Along-Wind Load Effects on Tall Buildings: Comparative

Study of Major International Codes and Standards

Yin Zhou1; Tracy Kijewski, S.M.ASCE2; and Ahsan Kareem, M.ASCE3

Abstract: Most international codes and standards provide guidelines and procedures for assessing the along-wind effects on tall
structures. Despite their common use of the ‘‘gust loading factor’’ 共GLF兲 approach, sizeable scatter exists among the wind effects
predicted by the various codes and standards under similar flow conditions. This paper presents a comprehensive assessment of the source
of this scatter through a comparison of the along-wind loads and their effects on tall buildings recommended by major international codes
and standards. ASCE 7-98 共United States兲, AS1170.2-89 共Australia兲, NBC-1995 共Canada兲, RLB-AIJ-1993 共Japan兲, and Eurocode-1993
共Europe兲 are examined in this study. The comparisons consider the definition of wind characteristics, mean wind loads, GLF, equivalent
static wind loads, and attendant wind load effects. It is noted that the scatter in the predicted wind loads and their effects arises primarily
from the variations in the definition of wind field characteristics in the respective codes and standards. A detailed example is presented to
illustrate the overall comparison and to highlight the main findings of this paper.
DOI: 10.1061/共ASCE兲0733-9445共2002兲128:6共788兲
CE Database keywords: Buildings, highrise; Building codes; Wind loads; Dynamics; Wind velocity.

Introduction recently, Kijewski and Kareem 共1998兲 compared different inter-

national codes and standards with experimental data obtained in a
Most international codes and standards utilize the ‘‘gust loading
wind tunnel using a high-frequency base balance. Their compari-
factor’’ 共GLF兲 approach for assessing the dynamic along-wind
son also included the across-wind and torsional responses. De-
loads and their effects on tall structures. The concept of the GLF
spite these studies, an in-depth investigation concerning the un-
for civil engineering applications was first introduced by Daven-
derlying sources of this scatter has not been conducted,
port 共1967兲, following the statistical treatment of buffeting in
motivating this study. Furthermore, with the globalization of the
aeronautical sciences 共Liepmann 1952兲. Several modifications
construction industry and the prospect of developing unified in-
based on the first GLF model by Davenport followed, which in-
ternational standards, it is becoming increasingly important to
clude Vellozzi and Cohen 共1968兲, Vickery 共1970兲, Simiu and
better understand the underlying differences.
Scanlan 共1996兲, and Solari 共1993a, b兲. Variations of these models
This paper presents a comprehensive comparative study of the
have been adopted by major international codes and standards.
along-wind loads and their effects on tall buildings utilizing major
Although a similar theoretical basis is utilized in these formu-
international codes and standards: ASCE 7-98 共ASCE 1999兲,
lations, considerable scatter in the predictions of codes and stan-
AS1170.2-89 共Australian Standards 1989; Holmes et al. 1990兲,
dards has been reported 共e.g., Loh and Isyumov 1985; Ferraro
NBC-1995 共NRCC 1995兲, RLB-AIJ-1993 共AIJ 1996兲, and
et al. 1989兲. Lee and Ng 共1988兲 compared some of the interna-
Eurocode-1993 共Eurocode 1995兲. In order to provide a self-
tional codes, mainly focusing on the GLF. Discrepancies in the
sufficient document, the derivation of GLF for the along-wind
definitions of wind spectrum and wind correlation among the
response is briefly reviewed to highlight the salient features and
various codes and standards were noted in this study. In a subse-
their respective roles in the overall framework used in the estima-
quent study, Jesien et al. 共1993兲 made similar observations. More
tion of wind load effects. The comparison in this study considers
1
the definition of wind characteristics, associated mean wind loads,
Engineer, Malouf Engineering International, Inc., 275 West Campbell GLF, equivalent static wind loads 共ESWL兲, and attendant wind
Rd Suite 611, Richardson, TX 75080; formerly, Rsearch Associate, Dept. load effects. An example is given to highlight this comparison and
of Civil Engineering and Geological Sciences, Univ. of Notre Dame,
the main findings of this study.
Notre Dame, IN 46556. E-mail: [email protected]
2
Graduate Research Assistant, Dept. of Civil Engineering and Geo-
Gust Loading Factor
logical Science, Univ. of Notre Dame, Notre Dame, IN 46556. E-mail:
[email protected] Following the concept of the GLF approach 共Davenport 1967兲,
3
Robert M. Moran Professor and Chair, Dept. of Civil Engineering the peak ESWL on tall buildings provided in codes and standards
and Geological Science, Univ. of Notre Dame, Notre Dame, IN 46556. is described by a product of the mean wind force and an appro-
E-mail: [email protected] priate amplification factor
Note. Associate Editor: Bogusz Bienkiewicz. Discussion open until
November 1, 2002. Separate discussions must be submitted for individual P̂ T 共 z 兲 ⫽G ␶ • P̄ ␶ 共 z 兲 (1)
papers. To extend the closing date by one month, a written request must
where P̂ T (z)⫽peak ESWL at height z during observation time T,
be filed with the ASCE Managing Editor. The manuscript for this paper
was submitted for review and possible publication on November 21,
usually one hour 共1 h兲 or 10 minutes 共10 min兲 for most civil
2000; approved on August 22, 2001. This paper is part of the Journal of engineering applications; superscript ␶⫽averaging time used to
Structural Engineering, Vol. 128, No. 6, June 1, 2002. ©ASCE, ISSN evaluate the mean wind velocity; and P̄ ␶ ⫽mean wind force with
0733-9445/2002/6-788 –796/$8.00⫹$.50 per page. averaging time ␶

