Engineering Structures 103 (2015) 53–62

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Wind load factors for dynamically sensitive structures

with uncertainties
Dae Kun Kwon a,⇑, Ahsan Kareem a, Rachel Stansel b, Bruce R. Ellingwood c
a
Department of Civil & Environmental Engineering & Earth Sciences, Univ. of Notre Dame, Notre Dame, IN 46556, United States
b
ABS Consulting, San Antonio, TX 78232, United States
c
School of Civil and Environmental Engineering, Colorado State University, Ft. Collins, CO 80517, United States

a r t i c l e i n f o a b s t r a c t

Article history: The current recommendations for the load factors on wind load in ASCE Standard 7 are based on an anal-
Received 5 February 2015 ysis of performance of rigid buildings, which may not be adequate for dynamically sensitive structures. In
Revised 10 August 2015 light of the uncertainties associated with dynamic characteristics of flexible buildings such as natural fre-
Accepted 19 August 2015
quency and damping ratio, the load factors for such buildings may deviate from that in ASCE 7. This study
investigates the efficacy of the current wind load factor for dynamically sensitive structures in the pres-
ence of uncertainties. A systematic analysis is performed using Monte Carlo simulations. Uncertainties
Keywords:
associated with each component of the wind load effects such as the wind speeds, natural frequency
Buildings (codes)
Design (buildings)
and damping ratio of a building are incorporated in the load effects based on both ASCE Standard 7-05
Hurricanes and ASCE Standard 7-10. In addition, a database-enabled design (DED) procedure is utilized to support
Structural engineering the analysis of the wind load factor, especially for the acrosswind case where ASCE 7 does not offer
Uncertainty any guidance. In addition, the effects of terrain conditions, amplitude-dependent frequency and damping,
Wind engineering and negative aerodynamic damping on the wind load factor are also discussed. Recommendations are
Wind load factors made for wind load factors for dynamically sensitive structures both in the alongwind and acrosswind
directions and for non-hurricane and hurricane winds.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction
The wind load factor recommended in ASCE Standard 7 [1–3] for
load combinations involving wind load are based on an analysis of
performance of rigid buildings [17,18]. It accounts for deviations in
the actual loads from the nominal loads and for the uncertainties
associated with the load effects. To estimate this factor realisti-
cally, studies have been carried out to account for the effects of
both aleatory and epistemic uncertainties, including those in the
estimation of extreme wind speeds, wind load effect model, wind
tunnel experiment results, etc., which significantly influence the
wind effects on rigid buildings (e.g., [37,15,16]).
Wind effects on flexible buildings are more significant than
those on rigid buildings, and the dynamic response parameters,
especially frequency and damping, contribute additional uncer-
tainty to the response (e.g., [19,5,10]). In light of these uncertain-
ties and the fact that the building response is no longer
proportional to the square of the wind velocity, a wind load factor
⇑ Corresponding author.
E-mail addresses: dkwon@nd.edu (D.K. Kwon), kareem@nd.edu (A. Kareem),
rstansel@eagle.org (R. Stansel), bruce.ellingwood@colostate.edu (B.R. Ellingwood).
(or load factors) that are more conservative than the factor cur-
rently used in ASCE 7 may be required for tall building design.
The aforementioned studies considered a sample building to deter-
mine wind induced effects such as wind pressures/loads, base
moments and top displacements based on wind tunnel datasets.
Uncertainties associated with wind speeds, frequency and damp-
ing, and other parameters such as errors in wind tunnel experi-
ments and measured datasets were taken into account. Overall
trends from the studies have indicated an increase of the load
factor.
A number of issues remain concerning wind load factors for
flexible buildings. First, ASCE 7-10 [3] has introduced new wind
maps based on a mean recurrence interval (MRI) of 700 years for
basic wind speed for the Category II buildings [44,45], which
replaced the 50-year MRI wind speed map in ASCE 7-05 [2]. Along
with this change, the wind load factor was reduced from 1.6 to 1.0
for the load combinations in which the wind load is the principal
action. Most previous studies of tall, flexible buildings have
focused on the wind loads and wind load factors in ASCE Standard
7-05. These earlier studies must be revisited to explore how these
recent changes might affect the wind load factors and other wind
load requirements for flexible buildings in the current edition of

http://dx.doi.org/10.1016/j.engstruct.2015.08.031
0141-0296/Ó 2015 Elsevier Ltd. All rights reserved.
54 D.K. Kwon et al. / Engineering Structures 103 (2015) 53–62
ASCE 7. Second, although past studies (e.g., [19,5,10]) have con-
cluded that the load factor for flexible buildings should be higher
than that for rigid buildings, the results among these studies were
inconsistent. In addition, given that the effect of the uncertainties
on the acrosswind response of tall buildings would be different
from their alongwind response due to more significant aerody-
namic interactions and negligibly small contribution of the mean
response, the load factor in the acrosswind direction may also be
different from that in the alongwind direction. For example, Gabbai
et al. [19] reported that overall wind load factors for selected mem-
bers of an example tall building using synchronous pressure mea-
surement data ranged from 1.9 to 3.5 for the alongwind and
acrosswind top displacements based on the probabilistic peak
wind load effects used in rigid buildings [37,16]. However, these
load factors for the case of rigid building ranged from 1.9 to 2.3,
which depart from the expected load factor of 1.5–1.7 in Elling-
wood and Tekie [18] and 1.51 in Bashor and Kareem [5]. In addi-
tion, the load factors reported in Gabbai et al. [19] were very
close in the alongwind and acrosswind cases, while Bashor and
Kareem [5] reported that load factors for the acrosswind response
were larger than those for the alongwind response. This suggested
that the current load factor might not be adequate to account for the
acrosswind load effects. Overall, the load factors suggested by
Gabbai et al. [19] were considerably higher than those suggested
by others (e.g., [5]). Chen and Huang [10] also examined wind load
factors, quantifying the probabilistic wind load effects with the
assumption of parametric variations of wind speed and extreme load
coefficient and finding that the structural response was proportional
to the wind velocity to the power of 2–3 for flexible buildings. Their
load factor was close to that reported in Bashor and Kareem [5] when
the power was 2.5. Third, there has been little consideration of the
impact of different terrain conditions (exposures) on response of
dynamically sensitive buildings, which may be one of the key param-
eters for proper assessment of the wind load factor. Finally, there is
limited information for the load factor associated with hurricane
winds, which might be higher due to inherently higher uncertainties
in hurricane wind speeds, as shown in the literature for rigid build-
ings (e.g., [18,37,15,16]).
To address the above issues, this study investigates the applica-
bility of the current wind load factor for dynamically sensitive
structures in the presence of uncertainties. A systematic analysis
is performed using Monte Carlo simulations in which uncertainties
associated with each parameter involved in the wind load effects,
such as wind speeds, natural frequency and damping ratio, are
incorporated using both ASCE Standard 7-05 and ASCE Standard
7-10 procedures. In addition, a database-enabled design (DED) pro-
cedure [48,33,34] is also included to validate the assessment of the
wind load factor, especially for the acrosswind direction where
ASCE Standard 7 does not offer any guidance. Finally, recommenda-
tions are made for wind load factors for dynamically sensitive
structures loaded in both the alongwind and acrosswind directions
as well as in non-hurricane and hurricane wind regimes.
2. Wind load factor in ASCE Standard 7
2.1. Background of the wind load factor for load combinations
in ASCE 7
The wind load factor (cw) in ASCE Standard 7 is defined using the
first-order reliability method (FORM) as [17,18]:


