film between the panes, and the glazing remains in the frame. Although 
the glass will need to be replaced, the costly interior water and wind damage is 
avoided. On North Captiva Island, a house with laminated glass was observed 
where one sliding glass door panel was broken by impact from porch furniture, 
but the laminate held without a penetration. However, the impact of debris 
knocked the glass doors out of their tracks and opened the home to wind and 
water.

Some power-operated roll-down shutter systems were also observed. In at least 
one case, the shutter system did not include a manual system for retracting the 
shutter. As a result, it was impossible for the owner to open the shutters and 
air out the home after the storm. To minimize the possibility of developing 
mold, power-operated shutters should have alternate means of operation to allow 
opening after a storm.

5.2.2	Commercial and Critical/Essential Facilities
Window damage was observed on commercial buildings and critical/essential 
facilities throughout the impact area. Figure 5-16 shows a broken window in a 
mid-rise hotel in the Orlando airport area. Two windows on the same floor were 
likely broken by the missing plastic lens covers on the hotel sign. Figure 3-33 
also showed glass breakage near the Orlando airport. In the Wauchula central 
business district, windborne debris broke glass in several adjacent buildings 
(Figure 5-17). The estimated wind speed in this area was 100 to 115 mph. In 
Punta Gorda, tempered glass in a door and several windows were broken by 
windborne debris (as shown earlier in Figure 5-2) and a nearby three-story 
office building had very extensive glass breakage on all sides of the building 
(Figure 5-18). At least some of the breakage in both buildings was caused by 
aggregate from a nearby built-up roof (BUR) (Figure 5-19). Other types of 
windborne debris also impacted the three-story building (one missile penetrated 
the stucco and underlying metal lath). All of the glazing, including glass 
spandrel panels, was broken on the long side of the building shown in Figure 5-
18. 

Figure 5-16. Photograph. Window most likely broken by missing plastic lens 
covers on hotel sign (see top of building) (Orlando airport area)

Figure 5-17. Photograph. Broken glass in windows and doors in this building. 
Buildings across the street also had several broken windows caused by windborne 
debris Wauchula).

Figure 5-18. Photograph. All of the glazing, including glass spandrel panels, 
was broken on the long side of the building (Punta Gorda).

Figure 5-19. Photograph. Windows broken by aggregate from a nearby BUR. Besides 
impact at the crack intersection, aggregate chipped the glass in three other 
locations (Punta Gorda).

At a Charlotte County government building in Punta Gorda, a few of the windows 
were broken by windborne debris. Figure 5-20 shows missing spandrel panels; in 
other locations, the glass was broken. The windows extended from the floor to 
the ceiling, so tempered glass was used for personnel protection. Had laminated 
glass been used instead, any damaged glass would likely have stayed in its frame 
and would have provided wind and water protection. Glass broken by windborne 
debris was also observed at a hospital in Arcadia (Figures 3-41 and 6-8) and 
some of the fire stations in Port Charlotte and Punta Gorda. 

Figure 5-20. Photograph. Plywood panels installed where aluminum spandrel panels 
were blown out of the curtain wall (Punta Gorda)

Skylights were not particularly common in the area, but a couple of failures 
were noted.  In one case, the skylight in the bathroom of a manufactured home 
was broken by flying debris, allowing water to flood the bathroom. In another 
case, the failure of a skylight and difficulties in getting it replaced resulted 
in building officials prohibiting the owners from inhabiting their house for 2 
months following the storm.

Skylights were observed at a fire station on Pine Island in the service garage 
area; one of these skylights blew out. It was an old plastic skylight that 
integrated with an R-panel roof covering. This skylight likely failed due to 
inadequate resistance to wind pressures rather than by missile damage. 


5.3	Roof Systems 

Historically, damage to roof coverings and rooftop equipment is the leading 
cause of building performance problems during hurricanes. Rains accompanying a 
hurricane can cause water to enter buildings through damaged roofs, resulting in 
major damage to the contents and interior (Figures 5-21 and 5-22). Unless quick 
action is taken to dry a building, mold bloom can quickly occur in the hot, 
humid Florida climate. Drying of buildings was hampered after Hurricane Charley 
by the lack of electrical power to run fans and dehumidifiers. These damages 
frequently are more costly than the roof damages themselves. Water leakage can 
also disrupt the functioning of critical and essential facilities and weaken 
ceilings and cause them to collapse. Although ceiling collapse is unlikely to 
result in death, it can cause injury to occupants and further frighten them as 
they ride out the hurricane. 

Figure 5-21. Photograph. After the attic vent failed, water entered this 
residence. Wet carpeting and a substantial amount of wet gypsum board had to be 
removed (Punta Gorda Isles).

Figure 5-22. Photograph. The attic vent to the right (temporarily covered with 
felt) on this foam-set tile roof lifted during the hurricane and allowed water 
to enter the residence shown in Figure 5-21. The failed vent is like the one on 
the left (Punta Gorda Isles).

Essentials to good high-wind roof system performance include selection of a 
suitable system; product testing to ensure sufficient factored strength to 
resist design wind loads; enhanced design of details; quality application; and 
timely maintenance and repair. In addition, for critical and essential 
facilities in hurricane-prone regions, it is important to design a roof system 
that is likely to avoid water infiltration if the roof is hit by windborne 
debris (guidance is given in FEMA 424, Design Guide for Improving School Safety 
in Earthquakes, Floods, and High Winds).

For steep-sloped roofs, a secondary water penetration barrier that minimizes the 
water infiltration through the sheathing, if the roof cover fails, offers 
important backup protection for shingle, tile, and metal panel installations. 
Figure 5-23 illustrates the installation of self-adhering modified bitumen tape 
at sheathing joints. The tape is installed at the joints to allow water to shed 
off the sheathing if the primary roof covering (e.g., shingles) and underlayment 
are blown off. 

In lieu of attaching metal panels directly to structural members, installation 
of a roof deck between the panels and structure is preferred in hurricane-prone 
regions. The deck provides increased protection from windborne debris in the 
event of roof panel blow-off, as well as an opportunity for a secondary 
membrane. 

Figure 5-23. Photograph. Installation of self-adhering modified bitumen tape at 
sheathing joints, as part of an enhanced underlayment system on a Fortified…for 
safer living house under construction (IBHS) 

5.3.1	Asphalt Shingles
Although damage was observed on several new roofs (Figures 3-14, 4-11, and 5-
24), in general it appeared that asphalt shingles installed within the past few 
years performed better than shingles installed prior to the mid-1990s. The 
enhanced performance is likely due to product improvements (e.g., availability 
of greater bond-strength of the self-seal adhesive and availability of greater 
adhesive surface area) and less degradation of physical properties due to 
limited weathering time. It is doubtful that any of the observed roofs had been 
designed in accordance with UL Standard 2390, which was published in 2003. This 
standard pertains to two main items: 1) it provides a lab test method that 
manufacturers use to establish pressure coefficients for specific types of 
shingles; and 2) it provides a calculation procedure for a designer to determine 
the design wind load on the shingles, which is based on the coefficient from the 
testing, ASCE 7 criteria such as basic wind speed, and factors developed 
specifically for shingles. FEMA Hurricane Recovery Advisories No. 1 and No. 2 
(Appendix D) provide recommended practices for asphalt shingles on roofs in 
hurricane-prone regions. FEMA Hurricane Recovery Advisories No. 1 and No. 2 were 
based on guidance given in the fact sheets for FEMA’s Home Builder’s Guide to 
Coastal Construction (to be published).

