Skip to main content Skip to footer

Field-Monitoring the Wind Performance of Commercial Roofs Part 1: Data from Ottawa Site

May 15, 2012

Phone: 613-990-3616 • Fax: 613-998-6802 • E-mail:
2 7 T H RC I I N T E R N A T I O N A L C O N V E N T I O N A N D T R A D E S H OW • MA R C H 1 5 2
0 , 2 0 1 2 B A S K A R A N E T A L . • 5 1
Building codes and standards specify pressure coefficient data for wind design of commercial
roofing systems. Data is derived primarily from wind tunnel studies. The roofing
community in North America has undergone much change over the last 30 years along with
advances in material science, computer-aided design, and engineering application.
However, the wind tunnel data that were developed over two decades ago can be said to be
less appropriate to quantify wind-induced loads of current roof coverings. Data from field
measurements can be used to benchmark the wind tunnel studies and current wind load
provisions. A strategic project has been initiated by SIGDERS at the NRC to collect the field
performance of commercial roofs in three locations across North America: (1) Ottawa,
Ontario, Canada (2) Mt. Pleasant, MI, USA, and (3) Rialto, CA, USA. This paper presents the
data from the Ottawa site.
A. “BAS” BASKARAN is a group leader and senior research officer at the National
Research Council of Canada, Institute for Research in Construction (NRC/IRC). He has
spent 25 years researching the effects of wind on building envelopes through wind-tunnel
experiments and computer modeling. Baskaran serves as adjunct professor at the
University of Ottawa. He is a member of RCI, ASCE, SPRI, RICOWI, ICBEST, and CIB technical
committees. His work in the area of wind engineering and building envelopes has
received national and international recognition. Baskaran has an extensive research record
with more than 150 publications in peer-reviewed journals and conference proceedings. A
professional engineer, Baskaran received his master’s degree in engineering and his doctorate
degree from Concordia University, Montreal, Canada. Both of his graduate research topics
focused on the effects of wind on buildings and earned “best dissertation” awards from
the Canadian Society of Civil Engineers.
5 2 • B A S K A R A N E T A L . 2 7 T H RC I I N T E R N A T I O N A L C O N V E N T I O N A N D T R A D E S H OW • MA R C H 1 5 2
0 , 2 0 1 2
Understanding wind performance of
commercial roofs and minimizing roof blowoffs
are critical in the service life of a building
as the roof protects its occupants and
contents. The roofing community in North
America has undergone major changes over
the last 30 years, along with advances in
material science, computer-aided design,
and engineering application. A companion
paper, “Evaluating Wind Effects on
Commercial Roofs – North American Advancements,”
presents the market share
data and classifies commercial roofs into
the following groups: Flexible Roofing
Systems (FRS) and Rigid Roofing Systems
(RRS) (Baskaran et al., 2010). In an FRS,
components such as insulation and cover
boards are integrated using mechanical fasteners.
In an RRS, components are integrated
using adhesives. Comparison of the
wind uplift performance
showed major differences in the
response between these systems.
During the last 16 years
(1994-2010), a North American
roofing consortium, Special
Interest Group for the Dynamic
Evaluation of Roofing Systems
(SIGDERS) developed test
methods and standards that
were published by the Canadian
Standards Association
(CSA A123.21-10, 2010). To
assist roof designers and manufacturers,
these new developments
are being referenced in
the National Building Code of
Canada (NBCC, 2010).
The wind tunnel methodology
that was developed over two
decades before is less appropriate
to quantify wind-induced
loads on roof coverings. Data
used in the building codes and
standards were derived primarily
from the wind tunnel studies,
without considering the
influence of the roof covering on
the induced loads. Field monitoring of roofs
can provide a basic understanding for wind
flow interactions with roofs. For asphalt
shingle roofs, Peterka et al. (1997) conducted
a field model study and the data were
used in the development of the ASTM
D7158 test method. Limited studies were
conducted on roof systems at a full-scale
test facility at Texas Tech University
(McDonald et al., 1991). However, the questions
have neither been investigated nor
addressed for commercial roofs. Some of
them are as follows:
• How should one quantify the difference
in the wind loads between rigid
versus flexible roofs?