788 / JOURNAL OF STRUCTURAL ENGINEERING / JUNE 2002

 P̄ ␶ 共 z 兲 ⫽q 共 z 兲 •C d •B (2) Table 1. Averaging Time in Codes and Standards
ASCE 7 AS1170.2 NBC RLB-AIJ Eurocode
in which C d ⫽drag force coefficient; B⫽width of the building
normal to the oncoming wind; and q(z)⫽1/2␳V̄(z) 2 ⫽mean wind Basic wind velocity or 3s 3 sa 1h 10 min 10 minb
velocity pressure, where ␳⫽air density and V̄(z)⫽mean wind ve- basic wind pressure
locity evaluated at height z above ground. The gust factor G ␶ is Gust-loading factor 3s 1h 1h 10 min 3s
given by Wind-induced 1h 1h 1h 10 min 10 min
response
G ␶ ⫽G TY /G ␶q 共 T 兲 (3) a
Although the basic wind velocity is defined as 3 s gust in AS1170.2, it is
converted to the 1 h mean wind velocity to evaluate the gust-loading
in which G TY ⫽GLF for displacement and G ␶q (T)⫽gust
factor factor and the wind-induced response of dynamic structures.
共GF兲 for wind velocity pressure. The displacement GLF takes into b
Using an exposure coefficient, the 10 min basic wind velocity or pres-
account the correlation structure of random wind field, wind- sure is transformed to the 3 s gust wind pressure, which is then multiplied
structure interaction, and the dynamic amplification introduced by with the 3 s gust-loading factor to obtain the equivalent static wind loads.
the structure. Following the current practice in the GLF approach
共Davenport 1967兲, it can be evaluated by
current major codes and standards are summarized in Table 2. It is
G TY ⫽Ŷ T 共 z 兲 /Ȳ T 共 z 兲 (4) noted that ASCE 7-98 and Eurocode are based on a similar back-
ground 共Solari 1993a; Solari and Kareem 1998兲, which has re-
where Ŷ T and Ȳ T ⫽peak and mean wind-induced displacement
sulted in very similar formulations in these codes. To facilitate a
response, respectively. For the sake of completeness and quick
convenient comparison, the expressions in Table 2 are rewritten
reference, a general derivation of GLF based on displacement is
from the original expressions in the codes and standards to make
provided in Appendix I. It is noted that, in this formulation, the
the displacement GLF follow the format presented in Eq. 共22兲 or
averaging time for the mean wind velocity is equal to the obser-
Eq. 共24兲 later in the paper and subsequently referred to as the
vation time. This implies that the mean wind velocity and the
standard form in Appendix I.
mean wind load or response have the same averaging time. Al-
though this is the case for some codes 共e.g., RLB-AIJ-1993;
NBC-1995兲, the mean wind velocity with a shorter averaging
Description of Wind Characteristics in Codes and
time, e.g., 3 seconds 共3 s兲, is also used in some codes 共e.g., ASCE
Standards
7-98; AS1170.2-89兲. The following GF is used to convert the
mean wind-velocity pressure to an appropriate averaging period
In the formulation of the GLF approach, the GLF, ESWL and
when the mean wind velocity or pressure with a shorter averaging
wind load effects depend on the mean wind velocity profile, tur-
time is used as an input in determining the mean wind load or
bulence intensity, wind spectrum, turbulence length scale, and
response
correlation structure of the wind field. An overview of the defini-
G ␶q 共 T 兲 ⫽q̄ ␶ /q̄ T (5) tion or description of these wind characteristics in codes and stan-
dards is provided in this section.
where q⫽wind velocity pressure. Several representative models
of GF for wind velocity or wind pressure are provided in Appen-
dix II. Using the result given by Durst 共1960兲 and assuming T⫽1 Basic Wind Velocity ÕPressure
h, GFs for the basic mean wind velocity at a 10-meter 共10 m兲 The basic wind velocity in most codes and standards is based on
height in the open country terrain are G 3␯ s 共1 h兲⫽1.51 and G 1␯ h wind measurements at 10 m height in an open terrain associated
共1 h兲⫽1.00, respectively. Using a simple square law, the corre- with different mean recurrence intervals and averaging times. The
sponding GFs for the wind-velocity pressure are G 3q s 共1 h兲⫽2.28 basic wind velocity is converted to the design reference wind
and G 1q h 共1 h兲⫽1.00. velocity for a particular site by introducing the influence of local
The above discussion clarifies the important role of averaging environment, directionality, mean recurrence interval, and signifi-
time in this comparative study. On the one hand, when ␶⫽T, the cance factors associated with the planned structure as shown
wind load model in Eq. 共1兲 reduces to the general GLF model by below
Davenport 共1967兲. On the other hand, when using the mean wind
V⫽V 0 •C direction•C shield•C importance•C return (6)
velocity with a shorter averaging time, G in Eq. 共1兲 may be sig-
nificantly less than the GLF in Eq. 共4兲. Therefore, it is important where V 0 ⫽basic mean wind velocity; C direction⫽directionality
to compare the results based on similar averaging times. A sum- factor; C shield⫽shielding factor; C importance⫽building importance
mary of the averaging time for the basic wind velocity or pres- factor; and C return⫽a factor for adjusting wind mean recurrence
sure, GLF, ESWL or wind-induced response employed in codes interval. For the sake of simplicity, these factors are ignored in
and standards is given in Table 1. this study.
All procedures for estimating GLF and ESWL in major codes
and standards are based on the preceding expressions, including
Mean Wind Velocity Profile
Appendix I and II, but differ in their modeling of the wind field
and structural dynamic characteristics. These details have led to a Generally, the wind velocity distribution along the height is influ-
large scatter in the predicted values of the GLF and wind load enced by the local topography, surrounding terrain, and averaging
effects based on distinct formulations. Furthermore, as a result of time. The local topography may have a speed-up effect on the
several mathematical manipulations that have been introduced by wind profile, which is ignored in this discussion for simplicity.
individual codes and standards, the final expressions for the GLF The averaging time also influences the wind-velocity profile. For
do not follow exactly the same form as outlined in the Appen- example, the 3 s gust has a flatter distribution than the 1 h mean
dixes. For a quick reference, the procedures for the GLF in the wind velocity. AS1170.2 and ASCE 7 provide mean wind-

JOURNAL OF STRUCTURAL ENGINEERING / JUNE 2002 / 789

Table 2. Calculation of Gust-Loading Factors in Codes and Standards
ASCE 7 AS1170.2 NBC RLB-AIJ Eurocode

冉 1⫹r 冑g Q
冊 1⫹r 冑 1⫹g f r 冑B⫹ R 1⫹g f r 冑B⫹ R
a
B⫹ g R2 R g V2 B(1⫹w) 2 ⫹g 2f Rd
2 a,c
G c
0.925
1⫹g ␯ •r
冑
1⫹g f r B⫹ R
1⫹3.5r

T 3,600 s 3,600 s 3,600 s 600 s 600 s

z̄ 0.6H H H H 0.6H

r⫽ 冑2K/C eH f
e g
r r⫽1.7I z̄ r⫽2•I z̄ r⫽(3⫹3␣)/(2⫹␣)•I z̄ r⫽2I z̄

g Q ⫽g V ⫽3.4 g V ⫽3.7; g f ⫽g R (T,␷); g g f ⫽ 冑2 ln(T•␷)⫹1.2 g f ⫽g R (T,␷) h

g
g R ⫽g R (T, f 1 ) h g f ⫽ 冑2 ln(T•f 1) ␷⫽ f 1 冑SE/(SE⫹␨B) ␷⫽ f 1 冑R/( B⫹ R) 冑
␷⫽ (␷ 20 B⫹ f 21 R)/( B⫹ R)