Many shingle roofs in the Port Charlotte and Punta Gorda area were undamaged, 
while others lost a few hip and/or ridge shingles or a few tabs. Other roofs, 
including roofs far inland, lost many shingles, as shown in Figures 5-24 and 5-
25. The shingles shown in Figure 5-24 were attached with 6 nails per shingle, 
but the nails were attached about 1½ inches above the nail line. In addition, 
the shingles were poorly bonded. Though continuous, the self-seal strip was 
narrow (approximately ½ inch). When shingles were pulled apart during the 
investigation, many granules from the underlying shingle were pulled up, thus 
indicating that the granules were not well embedded. The starter course was 
incorrectly applied, and the nails at the hip shingles were incorrectly located. 
Note the area where the deck is exposed. Water can flow between the deck and 
underlayment and leak into the building at the sheathing joints. 

Figure 5-24. Photograph. Asphalt shingle roof installed on a new residence about 
2 months before the hurricane hit; shingles were blown off several areas (Deep 
Creek).

Hip or ridge shingles were often blown off while the remainder of the shingles 
were undamaged. The fasteners on all of the hip and ridge shingles that were 
observed were located in or above the self-seal adhesive, rather than below the 
adhesive as recommended by the industry. However, the hip and ridge shingles 
were blown off because of lack of bonding of the adhesive. Sometimes a limited 
amount of bonding occurred as shown in Figure 5-26, but frequently none of the 
adhesive had bonded. Lack of bonding of hip and ridge shingles is common. Figure 
5-2 shows nails that were improperly installed through the adhesive strip; they 
should have been driven below it. Use of asphalt cement to bond hip, ridge, and 
rake shingles (as recommended in FEMA 55, Coastal Construction Manual) was 
observed on only one roof. 

Figure 5-25. Photograph. Residence with a significant number of asphalt shingles 
lost. The metal window shutters shown were not designed for windborne debris 
(Fort Meade).

Figure 5-26. Photograph. Only the portion of the self-seal adhesive that is 
indicated in yellow had bonded (within the red circle). No bonding occurred on 
the right side of the hip line (Deep Creek).

Only one of the observed starter courses complied with industry recommendations. 
A common practice was to turn the starter shingle 180 degrees, rather than cut 
off the tabs. By turning the starter 180 degrees, the tabs of the first course 
of shingles were not bonded to the starter course, thereby making them 
susceptible to lifting. Use of asphalt cement to bond the first course (as 
recommended in FEMA 55, Coastal Construction Manual) was not observed. 

On a few roofs with architectural shingles, instances of blow-off of laminated 
tabs were observed (Figure 5-27). This type of failure was due to an inadequate 
amount and/or strength of adhesive used in the manufacturing of the shingles.

Figure 5-27. Photograph. Two laminated tabs blown off (Deep Creek)

In many instances where shingles were blown off, the underlayment was not 
damaged and, therefore, provided some degree of protection from water 
infiltration. In other instances, the underlayment was also blown off. Rain was 
then able to enter the building at the sheathing joints. FEMA Hurricane Recovery 
Advisory No. 1 (Appendix D)provides recommended practices for underlayments on 
roofs in hurricane-prone regions, including the use of self-adhering modified 
bitumen tape at the sheathing joints as shown in Figure 5-23.

On many residences that had been re-covered (i.e., new shingles had been 
installed on top of old shingles), large numbers of the re-cover shingles were 
blown away and the underlying older shingles remained in place. Some of these 
blow-offs may have been due to use of nails that were too short, although on the 
building shown in Figure 5-28, the nails had adequate sheathing penetration, but 
the newer shingles were poorly bonded (likely due to substrate irregularities). 
When re-covering versus tearing off the old shingles down to the sheathing, more 
substrate irregularity occurs, which can interfere with bonding of the self-seal 
adhesive of the new shingles. Most of the re-cover blow-offs were likely due to 
bonding problems associated with substrate irregularities.

Figure 5-28. Photograph. Re-covered apartment building (the newer shingles are 
grey and the older shingles are brown) (Deep Creek).

The shingles on the roof of the elementary school shown in Figure 5-29 were 
installed over underlayment over two layers of gypsum board atop a steel deck. 
The shingles were attached with a split-shank self-locking nail. At the rakes, 
the shingles were set in asphalt roof cement over the metal edge flashing 
(somewhat similar to the detail shown in FEMA Hurricane Recovery Advisory No. 2 
(Appendix D)). The 4½-inch vertical flange of the edge flashing was not cleated. 
At this rake and another rake, the edge flashing lifted. Because the shingles 
were well bonded to the flashing, they progressively failed. The shingle end 
nails at the rake were well inward of the industry’s recommended 1-inch 
placement. One of the end nails was 4 inches in from the edge of the shingle. 
Adhering the shingles to the edge flashing was a good practice, but the end 
nails should have been much closer to the edge, and the edge flashing should 
have had a much shorter vertical flange or the flange should have been face-
fastened or cleated. Several of the laminated tabs at this school were blown off 
(similar to Figure 5-27). This type of failure was due to an inadequate amount 
and/or strength of adhesive used in the manufacturing of the shingles. 

Figure 5-29. Photograph. Edge flashing that caused a progressive failure of the 
shingles (Deep Creek)

A portion of the shingles at the fire station in Cape Coral (constructed in 
1991) also blew off. Water leaked into the room housing the Emergency Management 
Services (EMS) computer equipment, resulting in minor damage. Minor damage also 
occurred at a post office on Pine Island that was constructed in 1993. 
Performance was quite good except for the loss of a few hip shingles and 
laminated tabs (similar to Figure 5-27). At a fire station in Punta Gorda, many 
of the three-tab shingles were blown off, and many of the staples were 
incorrectly oriented. This was one of the few roofs observed that had been 
attached with staples.

Several instances of ridge vent blow-off were observed. The performance of ridge 
vents with respect to prevention of wind-driven rain infiltration during the 
hurricane was not evaluated. 

The use of a larger number of nails (six instead of four) to attach shingles may 
have also played a role in the improved resistance of some of the newer roofs, 
but this was not verified through detailed inspections of the installations 
because it would have required access to the attics. The fasteners on all of the 
damaged shingles that were observed were located too high above the nailing line 
(i.e., the line printed on the shingle by the manufacturer). Fasteners were 
typically located 1 to 2 inches above the nailing line. End fasteners were often 
2 to 3 inches from the end, rather than the industry-recommended 1 inch. Nails 
rather than staples were used to attach most of the shingle roofs that were 
investigated.

5.3.2	Tiles
Clay and concrete tiles were observed, with concrete being the most common. A 
variety of tile profiles (e.g., S-tile and flat) were also observed, but no 
significant wind performance differences were attributed to profile. Mortar-set, 
mechanically attached, and foam-set (adhesive-set) attachment methods for tile 
roofs were observed during the assessment. Tile damage was observed along the 
path of the hurricane from the Port Charlotte/Punta Gorda area to Orlando. For 
the areas east of Port Charlotte and Punta Gorda, damage was typically limited 
to blow-off of hip and ridge tiles and blow-off of tiles along eaves. In the 
areas of Port Charlotte and Punta Gorda that received very high winds, there 
were larger areas of blown-off tiles. Tile underlayments were generally not 
blown off, with few exceptions (Figure 5-30). Therefore, many buildings with 
significant tile damage likely experienced little, if any, water infiltration 
from the roof. 