lW
cw ¼ ð1 þ abV W Þ ð1Þ
Wn
where lW/Wn = the bias or ratio between the mean and the nominal
wind loads/pressures, where the nominal value (Wn) is determined
in accordance with the ASCE 7 criteria; b = the reliability index;
a = the sensitivity coefficient; and VW = the coefficient of variation
(COV) in the wind load/pressure. Using the values for a rigid build-
ing reported by Ellingwood and Tekie [18], i.e., lW/Wn = 0.78,
b = 2.5, a = 0.75, VW = 0.37, the load factor defined in Eq. (1) results
in cw = 1.32, which is equivalent to the load factor (1.3) used in ASCE
7-95. The mean wind load/pressure in that study implicitly included
a wind directionality factor of 0.85. When an explicit wind direc-
tionality coefficient, Kd = 0.85 was added to the ASCE 7 provisions
for rectangular buildings in 1998, the corresponding wind load fac-
tor (=1.3/0.85), excluding the directionality factor, was increased to
1.6. When Ellingwood and Tekie [18] later revisited the choice of
probability distribution for modeling the extreme wind speed and
the difference between wind speed models in hurricane zones ver-
sus non-hurricane zones, they concluded that the reliability index,
b, for wind loads should be approximately 3.0. A single wind load
factor of 1.6 was employed for editions of ASCE 7 between 1998
and 2005.
In ASCE Standard 7-10, the wind load factor in combinations 4
and 6 was reduced to 1.0 because of the change in the specification
of the design wind speed. The wind speeds for strength design
were re-mapped at much longer MRIs, which are 700–1700 years
depending on the Occupancy Category (Figure 26.5 in the ASCE
7-10), thus eliminating the need for an importance factor for differ-
ent building risk categories and the discontinuity in the risk
between the hurricane-prone coastal areas and the remainder of
the country, and better aligning the treatment of wind and earth-
quake effects (C2.3.2 in ASCE 7-10).
2.2. Relationship between the wind load factor and MRI of wind speed
in ASCE 7
The wind pressures/loads in ASCE Standard 7 are calculated
from the 3-s gust wind speed, V3-s. The ratio of this wind speed
for any MRI, T, (VT) to the 50-yr MRI wind speed (V50) in non-
hurricane prone regions is (e.g., [38]):
VT
¼ 0:36 þ 0:1 lnð12TÞ ð2Þ
V 50
When the wind speed maps in ASCE 7-2010 were developed, the
wind load factor (cw) was defined as the ratio between the point
estimates of the T-yr and 50-yr wind speeds as:

n
VT
cw ¼ ð3Þ
V 50
where n equals 2, which is based on the behavior of rigid buildings
in the absence of any dynamic amplification. Eq. (3) implies that all
uncertainty in the wind load is vested in the uncertainty in the wind
speed. Using the load factor of 1.6 defined in ASCE 7-05, Eqs. 2 and 3
yield T = 709 years for Risk Category II structures (where the impor-
tance factor (I) is 1.0). Accordingly, T was set equal to 700 years in in
developing the wind speed maps in ASCE 7-10 and cw was set equal
to 1.0. For example, V50 defined in ASCE 7-05 was 40 m/s in the Mid-
west region; using T = 700-yr MRI and V50 = 40 m/s in Eq. (2) for the
Risk Category II structures (I = 1.0), V700 becomes 51 m/s, which is
the new basic wind speed defined in ASCE 7-10.
For rigid buildings such as low-rise structures, the wind loads
are proportional to the square of the wind speed because the wind
responses are governed by mean (static) and background turbu-
lence effects (quasi-static). However, in the case of flexible build-
ings, n may exceed 2 because the response is dominated by
inertial effects (e.g., [10]). For example, assuming that n = 2.5 and
T = 700-yr, V700 = 51 m/s; with V50 = 40 m/s, as before, the implied
wind load factor based on Eq. (3) would increase from 1.6 to 1.84
for flexible buildings. Such increases will be considered in more
detail in the sequel.
 D.K. Kwon et al. / Engineering Structures 103 (2015) 53–62
3. Evaluation of wind load factor
3.1. Analysis methods
In order to determine appropriate wind load factors for dynam-
ically sensitive buildings, extensive Monte Carlo simulations were
performed for over 100 separate cases, in which the effects of rigid
versus flexible buildings and uncertainties associated with wind
speeds, frequency, damping, and other parameters such as pres-
sure/force coefficients and wind parameters (e.g., wind profile
and turbulence coefficients) were investigated. Two structural
responses – base moments and building top displacements – were
selected to assess the wind load factor. In an earlier study by
Ellingwood and Tekie [18], the load factor was derived based on
wind load/pressure of low-rise buildings in which displacement
response is negligible. However, for flexible buildings both the base
moments and top displacements have their respective significance.
Two different methods were employed to estimate the wind
load factors. The first method is the ASCE 7 analysis procedure
for rigid and flexible buildings, i.e., gust loading/effect factor
approach (e.g., [40,41]). Provisions for the wind load effects on
structures in most codes and standards, including ASCE 7, gener-
ally focus primarily on the alongwind (buffeting) response. How-
ever, this method fails to capture the loading mechanisms in the
acrosswind and torsional directions. Given the fact that the across-
wind response is generally dominant for flexible buildings, a
database-enabled design (DED) procedure, which is a promising
design procedure for assessing wind-induced response in a more
reliable way than is possible with conventional codes and stan-
dards, is employed as a alternative approach to load factor deter-
mination. The DED procedure used in this study employs a base
moment-based 3-D gust loading factor approach [31] drawing
from a high-frequency base balance (HFBB) database. This
approach has been adopted in cyberbased on-line DED modules
such as the NatHaz Aerodynamic Loads Database (NALD)
[48,33,35] and has been introduced in the Commentaries of both
ASCE 7-05 and 7-10. A more recent expanded e-module for high-
rise buildings, DEDM-HR has a unique capability, of hosting multi-
ple databases [34]. For reference, both e-modules are accessible via
an engineering virtual organization, VORTEX-Winds (https://
vortex-winds.org) [30].
Two HFBB databases are utilized in the DED procedure: the first
is from the NatHaz Modeling Laboratory at the University of Notre
Dame, USA, which was the basic HFBB database used in the NALD
[48,33], and the second is from the Wind Engineering Research
Center (WERC) at Tamkang University (TKU), Taiwan [12]. Both
databases are housed in e-module, DEDM-HR [34]. Both NatHaz
Table 1
Uncertain parameters used in ASCE 7 procedure.
Parameter Description
b Errors in the estimation of extreme wind speeds
c Wind speed conversion for different averaging time
q Observation error in wind speed
f Fundamental natural frequency
f Damping ratio
Cpw, Cpl Windward and leeward external pressure
coefficients, respectively
q Air density
K Mode shape
qB Building density
a; zg ; a^ ; b^ Wind parameters pertaining to wind profile,
 c; d; ‘; 
a ; b; e turbulence intensity, etc.*
*
Symbols used for wind parameters are the same with those defined in the ASCE 7, thus detailed explanation for each symbol is omitted here for brevity.
55
and TKU databases have two common terrain conditions (Expo-
sures A and C), but the TKU database also has a greater subdivision
in the datasets with an additional terrain condition (Exposure B).
We have included Exposure A in this paper, despite the fact that
ASCE 7 no longer includes this exposure, in order to examine the
effect of terrain exposure on flexible building response in later
sections. The databases consist of non-dimensional base moment/
torque power spectral density [CM(f)] and RMS base moment/
torque coefficients (rCM) in alongwind, acrosswind and torsional
directions in each dataset, which are the key parameters for esti-
mating wind responses using a 3-D gust loading factor approach.
For completeness, a brief summary of the DED procedure is dis-
cussed in the Appendix A. More detailed descriptions of the analy-
sis scheme, the datasets, wind tunnel test conditions, and validity
of the HFBB datasets can be found in Kareem and Zhou [31], Zhou
et al. [48], and Cheng and Wang [12], Kwon et al. [33], and Kwon
and Kareem [34].
3.2. Load factor definitions and uncertain parameters
The test cases for the wind load factor analysis are classified in
terms of: rigid or flexible buildings; uncertainties in wind speed;
uncertainties in fundamental natural frequency and damping ratio
of buildings; all uncertainties. Tables 1 and 2 describe the uncer-
tain parameters for the ASCE 7 and DED procedures, respectively,
based on a comprehensive review of the literature. The parameters
are used to estimate the base moment and top displacement of the
case study buildings. Note that the uncertain parameters b, c and q
are associated with the wind speed, which includes errors in the
estimation of extreme wind speeds such as a proper choice of
probability distribution of extreme winds, sampling errors, wind
speed conversion errors for different averaging time (e.g., 3-s to
1-h for wind response estimation), and observation error in wind
speed, respectively. In addition, note that the epistemic uncertain-
ties that arise from the use of HFBB aerodynamic loads, such as
that used in the DED procedure (Appendix A), may become impor-
tant when the fundamental mode shapes depart from the ideal
uncoupled linear form [7]. Although Zhou et al. [47] reported that
the influence of non-ideal mode shapes is negligible for base bend-
ing moments (and the DED procedure used in this study), such
uncertainties may also affect the load factors to some degree.
Two load factor definitions are utilized in this study:
(1) Load Factor 1 (LF 1): This load factor, defined in Eq. (1),
was developed from traditional First-Order reliability
analysis. The sensitivity coefficient, a = 0.8, reflects the
PDF COV Reference
Normal 0.075 Minciarelli et al. [37], Coles and Simiu [13], Diniz et al. [15], Gabbai
et al. [19]
Normal 0.05 Minciarelli et al. [37]
Normal 0.025 Minciarelli et al. [37]
Lognormal 0.05 Kareem [26], Kareem [27]; Bashor et al. [6]; Gabbai et al. [19]
Lognormal 0.4 Davenport and Hill-Carroll [14], Kareem [26], Kareem [27]; Bashor
et al. [6]; Gabbai et al. [19]
Normal 0.1 Bashor et al. [6]
Normal 0.05 Bashor et al. [6]
Normal 0.05 Bashor et al. [6]
Normal 0.05 Bashor et al. [6]
Normal 0.05 Minciarelli et al. [37]; Bashor et al. [6]
56 D.K. Kwon et al. / Engineering Structures 103 (2015) 53–62