B
1⫹0.63
1

冉 冊
B⫹H 0.63

1⫹
冑
1

36H 2 ⫹64B 2
2
3 冕 914/H

0 1⫹
1
xH
457
1⫹
1 x
xB 共 1⫹x 2 兲 4/3
122
dx 1⫺

再 冉冑 冊 冉 冊 冎
1⫹5.1
LH
1
1.3 B k 1/3
1⫹0.9
1

冉 冊
B⫹H 0.63

L z̄ LH HB H L z̄

i 9.5N 1 /(1⫹10.3N 1 ) 5/3 e

E 0.6N 1 /(2⫹N 21 ) 5/6 2N 21 /3(1⫹N 21 ) 4/3 4N 1 /(1⫹71N 21 ) 5/6 6.8N 1 /(1⫹10.2N 1 ) 5/3

1 1 0.84

冋 册冋 册 冋 册冋 册 冋 册冋 册
k
S R H R B (0.53⫹0.47R D ) j f 1H f 1B 8 f 1H 10f 1 B 2.1f 1 H 2.1f 1 B R HR B j
1⫹3.5 1⫹4 1⫹ 1⫹ 1⫹ • 1⫹
V̄ H V̄ H 3V̄ H V̄ H V̄ H V̄ H
a
Expressions for GLF in this table are not necessarily reproduced from the original codes and standards, but are rewritten in the standard form 关refer to
Eq. 共22兲 or Eq. 共24兲兴.
b
0.925 is an adjustment factor used to make the wind load in the updated code consistent with the former version.
c
Numerator is the displacement GLF and the denominator is the GF for the wind velocity pressure.
d
w is an approximate consideration of the quadratic wind velocity term 共Vickery 1995兲.
e
A 3 s low-pass filter has been included 共Solari and Kareem 1998兲.
f
K is provided for different terrains in NBC.
g
A 0.75 factor is used to account for nonuniform load distribution 共RLB-AIJ 1994兲.
h
g R (T, f 1 ); see Eq. 共23兲 by substituting relevant parameters.
i
E⫽ f 1 S ␯ ( f 1 )/␴ 2␯ and N 1 ⫽ f 1 L ¯Z /V̄ ¯Z .
j
R l ⫽1/␩⫺1/2␩ 2 (1⫺e ⫺2␩ ) for ␩⬎0; and R l ⫽1 for ␩⫽0. R H , ␩⫽4.6f 1 H/V ¯Z ; R B , ␩⫽4.6f 1 B/V ¯Z ; and R D , ␩⫽15.4f 1 D/V ¯Z .
k
Aerodynamic admittance function is equal to 0.84 at zero frequency.

velocity profiles based on both 3 s and 1 h averaging times, V 共 z 兲 ⫽V 0 •b• 共 z/10兲 ␣ (7)
whereas NBC, RLB-AIJ, and Eurocode utilize averaging times of
1 h, 10 min, and 10 min for the mean velocity profiles, respec-
tively. where b and ␣⫽constants depending on the terrain type. For an
There are two kinds of basic wind-velocity profile descrip- open terrain case 共exposure C兲 at 10 m height, b is equal to unity
tions, i.e., the logarithmic and the power law. AS1170.2 and Eu- since the basic wind velocity is defined for this exposure. Coef-
rocode use the logarithmic profile, whereas all others use a ficients b and ␣ for all exposures provided in codes and standards
power-law profile. Nevertheless, the wind profiles provided in the are summarized in Table 3.
codes and standards discussed here can be expressed in terms of When using 3 s reference velocity V̂ 30 s , the mean wind veloc-
the following general power law: ity profiles in codes and standards can also be expressed as

Table 3. Mean Wind Velocity Profiles in Codes and Standards 关Eq. 共7兲兴
ASCE 7 AS1170.2 共fitted兲 NBC RLB-AIJ Eurocode 共fitted兲
3s 1h 3s 1h 1h 10 min 10 min
b ␣ b ␣ b ␣ b ␣ b ␣ b ␣ b ␣
A 0.66 0.20 0.30 0.33 0.76 0.14 0.29 0.28 0.43 0.36 0.39 0.35 0.55 0.29
B 0.85 0.14 0.45 0.25 0.91 0.10 0.45 0.20 0.67 0.25 0.58 0.27 0.77 0.21
C 1.00 0.11 0.65 0.15 1.04 0.07 0.58 0.16 1.00 0.14 0.79 0.20 1.00 0.16
D 1.09 0.09 0.80 0.11 1.18 0.04 0.69 0.13 1.00 0.15 1.17 0.12
E 1.23 0.10
Note: Basic wind velocity refers to the condition where the coefficient b is equal to unity, which is shown in bold in this table.