Figure 5-30. Photograph. A large area of underlayment at this mortar-set flat 
tile roof blew away. The loss of tile underlayment was atypical (Punta Gorda).

5.3.2.1 Mortar-Set Tile Roofs
The size of the blow-off area of tile roofs attached using mortar-set systems 
was typically much greater than for tile roofs attached using foam-set and 
mechanically attached systems. Figure 3-31 showed a mortar-set roof with a large 
area of blown-off tiles. 

On the roof shown in Figure 5-31, some of the tiles debonded from the mortar 
patties; other tiles debonded from the underlayment; and, in other instances, 
the underlayment tore off with the mortar. Mixed failure modes also occurred on 
the roof shown in Figure 5-32. Mixed failure modes also occurred in Hurricane 
Andrew (FEMA FIA-22, 1992). The mortar patties at the roofs shown in Figures 5-
31 and 5-32 were incorrectly located, and most of them were too small. 

Figure 5-31. Photograph. Mixed failure modes occurred on this mortar-set tile 
roof (Port Charlotte).

Figure 5-32. Photograph. Most of the mortar-set hip and ridge tiles blew off 
this house (Port Charlotte).

On the roof shown in Figure 5-32, most of the mortar-set hip and ridge tiles 
blew off. Some of the mortar-set flat tiles were also blown off, and other field 
tiles were broken by windborne debris (likely other tiles from this roof). 
Figure 5-33 shows three tiles from the roof shown in Figure 5-32. The mortar 
paddy on the left debonded from the underlayment. For the other two tiles, the 
underlayment tore away. The paddies were incorrectly located near the head of 
the tiles, which offers reduced uplift resistance. 

Figure 5-33. Photograph. Tile debris from the roof shown in Figure 5-32 (Port 
Charlotte)

5.3.2.2 Mechanically Attached Tile Roofs
Both direct-to-deck and batten-attached systems were investigated. Figure 5-34 
shows a batten-attached system where the roof is attached with nails. According 
to 2001 FBC (Table 1507.4.7), the attachment method observed on the roof shown 
in Figure 5-34 is suitable for buildings with a mean roof height up to 40 feet 
in areas with a design wind speed of 100 mph. (see note) The building (which has 
a mean roof height of less than 15 feet) is located in an area that is now 
mapped with a wind speed of approximately 110 mph; therefore, the installed 
attachment at this older residence was inadequate to meet the current code. The 
estimated speed at this Exposure B location was in the range of 110 to 120 mph. 
If the speed was in the lower portion of this range, the tiles did not perform 
as predicted by 2001 FBC (i.e., the tiles should have been good for 100 mph at a 
roof height up to 40 feet).

Note: In this chapter, basic wind speeds cited from the 2001 FBC are 3-second 
peak gust wind speeds, Exposure B, unless otherwise noted.

Figure 5-34. Photograph. Each tile on this building was attached to battens with 
a single 31/8-inch long smooth shank nail (Arcadia)

The building shown in Figures 5-35 and 5-36 is located in an area with a basic 
wind speed of approximately 120 mph. The 2001 FBC, therefore, requires 
compliance with the calculation method given in Section 1606.3.3. Load and 
resistance data can be found in the March 1, 2003, Addendum to the 3rd edition 
of the Concrete and Clay Tile Installation Manual (published by the Florida 
Roofing, Sheet Metal and Air Conditioning Contractors Association [FRSA] and 
Roof Tile Institute [RTI]). The roof in Figures 5-35 and 5-36 was attached with 
one 2½-inch long screw per tile directly to the deck. According to Table 12 of 
the Addendum, the attachment of this roof is suitable for buildings with a mean 
roof height up to 40 feet in areas with a basic wind speed of 150 mph. The 
estimated speed at this Exposure B location was in the range of 125 to 140 mph; 
therefore, the tiles did not perform as predicted by the Concrete and Clay Tile 
Installation Manual.

At a residence near the one shown in Figure 5-35, missiles (likely tiles from 
its roof) broke a few field tiles. The field tiles were attached with one screw 
per tile directly to the deck. 

Figure 5-35. Photograph. Windborne debris (likely tiles from this roof) broke 
several of the field tiles (Deep Creek).

Figure 5-36. Photograph. Loss of mortar-set hip tiles and several of the field 
tiles. Some of the screws remained in the deck, while others had been pulled out 
(Deep Creek).

Figure 5-37 shows several areas where batten-attached tiles on a fire station 
were damaged. The tile debris on the lower roof is from the upper roof. The 
tiles were installed when the building was re-roofed in the mid- to late-1980s. 
According to the 2001 FBC (Table 1507.4.7), the attachment method observed on 
this roof is suitable for buildings with a mean roof height up to 40 feet in 
areas with a basic wind speed of 100 mph. The building (which has a mean roof 
height of less than 30 feet) is located in an area that is now mapped with a 
basic wind speed somewhat less than 110 mph; therefore, the installed attachment 
at this pre-2001 FBC building does not meet the current code. The estimated 
speed at this Exposure B location was in the range of 95 to 110 mph. The tiles 
did not perform as predicted by the code (i.e., the tiles should have been good 
for 100 mph at a roof height up to 40 feet).

Figure 5-37. Photograph. Fire station with at least three battens blown off. 
Some tiles remained attached (Fort Meade).

5.3.2.3 Foam-Set Tile Roofs
The foam-set attachment method was developed after Hurricane Andrew in response 
to the widespread poor performance of mortar-set systems. Hurricane Charley was 
the first hurricane to deliver at or near-design wind speeds to this new 
attachment method. One- and two-part specially formulated polyurethane foam tile 
adhesives are available. Depending upon design uplift pressures and tile 
profiles, a variety of proprietary paddy schemes are available, including single 
paddy placement (with either small, medium, or large paddies) and two paddy 
placements. Although large areas of blow-off were unusual with this attachment 
method, they were observed on some residences. 

A large number of damaged foam-set systems were observed as shown in Figures 5-
38 through 5-45. Significant installation problems were observed with the size 
and/or location of the foam paddies. The side of the residence shown in Figure 
5-38 was the side the damaging winds came from. Assuming the intent was to 
provide a small paddy placement, according to the foam manufacturer’s 
literature, this attachment would have been suitable for a basic wind speed of 
135 mph in Exposure B (assuming proper application). According to the 2001 FBC, 
the basic wind speed where this residence is located is approximately 125 mph; 
therefore, this small paddy placement meets code. The estimated Exposure B wind 
speed at this site was in the range of 125 to 140 mph. If the foam paddies had 
been properly sized and located according to the manufacturer’s literature, the 
tiles should not have blown off. 

Figure 5-38. Photograph. In addition to the damage shown in this photo, this 
one-story roof lost virtually all of the hip and ridge tiles (see Figures 5-22, 
5-39, and 5-40) (Punta Gorda Isles).

