Wind Pressure Measurements on Full Scale Flat Roofs

May 15, 2003

April 2003 Interface • 17
A flat roof surface is known to be subjected to unusually high
suction induced by a pair of “horseshoe” vortices caused by the
wind coming diagonally facing a corner of the building (Kind and
Wardlaw, 1979). This phenomenon can cause very serious damage
to the roofing system, such as dislocation of concrete “pavers” or
insulation boards. It is also known that some architectural features
of the building, such as parapets, varying in height, have
some influence on these phenomena (Baskaran, 1986). At the
same time, wind tunnel testing of this situation is a challenging
task because the extent of damage depends a
lot on the structural details, which can hardly
be modeled properly in a reduced scale.
Since the early 1970s, several wind tunnel
studies have focused on this issue, and there
are commonly observed difficulties in modeling
the structural details (Kind, Savage & Wardlaw
1988). Wind tunnel testing of a full-scale
building, on the other hand, would be nearly
impossible. By taking advantage of a very large
test section of the 9m x 9m wind tunnel at the
National Research Council of Canada (NRCC),
Kind and Wardlaw carried out a series of comprehensive
studies of the wind effects on a
variety of roof assemblies during the period of
1975-1990. These formed the basis for several
roofing standards internationally.
In wind tunnel studies, models had rigid
roofs, and their deformation due to wind suction
was assumed negligible. However, in a
mechanically attached single ply roof, the
membrane may oscillate once high suction is
applied, which in turn may give a different
wind-induced pressure distribution. Another
engineering concern is the structural detail of how the roofing
system is installed on the building. In order to examine these
points, a pilot study was carried out. This paper reports benefits
of the use of full-scale roof component materials for the wind tunnel
tests of such roof sections.
The models used for this series of tests are 1/10 in linear
scale and equipped with full-scale roof component materials, as
shown in Figure 1. All tests were conducted at the 9m x 9m wind
tunnel. Two different types of roofing membranes were used for
Figure 1: Roof assembly module with full-scale components.
A flat roof surface is known to be subjected to unusually high suction associated with a pair of vortices created under diagonally
skewed wind. Prediction of the wind load and its influence on the roof system is an important engineering topic, but wind tunnel
testing of this situation is not easy with a scaled model. Considering the physics of wind flow for the most critical case, a possibility
to use a full-roof configuration has been investigated. Results are found to be encouraging, allowing the use of much more accurate
structural details of roofing systems, such as flexible membranes.
this experiment; one is a reinforced
polyvinyl chloride (PVC) membrane and
the other is a non-reinforced membrane
made of ethylene propylene diene
monomer (EPDM). Physical, mechanical,
and chemical properties of these
two membranes are significantly different
(Baskaran, Paroli & Booth, 1997).
This is because PVC is a thermoplastic
polymer, whereas EPDM is a rubberbased
thermoset material.
A typical mechanical property of
these membranes is shown in Figure 2,
which compares the stress-strain
behavior of two materials under the
same tensile stress rate. The PVC membrane
has significantly higher breaking
load compared to EPDM. However, due to its reinforced nature, it
can stretch only about 1/8 as much as EPDM.
Differences between two roofing assemblies used in the wind
tunnel study are summarized in Table 1. The membranes were
equipped with 89 and 100 special pressure taps for PVC and
EPDM models, respectively, in a manner to minimize their influence
on the mechanical properties of membranes. For both roof
configurations, the influence of wind speed and direction and
building height on wind pressure were examined. Details of the
wind simulation are found elsewhere (Savage et al., 1997;
Baskaran et al. 1996).
Figures 3 and 4 compare the PVC and EPDM system response.
They present an overall pattern of the mean pressure distribution
for wind approaching normal to the building side and for the 45°
skewed wind for each of the PVC and EPDM cases. High suction
near the up-wind corner and moderate pressure distribution for
the rest of the roof are commonly observed for both systems. The
characteristic “horse-shoe” vortex zone for the 45° wind case is
again clearly identified. These features are similar to the previous
wind tunnel studies on rigid models.
Wind-induced suction over the assembly results in membrane
deflection between the attachments to the roof deck. As indicated
in Table 1, the PVC roof is composed of four sheets that are
attached to the structural deck using mechanical fasteners. These
four areas are referred to as S1, S2, S3, and S4 in Figure 3.
During wind tunnel investigations, four zones of membrane ballooning
were noticed. For the PVC roofs, maximum deflection was
observed at the sheet S3, which is wider than S1and S2 and it is
not restrained along its edges as it is in S4.