790 / JOURNAL OF STRUCTURAL ENGINEERING / JUNE 2002

Fig. 1. Mean wind velocity profiles in codes and standards 关Eq. 共8兲兴; Fig. 2. Turbulence intensity profiles in codes and standards 关Eq. 共9兲兴;
in terms of 3 s gust basic wind velocity. A: large city center; C: open A: large city center; C: open country terrain
country terrain. Note: Open country terrain corresponds to exposure
C in codes and standards, except for AIJ standard, in which exposure From Table 4 and Fig. 2, significant differences in the descrip-
D is used. Refer to Table 3. tion of the turbulence intensity can be noted. These variations will
affect both the background and the resonant GLF components,
without a significant influence on the mean wind loads.
C e 共 z 兲 ⫽V̄ 共 z 兲 /V̂ 30 s
(8)
Wind Spectra, Turbulence Length Scales,
This information is included in both ASCE 7 and AS1170.2. For
and Correlation Structure
other codes and standards that do not use 3 s gust as the reference
wind velocity, Eq. 共32兲 of the Appendix is used to relate the mean Table 5 lists wind spectra and definitions of turbulence length
wind velocity to the 3 s gust. For example, in the case of NBC, scales given in the codes and standards. It is interesting to note
T⫽1 h. Considering exposure C at the 10 m height and assuming that AS1170.2, NBC, and RLB-AIJ all prescribe a length scale
I⫽0.2, the velocity gust factor G 3V 0s 共1 h兲⫽1.58. Similarly, by formulation independent of terrain, though data in Counihan
setting T⫽10 min, G 3V 0s 共10 min兲⫽1.48 can be used to convert the 共1975兲 suggest that it is a decreasing function of terrain tough-
ness. Fig. 3 provides wind spectra for exposures A and C for the
10 min mean wind velocity in RLB-AIJ and Eurocode to 3 s gust.
case of a 200 m tall building. The reference height, turbulence
The mean wind profiles provided by each code and standard for
length scale, and wind velocity for each spectrum are given in
exposures C and A 共large city centers兲 are shown in Fig. 1.
Table 6. The difference in the definition of the wind spectrum
From both Table 3 and Fig. 1, considerable differences be-
primarily influences the resonant component of GLF and the as-
tween the mean wind-velocity profiles among different codes and
sociated acceleration response, whereas there is a marginal influ-
standards can be noted. Clearly, this will have a significant impact
ence on the background component of the GLF.
on the mean wind load estimates. The mean wind velocity will
As shown in the derivation of the displacement-based GLF in
also indirectly influence the GLF, thus impacting estimates of the
Appendix I, the correlation structure of the fluctuating wind ve-
overall wind load effects.
locity is reflected in the background factor and the aerodynamic
admittance function or size reduction factor. A comparison of size
Turbulence Intensity Profile reduction factors used in codes and standards is provided in Fig.
The turbulence intensity profile can also be expressed in terms of 4. Once again, the variations in size reduction factors are note-
a power law worthy.

I 共 z 兲 ⫽c• 共 z/10兲 ⫺d (9)

where c and d⫽constants depending on the terrain type. These Application of Gust Loading Factor to Wind Load
coefficients, as provided in the codes and standards, are summa- Effects
rized in Table 4. Fig. 2 illustrates a comparison of these profiles Although all current codes and standards follow the preceding
for exposures A and C. outline, it is interesting that the ESWL obtained using the tradi-

Table 4. Turbulence Intensity Profiles in Codes and Standards 关Eq. 共9兲兴

ASCE 7 AS1170.2 共fitted兲 NBC 共derived兲a RLB-AIJ Eurocode 共fitted兲
Terrain c d c d c d c d c d
A 0.450 0.167 0.453 0.300 0.621 0.360 0.402 0.400 0.434 0.290
B 0.300 0.167 0.323 0.300 0.335 0.250 0.361 0.320 0.285 0.210
C 0.200 0.167 0.259 0.300 0.200 0.140 0.259 0.250 0.189 0.160
D 0.150 0.167 0.194 0.300 0.204 0.200 0.145 0.120
E 0.162 0.150
a
Turbulence intensity profile is not explicitly available in NBC. The data herein are derived by rewriting the code procedure in the standard form.

JOURNAL OF STRUCTURAL ENGINEERING / JUNE 2002 / 791

Table 5. Wind Spectra and Turbulence Length Scales in Codes and Standards
Wind spectra 共unified兲 Turbulence length scale 共m兲
¯
fS␯共 f 兲 6.868x f •L z̄ 共 z̄ 兲 L z̄ ⫽l(z̄/10) ␧
ASCE 7 ⫽ , x⫽
␴2␯ 共 1⫹10.302x 兲 5/3 V̄ 共 z̄ 兲
l and ¯␧ depend on terrain

L H ⫽1,000(H/10) 0.25; A
fS␯共 f 兲 4x LH f
AS1170.2 ⫽ , x⫽
␴2␯ 6.677共 2⫹x 2 兲 5/6 V̄ H measure of turbulence scale

1,220; A length scale,

fS␯共 f 兲 2x2 1,200f
NBC ⫽ , x⫽
␴2␯ 3共1⫹x2兲4/3 V̄ H Davenport 共1967兲

fS␯共 f 兲 4x f LH
RLB-AIJ ⫽ , x⫽ L H ⫽100(H/30) 0.5
␴2␯ 共 兲
1⫹70.8x 2 5/6 V̄ H
¯
fS␯共 f 兲 6.868x f •L z̄ 共 z̄ 兲 L z̄ ⫽300(z̄/300) ␧
Eurocode ⫽ , x⫽
␴2␯ 共 1⫹10.302x 兲 5/3 V̄ 共 z̄ 兲 ¯␧ depends on terrain

tional GLF formulation in Eq. 共1兲 does not represent the actual base bending moment following the procedure provided by Zhou
ESWL acting on a tall building. Rather, the ESWL given in Eq. and Kareem 共2001兲. For example, the background ESWL compo-
共1兲 follows the distribution of the mean wind load. This is not nent is given by
consistent with the distribution of the inertial force that is propor-
tional to the mass distribution and the mode shape of the building. P̂B 共 z 兲 ⫽G Y B • P̄ 共 z 兲 (11)
Although the ESWL in Eq. 共1兲 can ensure accurate estimation of where G Y B ⫽g B •r 冑B; and for the resonant ESWL component
the first mode displacement, it may result in less reliable estimates
of other wind effects, e.g., the base shear 共Zhou et al. 1999a,b, m 共 z 兲 ␸ 1共 z 兲
P̂R 共 z 兲 ⫽ •M̂R (12)
2000兲. A more novel way to correctly use the traditional GLF is to 兰 0 m 共 z 兲 ␸ 1 共 z 兲 zdz
H
express it in terms of the base bending moment response. As
outlined in a new GLF formulation by Zhou and Kareem 共1999c, where M̂R ⫽G Y R •M̄ and G Y R ⫽g R •r 冑R. The RMS acceleration
2001兲, the actual peak base bending moment response can be can be computed by
estimated by
0 P̂R 共 z 兲 ␸ 1 共 z 兲 dz
兰H
M̂⫽G Y •M̄ (10) ␴ a共 z 兲 ⫽ •␸ 1 共 z 兲 (13)
g R• 兰 H
0 m 共 z 兲 ␸ 1 共 z 兲 dz
2

where M̄ ⫽ 兰 H0 P̄(z)•zdz⫽base bending moment under the mean The peak acceleration is obtained by multiplying the RMS accel-
wind load and M̂⫽peak base bending moment response. The dis- eration by g R . The acceleration response depends only on the
tribution of the ESWL components can be evaluated from the resonant component of GLF. It is noted that in this section the
difference in the averaging time is ignored. This difference can be
similarly treated as in the proceeding sections.