In Figure 5-39, to meet the small paddy placement criteria, the paddy should 
have been 3 inch by 3 inch minimum, with approximately 8 to 9 square inches of 
foam contact with the tile near the head. As shown in the photo, clearly there 
was very insufficient contact area. The paddies were rectangular rather than 
square; perhaps a medium paddy placement was intended. To meet the medium paddy 
placement criteria, the paddy should have been 2 inch by 7 inch minimum, with 
approximately 12 to 14 square inches of foam contact area. The small round 
paddies shown in Figure 5-39 were placed after the down-slope tiles were set. 
Foam from these paddies occurred between the tile end laps. These round paddies 
are not shown in the foam manufacturer’s installation instructions. Although 
failed tiles typically debonded from the paddies, in at least one location, the 
paddy debonded from the cap sheet underlayment. 

Figure 5-39. Photograph. Note the very small contact area of foam at the tile 
heads (left side of the tiles) and very small contact area at the tails. The 
long narrow paddies were intended to be underneath the pan portion of the tile 
(Punta Gorda Isles).

Figure 5-40 is a close-up of the eave area of the roof shown in Figure 5-38. The 
manufacturer’s installation instructions do not require screws, but they do 
require foam paddies. 

Figure 5-40. Photograph. View of the eave. The first row of tiles was attached 
with two screws per tile; foam was not used to adhere this row (Punta Gorda 
Isles).

The residence shown in Figures 5-41 and 5-42 was located in an area identified 
in the code with a basic wind speed of approximately 125 mph. Assuming the 
intent was to provide a small paddy placement, according to the foam 
manufacturer’s literature, this attachment would have been suitable for a basic 
wind speed of 135 mph in Exposure B (assuming proper application). The estimated 
speed at this Exposure B location was in the range of 125 to 140 mph. Therefore, 
had the foam paddies been properly sized, located, and installed according to 
the manufacturer’s literature, the tiles should not have blown off.

Figure 5-41. Photograph. In addition to field tile blow-off, most of the hip 
tiles and several ridge tiles were also blown off this house (Punta Gorda 
Isles).

To meet the small paddy placement criteria, the paddy should have been 3 inch by 
3 inch minimum, with approximately 8 to 9 square inches of foam contact with the 
tile near the head. The paddies were typically about the correct size, but they 
did not achieve the required contact area (see inset in Figure 5-42). The 
paddies were also typically located too close to the upslope end of the tile. In 
Figure 5-42, the first row of tiles at the eave was attached with one nail and a 
foam paddy. Most of the nails remained in the deck. The foam manufacturer’s 
instructions do not require nails at the eave. (The dark spots on the tile are 
rain drops.) An attic vent also rolled back and allowed water to enter the 
building.

Figure 5-42. Photograph. The paddy on the tile at the lower left debonded from 
the asphalt bleed-out near a cap sheet lap. Only the center portion of the 
paddies made contact with the tiles, as shown in the inset (Punta Gorda Isles).

Several foam-set tiles were blown off a one-story bank (Figures 5-43 through 5-
45), primarily due to insufficient contact area of the paddies. To meet the 
small paddy placement criteria, the paddies should have provided approximately 8 
to 9 square inches of foam contact with the tile near the head. The paddies were 
observed to be about the correct size, but they did not achieve the required 
contact area. The paddies were also typically located too close to the upslope 
end of the tile. Though not required, a very small paddy was placed at the tile 
overlaps (red arrow in Figure 5-43). The tiles typically debonded from the 
paddies, but the two paddies shown at the bottom of Figure 5-43 debonded from 
the cap sheet. Many mortar-set hip and ridge tiles were also blown off. The bank 
was located in an area identified in the code with a basic wind speed of 
approximately 125 mph. Assuming the intent was to provide a small paddy 
placement, according to the foam manufacturer’s literature, this attachment 
would have been suitable for a basic wind speed of 135 mph (assuming proper 
application). The estimated speed at this location was in the range of 125 to 
140 mph. Therefore, had the foam paddies been properly sized and located, 
according to the manufacturer’s literature, the tiles should not have blown off.

Figure 5-43. Photograph. This photo clearly shows insufficient contact area of 
foam-set paddies on the bank’s roof (Punt a Gorda Isles).

Figure 5-44. Photograph. In this photo, the portion of the paddy that made 
contact with the tile is clearly visible (Punta Gorda Isles).

Figure 5-45. Photograph. Tile remained bonded to the paddy, but, except where 
bonded, the tile blew away. A large portion of the paddies shown in Figure 5-43 
and this figure failed to make tile contact, which was a typical observation 
(Punta Gorda Isles).

5.3.2.4 Hip and Ridge Tiles 
Blow-off of hip and ridge tiles as shown in the previous figures was very 
common, even in areas with only moderate wind speeds. Most of the hip and ridge 
tiles that were investigated were attached with mortar, although a few were 
attached with mortar and a single nail. The installation of a nail near the head 
of the hip and ridge tiles did not greatly improve blow-off resistance. Because 
the hip and ridge tiles project several inches above the adjacent field tiles 
and form a transition between different roof surfaces, the raised hip/ridge line 
of tiles may be subjected to higher wind loads than expected on the field tiles 
due to turbulence. This research issue is worthy of future investigation. The 
vulnerability of hip and ridge tile blow-off was documented following Hurricane 
Andrew (FEMA FIA-22, 1992). It was reported that the current design guidelines 
were inadequate (T.L. Smith, 1994). 

5.3.2.5 Sprayed Polyurethane Foam
A few tile roofs that had been covered with sprayed polyurethane foam (SPF) were 
investigated by the MAT. Figure 5-46 shows one of these roofs. A missile had 
impacted the foam and gouged it in several locations, but no tile debris was 
blown off. The SPF appeared to provide some protection for the tiles. However, 
SPF applications may not improve the uplift resistance. Figure 5-47 shows a roof 
that lost SPF covered tiles. In this instance, the SPF bonded tiles together 
and, as a result, large sections of tiles were lifted off the roof. Although the 
larger fragments should not fly as far as smaller fragments, because they are 
more massive, they could be more damaging (depending upon their velocity) if 
they were to become windborne. 

Figure 5-46. Photograph. This residence had a tile roof that had been covered 
with SPF. A missile gouged the foam, but no tile debris was blown off (Punta 
Gorda Isles).

Figure 5-47. Photograph. The other side of the roof shown in Figure 5-46 with a 
portion of the underlayment and several tiles blown off (Punta Gorda Isles)

5.3.2.6 Tile Missiles
There were many reports of tiles or tile fragments hitting occupied buildings 
and flying through windows (shown previously in Figures 3-31 and 3-32). The 
owner of one residence reported that six of their windows were broken by tiles 
from a neighbor’s house (Figure 5-48). The homeowner’s metal roof was not 
damaged, but wind-driven rain forced through the broken windows caused extensive 
interior wind and water damage. 

Figure 5-48. Photograph. Tiles that flew through windows of an occupied 
residence (Deep Creek)

In addition to becoming windborne debris and further damaging the roof on which 
they were installed, many tile roofs were damaged by other types of windborne 
debris. One of the advantages of foam-set, according to one of the 
manufacturer’s literature, is that foam-set installation is supposed to result 
in “high resistance to damage from missile impact,” meaning that the “tile may 
break but remains adhered to the roof.” Although this may be true for the 
portion of the tile that is adhered, the MAT observed that broken portions that 
are not adhered are vulnerable to being blown away as shown in Figure 5-49. 

Figure 5-49. Photograph. A view of the roof on the back side of the residence 
shown in Figure 5-41. Tiles (including a hip tile) from the front garage roof 
landed in this area and broke several field tiles (Deep Creek).