Table 2
Uncertain parameters used in DED procedure.

Parameter Description PDF COV Reference

b Errors in the estimation of extreme wind speeds Normal 0.075 Minciarelli et al. [37], Coles and Simiu [13], Diniz et al. [15], Gabbai et al. [19]
c Wind speed conversion for different averaging Normal 0.05 Minciarelli et al. [37]
time
q Observation error in wind speed Normal 0.05 Minciarelli et al. [37]
f Fundamental natural frequency Lognormal 0.05 Kareem [26]; Bashor et al. [6]; Gabbai et al. [19]
f Damping ratio Lognormal 0.40 Davenport and Hill-Carroll [14], Kareem [26]; Bashor et al. [6]; Gabbai et al.
[19]
e1, e2  and a
Uncertainty of b * Normal 0.05 Bashor et al. [6]
e3 Uncertainty of CM(fr) ** Normal 0.15 Bashor et al. [6]
e4 Uncertainty of rCM** Normal 0.25 Bashor et al. [6]
q Air density Normal 0.05 Bashor et al. [6]
K Mode shape Normal 0.05 Bashor et al. [6]
qB Building density Normal 0.05 Bashor et al. [6]
CD Drag coefficient Lognormal 0.10 Kareem [26]; Bashor et al. [6]
*
Power law coefficients in the wind profile definition.
**
CM(fr) = Non-dimensional power spectral density; rCM = Non-dimensional RMS base moment coefficient.

skew-positive nature of the Type I extreme value distribu-
tion used to model the wind response. The reliability index,
b = 3.0, is used [18].
(2) Load Factor 2 (LF 2): Generally when a variable is a function
of a product of other independent random variables, the
resulting distribution is similar to a lognormal distribution.
Accordingly, the second load factor definition is introduced
here which assumes that the resulting wind response has a
Lognormal distribution:
 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

 exp be‘ lnð1 þ V 2W Þ
lW
cw;LF2 ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Wn
1 þ V 2W
where e‘ ¼ 0:72 (e.g., [36]) and the other variables are the same as
defined in LF1.
In both LF1 and LF2 definitions, the nominal wind response
(Wn), either the base moment or displacement, is determined by
ASCE 7 or the DED procedure assuming that there are no uncertain-
ties involved, while lW and VW are determined by Monte Carlo
simulation (100,000 samples) taking uncertainties in the parame-
ters into account. Note that the effect of wind directionality is
excluded in this study, as in Ellingwood and Tekie [18], in which
they recommended a separation between the wind directionality
effects and the load factor.
3.3. Example buildings
Two example buildings were utilized in the assessment of the
wind load factor: (a) Building 1, the CAARC (Commonwealth Aero-
nautical Advisory Research Council) standard tall building (e.g.,
[21]): width B = 46 m; depth D = 30 m; height H = 183 m; funda-
mental natural frequencies in alongwind and acrosswind direc-
tions, fx = 0.17 Hz and fy = 0.177 Hz, respectively; linear mode
shape; building bulk density qB = 157 kg/m3; damping ratio
f = 0.01; air density qA = 1.22 kg/m3. The building is assumed to
be located in an urban environment within the continental US,
which corresponds to Exposure A defined in ASCE 7-98 [1]; b)
Building 2: B = D = 40 m (square cross-section); H = 200 m; funda-
mental natural frequencies in alongwind and acrosswind direc-
tions, fx = fy = 0.2 Hz, respectively; linear mode shape; building
bulk density qB = 245 kg/m3; damping ratio f = 0.02; air density
qA = 1.22 kg/m3. This building is located in a suburban area (Expo-
sure B in ASCE 7-10).
Various magnitudes of 3-s gust wind speeds were utilized to
compare the 2005 and 2010 versions of ASCE Standard 7 and to
consider non-hurricane and hurricane wind climatologies. This
ð4Þ
provides overall trends in the wind load factors associated with
the recent changes in the ASCE 7 wind load provisions, especially
for the different basic wind speeds.
4. Wind load factors for flexible building response
4.1. Load factors using basic wind speed defined in ASCE -05 – non-
hurricane winds
The basic wind speed,V3-s, was assumed to be 40 m/s for Build-
ing 1 and 63 m/s for Building 2 for the estimation of alongwind
load factors. No special consideration for hurricane winds was con-
sidered in this case study; the hurricane case will be examined sep-
arately in a later section. The DED procedure utilized the TKU
database for Building 2 because the NatHaz database does not con-
tain datasets for Exposure B. A comparison of the alongwind
responses computed from the ASCE 7 and DED procedures, e.g.,
the base moments, top displacements and top root-mean-square
(RMS) accelerations, is presented in Table 3. Some differences
can be seen, with the DED results being slightly higher than those
given by ASCE 7. This may be attributed to the building-specific
aerodynamic load introduced in DED, whereas ASCE 7 is based
on a simplified procedure, i.e., gust loading/effect factor. In view