For the roofs with non-reinforced EPDM membrane, measured
deflections were much higher than the PVC roofs. It is critical to
note that the observed pressure distribution pattern was totally
45 mil (≈1.14 mm) thick 45 mil (≈1.14 mm) thick
Reinforced Non-reinforced
Membrane Spot fastener Bar attachment
Two perimeter (900mm) sheets One sheet
Two field (1800mm) sheets
Seams hot air weld Factory seam
Insulation 25 mm ISO 25 mm XEPS
Underlayment 6 mil (0.2 mm) vapor barrier 6 mm support board
Deck 22-gauge steel 22-gauge steel
Total of tested configurations 30 48
18 • Interface April 2003
Table 1: Differences between PVC and EPDM roof assemblies
Figure 2: Mechanical property of typical membranes.
April 2003 Interface • 19
different from the deflected shape of the membrane. This indicates
that neither the number of sheets used nor the way the membranes
are attached to the roof deck had any significant influence
on the wind-induced mean pressure distribution patterns. This
also indicates that the deflected shape stays inside of the low
pressure separation bubble, and it does not influence the mean
pressure distribution.
EPDM roofs generally experienced higher mean suction than
the PVC roofs for the same wind conditions. Maximum measured
mean suction coefficient for EPDM was -4.5 compared to that of
-2.4 for the system with PVC membranes. Also, the spatial distribution
is different between the two cases. For the normal wind, as
shown in Figure 3, the measured pressure is about -0.5 when x/L
= 0.4 in the case of the PVC roofs. For the same location, the
EPDM roof experienced -0.9 (Figure 4). Positive pressure on the
roof assembly indicates reattachment of the separated flow over
the roof surface. Figures 3 and 4 indicate longer distance before
the reattachment for the case of PVC roofs compared to EPDM roofs.
Figure 3: Mean pressure distribution with
PVC roof membranes.
Figure 4: Mean pressure distribution with
EPDM roof membranes.
In order to help understand the difference of wind effects
between the two roofs, spectral analyses were performed on the
pressure time histories. Figure 5 shows the normalized spectral
density of typical pressure records with PVC and EPDM roofs,
measured at the perimeter region. A low-pass filter with cut-off
frequency of 50 Hz was used to eliminate high-frequency noise
that existed with the pressure records. The spectral density function
of the approaching wind is also included for comparison.
Three distinct response regions have been noticed at three different
frequency bands as follows:
1) Up to a frequency of about 2 Hz, the approaching wind
and PVC record show similar energy distribution. In comparison
to this, the EPDM record has less energy in the
same frequency range.
2) For the frequency of 2 to 10 Hz, both PVC and EPDM have
similar energy distribution, which is also higher than the
energy of the approaching wind.
3) For frequencies above 10 Hz, the energy content of EPDM
is much higher than that of the approaching wind or the
PVC. This is due to the fact that a flexible membrane of
EPDM exhibits more fluttering compared to less flexible
PVC membrane. The fluttering vibration is caused by the
approaching wind.
To identify the effects of membrane motion on the induced
surface pressures, the above analysis could be extended to develop
the concept of a transfer function. This provides the ratio of
membrane response pressure to the applied flow pressure fluctuations
as measured on rigid roof pressure models at each particular
The transfer function developed as such could be used for the
correction of pressure readings measured on rigid
roof models to include the membrane vibration
effects. The success of this approach could open
up a much larger and more general database for
analyses useful for the design of roofing systems.
Wind tunnel studies were also carried out
with the models on full-scale roof component
materials. Flexible roof membranes could change
the geometrical shape of the roof under suction
and the resultant shape varies depending on the
direction of approaching wind and the materials,
and layout of the membrane attachments to the
deck. However, the overall mean pressure distribution
pattern on flat roofs was found to not be
significantly influenced by the materials. This
provides the possibility of introducing a transfer
function to take the dynamic effects of membrane
vibration into account as a correction factor
applicable to the measured results from rigid flat
roof models. 
Baskaran, A., Wind Loads on Flat Roofs with
Parapets, M.Eng. thesis, Concordia University,
Canada, 1986.
Baskaran, A., R.M. Paroli, and R.J. Booth, “Wind Performance
Evaluation Procedures for Roofing Systems: Current
Status and Future Trends,” Proceedings of the 5th International
Conference on Building Envelope Systems &
Technology, Bath, U.K: 37-54, 1997.
Baskaran, A., M.G. Savage, F. Alfawakhiri, and K.R. Cooper,
“Pressure Distribution Data Measured During the October
1995 Wind Tunnel Tests on a Mechanically Attached
EPDM Single Ply Roofing Systems,” LTR-A-004, National
Research Council Canada, 1996.
Kind, R.J., M.G. Savage, and R.L. Wardlaw, “Prediction of
Wind-induced Failure of Loose-laid Roof Cladding
Systems,” Journal of Wind Engineering & Industrial Aerodynamics,
29: 29-37, 1988.
Kind, R.J. and R.L. Wardlaw, “Model Studies of the Wind
Resistance of Two Loose-laid Roof Insulation Systems,”
LTR-LA-234, National Research Council Canada, 1979.