It is important to note that other types of roofing systems are also capable of 
generating windborne debris (Figure 3-34); however, missiles are most 
problematic with tiles. FEMA Hurricane Recovery Advisory No. 3 (Appendix D) 
provides recommended practices for tiles on roofs in hurricane-prone regions. 
This Advisory was based on observations from Hurricanes Charley, Frances, and 
Ivan.

5.3.3	Metal Panel Roofs
Although small in number compared to houses with asphalt shingle and tile roofs, 
several residences in the Port Charlotte, Punta Gorda, Pine Island, and Sanibel 
and North Captiva Island areas had metal roof coverings. Many of these coverings 
were 5V-Crimp metal panels. This type of panel uses exposed fasteners (Figure 5-
50). The majority of the 5V-Crimp metal panel roofs observed were not damaged, 
or only experienced hip or ridge flashing damage. However, significant panel 
loss was observed at a few residences. At a fire station on Pine Island, the 5-V 
Crimp metal panels blew off the main building and the plywood panels blew off 
with the panels. Furring strips (1x) occurred between the plywood and the 
trusses. The furring strips, which had been stapled to the trusses, were lifted 
off with the plywood and likely were the cause of the panel loss. There was 
significant water infiltration; however, a temporary roof had been installed 
after the hurricane, and the station was occupied at the time of the 
investigation. 

Success or failure of the 5-V Crimp metal roof coverings was likely primarily 
dependent upon fastener spacing and type, although panel gauge may have had some 
influence (panels are available in 24 to 29 gauge). Screws provided greater 
pull-out resistance than ring-shank nails and were more resistant to dynamic 
loading. One of the failed roofs that were investigated was attached with ring-
shank nails. 

Another key element of good performance is the spacing of fasteners along the 
eave and at hip and ridge flashings. Only a single fastener occurred at the eave 
between the rib fasteners shown in Figure 5-50; considering the basic wind speed 
of 125 mph 3-second peak gust in this location, use of two fasteners between the 
ribs would have been prudent. Note that the hip flashing is bowed; two fasteners 
between the ribs would have also been prudent at the flashings. Close spacing at 
the flashings and eave is important to keep the flashings and panel ends from 
billowing during high winds. Although the roof in Figure 5-50 did not fail, the 
flashing and eave fasteners were too far apart. 

Figure 5-50. Photograph. The number of fasteners was not increased at the 
corner, perimeter, hip, or ridge areas (close-up of the residence shown in 
Figure 5-5). Also note that several of the soffit panels were blown away (Deep 
Creek).

Most of the 5V-Crimp panels that blew off failed as a result of the panel 
fasteners pulling out of the sheathing. However, plywood substrate blow-off and 
wood nailer failures were also observed (Figure 5-51). The upper asphalt shingle 
roof shown in Figure 5-51 had been re-covered with 5V-Crimp panels attached to 
nailers. The nailers were inadequately attached to the sheathing. Note that the 
hip flashing on the lower roof blew off.

Figure 5-51. Photograph. These panels blew off the upper roof and landed on the 
lower roof of this house (Bokeelia, north end of Pine Island).

All of the 5V-Crimp roofs that were observed were unpainted galvanized or 
aluminum-zinc alloy (“Galvalume”) panels. Aluminum-zinc alloy panels are very 
resistant to corrosion. No significant corrosion problems were observed. An 
advantage of 5V-Crimp (and other types of exposed fastener) panels (versus 
panels with concealed clips) is that, after installation, it is easy to verify 
that the correct number of fasteners were installed.

A variety of architectural metal panels were also observed. As with the 5V-Crimp 
panels, some of the roofs were undamaged, others had lost hip or ridge 
flashings, and others lost a large number of panels. Performance of 
architectural panels is a function of the strength of the panels and their 
interlock with the clips, clip spacing and attachment, and strength of the 
flashing attachments. Some of the failed hip and ridge flashings were attached 
with cleats rather than exposed fasteners. Cleat attachment is not as reliable 
as exposed fasteners.

When metal panels or hip/ridge flashings blow off, they can become high-energy 
windborne debris that can damage buildings and other property and cause injury. 
These types of windborne debris can travel a considerable distance.

A variety of exposed fastener and architectural and structural metal panels were 
observed on commercial and critical/essential facilities. Figures 5-52 through 
5-54 show a medical office building that lost approximately 75 percent of the 
superstructure supporting the architectural metal panel roof that encircled the 
perimeter of the building. This building experienced significant water damage. 
Although much of the aggregate roof covering remained in the center portion of 
the roof, temporary roof covering (Figure 5-53) was installed to minimize water 
intrusion after the metal panel structure blew away.

Figure 5-52. Photograph. Medical office building (Port Charlotte).

Figure 5-53. Photograph. The wood and metal framed superstructure blew away and 
exposed the lightweight insulating concrete roof deck (Port Charlotte).

Figure 5-54. Photograph. View of the canopy ridge at the building shown in 
Figure 5-52. The ridge flashing fasteners were placed too far apart. A 
significant amount of water leakage can occur when ridge flashings are blown 
away (Port Charlotte).

Figure 5-55 shows a roof on a school in Arcadia that performed fairly well; the 
building was located inland in an area that experienced approximately 110 to 120 
mph wind speeds. Temporary repairs to the roof covering had been made prior to 
this photo taken by the MAT.

Figure 5-55. Photograph. This standing seam metal roof had a 16-inch rib 
spacing. There was some rake flashing damage, and a few rake panels were also 
damaged (Arcadia).

Figures 5-56 and 5-57 show an architectural panel roof on a fire station in the 
Deep Creek area that performed poorly, with metal panels that were blown off. At 
this station, the 2-inch high ribs had a 16-inch spacing. The panels had a 
single-lock fold. There were two screws per clip. Typically the clips remained 
attached to the deck, but some did not. Clip spacing varied widely across the 
roof and from panel to panel with spacings ranging from 2 feet 4 inches to 3 
feet 3 inches. The eave clips shown in Figure 5-57 should have been located near 
the edge. It would have been prudent to install double clips along the eave. 

Figure 5-56. Photograph. Several of the architectural panels and hip flashings 
blew off this fire station (Deep Creek).

Figure 5-57. Photograph. This photo provides a view of the eave of the building 
shown in Figure 5-56. The clip at the left was 13 inches from the edge of the 
deck. The other clip was 17 inches from the edge (Deep Creek).

At the headwall flashing, the flashing was pop-riveted to the panels at 16 
inches on center. Along one side of the hip, the flashing was attached at 2 feet 
2 inches; on the other side of hip line, they were at 1 foot 10 inches. The hip 
and headwall flashing fastener spacing was excessive; for the building code 
design requirements in Charlotte County, a fastener spacing of 4 to 6 inches on 
center is typically used, depending on the design wind speed at one site. Some 
panels on another roof area were damaged by windborne debris (OSB panels), and 
water entered the building at the penetration location. 

Several structural standing-seam trapezoidal panel system failures were 
observed, including the panels on the Turner Agri-Civic Center in Arcadia, which 
partially collapsed (see Section 6.5.1.1. and Figures 6-17 and 6-18). It was 
reported that the roof covering was lifting prior to the collapse. The panels 
were installed over fiberglass batts over a vapor retarder atop the light-gauge 
purlins. 