• What is the role of roof membrane
flexibility on wind loads?
• How should one model the roof covering
flexibility and maintain boundary
layer flow in the wind tunnel
Figure 1 – EPDM roof mock-up construction.
without introducing the Reynolds
number effect?
As part of the SIGDERS, a long term
project has been initiated at the NRC to collect
field performance of commercial roofs
in three locations across North America: (1)
Ottawa, Ontario, Canada, (2) Mt. Pleasant,
MI, USA, and (3) Rialto, CA, USA. This
paper presents data from the Ottawa Site.
In Ottawa, the field performance study
was conducted on two roof mock-ups. The
roof mock-ups were 3.7 x 3.7 m (12 ft. x 12
ft.) and were constructed to replicate the
field construction of the FRS. Mock-up 1
had PVC flexible roof membrane and mockup
2 had reinforced EPDM. In addition to
the membrane, both the mock-ups incorporated
50-mm (2-in.) thick polyisocyanurate
insulation boards that were
fastened to the 22-Ga steel
decks. The EPDM membrane
layout had a fastener row spacing
(Fr) of 1470 mm (58 in.),
while the PVC was designed
with Fr of 1670 mm (66 in.).
Both the EPDM and the PVC
membranes were fastened to
the steel deck with fasteners
and plates at a fastener spacing
(Fs) of 610 mm (24 in.).
However, the EPDM membrane
seams were overlapped with
primed, dual-sided secure
tape, while the PVC seams were
one-sided hot-air weld. Figure 1
shows the typical construction
of EPDM mock-up.
The two constructed mockups
were placed on the southside
roof of an NRC office building
with a roof height of 13 m
(42 ft.), resulting in a mock-up
roof elevation of 14 m (47 ft.)
from ground level. Figure 2
gives the impression of the surrounding
exposure of NRC and
2 7 T H RC I I N T E R N A T I O N A L C O N V E N T I O N A N D T R A D E S H OW • MA R C H 1 5 2
0 , 2 0 1 2 B A S K A R A N E T A L . • 5 3
was verified with
a preset threshold
value. If the
maximum wind
speed was less
than the threshold
wind speed,
the data for that
particular day
were discarded;
otherwise, the
sampled data
were saved for
further anaysis as
minute averages
for the 24-hr period.
The raw data
Figure 2 – Site exposure of roof rocks and wind instrumentation. (0.01 Hz) for
the wind measurement coordinates. As per
the NBCC 2010, the exposure can be classified
as “rough or urban” exposure (which is
equivalent to Exposure B in ASCE 7-2010).
As shown in Figure 2, wind measurements
were made in two locations: one at a freestanding
meteorological tower that is located
farther west of the building at a distance
of about 61 m (200 ft.) in the upstream. The
meteorological tower has a propeller-type
wind anemometer at two different elevations
of 6 m (20 ft.) and 9 m (30 ft.). The second
location was on the rooftop, where two
types of anemometers—propeller and ultrasonic—
were placed side by side at a height
of 2.7 m (9 ft.) above the roof. Each mockup
was equipped with two differential-pressure
transducers to quantify the windinduced
suctions on the membrane surface.
To measure the membrane deformation,
each mock-up was fitted with an ultrasonic
deflection sensor; and to measure the fastener
load transferred by the deflected
membrane to the steel deck, load cells were
installed on each mock-up. Details of the
instrumentation, sensor accuracy, and
data-collection software were documented
elsewhere (Usama et al., 2010).
At the freestanding meteorological
tower, the wind data were recorded at a
sampling rate of 0.1 Hz (10 samples per second).
On the building top where the roof
mock-ups were placed, the wind and the
wind-induced response on the roofing system
(i.e., pressure), membrane deflection,
and fastener load were sampled at a frequency
of 0.01 Hz (100 samples per second).
The sampled data of 0.01 Hz were
saved on an hourly basis, and at the end of
those particular
hours where the wind speed was greater
than the threshold were also saved for further
analysis. This threshold was set to 15.6
m/s (35 mph), which was assumed low
enough to gather a sufficient number of data.