Table 3
Comparison of alongwind responses based on ASCE 7-05 and DED procedures (V3-s = 40 m/s).

Procedure* Base moment (106 kN-m) Top displacement (m) Top RMS acceleration (milli-g)
Building 1 ASCE 7 1243 0.39 4.46
DED  NatHaz 1241 0.45 5.40

Building 2 ASCE 7 4257 0.46 4.25

DED  TKU 4711 0.57 5.94
⁄
ASCE 7 = ASCE 7 procedure; DED  NatHaz = DED procedure using NatHaz database; DED  TKU = DED procedure using TKU database.
 D.K. Kwon et al. / Engineering Structures 103 (2015) 53–62 57

Table 4
Alongwind load factors for example buildings 1 and 2 based on ASCE 7-05 basic wind speed for non-hurricane case using ASCE 7 and DED procedures.

Uncertainty cases* Example building and procedure** Base moment Displacement

LF1 LF2 LF1 LF2
Rigid None B1, ASCE 7 1.00 1.00 1.00 1.00
B1, DED–NatHaz 1.00 1.00 1.00 1.00
B2, ASCE 7 1.00 1.00 1.00 1.00
B2, DED–TKU 1.00 1.00 1.00 1.00
Flexible None B1, ASCE 7 1.00 1.00 1.00 1.00
B1, DED–NatHaz 1.00 1.00 1.00 1.00
B2, ASCE 7 1.00 1.00 1.00 1.00
B2, DED–TKU 1.00 1.00 1.00 1.00
Rigid bcq B1, ASCE 7 1.49 1.52 1.49 1.52
B1, DED–NatHaz 1.49 1.49 1.49 1.49
B2, ASCE 7 1.49 1.52 1.49 1.52
B2, DED–TKU 1.49 1.52 1.49 1.52
Flexible bcq B1, ASCE 7 1.61 1.67 1.61 1.66
B1, DED–NatHaz 1.59 1.63 1.59 1.63
B2, ASCE 7 1.59 1.64 1.59 1.64
B2, DED–TKU 1.56 1.60 1.56 1.60
Flexible bcq, f B1, ASCE 7 1.61 1.67 1.71 1.79
B1, DED–NatHaz 1.59 1.64 1.69 1.76
B2, ASCE 7 1.59 1.64 1.69 1.76
B2, DED–TKU 1.56 1.60 1.65 1.71
Flexible bcq, f & f B1, ASCE 7 1.71 1.78 1.81 1.91
B1, DED–NatHaz 1.68 1.74 1.77 1.86
B2, ASCE 7 1.67 1.73 1.76 1.86
B2, DED–TKU 1.66 1.72 1.75 1.83
Flexible All B1, ASCE 7 1.74 1.83 1.93 2.08
B1, DED–NatHaz 1.98 2.14 2.08 2.28
B2, ASCE 7 1.71 1.78 1.89 2.01
B2, DED–TKU 1.93 2.08 2.02 2.20
*
Rigid = rigid building; Flexible = flexible building; None = no uncertainties; bcq = wind speed uncertainties; f and f = fundamental natural frequency and damping ratio
uncertainties, respectively; All = all uncertainties (e.g., pressure coefficients, etc.) were considered. See Tables 1 and 2.
**
B1 and B2 = building examples 1 and 2, respectively.

of the inherent differences between the two procedures, the build-
ing responses agree reasonably well. Similar conclusions have been
drawn in other studies [48,33,34].
Table 4 shows the load factors for both example buildings using
the ASCE and DED procedures for the alongwind response. Note
that the response for rigid buildings was obtained by excluding
the resonant response component, i.e., including only mean and
background components. For the cases in which no uncertainties
were involved (described as None in the table), the load factors
both for the rigid and flexible buildings reduced to unity (1.0). This
indicates that the bias (lW/Wn in Eqs. 1 or 4) between the mean
and the nominal wind loads/displacement was unity. The load fac-
tors for the two example buildings using ASCE 7 and DED showed
very good agreement despite different wind characteristics (e.g.,
basic wind speeds, exposures) and building properties except
when all uncertainties are involved (Flexible All in Table 4), the lat-
ter due to the level of uncertainties (i.e., COV) assumed in the non-
dimensional base moment spectrum [CM(fr)] and RMS base
moment coefficients (rCM) in Table 2. The displacement-based load
factors were slightly higher than the base moment-based load fac-
tors especially when frequency and damping uncertainties were
involved. This is attributed to the involvement of frequency and
damping multiple times in displacement calculations. For example,
the displacement calculation procedure in ASCE 7 is explicitly
expressed as a function of frequency and gust effect factor (G)
and the latter is also associated with frequency and damping
parameters.
Overall, the wind load factors LF2 were slightly higher than LF1,
but they generally showed good agreement with less than 5% dif-
ference between them. With regard to rigid versus flexible build-
ings, there was an increase in the load factors for the flexible
building, especially when uncertainties associated with wind
speed, frequency and damping (Flexible bcq, f & f case) were intro-
duced. The effects of frequency uncertainty (Flexible bcq, f) for the
base moment-based load factors turned out to be very small as
compared to the effect of wind speed uncertainty (Flexible bcq);
however, such effects were more apparent in the displacement-
based load factors. Given the fact that there exists large uncer-
tainty associated with damping (COV = 0.40 in Tables 1 and 2), it
is not surprising that damping effects are dominant in the esti-
mated load factors for the Flexible bcq, f & f case. In addition, the
scatter between the Flexible bcq, f & f and all uncertainties involv-
ing wind profile, turbulence intensity, external pressure coeffi-
cients parameters, etc. using the ASCE 7 procedure tends to be
negligible. This shows that uncertainties associated with wind
speed, and frequency and damping (Flexible bcq, f & f case) are
the key factors that contribute most to uncertainty in flexible
building response. Additional simulations [4], not reported here
for brevity, include an uncertainty analysis with wind-related
parameters such as wind speed, wind profile, and turbulence
intensity parameters. The results were virtually the same when
compared to the case involving only wind speed uncertainty
parameters (Flexible bcq).
For the acrosswind case, only the DED procedure could be uti-
lized to assess wind load factors because there are no comparable
provisions in ASCE 7 for this case. The basic wind speed was
assumed to be 40 m/s both for Building 1 and 2. The results for
the acrosswind analysis are given in Table 5. The overall trends
are similar to those for the alongwind case given in Table 4: the
load factors for flexible buildings are notably affected by the wind
speed, frequency and damping and the displacement-based factors
are larger than the base moment-based factors. It is observed that
58 D.K. Kwon et al. / Engineering Structures 103 (2015) 53–62

Table 5
Acrosswind load factors for building 1 and 2 examples based on ASCE 7-05 basic wind speed for non-hurricane case using DED procedure (V3-s = 40 m/s in all cases).