Figure 5-58 shows an exposed fastener R-panel roof on two old pre-engineered 
metal buildings that had been re-covered with SPF. A large wall section blew out 
of one of the buildings, and the edge flashing was torn away, but the metal roof 
panels remained in place. At an adjacent building with a similar roof, the metal 
panels on a canopy were blown away, but the failure did not propagate into the 
roof panels on the main building. The SPF covering likely prevented progressive 
failure at both of these buildings due to the stiffness that it imparted to the 
panels.

Figure 5-58. Photograph. The metal wall panels and metal edge flashing on this 
building blew away, but the exposed fastener R-panels with an SPF covering did 
not progressively fail (Wauchula).

Figure 5-59 shows a mansard with metal shingles simulating tiles. The metal 
shingles performed well. However, metal shingles can also experience significant 
damage as discussed in FEMA 489, Hurricane Ivan in Florida and Alabama. Note 
that the rooftop mechanical equipment in Figure 5-59 remained attached to the 
support stands. Also note the lightning protection system on the parapet in the 
foreground. One of the conductor connectors detached from the roof and the 
conductor pulled out of some of the connectors.

Figure 5-59. Photograph. Metal shingles (simulating tile) that performed well 
(Port Charlotte)


5.3.4	Low-Slope Membrane Systems
The MAT observed several types of low-slope roof systems. These systems included 
BURs, modified bitumen, and single-ply, which are described below. 

5.3.4.1 Built-Up Roof (BUR) and Modified Bitumen
A BUR failure was observed at one of the terminals at the Orlando International 
Airport. Portions of its BUR blew off, resulting in water infiltration. To dry 
out the interior and to avoid mold growth, the airport used large air dryers to 
remove the moisture. 

At a hospital in Arcadia, several windows at the intensive care area were broken 
by windborne debris (Figure 6-8). Most, if not all, of the windborne debris was 
aggregate from the hospital’s roofs. Three of the eight intensive care rooms 
were taken out of service due to the glass breakage and windows were broken in 
other patient rooms. Gutters and walkway pads were also blown off (Figure 5-60). 
The gutters and pads possessed sufficient mass to be very damaging missiles and 
may have caused some of the glass breakage observed.

Figure 5-60. Photograph. This view of the back side of the upper roof of a 
hospital (see Figure 6-8) shows that the missing gutter and asphalt plank 
walkway pad were blown away (Arcadia).

Aggregate commonly used on BUR systems was blown off many buildings. One example 
is shown in Figure 5-61 where the school was being used as a shelter. There was 
no other apparent damage to the roofs at the school, including the mechanically 
attached single-ply membrane on a courtyard building. The boarded-up broken 
windows at the residence across from the school in Figure 5-61 were likely 
broken by aggregate from this roof. Roofing aggregate was found at the far side 
of the street in front of the house. The inner leg of the coping in Figure 5-61 
was attached with screws spaced at 3 feet 5 inches, 2 feet 11 inches, and 3 feet 
1 inch; the coping was not damaged. Aggregate also blew off a new portion of a 
hospital in Port Charlotte, but no missile damage was observed. At another roof 
area of the hospital, a portion of the mineral surface cap sheet roof was blown 
off. The metal edge flashing had improperly been installed underneath the 
membrane; therefore, the flashing was unable to clamp the roof edge. Wind lifted 
the gutter and metal edge flashing and peeled the roof membrane. Figure 5-62 
shows an area of the hospital roof that nearly failed. With the flashing in a 
lifted position, the membrane was very susceptible to peeling. Apparently the 
winds subsided before this occurred.

Figure 5-61. Photograph. Although this roof had an 11-inch high parapet, 
aggregate was blown off (Port Charlotte).

Figure 5-62. Photograph. The edge flashing at this mineral surface cap sheet 
roof lifted (Port Charlotte).

An edge flashing failure also occurred at a middle school in Cape Coral, first 
occupied in 1998 (Figure 5-63). This metal edge flashing had also been 
improperly installed underneath the membrane. The flashing should have been 
installed over the modified bitumen cap sheet and then stripped in to clamp the 
edge of the membrane. Wind lifted the gutter and metal edge flashing and peeled 
the modified bitumen membrane. The gutter was not designed for uplift 
resistance. A portion of a middle school in Port Charlotte also had a mineral 
surface BUR cap sheet roof, and the metal edge flashing had also been improperly 
installed underneath the membrane. However, none of the edge flashings lifted. 
Except for some missile damage, the BUR on this roof performed very well.

Figure 5-63. Photograph. The edge flashing had a 2-inch vertical flange that 
extended into the gutter. The flashing was not cleated (Cape Coral).

A high school in Arcadia had an aggregate surface BUR over lightweight 
insulating concrete (LWIC) over steel form deck that had been  installed in the 
mid-1970s. Two areas over the cafeteria blew off, as did a portion over the gym. 
Repairs had been made by the time the MAT inspected, so it was not possible to 
definitively determine the cause of the failures. The failures at the cafeteria 
occurred several feet from the parapet. These failures may have been due to base 
sheet rupture around the fasteners (which may have been due to spacing problems, 
fastener corrosion, or deterioration of the base sheet), or deformation or 
cracking of the LWIC. At the gym roof, the blow-off area extended to the 
parapet, but it was unclear if the blow-off originated at the parapet. This roof 
may have failed for the reasons given at the cafeteria, or this failure may have 
been related to the coping or base flashing attachment. The 13½-inch-wide coping 
was attached only at each coping joint with three nails in the horizontal flange 
and one in the vertical flange. There was significant water infiltration in the 
cafeteria and gym.

5.3.4.2 Single-Ply
One aggregate ballasted system was observed in the Orlando airport area. Some 
aggregate was blown off the roof, but this may have been due to gutter blow-off. 
A detailed investigation was not performed. In addition to the BURs discussed 
above, one of the hospitals in Port Charlotte had single-ply membranes at two 
different areas. There was no apparent damage to the mechanically attached 
ethylene propylene diene monomer (EPDM) membrane roofs on the lower levels. 
However, there was extensive damage to the mechanically attached polyvinyl 
chloride (PVC) membrane (with 6-foot 3-inch row spacing) on the fourth floor 
roof (the highest roof), as shown in Figure 5-64. Emergency repairs had been 
made, so it was not possible to definitively determine the cause of the failure. 
Mechanical equipment was missing and portions of the lightning protection system 
(LPS) had become detached. It is possible that a piece of equipment or LPS 
conductor cut the membrane, and a progressive failure occurred. Extensive water 
damage was observed on the fourth floor and some on the third floor. The fourth 
floor was evacuated after the roof membrane blew off. 

Figure 5-64. Photograph. View of a portion of the fourth floor roof of a 
hospital after installation of an emergency roof (the black area). The deck was 
concrete (Port Charlotte).

In addition to the BUR discussed above, the middle school in Port Charlotte also 
had single-ply membranes on two roof areas. Wind likely lifted the gutter and 
metal edge flashing and peeled the membrane on the gym (Figure 6-21). At a lower 
roof, the mechanically attached PVC alloy membrane with 4-foot 3-inch row 
spacing was installed over polyisocyanurate insulation over an old BUR over LWIC 
over a steel form deck (shown previously in Figure 3-59). The failure of this 
roof was also likely initiated by gutter failure. However, it may have been 
initiated by progressive tearing after missile impact (there were numerous 
missile tears), or by pull-out of membrane fasteners. Several membrane fasteners 
near the edge of the roof had been pulled out, which is not surprising. Metal 
form decks are typically thinner and therefore offer less pull-out resistance 
than standard steel decks.