Application of Codes and Standards to an Example

Tall Building

A tall building is used as an example to compare the estimates of

wind load effects based on the codes and standards considered.
The building particulars are H⫽200 m, B⫽D⫽33 m; f 1 ⫽0.2
Hz, and linear mode shape in two translation directions; ␨
⫽0.01; C d ⫽1.3; and building density ⫽180 kg/m3. The building
is located at the edge of a central business district with exposure
A on one side and exposure C on the other; and the basic 3 s gust
wind velocity ⫽40 m/s. For simplicity, the effects of the wind
direction, topography, shielding, importance, and return period
are ignored in the following discussion.
The results obtained by using the procedures in the codes and
Fig. 3. Wind spectra in codes and standards; Refer to Tables 5 and 6
standards selected here are listed in Table 6. As expected, the
for involved parameters. A: large city center; C: open country terrain.
averaging time shown in Table 1 manifests significant influence

792 / JOURNAL OF STRUCTURAL ENGINEERING / JUNE 2002

Table 6. Results of Computation using Codes and Standards
ASCE 7 AS1170.2a NBC RLB-AIJ Eurocode
A C A C A C A C A C
V 0 共m/s兲 40 共3 s兲 40 共3 s兲 26 共1 h兲b 27 共10 min兲b 27 共10 min兲b
z̄ 共m兲 120 200 200 200 120
V̄ z̄ 共m/s兲 27.5 38.1 26.7 37.3 32.6 39.5 30.4 42.3 30.7 39.3
r 0.506 0.225 0.368 0.210 0.423 0.303 0.276 0.180 0.422 0.254
L z̄ 共m兲 190 250 2115 2115 1220 1220 258 258 197 236
B 0.583 0.624 0.633 0.633 0.300 0.300 0.582 0.582 0.500 0.529
E 0.140 0.144 0.094 0.117 0.170 0.191 0.080 0.100 0.106 0.109
S 0.048 0.079 0.080 0.123 0.077 0.101 0.154 0.212 0.087 0.121
R 0.525 0.889 0.596 1.138 1.031 1.524 0.967 1.655 0.726 1.039
gf g R ⫽3.79; g ␯ ⫽3.40 g R ⫽3.63; g ␯ ⫽3.70 3.759 3.768 3.209 3.235 3.208 3.225
3s 0.447 0.316 0.386 0.315
GB 10 min 0.676 0.443 0.958c 0.596c
1h 1.214c 0.559c 1.083 0.618 0.870 0.626
3s 0.472 0.421 0.466 0.442
GR 10 min 0.872 0.747 1.154c 0.835c
1h 1.283c 0.742c 1.030 0.813 1.614 1.411
3s 0.990 1.051 1.009 1.073
G 10 min 2.103 1.868 2.500c 2.026c
共78.2%兲g 共100.8%兲g 共92.9%兲g 共109.2%兲g
1h 2.691c 1.854c 2.495 2.021 2.833 2.544
共92.8%兲g 共109.2%兲g 共105.3%兲g 共137.3%兲g
M̄ 3s 1,035,400 1,465,400
共kN•m兲 10 min 367,810 833,050 528,250 837,510
共86.3%兲g 共105%兲g 共124%兲g 共106%兲g
1h 425,980d 790,360d 297,600 644,490 417,880 735,690
共69.9%兲g 共81.5%兲g 共98.1%兲g 共93.1%兲g
M̂ 10 min 773,410 1,556,400 1,320,400 1,696,700
共kN•m兲 共75.5%兲g 共101.1%兲g 共128.6%兲g 共110.3%兲g
1h 1,024,808 1,539,848 742,420 1,302,400 1,183,900 1,871,300
1,146,260e 1,465,015e 共72.4%兲g 共84.5%兲g 共115.7%兲g 共121.1%兲g
␴a 10 min 3.82 7.36 7.27 8.30
共milli-g兲 共77.3%兲g 共117.8%兲g 共147%兲g 共133%兲g
1h 4.94 6.23 3.23 5.52 6.86 10.54
5.52f 5.93f 共65.4%兲g 共88.5%兲g 共138.4%兲g 共168.6%兲g
a
Neglect the correction for the quadratic term.
b
Wind velocity GFs of 0.65 and 0.676 are used to transfer the 3 s gust wind velocity to those in 1 h and 10 min mean, respectively.
c
Displacement GLF, computed by neglecting the dominator in the GLF expression.
d
Computed based on the 1 h mean wind profile in ASCE 7.
e
Product of the GLF and the mean base bending moment both based on the 1 h averaging time.
f
Correspond to the base bending moment in e.
g
Comparing with the result by ASCE 7. For simplicity, the effect of a shorter averaging time 共10 min兲 in RLB-AIJ and Eurocode is not considered.

on both the GLF and the mean wind load estimates. Since the With the same input of 3 s basic gust wind velocity, the mean
basic wind-velocity pressure GF for the 3 s gust is G 3q s (1 h) wind loads obtained by each code and standard are quite varied
⫽(1.58) 2 ⫽2.50 or G 3q s 共10 min兲⫽(1.48) 2 ⫽2.19, the GLFs ob- due to their distinct mean wind velocity profiles. This results in
tained by using ASCE 7 and Eurocode, based on 3 s averaging disparities in the mean base bending moments. For example, the
time, are significantly less than those by other codes and stan- mean base bending moments in terms of a 1 h averaging time by
dards, which are based on a longer averaging time. However, in AS1170.2 are 70 and 81.5% of those given by ASCE 7 for expo-
terms of peak base bending moments and displacement GLFs, sures A and C, respectively.
which are based on longer averaging time, the difference in the Regarding the GLF, it is noted that estimates based on ASCE 7
estimates among the codes and standards is considerably reduced. are distinct from those given by Eurocode due to their unique
Since the codes and standards employ quite varied definitions definitions of wind characteristics, although these two codes are
of wind characteristics, it is not surprising to find considerably both based on a similar closed-form formulation 共Solari 1993a;
different results at each step of the wind load effects analysis even Solari and Kareem 1998兲. Eurocode neglects the correlation be-
when considering a similar averaging time. tween the windward and leeward wind pressures, which results in

JOURNAL OF STRUCTURAL ENGINEERING / JUNE 2002 / 793

 theories兲. These nuances in the wind field characteristics have
resulted in discrepancies not only in the GLF estimates, but also
in the mean wind loads, which correspondingly lead to significant
variations in the estimates of the ESWL and associated wind-
induced load effects.

Acknowledgments

The writers gratefully acknowledge partial support from NSF

Grants No. CMS95-03779, CMS95-22145, and CMS00-85019
for this study.