Uncertainty cases Example building, procedure and exposure Base moment Displacement
LF1 LF2 LF1 LF2
Rigid and flexible None All cases 1.00 1.00 1.00 1.00
Rigid bcq B1, DED–NatHaz, Exp. A 1.49 1.52 1.49 1.52
B2, DED–TKU, Exp. B 1.49 1.52 1.49 1.52
B2, DED–NatHaz, Exp. A 1.49 1.52 1.49 1.52
Flexible bcq B1, DED–NatHaz, Exp. A 1.78 1.89 1.78 1.89
B2, DED–TKU, Exp. B 1.70 1.78 1.70 1.78
B2, DED–NatHaz, Exp. A 1.79 1.93 1.79 1.93
Flexible bcq, f B1, DED–NatHaz, Exp. A 1.81 1.94 1.93 2.10
B2, DED–TKU, Exp. B 1.72 1.80 1.82 1.94
B2, DED–NatHaz, Exp. A 1.81 1.95 1.93 2.11
Flexible bcq, f & f B1, DED–NatHaz, Exp. A 2.04 2.23 2.15 2.40
B2, DED–TKU, Exp. B 1.93 2.07 2.04 2.21
B2, DED–NatHaz, Exp. A 2.00 2.19 2.11 2.35
Flexible All B1, DED–NatHaz, Exp. A 2.63 3.13 2.75 3.32
B2, DED–TKU, Exp. B 2.37 2.71 2.48 2.88
B2, DED–NatHaz, Exp. A 2.57 3.06 2.69 3.23

unlike the alongwind case, discrepancies between Buildings 1 and
2 are more notable for the flexible cases. However, such discrepan-
cies are less significant when the acrosswind load factors for Build-
ing 2 are estimated in Exposure A, included for comparison in
Table 5 (‘‘B2, DED–NatHaz, Exp. A”). This trend is due to the effect
of different exposures for each building example, i.e., Exposure A
for Building 1 vs. Exposure B for the Building 2, which will be
examined further in the next section. The load factors for the
acrosswind response are larger than those for the alongwind
response for flexible buildings, which indicates that the uncer-
tainty in the acrosswind response is higher than that in the along-
wind response. This trend will also be investigated in the next
section.
In summary, for non-hurricane wind speeds in ASCE 7-05 in the
alongwind direction, uncertainties in response of flexible buildings
(‘‘Flexible All” in the Table 4) yield wind load factors of about 1.9
and 2.1 for the base moment and displacement using the ASCE 7
procedure, respectively, while the factors using the DED procedure
result in slightly higher values, about 2.1 (base moment) and 2.3
(displacement), due to the aerodynamic uncertainty. The factors
are much higher in the acrosswind case and depend on the expo-
sure conditions as shown in Table 5. Based on these observations,
wind load factors in ASCE 7-05 for flexible buildings should be
higher than the factor 1.6 used for rigid buildings if the same level
of reliability is to be achieved in both cases; moreover, the along-
wind wind load factors are not adequate if the acrosswind load
effects are predominant.
4.2. Load factors using basic wind speed defined in ASCE 7-10 – non-
hurricane winds
ASCE Standard 7-10 reduced the wind load factor from 1.6 to 1.0
when wind is treated as the principal action in the load combina-
tion as a result of increasing the MRI for nominal wind speed to
700 years from the 50-year MRI used previously. However, any
adjustments to the load factor to account for flexible or across-
wind response have yet to be considered. The load factors
determined in the previous section were relatively insensitive to
different building properties including frequency and damping
values and to different load factor definitions (LF1 and LF2).
Accordingly, only Building 1 and LF2 were utilized to evaluate
the wind load factors for ASCE 7-10 cases. Three exposures, defined
in ASCE 7 as Exposures A, B and C, were employed to investigate
the effects of exposures both in the alongwind and acrosswind
directions. The nominal wind effect (Wn) was determined using
both ASCE 7-10 and DED procedures with the new basic wind
speed for non-hurricane winds, assuming that building 1 was
located in the Midwest region. The 3-s gust speed increased from
40 m/s (ASCE 7-05) to 51 m/s using the new ASCE 7-10 wind
map. The method for establishing the new wind map in ASCE 7-
10 implies that the uncertainties (e.g., wind speeds) considered
in ASCE 7-05 (50-yr MRI) are implicitly encompassed in ASCE 7-
10 (700-yr MRI). For this reason, uncertainties associated with fre-
quency and damping were examined only to compare wind load
factors for the flexible building case. Fig. 1 shows probability of
exceedance curves for alongwind base moments and displace-
ments of the building 1 example using both ASCE 7-05 and 7-10
winds. It is clearly seen that at the same level of exceedance prob-
ability, the effects of two different return periods (50-yr in ASCE 7-
05 and 700-year in ASCE 7-10) for both rigid and flexible cases are
significant.
The alongwind and acrosswind load factors using ASCE 7-10
basic wind speeds under three exposures (Exposure A, B and C)
are summarized in Table 6(a); the values in parentheses were
obtained through the DED procedure. The load factors for the
alongwind direction obtained by the ASCE 7 and DED procedures
for the same exposure agree reasonably well. The discrepancies
among exposures become more noticeable when the terrain
roughness becomes smoother, i.e., moving toward Exposure C.
For the alongwind case, the increase in the load factors from Expo-
sure A to C for both base moment and displacement were marginal,
while the increases were more apparent in the acrosswind cases
which are more sensitive to inflow conditions.
Terrain effects tend to be different for load factors on alongwind
and acrosswind responses. When the terrain roughness becomes
smoother, the alongwind load factor decreases, while the across-
wind load factor increases. An examination of the wind load spec-
tra sheds further light on this observation. Fig. 2 shows non-
dimensional base moment spectra for Building 1 under different
exposures, which represent typical alongwind and acrosswind
moment spectra for tall buildings. The vertical lines shown in
Fig. 2 denote corresponding reduced frequencies (fr1 = f1B/UH, f1
= natural frequency of building) for different exposures, where fr1
is inversely proportional to the mean hourly wind speed at the
top of the building (UH). The reduced frequencies decrease when
the terrain becomes smoother due to an increase in the mean
hourly wind speed at the top of the building. The gust effect factor
in the ASCE 7 procedure and 3-D gust loading factor in DED
 D.K. Kwon et al. / Engineering Structures 103 (2015) 53–62 59