At a county building in Punta Gorda, the mechanically attached PVC alloy 
membrane was punctured in several areas. Because the roof was much taller than 
surrounding buildings, the punctures were likely caused by rooftop equipment 
that was blown away and by the LPS components that became detached. The membrane 
peeled back at a corner area, but since emergency repairs had been made at the 
time the MAT visited, it was not possible to definitively determine the cause of 
the failure. The membrane fastener rows were at 4 feet 6 inches on center in the 
field of the roof. At the perimeter, the rows were 1 foot 11 inches on center. 
The perimeter width was 11 feet 8 inches. The corners appeared to be attached in 
the same manner as the perimeter.

5.3.5	Gutters and Downspouts
Gutters and/or downspouts were blown off many buildings. In most cases, loss of 
gutters caused little or no damage to the steep-slope roof coverings to which 
they were attached; however, the gutters and downspouts that were blown off 
became windborne debris. As discussed in Section 5.3.4, loss of gutters often 
resulted in lifting and progressive failure of low-slope membrane systems.


5.4	Wall Coverings, Non-Load Bearing Walls, and Soffits 

Hurricane Charley caused wall covering, non-load bearing walls, and a 
significant amount of soffit damage throughout the hurricane path. The following 
factors are essential to resist high winds: product testing to ensure sufficient 
factored strength to resist design wind loads; suitable anchoring of the wall 
coverings, non-load bearing walls, and soffits to the building; use of moisture 
barriers (e.g., asphalt saturated felt or housewrap) where appropriate; and 
proper flashing, sealants, and drainage to minimize water intrusion into wall 
cavities or into occupied space.

5.4.1	Wall Coverings
Wall covering damage was observed by the MAT primarily on houses with vinyl 
siding. There were several instances of vinyl siding failure as shown in Figures 
5-65 and 5-66 (and previously in Figure 3-24). Wall covering failure was more 
commonly observed in manufactured home parks than elsewhere in the hurricane's 
path. When vinyl siding was blown off, the underlayment (either asphalt-
saturated felt or housewrap) was also typically blown away. With loss of the 
siding and underlayment, wind-driven rain was then able to enter the wall 
cavity, causing water damage and initiating mold growth. Vinyl sidings that 
became windborne debris were capable of breaking unprotected windows. 

Vinyl siding that was blown off typically tore around the fastener points. Vinyl 
siding manufactured for high-wind areas is available. With high-wind siding, the 
nailing flange is folded over, so there is a double thickness of vinyl at the 
fastener points. None of the failures that were observed used high-wind siding. 

In some cases, the MAT believes that the blow-off was triggered by unlatching of 
the bottom portion of the panel (Figure 5-65). Once the panel unlatches from the 
retainer slot just below the nailing flange, the panel is free to rotate outward 
where it can be caught by the wind and blow off. The magnitude of the unlatching 
issue, compared to the strength of the nailing flange and fastener spacing, is 
unknown. When unlatched, panels are very susceptible to blow-off. 

Figure 5-65. Photograph. The vinyl siding panel with the red arrow is unlatched. 
The panel above and several others are also unlatched (Zolfo Springs).

Vinyl siding is quite susceptible to windborne debris damage as shown by Figure 
5-66. Because the vinyl siding cannot resist debris impact, resistance to debris 
impact is provided by the wall sheathing (if any) between the siding and the 
wall studs. On some of the residences, plastic foam sheathing was used instead 
of wood sheathing between the vinyl and the studs. The walls of these buildings 
offered very little resistance to windborne debris penetration, as they were 
composed only of vinyl siding, underlayment, foam sheathing, fiberglass batt 
insulation in the wall cavity, and gypsum board on the interior side of the 
studs. Residents who rode out the hurricane in their homes were quite 
susceptible to injury from windborne debris penetrating the light exterior 
walls.

Figure 5-66. Photograph. The vinyl siding on this manufactured house was 
ruptured in several locations by windborne debris (most of which were likely 
building envelope components from other nearby manufactured houses). Note the 
missing skirt and loose foundation anchor straps (Zolfo Springs).

Underlayment had not been installed at all on some residences and at the 
Bokeelia Post Office on Pine Island (constructed in 1993). Not installing 
underlayment is a poor practice because vinyl siding (like many other types of 
wall coverings) does not prevent water from getting behind the siding. 
Underlayment should always be installed to intercept the leakage and drain it 
out of the wall. The 2001 FBC does not currently require underlayment underneath 
vinyl siding. Further discussion and analysis of vinyl siding is presented in 
FEMA 489, Hurricane Ivan in Florida and Alabama.

A variety of wall coverings other than vinyl siding were observed. They also 
typically performed well, but there were exceptions. There were several 
instances of metal wall panel failures; these typically occurred on older pre-
engineered metal buildings. The key to achieving good performance of metal 
panels is selecting an appropriate panel system and installing an adequate 
number and type of fasteners. Figure 5-67 shows good attention to attachment of 
metal fascia panels on a school. Stitching the termination of panels with 
closely spaced fasteners as shown in Figure 5-67 prevents the end of the panel 
from billowing and becoming detached from the concealed clips. 

Figure 5-67. Photograph. Standing seam metal panels with a 16-inch rib spacing 
were used at the fascia and secured with closely spaced exposed fasteners  
(Arcadia).

5.4.2	Non-Load Bearing Walls
Exterior non-load bearing walls generally performed well, but there were notable 
exceptions. Figure 3-25 showed a collapsed unreinforced masonry wall. Figures 3-
33 and 5-68 show extensive EIFS failure on a hotel near the Orlando airport. 
Other EIFS damage was shown in Figure 3-26. Further discussion and analysis of 
EIFS failures is given in FEMA 489, Hurricane Ivan in Florida and Alabama, where 
this type of damage was prevalent.

Figure 5-68. Photograph. This hotel experienced significant EIFS failure on 
several sides (Orlando). EIFS debris broke several windows (Figure 3-33).

5.4.3	Soffits
Many buildings lost some or all portions of their soffits (shown previously in 
Figure 3-21). The damaged soffits were typically vinyl or aluminum. Some of the 
soffits failed by suction (i.e., downward pressure), while others failed by 
positive pressure (i.e., they were pushed upward). In many instances where 
soffits were lost on residences, water was driven into the attics and ultimately 
into living spaces. The wind also displaced attic insulation and blew it out of 
attics (much of the insulation was blow-in insulation, rather than insulation 
batts). Figure 5-69 shows ceiling damage adjacent to soffit loss at a residence 
on North Captiva Island.

Figure 5-69. Photograph. An exterior eave with soffit failure, which resulted in 
water intrusion (North Captiva Island)

Figure 5-70 shows a damaged soffit at a bank drive-through canopy. Figure 5-71 
shows damaged soffits at the Aqui Esta Fire Station. Most of the gutters and 
downspouts at the fire station blew away, but the 5V-Crimp metal roof had only 
minor damage at a hip flashing lap. The soffit panels were connected to the 
building only at their ends. A substantial quantity of wind-driven rain blew 
into the attic space and caused ceiling boards to collapse. Because of the water 
infiltration, this station was taken off-line after the storm. 