Verification of the measured wind data
with the theoretical calculation was completed
following the conventional power law
equation as shown in Equation 1.
V1 Z =( 1)α
V2 Z2
Equation 1.
As shown in Figure 3, the freestanding
meteorological tower was comprised of two
wind propellers at 6 m (20 ft.) [Z1] and 9 m
(30 ft.) [Z2], respectively. With the known
wind speeds at these heights (V1 and V2),
using Equation 1, the flow exponent (α) was
determined as 0.36. This confirms that the
building exposure falls into the category of
“urban or rough exposure.” Using the estathe
24-hr period, the maximum wind speed Figure 3 – Wind data at the freestanding weather tower
5 4 • B A S K A R A N E T A L . 2 7 T H RC I I N T E R N A T I O N A L C O N V E N T I O N A N D T R A D E S H OW • MA R C H 1 5 2
0 , 2 0 1 2
Figure 4 – Comparison of the wind speed: Rooftop versus free stream at z = 16 m
(52 ft.).
blished gradient height (Zg) of 500 m (1640
ft.) for the urban exposure, the gradient
wind speed Vg was obtained. Then the free
stream mean wind speed V52 at 16 m (52 ft.)
was calculated. Figure 3 shows the wind
data (measured and calculated). At the freestanding
meteorological tower, the data
clearly show the relationship between the
wind speed and height (i.e., increasing wind
speed with height).
Figure 4 compares the calculated free
stream wind speed at Z=16 m (52 ft.) with
the measured wind speed at the roof top
Z=16 m (52 ft.). The comparison clearly
indicates that calculated free stream wind
velocity is lower than the measured data;
however, it follows the trend of the measured
data with statistical data showing a
variation of 15% in the average. Therefore, it
can be said that the measured rooftop wind
speed at 52 ft. is not influenced by the
building flow separations. In other words, a
tower height ranging from 1.2 to 1.5h
(where h is the building height) was found
to be appropriate for measuring undisturbed
wind at the building roof.
The field mock-ups were monitored for a
period of one year, starting from January
2009. With the preset threshold limit of
15.6 m/s (35 mph), 13 days recorded a
maximum wind speed greater than the
threshold. In other words, for the remaining
94% of the time, the winds were less than
15.6 m/s (35 mph). As per the NBCC 2010,
the reference wind speed for wind load calculation
is the hourly mean wind. Following
that norm, the daily hourly mean wind
speed was calculated by taking a 60-minute
segment from the daily time history data.
For each day, the time at which the maximum
wind speed measured was used as the
reference and the one-hour time segment
were selected by taking 30-minute data on
either side of the reference. Figure 5 shows
the statistical data of the peak and hourly
mean wind. The highest wind speed measured
during the monitoring period was 30
m/s (67 mph) with hourly mean of 19.9
m/s (44 mph). Figure 5 also clearly shows
that the wind comes between 270 and 360
degrees (i.e., between W-N directions), with
dominant wind direction being NNW.
Figure 6 shows a typical 24-hr time history
of the measured wind-induced
response on the EPDM and PVC mock-ups.
The horizontal axis represents time. For
this particular day, the winds were mostly
from a NW-N direction with a recorded peak
speed of 30 m/s (67 mph). Note that the
average wind speed was 19.9 m/s (44 mph),
resulting in a gust factor of 1.5. On the
EPDM system, the measured peak suction
pressure was 1 kPa (21 psf), while the PVC
system recorded a peak suction pressure of
0.7 kPa (14 psf). The wind approaches from
the NW-N direction and the induced suctions
over the two mock-ups show that the
roof with EPDM membrane experienced
higher pressure fluctuations than the roof
with PVC membrane. These higher pressure
fluctuations on the EPDM membrane response
clearly indicate the influence of
membrane material stiffness on the windinduced
response. A similar observation is
noticed in the membrane deflections plotted
in Figure 6.
The measured suction pressures were
converted into pressure coefficients referenced
at the building roof height. The pressure
coefficients’ time series is defined in
Equation 2.
Figure 7 shows the peak pressure coeffi-
Figure 5 – Wind parameters for the 1-year monitoring period.
