You are on page 1of 13

Ultimate Wind Load Design Gust Wind Speeds

in the United States for Use in ASCE-7


Peter J. Vickery1; Dhiraj Wadhera2; Jon Galsworthy3; Jon A. Peterka4; Peter A. Irwin5; and
Lawrence A. Griffis6
Downloaded from ascelibrary.org by Universidad Politecnica De Valencia on 06/09/15. Copyright ASCE. For personal use only; all rights reserved.

Abstract: This paper presents an overview of the approach employed in the development of the wind speed maps for use in ASCE 7-10.
The reason for a reduction in the wind speeds in the new standard as compared to those given in the ASCE 7-98 through 7-05 standards
is presented, as well as the reason for the reintroduction of Exposure D along the hurricane coastline. The most significant change in the
wind speed maps from the previous version is the shift from a single map for an importance factor for buildings and other structures of
1.0 to three separate maps, one for each category of occupancy, thus eliminating the need for importance factors that vary between
hurricane and nonhurricane regions.
DOI: 10.1061/共ASCE兲ST.1943-541X.0000145
CE Database subject headings: Hurricanes; Risk management; Wind loads; Structural design; Wind speed; Wind gusts.
Author keywords: Hurricane; Risk; Wind loads; ASCE-7; Design.

Introduction roughness of sea surface does not continue to increase with in-
creasing wind speed, and thus Exposure D should be used along
Predicted wind speeds resulting from the hurricane hazard model the hurricane coastline.
described in Vickery et al. 共2009a,b兲 are used to develop design
wind speed maps for the United States. The hurricane wind
speeds are combined with the statistical model for nonhurricane Review of Major Changes in Hurricane Hazard
wind speeds given in Peterka and Shahid 共1998兲 to develop maps Model
that consider both hurricane and nonhurricane winds. We note
that the wind speeds produced by the hurricane hazard model The details of the development and validation of the hurricane
described in Vickery et al. 共2009a,b兲 are lower than those ob- simulation model used to derive the wind speeds presented here
tained from the Vickery et al. 共2000a,b兲 model. The reasons be- are given in Vickery et al. 共2009a,b兲, and are not repeated here.
hind the lower estimated wind speeds are discussed. Following Here we discuss some of the major differences between the char-
the procedures discussed in the Commentary of ASCE 7 共ASCE acteristics of the modeled storms at the time of landfall as pre-
2005兲, we develop wind speed maps for return periods of 300 dicted using the Vickery et al. 共2000a,b兲 model 共used to define the
years 共Category I Buildings兲, 700 years 共Category II Buildings兲 design wind speeds in ASCE 7-98 through ASCE 7-05兲 to those
and 1700 years 共Category III and IV buildings兲 to be used for the predicted using the current model 共Vickery et al. 2009a,b兲, and
wind design of buildings and other structures in conjunction with their effect on wind speeds.
a wind load factor of 1.0. The paper also discusses recent research
examining the behavior of the sea surface drag coefficient at high
wind speeds in hurricanes that indicates that the aerodynamic Holland B Parameter
As discussed in Vickery et al. 共2009b兲, the primary reason for the
1
Principal Engineer, Applied Research Associates, Inc., 8357 Six reduction in the wind speeds produced by the 2009 model com-
Forks Rd., Suite 600, Raleigh, NC 27615 共corresponding author兲. E-mail: pared to the model used to develop the design wind speed maps
pvickery@ara.com given in ASCE 7-98 through ASCE 7-05, is due to the modeling
2
Senior Scientist, Applied Research Associates Inc., 8357 Six Forks of the Holland B 共Holland 1980兲 parameter. The Holland B pa-
Rd., Suite 600, Raleigh, NC 27615. rameter is used to define the pressure field as
3
Project Director, RWDI Inc., 650 Woodlawn Rd. West, Guelph, ON,
Canada N1K 1B8.