Figure 5-70. Photograph. Loss of soffit at a bank drive-through. Note the coping 
damage (Port Charlotte).

Figure 5-71. Photograph. Essentially all of the perforated aluminum soffit on 
this fire station was blown away (Aqui Esta, east of Punta Gorda Isles).


5.5	Exterior Mechanical and Electrical Equipment Damage

The MAT observed many damages to mechanical and electrical devices mounted on 
the exterior of buildings. The devices attached to residential, commercial, and 
critical/essential facilities are typically different from each other; for this 
reason, the following section presents information according to building type. 

The following factors are essential to good high-wind performance of exterior 
mechanical and electrical equipment: determining design wind loads on equipment 
and designing suitable attachments to resist the loads; special anchoring of fan 
cowlings and access panels; and special design of LPS anchorage. Guidance for 
these design factors is provided in FEMA 424, Design Guide for Improving School 
Safety in Earthquakes, Floods, and High Winds.

5.5.1	Damage to Exterior Equipment Attached to Residential Buildings
Typically, the types of exterior equipment attached to residential buildings 
included air-conditioning condenser units and TV satellite dishes; however, this 
report focuses on condensers.

Condenser units were generally not anchored to their support pad, which resulted 
in their being displaced off the support pad by wind (Figure 5-72). In some 
instances, the condensers broke free from the electrical and copper tube 
connections and were blown away entirely. 

Figure 5-72. Photograph. This condenser was not anchored to the concrete pad. 
The electrical and copper tube connections kept it from blowing farther away 
(Deep Creek).

In several cases, the condensers were fastened and remained anchored throughout 
the hurricane (Figure 5-73). Typically, where anchors were used, the clips were 
often very thin and the screws quite small. Although the condensers did not move 
during this hurricane, additional precautions to prevent wind damage should be 
taken. In some cases, corrosion of fasteners was observed; this can result in 
failure in a future hurricane event. In high-wind areas, clips and screws with 
high strength and corrosion resistance should be used.

Figure 5-73. Photograph. Condenser on the elevated platform attached with four 
angle brackets. The other condenser, located adjacent to it on the ground, 
should also have been on an elevated platform to account for storm surge (Pine 
Island).

5.5.2	Damage to Exterior Equipment Attached to Commercial and Critical/Essential 
Facilities 
Commercial and critical/essential facilities typically have a wide variety of 
mechanical and electrical equipment attached to their rooftops and elsewhere. 
Equipment lost included fan units and HVAC units, electrical and communications 
equipment, and LPS systems. There are several effects due to loss of this 
equipment: in many instances, the displaced equipment left large openings 
through the roof and/or punctured the roof membrane; equipment loss often 
affected the operational functions of the facilities; and blown-off equipment 
became high-energy windborne debris in some cases. The equipment observed on 
hospitals, fire stations, and schools was not anchored more effectively than the 
equipment on common commercial buildings.

5.5.2.1 Condensers
Condenser problems like those discussed in Section 5.5.1 were also observed at 
commercial and critical/essential facilities (Figures 5-74 and 5-75). A complete 
lack of anchor systems or inadequate or deteriorating fasteners resulted in the 
loss of many compressors. Installation methods observed were not standardized. 
In Figure 5-75, although the condenser did not move off its rail, it would have 
been prudent to use two side-by-side screws, with more edge distance between the 
screw and strap end.

Figure 5-74. Photograph. Condenser unit displaced from the elevated platform 
(Port Charlotte)

Figure 5-75. Photograph. Rooftop condenser anchored to a support rail, but with 
only one small screw (which was corroded) used to connect the strap (Port 
Charlotte).

5.5.2.2 Fan Units and HVAC Units
Figure 5-76 shows the loss of fan cowlings on a roof. Two of the three cowlings 
had blown off. At one curb, which was 2 feet 4 inches square, the fan unit was 
attached to the curb with two small screws at two sides and three small screws 
on the other two sides (total of 10 screws). Attachment was not checked at other 
fans. No fans were blown off this building. This success was likely the result 
of using multiple screws to secure the fasteners (unlike many other buildings, 
where often only two screws per fan were used). 

Figure 5-76. Photograph. Cowlings blown off two exhaust fans in the foreground. 
Note also he loose LPS conductors and missing walkway pad (Punta Gorda).

Loss of HVAC units was also observed as shown in Figure 5-77. A number of these 
large units were blown off their supports. Additional damage to equipment 
included loss of access panels on package units, debris impact damage to relief 
air hoods, and damaged rooftop ductwork. Figure 5-78 shows a unit that was 
marginally anchored.

Figure 5-77. Photograph. A large HVAC unit blew off this curb. Note the loose 
LPS conductors (this side of the curb). This school had significant damage to 
several pieces of rooftop equipment (Port Charlotte).

Figure 5-78. Photograph. A thick angle bracket was used to anchor this unit. 
Although two screws attached the angle to the support beam, only one screw was 
used at the unit (Port Charlotte).

5.5.2.3 Electrical and Communications Equipment
Rooftop electrical and communications equipment were also observed to be 
inadequately anchored. Problems included blow-off of satellite dishes (Figures 
5-79 and 5-80), antenna collapse (shown previously in Figure 3-44), and 
displacement of LPS (Figures 5-59 and 5-81 through 5-85). Four buildings with 
LPS were investigated, and the systems on all four buildings were damaged. Three 
of the buildings had two or more roof levels and damage occurred at several of 
the different levels. Consequences of the damage included loss of 
communications, damage to the roof covering, and loss of lightning protection, 
the latter of which is significant, considering the frequency of lightning 
storms in Florida.

LPS failures were typically the result of poorly anchored systems. Connectors 
often fail by opening up and releasing the conductor cable or they debond from 
the roof (Figure 5-82). In other cases, the air terminal base plates debond from 
the roof (Figure 5-83). In Figure 5-84, a prong-type conductor splice connector 
(approved for roof heights up to 75 feet) failed. Bolted-type connectors are 
prudent in hurricane-prone regions because they are less likely to pull apart 
and cause damage to the roof (Figure 5-85).

Figure 5-79. Photograph. This satellite dish at a hospital was held down only 
with CMU. Note the loose LPS conductors and displaced air terminal at the corner 
(Arcadia).

Figure 5-80. Photograph. A satellite dish previously sat in this location. It 
was held down only with CMU and blew off the five-story building (Punta Gorda).

Figure 5-81. Photograph. The LPS conductor on this hospital blew away, but the 
air terminal was still attached. A lightning strike to this air terminal would 
not be safely dissipated (Port Charlotte).

Figure 5-82. Photograph. The LPS conductor pulled away from the conductor 
connector at the top of the photo. The conductor was also attached to the 
membrane with poorly welded strips of PVC (Port Charlotte).

Figure 5-83. Photograph. The conductor connectors detached from the cap sheet on 
a hospital’s BUR. The air terminal was also displaced (Port Charlotte).

Figure 5-84. Photograph. A failed prong-type splice connector with prongs 
permitted for roof heights up to 75 feet caused roof damage at this facility 
(Cape Coral).

Figure 5-85.  Photograph. When LPS conductors detach, the conductor ends can 
whip around and puncture and tear the roof membrane. The patch near this frayed 
conductor is likely a repair of damage caused by a whipped conductor (Punta 
Gorda).