Appendix I: Derivation of GLF for displacement

Assuming that the wind-induced displacement response is a

Fig. 4. Size factors in codes and standards; refer to Tables 2 and 6 Gaussian process, the displacement GLF in Eq. 共4兲 can be ex-
for involved parameters. A: large city center; C: open country terrain pressed as 共Davenport 1967兲
G Y ⫽1⫹g•␴ Y 共 z 兲 /Ȳ 共 z 兲 (14)
larger size reduction factors. This effect, although partly compen- For the sake of brevity, the superscript T is omitted in this discus-
sated by the decreased gust energy factor and turbulence intensity, sion, implying that the mean wind velocity corresponds to the
combined with higher estimates of mean wind loads, leads to same averaging time as the observation time. Usually, the mean
28.6% and 47% higher estimates of base bending moment and structural displacement can be approximated in terms of the first
acceleration, respectively, for exposure A, if the difference in av- mode mean displacement response
eraging times between these two codes is ignored.
Comparing with other codes, ASCE 7 prescribes higher turbu- Ȳ 共 z 兲 ⫽ 共 P̄ * * 兲 •␸ 1 共 z 兲
1 /k 1 (15)
lence intensity, especially for exposure A, which results in greater
1 ⫽ 兰 0 P̄(z)␸ 1 (z)dz; 1 ⫽(2␲ f 1 ) m *
H 2
where P̄ * k* 1 ; m*1
estimates in both background and resonant GLF components. ⫽ 兰 0 m(z)␸ 1 (z)dz⫽generalized load,
H 2
stiffness, and mass of the
Coupled with the relatively higher mean wind load, ASCE 7 pro- first mode, respectively; f 1 ⫽natural frequency of the first mode;
duces the second largest peak estimates of the base bending mo- m(z)⫽mass per unit height; and the fundamental mode shape of
ment and acceleration for exposure A. In comparison to ASCE 7 a tall building can be approximated by ␸ 1 (z)⫽(z/H) ␤ , where
and NBC, AS1170.2 provides the lowest estimates of the GLF, ␤⫽mode shape exponent. This has been assumed to be equal to
base bending moment, and the acceleration since it prescribes unity in most codes and standards, since the effect of a nonlinear
lower turbulence intensity, gust energy factor, and mean wind mode shape on the derivation of the GLF can usually be neglected
load. However, NBC employs the highest gust energy factor, 共Vickery 1970; Zhou et al. 1999b兲.
which yields the largest value of GLF, base bending moment, and Using the quasi-steady and strip theories and neglecting the
acceleration for the exposure C case. Meanwhile, the background contribution of the quadratic wind-velocity term, the fluctuating
factor in NBC is apparently less than those by other codes. RLB- aerodynamic wind force acting on the surface of a tall building
AIJ employs the lowest turbulence intensity definition, which can be estimated by
leads to the lowest GLF estimates despite the fact that the size
reduction factor in this standard is the greatest. P̃ 共 z,t 兲 ⫽␳V̄ 共 z 兲 •␯ 共 z,t 兲 •C d •B (16)
where ␯(z,t) is the fluctuating wind velocity. Under the action of
this fluctuating wind pressure, the fluctuating displacement in the
Concluding Remarks first mode can also be computed by
All major international codes and standards are based on the GLF
approach for estimating the maximum wind load effects in the
along-wind direction; however, each employs unique definitions
␴Y共 z 兲⫽ 冉冕 0
⬁
S ˜P * 共 f 兲 d f • 兩 H 1 共 f 兲 兩 2 /k *
1 1
2
冊 1/2
•␾ 1 共 z 兲 (17)

of wind field characteristics, including the mean wind-velocity where 兩 H 1 ( f ) 兩 2 ⫽„关 1⫺( f / f 1 ) 2 兴 2 ⫹(2␨ f / f 1 ) 2 …⫺1 ⫽first mode
profile, turbulence intensity profile, wind spectrum, turbulence structural transfer function; ␨⫽critical damping ratio in the first
length scale, and the wind correlation structure 共related to the mode; and S ˜P * ( f )⫽power spectral density of the fluctuating gen-
1
aerodynamic admittance function using strip and quasi-steady eralized wind load, which can be expressed as

S ˜P * 共 f 兲 ⫽
1
冕冕冕冕
H

0
H

0
B

0
B

0
␳ 2 C 2d B 2 V̄ 共 z 1 兲 V̄ 共 z 2 兲 •R 共 x 1 ,x 2 ,z 1 ,z 2 , f 兲 S ␯ 共 z 1 ,z 2 , f 兲 ␸ 1 共 z 1 兲 ␸ 1 共 z 2 兲 dx 1 dx 2 dz 1 dz 2 (18)

where R(x 1 ,x 2 ,z 1 ,z 2 , f )⫽correlation function of the fluctuating wind pressures and S ␯ (z 1 ,z 2 , f )⫽cross-PSD of the fluctuating wind
velocity.
Using Eqs. 共15兲 and 共17兲, the fluctuating component of the GLF can be derived as

794 / JOURNAL OF STRUCTURAL ENGINEERING / JUNE 2002

 ␴ Y 共 z 兲 /Ȳ 共 z 兲 ⫽ 冉冕
0
⬁
S ˜P * 共 f 兲 兩 H 1 共 f 兲 兩 2 d f
1 冊冒
1/2
P̄ *
1 (19)

which shows that the GLF is independent of the mass. The integration in Eq. 共19兲 is usually performed in the background and resonant
portions. After some mathematical manipulations, the fluctuating component of the GLF can be derived
␴ Y /Ȳ ⫽2•I z̄ • 冑B⫹R (20)
where I z̄ ⫽ 关 (2⫹2␣)/(2⫹␣) 兴 I H in which I H ⫽␴ ␯ /V̄ H ⫽turbulent intensity evaluated at the top of the building; z̄⬇2H/3⫽reference
⬁
height; B⫽ 兰0 ␹( f )•S *
␯ (z̄, f )d f ⫽background factor; R⫽␲SE/4␨⫽resonant factor; S * ␯ ( f )⫽normalized wind velocity spectrum with re-

spect to the mean-square fluctuating wind velocity, ␴ 2␯ ; E⫽ f 1 S *

␯ (z̄, f 1 )⫽gust energy factor

␣ ␣
0 兰 0 兰 0 兰 0 共 z 1 /H 兲 共 z 2 /H 兲 共 z 1 /H 兲共 z 2 /H 兲 •R 共 x 1 ,x 2 ,z 1 ,z 2 , f 兲 dx 1 dx 2 dz 1 dz 2
兰H H B B
␹共 f 兲⫽ ␣ ␣
(21)
兰H
0 兰 0 兰 0 兰 0 共 z 1 /H 兲 共 z 2 /H 兲 共 z 1 /H 兲共 z 2 /H 兲 dx 1 dx 2 dz 1 dz 2
H B B