cies associated with different exposures for the given building

Fig. 1. Probability of exceedance curves of the alongwind response by ASCE 7-05
(50-yr wind) and 7-10 (700-yr wind) procedures (non-hurricane winds): (a) base
moment and (b) displacement.
procedure are key parameters for estimating response of flexible
structures to wind; both utilize a non-dimensional spectral value
corresponding to a reduced frequency (i.e., a point value). Consid-
ering the features of the alongwind spectra and reduced frequen-
example, uncertainty associated with wind speed results in varia-
tions in the wind spectral value of interest corresponding to the
reduced frequency that are dependent on building exposure.
Accordingly, Building 1 in Exposure A experiences higher uncer-
tainty in the response than in Exposure C because the spectrum
is more variable at the relevant reduced frequency in that Expo-
sure. This explains the trends in alongwind load factors for differ-
ent exposure conditions observed in Table 6a since the load factors
defined here (Eq. (4)) depend on the COV of the building response.
The effect is the opposite in the acrosswind direction. A typical
acrosswind spectrum exhibits a sharp peak around the Strouhal
number due to vortex shedding that strongly depends on the
intensity of the incident turbulence. Accordingly, the building in
Exposure C has a sharper (or narrow-band) peak around the Strou-
hal number than the same building in Exposure A because the vor-
tex shedding is vitiated by the presence of high levels of turbulence
intensity in the incident flow (e.g., [25]). Unlike the alongwind
spectra, the slope of the acrosswind spectra increases with the
decrease of the frequency, i.e., as the Strouhal number is
approached, and this trend is more predominant in smoother ter-
rain due to the influence of turbulences on vortex-shedding. Since
the slope changes around the reduced frequencies of interest are
more rapid in the acrosswind spectrum, the acrosswind load fac-
tors for different exposures exhibit larger scatter than the along-
wind factors.
Table 6(a) suggests that wind load factors for flexible buildings
in the alongwind case should be 1.3 (base moment) and 1.5 (dis-
placement) in Exposure B to achieve the same reliability as in rigid
buildings. For the acrosswind case, the load factors of about 1.6
(base moment) and 1.8 (displacement) in Exposure B are reason-
able but this factor may further increase if the tall building is
located in a smoother terrain, e.g., Exposure C or D, as noted in
Table 6a.
The current analysis is based on the assumption that the build-
ing dynamic characteristics remain constant under the design
loading condition. Recent full-scale studies reveal changes in both
frequency and damping are possible, even at below-design wind
speeds; damping in particular may exceed nominal values used
in design when buildings experience extreme excursions under
design loads or local inelastic action within the structural system
(e.g., [32,5]). Full-scale studies have revealed that with increasing
amplitudes of motion, natural frequency decreases and damping
ratio increases (e.g., [43,22,28,29,39,42]). The increase in wind
speeds in ASCE 7-10 (700-yr MRI) for flexible buildings would
result in higher building response than calculated using the 50-
year winds in ASCE 7-05 because the gust factor is velocity-

Table 6
Alongwind and acrosswind load factors based on ASCE 7–10 basic wind speed for non-hurricane and hurricane cases using ASCE 7–10 and DED procedures.