2 7 T H RC I I N T E R N A T I O N A L C O N V E N T I O N A N D T R A D E S H OW • MA R C H 1 5 2
0 , 2 0 1 2 B A S K A R A N E T A L . • 5 5
Cp(t) Dp(t) =(1
2 *ρ*V2)
Equation 2.
Dp(t) is the differential pressure measured by the membrane pressure tap;
the denominator being the dynamic pressure with ρ is the air density; and
V is the hourly mean wind measured at the rooftop.
cients of the EPDM and PVC systems for all 13 days. The highest
Cp calculated on the EPDM system was -4.1, while on the PVC system,
it was -3.4. For roof slopes less than 7º and building heights
less than 60 ft. (18 m), NBCC 2010 specifies design pressure coefficients
of -5.4, -2.5, and -1.8, respectively for the corner, edge, and
field zones of the roof. Based on the NBCC’s roof zone calculation,
the EPDM mock-up can be said to be placed in the corner zone of
the building, while the adjacent PVC mock-up moves into the edge
zone of the building. For comparison purposes, the respective corner
and edge coefficients are also plotted in Figure 7, which clearly
shows the conservatism of the code specifications in comparison
to the measured EPDM and PVC data.
Further analysis regarding data reproducibility and roofing systems’
response to wind loads is discussed by organizing the data
into four scenarios: S1, S2, S3, and S4 as shown in Figure 7.
• S1 is the scenario for data reproducibility. It would be ideal
to have a constant wind speed from a particular direction
for different days, and S1 represents this case for April 17
and May 3. The peak wind for both of these days was
approximately 18.7 m/s (42 mph), with mean speeds of 12
m/s (27 mph), and the prevailing wind direction was WNWNW-
NNW. The calculated pressure coefficients were almost
the same for the PVC system, while there was a minimum
Figure 6 – Typical wind-induced suction on the mock-ups and
their response.
Figure 7 – Wind-induced response of roof mock-ups: design pressure coefficients.
5 6 • B A S K A R A N E T A L . 2 7 T H RC I I N T E R N A T I O N A L C O N V E N T I O N A N D T R A D E S H OW • MA R C H 1 5 2
0 , 2 0 1 2
difference for the EPDM system.
• S2 is the scenario where the wind
was streaming the windward side of
the building at 1350, (i.e., in the
ESE-SE-SSE direction). As the PVC
system was in the immediate locality
of the approach wind, it recorded
higher pressures compared to the
EPDM system for both the days,
which is reflected through the higher
pressure coefficients.
• In S3, the approach wind is from the
N, which is parallel to the membrane
seam. The peak wind speed
was close to the threshold of 15.7
m/s (35 mph) with an average of
11.2 m/s (25 mph). Both EPDM and
PVC systems measured similar
pressure coefficients. Changing the
wind angle to almost perpendicular
(W-WNW-NW) to the membrane
seams as in scenario S4 (April 2),
the EPDM system measured higher
pressures compared to the PVC system,
thus affecting the influence of
membrane sheet orientation on
induced pressures.
The pressure coefficients reflect the roof
covering response under the wind dynamics.
It should be understood that membrane
flexibility can also influence the induced
pressures. Figure 8 plots the measured
membrane deflection for the 13 windy days.
Higher wind speeds caused higher membrane
deflections on both mock-ups. The
data also show that apart from the wind
speed, the wind direction was an influencing
parameter in the membrane response.
For example, take the scenario of S1 where
the approach wind is coming from the
WNW-NW-NNW. The maximum deflection
measured on the EPDM and PVC mock-up
was 57 mm (2.25 in.) and 35 mm (1.5 in.)
respectively. With the same speed in S2, the
wind changed the direction to ESE-SE-SSE
(i.e., to 180º). This caused higher deflection
of the PVC membrane compared to the
EPDM mock-up. Though both the systems
had different Fr spacing or sheet width [Fr
of EPDM = 1670 mm (66 in.), Fr of PVC =
1470 mm (58 in.)], the measured deflection
data indicate very minimal changes. Thus,
it is clear that the membrane response is
more dependent on the flexible nature of
EPDM membrane, which had higher peak
deflections compared to the stiffer PVC
As part of the ongoing SIGDERS consortium,
a major project to quantify the wind
uplift performance of low-slope commercial
roofs is in progress at NRC. This paper presents
a summary of the study conducted on
two roof mock-ups placed over the roof of an
NRC office building in Ottawa. This study
provides knowledge on field instrumentation,
data collection process, sensor accuracy,
and their sensitivity.