冋册
4
Principal, CPP Inc., 1415 Blue Spruce Dr., Fort Collins, CO 80524.
5 A
Chairman, RWDI Inc., 650 Woodlawn Rd. West, Guelph, ON, p共r兲 = pc + ⌬p ⫻ exp − 共1兲
Canada N1K 1B8. rB
6
President, Structures Div., Walter P Moore and Associates, Inc., 221
West 6th St., Austin, TX 78701.
where p共r兲 = surface pressure at a distance r from the storm cen-
Note. This manuscript was submitted on September 9, 2008; approved
on November 2, 2009; published online on November 4, 2009. Discus- ter; pc = central pressure; ⌬p = difference between the peripheral
sion period open until October 1, 2010; separate discussions must be pressure and the central pressure; A = location parameter; and B
submitted for individual papers. This paper is part of the Journal of = Holland’s pressure profile parameter. Holland 共1980兲 showed
Structural Engineering, Vol. 136, No. 5, May 1, 2010. ©ASCE, ISSN that the radius to maximum winds, RMW= A1/B, and thus Eq. 共1兲
0733-9445/2010/5-613–625/$25.00. can be expressed as

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2010 / 613

J. Struct. Eng. 2010.136:613-625.


Downloaded from ascelibrary.org by Universidad Politecnica De Valencia on 06/09/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Example modeled and observed pressure profiles 共left-hand plots兲 corresponding to computed gradient balance wind speeds and measured
wind speeds 共right-hand plots兲

614 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2010

J. Struct. Eng. 2010.136:613-625.


Downloaded from ascelibrary.org by Universidad Politecnica De Valencia on 06/09/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Comparisons of B values derived from landfalling storm re-


construction studies 共wind modeling兲 to the statistical models given
in Vickery and Wadhera 共2008b兲 and Vickery et al. 共2000b兲
Fig. 4. Effect of Holland B modeling on predicted wind speeds

p共r兲 = pc + ⌬p ⫻ exp − 冋 册
RMW
r
B
共2兲

The gradient balance velocity, VG, for a stationary storm may then
be shown to be

VG = 冤冉 RMW
冊 B B⌬p 冋 冉 冊册
exp −
RMW
r
B

+
2 2
r f

1/2


fr
共3兲
r ␳ 4 2
where ␳ = density of air; f = Coriolis parameter equal to 2⍀ sin ␸,
where ⍀ is the earth’s angular velocity; and ␸ = latitude. The
maximum gradient wind speed at the RMW is

VG max ⬇ 冑 B⌬p
e␳
共4兲

and e = 2.718 and is the base of natural logarithms. A review of


the statistical model used to define B and the importance of B in
estimating hurricane wind speed risk is discussed in the following
sections.

Holland B Modeling: Comparison of Statistical Models


B Methods of Estimating B
As indicated above, and in Holland 共1980兲, Georgiou 共1985兲, and
Vickery and Wadhera 共2008b兲, among others, the maximum wind
speed, Vmax in a stationary hurricane is proportional to 冑B⌬p.
Thus, with all else equal, a 10% decrease in B corresponds to a
5% decrease in the maximum wind speed in the hurricane 共for a
stationary hurricane兲. As discussed in Vickery and Wadhera
Fig. 3. Comparison of mean 共upper map兲 and 99th percentile 共lower 共2008b兲, B is correlated with RMW 共decreasing RMW yields in-
map兲 values of the Holland B parameter derived from a 100,000-year creasing B兲, latitude 共B decreases with increasing latitude兲 and sea
simulation of hurricanes. Left number indicates the value derived surface temperature 共SST兲 共B decreases with decreasing SST兲. In
from the Vickery et al. 共2000b兲 model given here in Eq. 共6兲. Numbers an attempt to incorporate the relationship between latitude, SST,
to the right from the Vickery and Wadhera 共2008b兲 model, given here RMW, and B in a single parameter model, Vickery and Wadhera
as Eq. 共5兲. 共2008b兲 model B over open water as

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2010 / 615

J. Struct. Eng. 2010.136:613-625.


Downloaded from ascelibrary.org by Universidad Politecnica De Valencia on 06/09/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. Comparison of modeled and observed annual landfall rates of IH 共simulated and observed, 1900 through 2007, IH intensity based on
hurricane central pressure at time of landfall兲

B = 1.7642 − 1.2098冑A + ␧; r2 = 0.345,␴B = 0.226 共5a兲 Florida Peninsula兲 are often 共but not always兲 associated with a
reduction in B over the last 12–24 h before landfall. This apparent
RMW ⫻ f reduction in the magnitude of B from the nominal overwater val-