Uncertainty cases ASCE 7 procedure and exposure (DED procedure)* Alongwind Acrosswind
Base moment Displacement Base moment Displacement
(a) Non-Hurricane case (V3-s = 51 m/s)
Flexible f & f Exp. A (DED–NatHaz) 1.31 (1.34) 1.49 (1.55) (1.57) (1.75)
Exp. B (DED–TKU) 1.30 (1.33) 1.48 (1.48) (1.60) (1.77)
Exp. C (DED–NatHaz) 1.27 (1.32) 1.45 (1.44) (1.69) (1.94)
(b) Hurricane case (V3-s = 72 m/s) in Exposure B
Flexible f & f COV = 0** ASCE 7 1.36 1.53 – –
DED–TKU 1.36 1.49 2.20 2.67
Flexible f & f COV = 0.05 ASCE 7 1.53 1.68 – –
DED–TKU 1.49 1.61 2.83 3.20
Flexible f & f COV = 0.1 ASCE 7 1.91 2.03 – –
DED–TKU 1.82 1.92 3.68 3.91
*
Values in the parentheses of the non-hurricane case represent the results from DED procedure using NatHaz or TKU database.
**
COV value indicates additional uncertainty with regard to hurricane wind speeds.
60 D.K. Kwon et al. / Engineering Structures 103 (2015) 53–62
Fig. 2. Non-dimensional base moment spectra for building 1 example: (a)
alongwind and (b) acrosswind – values in the parentheses represent 3-s gust
speed (V3-s).
dependent. For an independent change either in frequency or
damping, it is obvious that a decrease in natural frequency would
result in an increase in the response; similarly an increase in the
damping ratio would lead to a reduction in the response. Correla-
tion between changes in frequency and damping remains to be
established.
4.3. Load factors using basic wind speed defined in ASCE 7-10 –
hurricane case
Although the new hurricane model adopted in ASCE 7-10 is an
improvement over previous editions of the standard, estimates of
hurricane wind speeds inevitably have higher uncertainty than
non-hurricane wind speeds due to errors in such factors as wind
speed conversion for different averaging times, observation errors,
and hurricane wind field modeling especially when much higher
MRI wind speeds (700-yr) are considered. For example, Coles and
Simiu [13] noted an increase in the sampling uncertainty of hurri-
cane wind speeds for higher MRI, which was due to limitations in
the data from which the hurricane wind model used in the simula-
tion was inferred and led to increase in wind speed uncertainty
when the model was extrapolated to long return periods [13].
Vickery et al. [44] investigated uncertainties in the new hurricane
model and reported that the uncertainty in the estimated 100-yr
MRI wind speed was about 6% along the Gulf of Mexico coastline,
increasing to about 16% along the coast of Maine. Considering
these observations, three different levels of uncertainty (COVs)
regarding hurricane winds such as 0, 0.05 and 0.1 were assumed
to examine the wind load factors.
Wind load factors using Building 1 with a hurricane wind speed
of V3-s = 72 m/s are shown in Table 6(b) for Exposure B. The
assumption of Exposure B is based on the following premises: (a)
the load factor in Exposure B is likely to be an average value of
all three exposures (Exposure A, B and C) as observed in the non-
hurricane case (Table 6a); (b) practically all tall buildings in the
United States have been constructed in urban/suburban areas
rather than open terrains, with the exception of coastal develop-
ments. The DED procedure produces reasonably good agreement
with the ASCE 7 procedure, a trend noted earlier in Table 6(a).
The acrosswind load factors are much larger than those in the
non-hurricane case (Table 6) and also are larger than the along-
wind load factors in the hurricane case because the reduced fre-
quencies corresponding to hurricane wind speed (V3-s = 72 m/s)
tend to be closer to the Strouhal number (see Fig. 2b). However,
such a high wind speed in tall buildings may introduce negative
aerodynamic damping, which may lead to a reduction in total
damping, especially for the acrosswind response
[23,24,8,46,11,20,9]. It is anticipated that aerodynamic damping
in the alongwind case may be negligibly small for the building
example considered herein; however, for the acrosswind case, it
may reduce the total damping. Thus, the effect of amplitude-
dependency may be partially offset by aerodynamic damping for
the acrosswind case.
Assuming that the expected level of uncertainty regarding hur-
ricane winds is 0.05 (‘‘COV = 0.05” in Table 6b), the alongwind load
factors (consistent with ASCED 7-10) for flexible buildings should
be approximately 1.5 (based moment) and 1.7 (displacement) to
achieve the same reliability as rigid buildings. For the acrosswind
case, load factors of about 2.8 (base moment) and 3.2 (displace-
ment) may be reasonable values. Caution should be exercised in
the case of super-tall buildings in different terrain conditions,
especially for the acrosswind case under Exposures C or D, where
the wind load factors may be impacted by the presence of negative
aerodynamic damping due to low turbulence intensity and high
wind speed conditions (high reduced velocity).
5. Concluding remarks
This study presented an assessment of the alongwind and
acrosswind wind load factors in ASCE Standard 7 for LRFD of
dynamically sensitive structures, in which uncertainties associated
with wind speeds, natural frequency and damping ratio of build-
ings are taken into account. A systematic analysis was performed
using extensive Monte Carlo simulations in which uncertainties
in the wind load effects were incorporated based on ASCE 7-05
and 7-10 procedures as well as a database-enabled design (DED)
procedure using HFBB measurements.
Overall, the analysis suggests that the wind load factor defined
in ASCE 7 based on rigid buildings is not adequate for flexible
buildings for either alongwind or acrosswind directions due to
the sensitivity of flexible buildings to dynamic load effects and
associated uncertainties related to frequency and damping. In
addition, the effects of terrain conditions on the wind load factor
for flexible buildings are significant for the acrosswind case. The
displacement-based load factors were generally higher than the
base moment-based factors, especially when frequency and damp-
ing uncertainties were considered. The load factors for hurricane
winds are larger than for non-hurricane winds, and increase when
uncertainty associated with hurricane wind speed estimation is
taken into account. Furthermore, the acrosswind load factors were
 D.K. Kwon et al. / Engineering Structures 103 (2015) 53–62
generally higher than the alongwind load factors. The acrosswind
response of tall buildings results from complex aerodynamic inter-
actions, whereas the alongwind response follows the buffeting
action of gusts; the uncertainties associated with acrosswind aero-
dynamics (e.g., wind tunnel test conditions) are more pronounced.
In addition, the contribution of the mean response is considerable
in the alongwind response, which tends to negate the effects of the
aerodynamic uncertainties, while it is negligible in the acrosswind
response. Hence, since the design is often controlled by the across-
wind direction, the relative significance of the alongwind load fac-
tor is diminished.
Note that these observations, which delineate the difference
between the rigid and flexible buildings, presume that the building
dynamic features remain unchanged under the design loading con-
dition. Recent full-scale studies reveal that changes in both fre-
quency and damping are possible at even below-design wind
speeds. In addition, in high wind speed conditions like in hurri-
canes, negative aerodynamic damping may reduce total damping,
especially for the acrosswind response. Thus, there is a possibility
that amplitude-dependency of frequency and damping and nega-
tive aerodynamic damping may offset one another to some degree.
Accordingly, caution should be exercised in the assessment of
appropriate wind load factors for super-tall buildings in different
terrain conditions.
Acknowledgements
The authors are grateful for the financial support provided in
part by a collaborative research project between the NatHaz
Modeling Laboratory and the Global Center of Excellence (GCOE)
at Tokyo Polytechnic University funded by MEXT, Japan and the
NSF Grant # CMMI 1301008. Dr. Seymour M.J. Spence of
the NatHaz Modeling Laboratory is thanked for his comments on
the manuscript.
Appendix A. Theoretical background of DED procedure
The theory and procedure employed in the DED procedure are
basically similar to NALD [48,33] and DEDM-HR [34] in which aero-
dynamic base moment/torque spectra obtained from HFBB experi-
ments are utilized for evaluating the base moment/torque, the
mean/background/resonant equivalent static wind loads (ESWL),
the maximum displacements and the peak/RMS (root-mean-
square) acceleration of a building. For the sake of completeness,
the underlying theoretical background is briefly described here.
Assuming a stationary Gaussian process, the expected maxi-