To determine these wind loads, field
monitoring of in-service roofing systems at
the Mt. Pleasant and Rialto sites are in
progress. Data measured from 2010 offer
insight on the wind dynamics on commercial
roofing systems. Preliminary comparison
with ASCE 7-2010 raises question
about the roof zone dimensions (i.e., corner,
edge, and field) as specified by ASCE.
Additionally, the validity of the tributary
area in transferring tensile loads to membrane
fasteners in mechanically attached
roofs is well below the current design practice
of the roofing community. Field-measured
wind loads and their verification with
wind tunnel studies may prove useful for
evaluating existing wind load provisions
and may verify the validity of the existing
roof cladding wind uplift tests. These data
are currently being reviewed by the
SIGDERS members. Upon approval, this
will be presented at the 2013 RCI convention
(Dregger, 2011).
The presented research was being carried
out for a SIGDERS, which was formed
from a group of partners who were interested
in roofing design. These partners
included the following:
Atlas Roofing Corporation, Canadian
General-Tower Ltd., Canadian Roofing
Contractors’ Association, Carlisle SynTec
Incorporated, Dow Roofing Systems, Duro-
Last® Roofing, Inc., Firestone Building
Products Company, GAF-Elk Materials
Corporation, IKO Industries Ltd., Johns
Manville Inc., National Roofing Contractors
Association, OMG Roofing Products, Public
Works and Government Services Canada,
RCI, Inc., Sika Sarnafil, Soprema Canada
Inc., Tremco Inc., and Trufast Corporation.
Aversion of this paper was also presented at
the 13th International Wind Engineering
Conference held at Amsterdam, Netherlands
American Society of Civil Engineers
Standard ASCE 7-2010, Minimum
Design Loads for Buildings and
Other Structures, 2010.
ASTM D7158, 2011, Standard Test
Method forWind Resistance of Asphalt
Shingles. American Society of Testing
Materials, West Conshohocken, PA.
CSA A123.21-10, 2010, Standard Test
Method for the Dynamic Wind Uplift
Resistance of MembraneRoofing
Systems. IRC 891, Canadian Standards
Association (CSA), Toronto,
Ontario, Canada.
2 7 T H RC I I N T E R N A T I O N A L C O N V E N T I O N A N D T R A D E S H OW • MA R C H 1 5 2
0 , 2 0 1 2 B A S K A R A N E T A L . • 5 7
Figure 8 – Wind-induced response of roof mock-ups: membrane deflection.
A. Baskaran, B. Murty, D. Pravett,
2010, “Wind Effects on Commercial
Roofs – North American Advancements.”
13th IWEC Conference Proceedings,
Amsterdam, Netherlands.
P. Dregger, “SIGDERS Data Raise
Questions About the Edge Zones,”
Interface, December 2011, Page 29-
J.R. MacDonald,1991, “Full-Scale Testing
of Roofing Performance,” Proceedings
of the Third International
Symposium on Roofing Technology,
National Roofing Contractors Association
(NRCA), Gaithersburg, MD.
J.A. Peterka, J.E. Cermak, L.S. Cochran,
B.C. Cochran, N. Hosoya, R.G.
Derickson, C. Harper, J. Jones, B.
Metz, “Wind Uplift Model for Asphalt
Shingles,” Journal of Architectural
Engineering 3(4), 147-155, 1997.
NBCC, 2010, National Building Code of
Canada. National Research Council
of Canada, Ottawa, Canada.
5 8 • B A S K A R A N E T A L . 2 7 T H RC I I N T E R N A T I O N A L C O N V E N T I O N A N D T R A D E S H OW • MA R C H 1 5 2
0 , 2 0 1 2