冑 冉 冊
A= 共5b兲 ues was not incorporated in the hurricane model used to develop
⌬p the wind speeds presented here and in Vickery et al. 共2009b兲,
2RdTs ⫻ In 1 +
pc ⫻ e primarily because it was not clear that this reduction in B would
necessarily occur for small intense hurricanes such as Hurricane
where r2 = fraction of the variance explained by the model; ␴B
Camille 共1969兲. The reduction in B was not evident for hurricanes
= standard deviation of the error associated with the prediction
making landfall along the Atlantic or Florida Peninsula coastlines.
model; and ␧ = random error term. The numerator of A = product
Another approach for developing a model for B is to use the
of the radius to maximum winds 共RMW in meters兲 and the Cori-
values of B used in the reconstruction of the wind fields associ-
olis parameter, f. The denominator of A is an estimate of the
ated with landfalling storms 共Vickery et al. 2009a兲. Here we use
maximum potential intensity 共as defined by wind speed兲 of a hur-
the B values associated with the reconstruction of wind fields
ricane. Rd is the gas constant for dry air, and Ts is the sea surface
from 31 different hurricane-landfall location combinations and
temperature in degrees C. Both the numerator and denominator of
compare these data to the statistical model described in Eq. 共5兲.
A have the units of velocity, and A is nondimensional. This model
The presentation of the comparison of the B values derived from
for B is used in Vickery et al. 共2009b兲 to estimate hurricane wind
the two methods that follows is an extension of that given in
speeds in the United States and subsequently those presented
Vickery and Wadhera 共2008b兲. Here the landfall set has been
here. The models for B described in Vickery and Wadhera
extended by five more hurricanes 共1983 Hurricane Alicia, 1991
共2008b兲 and implemented in the hurricane wind speed simulation
Hurricane Bob in New England, 1999 Hurricane Bret, and 2008
model described in Vickery et al. 共2009b兲 were developed using
Hurricanes Ike and Gustav兲. The comparisons are given in two
fits to hundreds of radial profiles of surface pressures derived
groups, with the first group comprised of Florida Peninsula and
from reconnaissance aircraft flights obtained from 37 different
Atlantic Coast landfalling hurricanes and the second comprised of
hurricanes in the Atlantic Ocean and Gulf of Mexico. Fig. 1 pre-
Gulf of Mexico landfalling hurricanes. In both groups, B values
sents some examples of the fits to the radial pressure profiles. In
Fig. 1, the fits are made using the pressure data 共left hand plots兲,
minimizing the square of the differences between the modeled
关Eq. 共2兲兴 and observed pressures over the range 0.5 RMW to 1.5
RMW. The plots on the right hand side are simply the wind
speeds obtained from Eq. 共3兲, using the estimated values of B and
RMW from the pressure profile and compared to the measured
winds. All wind speeds are given in storm relative coordinates
such that the effect of storm translation has been removed.
Note that the model used in Vickery et al. 共2000b兲 for B is
given as
B = 1.38 + 0.001 84⌬p − 0.003 09RMW + ␧;
r2 = 0.026,␴B = 0.381 共6兲
where RMW is expressed in kilometers, was also developed using
aircraft from reconnaissance data, but they used wind speed pro-
files instead of pressure profiles to estimate B and data was avail- Fig. 6. Gust wind speed ratio 共gust wind speed over water divided by
able from only 15 hurricanes. gust wind speed over land兲. A gust wind speed ratio of 1 indicates no
As discussed in Vickery and Wadhera 共2008b兲, hurricanes reduction in wind speed as wind moves from sea to land 共i.e., marine
making landfall along the Gulf of Mexico coastline 共excluding the roughness is the same as open terrain兲.

616 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2010

J. Struct. Eng. 2010.136:613-625.


Downloaded from ascelibrary.org by Universidad Politecnica De Valencia on 06/09/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 7. Three-hundred-year return period peak gust wind speeds at a height of 10 m in open terrain