which represents the aerodynamic admittance function; S The first model for the wind velocity GF is based on the sta-
⫽␹( f 1 )⫽size reduction factor, which is a measure of the overall tistical analysis of the meteorological wind-velocity records
effect of the wind pressure correlation. For an ideal pointlike 共Durst 1960兲. For an open terrain, Durst 共1960兲 provided GFs for
structure, or when f →0, the velocity field is fully correlated or some typical averaging times with regard to the mean wind ve-
R⫽1. Thus, ␹( f ) approaches unity and both S and B are also locity with an observation time of both 10 min and 1 h. Durst’s
equal to unity. Meanwhile, since the length scale of turbulence is results were employed by ASCE 7-98 in defining the 3 s gust and
finite in size, the correlation of wind pressure decreases as the the 1 h mean wind velocity 共Simiu and Scanlan 1996兲. For expo-
distance increases. Theoretically, when the building size becomes sure C at 10 m height, G 3V s 共1 h兲⫽1.51 in terms of 1 h mean wind
infinitely large, the lack of correlation diminishes the effects of velocity.
wind, thus S and B concomitantly approach zero. The second model is based on wind-velocity spectrum analy-
Using Eqs. 共14兲 and 共20兲, the GLF can be expressed as sis, which introduces a low-pass filter corresponding to the aver-
aging time 共Greenway 1979; Solari 1993a兲. Considering that
G⫽1⫹g•r• 冑B⫹R (22) thewind-velocity fluctuations are a stationary Gaussian process,
in which r⫽2I z̄ , and the displacement peak factor g can be com- the gust factor in Eq. 共26兲 can be given by
puted by
G V␶ 共 T 兲 ⫽ 关 V̄ T ⫹g ␯ 共 ␶ 兲 ␴ ␯ 共 ␶ 兲兴 /V̄ T ⫽1⫹g ␯ 共 ␶ 兲 •I ␯ 共 ␶ 兲 (27)

g⫽ 冑2 ln共 ␷T 兲 ⫹0.5772/冑2 ln共 ␷T 兲 (23) in which

where ␷⫽mean up-crossing rate. Alternatively, Eq. 共22兲 can be

expressed in terms of the peak factors associated with the back- ␴ 2␯ 共 ␶ 兲 ⫽ 冕 0
⬁
S V 共 f 兲 ␬ 共 f ,␶ 兲 d f (28)
ground and resonant response components as given in ASCE 7-98
␬ 共 f ,␶ 兲 ⫽sin2 共 ␲ f ␶ 兲 / 共 ␲ f ␶ 兲 2 (29)
G⫽1⫹r• 冑 g B2 •B⫹g R2 •R
where background peak factor g B ⫽g ␯ ⫽wind velocity peak factor
(24)
n 0共 ␶ 兲 ⫽ 冉冕 ⬁

0
f 2 S ␯ 共 f 兲 ␬ 共 f ,␶ 兲 d f 冒冕 0
⬁
S ␯ 共 f 兲 ␬ 共 f ,␶ 兲 d f 冊 1/2

and g R ⫽resonant peak factor approximated by setting ␷⫽ f 1 in (30)

Eq. 共23兲.
It is noted that in this study the GLF in Eqs. 共22兲 or 共24兲 is g ␯ 共 ␶ 兲 ⫽ 冑2 ln关 n 0 共 ␶ 兲 T 兴 ⫹0.5772/冑2 ln关 n 0 共 ␶ 兲 T 兴 (31)
referred to as the standard form in which the following conditions A closed-form solution of the wind velocity gust factor was pro-
are satisfied: vided by Solari 共1993a兲
B→1 when B and H→0
B→0 when B and H→⬁ G V␶ 共 T 兲 ⫽1⫹g ␯ 共 ␶ 兲 ⫻I⫻ 冑P0 共 ␶ 兲 (32)
(25)
S→1 when B and H→0 or f →0
in which I⫽␴ T␯ /V̄ T ⫽turbulence intensity; g ␯ (␶)⫽ 兵 1.175
S→0 when B and H→⬁ or f →⬁
⫹2 ln(T̃冑P1 /P0 ) 其 1/2; P0 ⫽1/(1⫹0.56␶ ˜ 0.74); P1 /P0
⫽1/(31.25␶ ˜ ); ˜␶ ⫽␶V̄ z /L z ; T̃⫽TV̄ z /L z . For ASCE 7 use, ␶
1.44

⫽3 s, T⫽3,600 s, thus g ␯ (␶)⫽3.41 and P0 ⫽0.723 共Solari and

Appendix II: Gust Factors for Wind VelocityÕ
Kareem 1998兲. Applying this relationship to the reference wind
Pressure
velocity where I is close to 0.2, the gust factor G 3V s 共1 h兲⫽1.58.
The difference between this approach and Durst’s result is 4.6%.
Gust Factor for Wind Velocity Nevertheless, some inconsistencies have been found in other wind
exposures 共Zhou and Kareem 2002兲. In ASCE 7-98 code, the gust
The gust factor for the wind velocity can be defined as
factor in Eq. 共32兲 is incorporated in the overall GLF as outlined in
Eq. 共3兲. The inconsistency due to the application of both spectral
G V␶ 共 T 兲 ⫽V̄ ␶ /V̄ T (26) model 共Solari 1993a兲 and statistical model 共Durst 1960兲 in ASCE