Melbourne, W.H., “Wind Tunnel Blockage Effects and
Corrections,” Wind Tunnel Modeling for Civil Engineering
Applications, Cambridge University Press: 197-216, 1982.
Savage, M.G, K.R. Cooper, and A. Baskaran, “Wind Tunnel
Investigation of the Wind Loads on a Single Ply Mechanically
Attached PVC Roofing System,” LTR-A-012, National
Research Council Canada, 1997.
The authors acknowledge the technical insight provided by Dr.
D. Surry on Figure 5. The discussion and encouragement toward
the project by Mr. R.L. Wardlaw are greatly appreciated. The presented
research is being carried out for a consortium: Special
20 • Interface April 2003
Figure 5: Comparison of pressure spectra from the PVC and EPDM roof assemblies.
April 2003 Interface • 21
Interest Group for Dynamic Evaluation of Roofing Systems
(SIGDERS). SIGDERS was formed from a group of partners who
were interested in roofing design. These partners included:
Manufacturers – Atlas Roofing Corporation, Canadian General
Tower Ltd., GAF Materials Corporation, GenFlex Roofing Systems,
Firestone Building Products Co., IKO Industries Ltd., Johns
Manville, Sarnafil, Soprema Canada, and Stevens Roofing
Systems. Building Owners – Canada Post Corporation, Department
of National Defence, Public Works and Government Services
Canada. Industry Associations – Canadian Roofing Contractors’
Association, Canadian Sheet Steel Building Institute, Industrial
Risk Insurers, National Roofing Contractors’ Association, and Roof
Consultants Institute. Research Agencies – Institute for Research
in Construction, Institute for Aerospace Research, and Canadian
Construction Material Centre.
Part of this paper was published originally at the 10th
International Wind Engineering Conference.
Dr. A. “Bas” Baskaran is a Group
Leader for the Roofing Sub-Program at
the National Research Council of
Canada, Institute for Research in
Construction (NRC/IRC). At the NRC,
he is researching the wind effects on
building envelopes through experiments
and computer modeling. He also
serves as adjunct professor at the
University of Ottawa. Baskaran is the
vice-chairperson for the Roofing
Committee on Weather Issues (RICOWI)
and a member of ASCE, SPRI, and CIB technical committees.
He has authored and/or co-authored over 150 research articles
in the area of wind effects on buildings. Being a professional
engineer, Baskaran received his bachelor’s degree in engineering
from Annamalai University, Madras, India. His master’s
degree in engineering and Ph.D. were from Concordia
University, Montreal, Canada. Both research topics focused on
the effects of wind on buildings and earned best dissertation
awards from the Canadian Society of Civil Engineers.
M.G. Savage is a Research Officer
for the Separated-flow Aerodynamics
group at the National Research
Council of Canada, Institute for
Aerospace Research (NRC/IAR). His
experience is in the field of wind engineering
and his activities have
involved wind tunnel studies to determine
the effects of wind on several
engineering structures. His focus has
been the aerodynamics of bluff bodies
and the effects of these aerodynamic
forces on the dynamic stability and response to
unsteady wind loads over a broad range of applications.
Research projects have included experimental studies on building
aerodynamics, wind uplift loads on flat roofs, power lines,
and aeroelastic stability of long-span bridges. He has been
involved in international collaborative research programs that
have involved working in wind tunnels in Japan. Savage
received his bachelor’s degree in mechanical engineering from
the University of Ottawa, Canada.
Over 40 speakers from eight different countries will present
at the first North American Green Roof Infrastructure
Conference, Awards, and Trade Show May 29-30 at Chicago’s
Congress Plaza Hotel. Training on green roof design and implementation
techniques will include waterproofing, structural
loading, plant selection, and design. It will cover case studies
from Nashville, Tennessee, to Sechelt, British Columbia, and
London, England. The latest green roof research findings from
Europe and North America will be presented in the areas of
energy efficiency, plant survivability, biodiversity, stormwater
retention, and integrated design.
Awards will be presented to designers of six exceptional
green roof projects. Entries are being accepted now, with submission
due by April 15. Judges are: William Thompson,
FASLA, Editor, Landscape Architecture Magazine; Monica E.
Kuhn, BES, B. Arch, OAA – Architect, Founder, Rooftop
Garden Resource Group; N. Marcia Jiménez, Commissioner,
City of Chicago Department of the Environment; Michael
Gibbons, FCSI, CCPR, Head of Green Roof Task Group, ASTM,
President, Architectural Systems; and Steven W. Peck,
Executive Director, Green Roofs for Healthy Cities.
For more information on evaluation criteria and how to
submit an entry, visit or e-mail Ireen
Wieditz, Director, Green Roofs for Healthy Cities, at
For information to attend the conference, visit www.greenroofs.
Enter Your Project