derived using Eq. 共6兲 共Vickery et al. 2000b兲 and those from Eq. tion noted in Vickery and Wadhera 共2008b兲 that B associated with
共5兲 共Vickery and Wadhera 2008b兲 are also presented. Gulf of Mexico Hurricanes often decreases in the last 12–24 h
Fig. 2 presents comparisons of the B values derived from the before landfall. Eqs. 共5兲 and 共6兲 overestimate the mean B derived
three approaches 共wind field modeling, Eqs. 共5兲 and 共6兲兲. The from the landfall analyzes of Gulf of Mexico hurricanes by 0.27
upper graph in Fig. 2 presents the comparison of the B values and 0.39, respectively. Note that as shown in Vickery and Wad-
derived from the three approaches for the 18 Atlantic Coast and hera 共2008b兲, B did not reduce before landfall in the case of
Florida Peninsula hurricanes. The B values associated with each hurricanes Elena 共19585兲 and Frederic 共1979兲. Hurricane Ike, the
of the 18 hurricanes predicted using Eqs. 共5兲 and 共6兲 are presented other apparent outlier shown in Fig. 2, however; did exhibit a
along with their 95% confidence bounds 共i.e., model mean⫾ 1.96 reduction in B before landfall reducing from about 1.3 to 1.06
␴B兲. The linear regression model resulting from a regression of B over the last 12 to 24 h before landfall. The comparisons indicate
derived from the wind modeling analysis as a function of 冑A is
that the design wind speeds along the Gulf of Mexico coastline
also shown. Note that the B values fit within the 95% confidence
presented in this paper and in Vickery et al. 共2009b兲 may be
bounds from either statistical model, but the mean B derived from
conservative, however; it is not clear if this reduction in B will
Eq. 共6兲 has a high bias of 0.16, whereas the mean B computed
using Eq. 共5兲 is unbiased. The lower graph in Fig. 2 presents the take place for the smaller intense hurricanes that contribute to the
comparison of the B values derived from the three approaches for rare 共⬎500-year return period兲 events.
the 13 Gulf of Mexico hurricanes. The B values associated with Note that the relatively low correlation 共low r2兲 between B and
each of the 13 hurricanes predicted using Eqs. 共5兲 and 共6兲 are A is consistent with the relatively low, but important, correlation
presented along with their 95% confidence bounds. The linear between RMW and central pressure presented in Vickery and
regression model resulting from a regression of B from the wind Wadhera 共2008b兲, and in most other models 共e.g., Batts et al.
modeling analysis as a function of 冑A is also shown, although the 1980; Georgiou 1985; Powell et al. 2005兲. These low correlations
negative slope noted in the graph is not statistically significant. are a result of both a lack of understanding and an inability to
Here the B values derived from the landfall analyzes do not all fit model the complex physics of hurricane development and inten-
within the 95% confidence bounds from model from Eq. 共5兲, and sification, including the effects of variations in sea surface tem-
there is a clear bias for both models to overestimate the B values perature, humidity, wind shear, etc., over the domain of the storm
derived from the landfall analyzes, consistent with the observa- in addition to size of the initial disturbance. This lack of under-

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2010 / 617

J. Struct. Eng. 2010.136:613-625.


Downloaded from ascelibrary.org by Universidad Politecnica De Valencia on 06/09/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. Seven-hundred-year return period peak gust wind speeds at a height of 10 m in open terrain

standing is clearly indicated in the inability to accurately forecast Fig. 4 presents a comparison of predicted hurricane wind
changes in hurricane intensity and size e.g., DeMaria and Kaplan speeds as a function of return period for locations in Wilmington,
共1999兲. North Carolina and Miami derived using B modeled with Eqs. 共5兲
and 共6兲. The wind speed versus return period curves clearly show
the effect of the modeling of B on the predicted wind speeds. In
the Wilmington, N.C. case, the use of the revised model for B
Holland B Modeling—Effect on Predicted Wind Speeds
reduces the estimated 100-year return period peak gust wind
Fig. 3 presents a comparison of the mean and 99th percentile speed from 128 to 116 mph, a 9.4% reduction in wind speed. In
values of B at the time of landfall derived from a 100,000-year the Miami case, the 100-year return period wind speed is reduced
simulation with B computed using Eqs. 共5兲 and 共6兲. As indicated from 148 to 138 mph, a 6.8% reduction in wind speed. For a
in Fig. 3 there is an overall decrease in the mean value of B along 700-year return period, modeling B using Eq. 共5兲 instead of Eq.
the entire hurricane prone coastline using Eq. 共5兲 共current model兲 共6兲 results in 10.9% and 9.3% reductions in wind speeds at Wilm-
compared to using Eq. 共6兲 共previous model兲. As an example, con- ington and Miami, respectively.
sider the mean values of B derived from Eqs. 共5兲 and 共6兲 at
Wilmington NC of 1.01 and 1.23, respectively, indicating an 18%
reduction in the mean value of B, corresponding to a 9% reduc- Landfall Intensity Defined by Central Pressure
tion of the maximum wind speed 共for a stationary storm兲. In the
case of the 99th percentile values of B a 24% reduction in B is Fig. 5 presents a comparison of the landfall frequencies of intense
evident, corresponding to a 12% reduction in wind speed 共for a hurricanes 共IH兲 as defined by central pressure at the time of land-
stationary storm兲. The relative importance of changes in B de- fall derived from the Vickery et al. 共2009b兲 model, the Vickery et
creases as the translation speed of a hurricane increases, thus for al. 共2000b兲 model, and the historical data. The historical data
hurricanes in the New England area, where the average transla- covers the period 1900 through 2007. As indicated in Fig. 5, the
tion speed of a hurricane is about 12 m/s, the reduction in B has newer model yields more IH landfalls per year than does the
less effect on the maximum wind speed than for a hurricane in the Vickery et al. 共2000a,b兲 model. In all cases the modeled landfall
Gulf of Mexico where the average translation speed of a hurri- rates from either model fit within the 95% uncertainty bounds
cane is about 6 m/s. associated with the observations.