JOURNAL OF STRUCTURAL ENGINEERING / JUNE 2002 / 795

7 manifests itself in the numerical example. More detailed infor- Australian Wind Engineering Society.
mation can be found in Zhou and Kareem 共2002兲. Kijewski, T., and Kareem, A. 共1998兲. ‘‘Dynamic wind effects: A compara-
AS1170.2-89 and Eurocode employed a similar simplified pro- tive study of provisions in codes and standards with wind tunnel
cedure data.’’ Wind Struct., 1共1兲, 77–109.
Jesien, W., Stathopoulos, T., and Kinh Ha, H. 共1993兲. ‘‘Dynamic along-
G V ⫽1⫹g ␯ ⫻I (33) wind response of building: Comparative study.’’ J. Struct. Eng.,
involving peak factors of 3.7 and 3.5, respectively. 119共5兲, 1498 –1515.
Lee, B. E., and Ng, W. K. 共1988兲. ‘‘Comparisons of estimated dynamic
along-wind responses.’’ J. Wind. Eng. Ind. Aerodyn., 30, 153–162.
Gust Factor for Wind Velocity Pressure Liepmann, H. W. 共1952兲. ‘‘On the application of statistical concepts to the
buffeting problem.’’ J. Aeronaut. Sci., 19共12兲, 793– 800.
Similarly, when neglecting the contribution of the quadratic fluc- Loh, P., and Isyumov, N. 共1985兲. ‘‘Overall wind loads on tall buildings
tuating wind-velocity term, the wind-velocity pressure is also and comparisons with code values,’’ Proc., 5th U.S. National Conf. on
Gaussian. The gust factor for the wind-velocity pressure can then Wind Engineering, R. Dillingham and K. Mehta, eds., Wind Engineer-
be related to the wind velocity GF by 共Solari 1993a兲 ing Research Council, Lubbock, Tex.
G q ⫽2G ␯ ⫺1 (34) NRCC. 共1996兲. ‘‘Commmentary B—wind loads.’’ User’s Guide-NBC
1995 Structural Commentaries, Canadian Commission on Building
This formulation was employed by ASCE7-98 and Eurocode. and Fire Codes, National Research Council of Canada, Ottawa, Part 4,
On the other hand, when simply using the square rule, the 9– 42.
wind-velocity pressure GF can be computed by Simiu, E., and Scanlan, R. 共1996兲. Wind effects on structures: Fundamen-
tals and applications to design, 3rd Ed., Wiley, New York.
G q ⫽G 2␯ (35) Solari, G. 共1993a兲. ‘‘Gust buffeting. I: Peak wind velocity and equivalent
This relationship is used in AS1170.2. The contribution of the pressure.’’ J. Struct. Eng., 119共2兲, 365–382.
quadratic wind velocity was included in AS1170.2 共Vickery Solari, G. 共1993b兲. ‘‘Gust buffeting. II: Dynamic along-wind response.’’
1995兲. A detailed consideration of the contribution of the qua- J. Struct. Eng., 119共2兲, 383–397.
dratic wind velocity can be found in Zhou and Kareem 共unpub- Solari, G., and Kareem, A. 共1998兲. ‘‘On the formulation of ASCE 7-95
gust effect factor.’’ J. Wind. Eng. Ind. Aerodyn., 77 and 78, 673– 684.
lished兲.
Vellozzi, J., and Cohen, E. 共1968兲. ‘‘Gust response factors.’’ J. Struct.
Div., ASCE, 94共6兲, 1295–1313.
Vickery, B. J. 共1970兲. ‘‘On the reliability of gust loading factors.’’ Proc.,
References Technical Meeting Concerning Wind Loads on Buildings and Struc-
tures, Building Science Series 30, National Bureau of Standards,
Architectural Institute of Japan. 共1996兲. Recommendations for loads on Washington, D.C., 296 –312.
buildings, Architectural Institute of Japan, Tokyo. Vickery, B. J. 共1995兲. ‘‘The response of chimneys and tower-like struc-
ASCE. 共1999兲. ‘‘Minimum Design Loads for Buildings and Other Struc-
tures to wind loading.’’ A State of the Art in Wind Engineering, Proc.,
tures.’’ ASCE 7-98, Reston, Va.
9th Int. Conf. on Wind Engineering, Wiley Eastern, New Delhi, India,
‘‘Australian Standards: SAA Loading code, Part 2 - wind loads.’’ 共1989兲.
205–233.
AS1170.2-89, Australia.
Zhou, Y., Gu, M., and Xiang, H. F. 共1999a兲. ‘‘Along-wind static equiva-
Counihan, J. 共1975兲. ‘‘Adiabatic atmospheric boundary layers: A review
lent wind loads and response of tall buildings. Part I: Unfavorable
and analysis of the data from the period 1880–1972.’’ Atmos. Envi-
ron., 9, 871–905. distributions of static equivalent wind loads.’’ J. Wind. Eng. Ind. Aero-
Davenport, A. G. 共1967兲. ‘‘Gust loading factors.’’ J. Struct. Div., ASCE, dyn., 79共1–2兲, 135–151.
93共3兲, 11–34. Zhou, Y., Gu, M., and Xiang, H. F. 共1999b兲. ‘‘Along-wind static equiva-
Durst, C. D. 共1960兲. ‘‘Wind speeds over short periods of time.’’ Meteorol. lent wind loads and response of tall buildings. Part II: Effects of mode
Mag., 89, 181–186. shape.’’ J. Wind. Eng. Ind. Aerodyn., 79共1–2兲, 151–158.
Eurocode 1. 共1995兲. ‘‘Basis of Design and Actions on Structures–Part Zhou, Y., and Kareem, A. 共2001兲. ‘‘Gust loading factor: A new model.’’ J.
2-4: Actions on Structures–Wind actions.’’ European Prestandard Struct. Eng., 127共2兲, 168 –175.
ENV 1991-2-4. Zhou, Y., and Kareem, A. 共2002兲. ‘‘On the wind profiles definitions in
Ferraro, V., Irwin, P. A., and Stone, G. K. 共1989兲. ‘‘Wind-induced build- ASCE 7.’’ J. Struct. Eng., in press.
ing accelerations.’’ Proc., 6th U.S. National Conf. on Wind Engineer- Zhou, Y., Kareem, A., and Gu, M. 共1999c兲. ‘‘Gust loading factors for
ing, A. Kareem, ed., Univ. of Houston, Houston. design applications.’’ Wind Engineering into the 21st Century, Proc.,
Greenway, M. E. 共1979兲. ‘‘An analytical approach to wind velocity gust 10th Int. Conf. on Wind Engineering, A. Larsen, G. L. Larose, and F.
factors.’’ J. Indust. Aerodyn., 5, 61–91. M. Livesey, eds., A. A. Balkema, Rotterdam, 169–176.
Holmes, J. D., Melbourne, W. H., and Walker, G. R. 共1990兲. ‘‘A commen- Zhou, Y., Kareem, A., and Gu, M. 共2000兲. ‘‘Equivalent static buffeting
tary on the Australian Standard for wind loads AS1170 part 2, 1989.’’ loads on structures.’’ J. Struct. Eng., 126共8兲, 989–992.

796 / JOURNAL OF STRUCTURAL ENGINEERING / JUNE 2002