mum base moment/torque response in the alongwind, acrosswind
and torsional directions can be expressed as:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b ¼ M þ g  rM  M þ M 2 þ M 2
M [19] Gabbai RD, Fritz WP, Wright AP, Simiu E. Assessment of ASCE 7 standard wind
B R
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2
2 p
¼ M þ ðg B  rCM  M 0 Þ þ g R  rCM  M 0  C M ðf r1 Þ
4f1
where M = mean base moment that becomes zero in the acrosswind
and torsional responses; MB, MR = peak background and resonant
base moment or torque component, respectively; g, gB, gR = peak
factors for total, background and resonant moments, respectively;
rM, rCM = RMS of the fluctuating base moment/torque response
and non-dimensional base moment/torque response coefficient
(=rM =M0 ), respectively; M 0 = reference moment or torque; f1 =
damping ratio of a building in the first mode; CM(fr1) = non-
dimensional base moment spectrum value corresponding to fr1
[CM(fr) = f r  SM ðf r Þ=r2M ]; fr1 = reduced frequency (fr = fB/UH); f =
frequency [Hz]; UH = mean wind velocity at building height H; fr1 =
reduced frequency according to f1 (fr = fB/UH); f1 = natural
61
frequency of a building in the direction of motion; SM(f) = power
spectral density (PSD) of the fluctuating base moment or torque
response; Since CM(fr) and rCM are obtained from the HFBB experi-
ments, the mean, background and resonant base moments/torques
can be computed using the building properties.
Using Eq. (1), the equivalent static wind loads (ESWL) on a
building can be computed by distributing the peak base moments
to each floor in a similar fashion to how the base shear is dis-
tributed to each floor in earthquake engineering applications. In
addition, the displacement response calculation can be computed
by invoking a modal analysis procedure. Details can be found in
Zhou et al. [48], Kareem and Zhou [31], and Kwon et al. [33].
References
[1] American Society of Civil Engineers (ASCE). Minimum design loads for
buildings and other structures. ASCE 7-98, Reston (VA): ASCE; 1998.
[2] American Society of Civil Engineers (ASCE). Minimum design loads for
buildings and other structures. ASCE 7-05, Reston (VA): ASCE; 2005.
[3] American Society of Civil Engineers (ASCE). Minimum design loads for
buildings and other structures. ASCE 7-10, Reston (VA): ASCE; 2010.
[4] Bashor R. Dynamics of wind sensitive structures. Ph.D. Dissertation, University
of Notre Dame; 2011.
[5] Bashor R, Kareem A. Load factors for dynamically sensitive structures. In: 11th
Americas conference on wind engineering, San Juan, Puerto Rico; 2009.
[6] Bashor R, Kijewski-Correa T, Kareem A. On the wind-induced response of tall
buildings: The effect of uncertainties in dynamic properties and human
comfort thresholds. Proc. Americas Conference on Wind Engineering, Baton
Rouge, LA; 2005.
[7] Bernardini E, Spence SMJ, Kareem A. A probabilistic approach for the full
response estimation of tall buildings with 3D modes using the HFBB. Struct
Safety 2013;44:91–101.
[8] Boggs DW. Validation of the aerodynamic model method. J Wind Eng Ind
Aerodyn 1992;42:1011–22.
[9] Cao H, Gu M. Alongwind aerodynamic damping of tall buildings. Personal
communication; 2013.
[10] Chen X, Huang G. Estimation of probabilistic extreme wind load effects:
combination of aerodynamic and wind climate data. J Eng Mech 2010;136
(6):747–60.
[11] Cheng C-M, Lu P-C, Tsai M-S. Acrosswind aerodynamic damping of isolated
square-shaped buildings. J Wind Eng Ind Aerodyn 2002;90:1743–56.
[12] Cheng C-M, Wang J. Wind tunnel database for an intermediate wind resistance
design of tall buildings. In: Proc. 1st international symposium on wind effects
on buildings and urban environment. Tokyo (Japan): Tokyo Polytechnic
University; 2004.
[13] Coles S, Simiu E. Estimating uncertainty in the extreme value analysis of data
generated by a hurricane simulation model. J Eng Mech 2003;129
(11):1288–94.
[14] Davenport AG, Hill-Carroll P. Damping in tall buildings: its variability and
treatment in design. In: Building Motion in Wind. Seattle (USA): ASCE Spring
Convention; 1986. p. 42–57.
[15] Diniz SMC, Sadek F, Simiu E. Wind speed estimation uncertainties: effects of
climatological and micrometeorological parameters. Probabilist Eng Mech
2004;19:361–71.
[16] Diniz SMC, Simiu E. Probabilistic descriptions of wind effects and wind-load
factors for database-assisted design. J Struct Eng 2005;131(3):507–16.
[17] Ellingwood B, MacGregor JG, Galambos TV, Cornell CA. Probability-based load
criteria: Load factors and load combinations. J Struct Div 1982;108(5):978–97.
[18] Ellingwood B, Tekie P. Wind load statistics for probability-based structural
design. J Struct Eng 1999;125(4):453–63.
loaf factors for tall building response estimates. J Struct Eng 2008;134
(5):842–5.
ð5Þ [20] Gabbai RD, Simiu E. Aerodynamic damping in the along-wind response of tall
buildings. J Struct Eng 2010;136(1):117–9.
[21] Holmes JD, Melbourne WH, Walker GR. A commentary on the Australian
standard for wind loads. AS 1170 Part 2, Australian Wind Engineering Society;
1990.
[22] Jeary AP. Damping in tall buildings – a mechanism and a predictor. Earthquake
Eng Struct Dyn 1986;14:733–50.
[23] Kareem A. Wind excited motion of buildings. Ph.D. dissertation, Fort Collins
(Colorado): Colorado State University; 1978.
[24] Kareem A. Acrosswind response of buildings. J Struct Div 1982;108
(ST4):869–87.
[25] Kareem A, Cermak JE. Pressure fluctuations on a square building model in
boundary-layer flows. J Wind Eng Ind Aerodyn 1984;16:17–41.
[26] Kareem A. Aerodynamic response of structures with parametric uncertainties.
Struct Safety 1988;5(3):205–25.
[27] Kareem A. Effect of parametric uncertainties on wind excited structural
response. J Wind Eng Ind Aerodyn 1988;30(1–3):233–41.
[28] Kareem A, Sun W-J. Dynamic response of structures with uncertain damping.
Eng Struct 1990;12(1):2–8.
62 D.K. Kwon et al. / Engineering Structures 103 (2015) 53–62
[29] Kareem A, Gurley K. Damping in structures: its evaluation and treatment of
uncertainty. J Wind Eng Ind Aerodyn 1996;59(2–3):131–57.
[30] Kareem A, Kwon DK, Tamura Y. Cyberbased analysis, modeling and simulation
of wind load effects in VORTEX-Winds. In: 3rd American association for wind
engineering workshop, Hyannis, Massachusetts, USA, August 12–14; 2012.
[31] Kareem A, Zhou Y. Gust loading factor – past, present and future. J Wind Eng
Ind Aerodyn 2003;91(12–15):1301–28.
[32] Kijewski-Correa T. Full-scale measurements and system identification: A time-
frequency perspective. Ph.D. dissertation, Notre Dame (IN): Univ. of Notre
Dame; 2003.
[33] Kwon D, Kijewski-Correa T, Kareem A. E-Analysis of high-rise buildings
subjected to wind loads. J Struct Eng 2008;134(7):1139–53.
[34] Kwon DK, Kareem A. A multiple database-enabled design module with
embedded features of international codes and standards. Int J High-Rise
Build 2013;2(3):257–69.
[35] Kwon D, Spence S, Kareem A. Performance evaluation of database-enabled
design frameworks for the preliminary design of tall buildings. J Struct Eng
2014. http://dx.doi.org/10.1061/(ASCE)ST.1943-541X.0001229 [04014242].
[36] Lind NC. Consistent partial safety factors. J Struct Div 1971;97(ST6):
1651–70.
[37] Minciarelli F, Gioffrè M, Grigoriu M, Simiu E. Estimates of extreme wind effects
and wind load factors: influence of knowledge uncertainties. Probabilist Eng
Mech 2001;16(4):331–40.
[38] Peterka JA, Shahid S. Design gust wind speeds in the United States. J Struct Eng
1998;124(2):207–14.
[39] Satake N, Suda K, Arakawa T, Sasaki A, Tamura A. Damping evaluation using
full-scale data of buildings in Japan. J Struct Eng 2003;129(4):470–7.
[40] Solari G. Gust buffeting. I: Peak wind velocity and equivalent pressure. J Struct
Eng 1993;119(2):365–82.
[41] Solari G, Kareem A. On the formulation of ASCE 7-95 gust effect factor. J Wind
Eng Ind Aerodyn 1998;77–78:673–84.
[42] Spence SMJ, Kareem A. Tall buildings and damping: a concept-based data-
driven model. J Struct Eng 2014;140(5):04014005.
[43] Trifunac MD. Comparisons between ambient and forced vibration
experiments. Earthquake Eng Struct Dyn 1972;1(2):133–50.
[44] Vickery PJ, Wadhera D, Twisdale Jr LA, Lavelle FM. United States hurricane
wind speed risk and uncertainty. J Struct Eng 2009;135(3):301–20.
[45] Vickery PJ, Wadhera D, Galsworthy J, Peterka JA, Irwin PA, Griffis LA. Ultimate
wind load design gust wind speeds in the United States for use in ASCE-7. J
Struct Eng 2010;136(5):613–25.
[46] Watanabe Y, Isyumov N, Davenport AG. Empirical aerodynamic damping
function for tall buildings. J Wind Eng Ind Aerodyn 1997;72:313–21.
[47] Zhou Y, Kareem A, Gu M. Mode shape corrections for wind load effects. J Struct
Eng 2002;128(1):15–23.
[48] Zhou Y, Kijewski T, Kareem A. Aerodynamic loads on tall buildings: interactive
database. J Struct Eng 2003;129(3):394–404.