618 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2010

J. Struct. Eng. 2010.136:613-625.


Downloaded from ascelibrary.org by Universidad Politecnica De Valencia on 06/09/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 9. One-thousand-seven-hundred-year return period peak gust wind speeds at a height of 10 m in open terrain

waves, then the behavior of Cd as a function of wind speed near


Sea-Land Transition and the Need for Exposure D the coast would be expected to be similar to that observed in the
One of the more significant differences between the hurricane open ocean.
wind field model used in Vickery et al. 共2000a兲 compared to that The model for Cd as implemented in Vickery et al. 共2009a,b兲 is
used in Vickery et al. 共2009b兲 is the modeling of the sea surface Cd = 共0.49 + 0.065U10兲10−3 ; Cd ⱕ Cdmax 共7a兲
drag coefficient. In Vickery, et al. 共2009a兲 the boundary layer
model used within the hurricane wind field model uses a repre-
sentation of the sea surface drag coefficient, Cd, which reaches a Cdmax = 共0.0881r + 17.66兲10−4 ; 0.0019 ⱕ Cdmax ⱕ 0.0025
maximum value at mean wind speeds over water of around 20 共7b兲
m/s. This treatment of the sea surface drag coefficient varies
markedly from that used in Vickery et al. 共2000a兲, where Cd in- where U10 = mean wind speed 共m/s兲 at a height of 10 m and
creased monotonically with wind speed. The change in the mod- r = radial distance from the storm center 共km兲, but r is constrained
eling of the drag coefficient is due to the work described in to have a minimum value equal to the RMW. In Vickery et al.
Powell et al. 共2003兲. A limiting value of Cd contradicts the find- 共2000a兲, Cd is modeled as
ings given in the earlier works of Powell 共1980兲 and Garratt
Cd = 共1.1 + 0.040U10兲10−3 共8兲
共1977兲, but is supported by the recent work such as Donelan et al.
共2004兲, and Black et al. 共2007兲. Powell et al. 共2003兲 note that the When implemented in the hurricane wind field model de-
data used to develop the drag coefficient are associated with scribed in Vickery et al. 共2009a兲, Eq. 共7兲 yields ratios of the
waves in the open ocean, and it is possible that the behavior of Cd overwater to overland gust wind speed that vary with storm size
as a function of wind speed near the coast 共where shoaling and and intensity as well as location within the storm. Fig. 6 presents
breaking waves occur兲 differs from that in the open ocean. How- the water-land gust wind speed ratios derived from the Vickery et
ever, if as suggested by Makin 共2005兲, the limiting drag is caused al. 共2009a兲 wind field model 共near the RMW兲, the ratios derived
by sea spray inhibiting momentum transfer as a result of sea spray from the Vickery et al. 共2000a兲 model 共also near the RMW兲, and
produced by the mechanical tearing by the wind from short steep those implied in ASCE 7. The wind speed ratio is defined as the

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2010 / 619

J. Struct. Eng. 2010.136:613-625.


Downloaded from ascelibrary.org by Universidad Politecnica De Valencia on 06/09/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 10. Ten-year return period peak gust wind speeds at a height of 10 m in open terrain

marine gust wind speed divided by the open terrain 共defined as speed versus return period curves in hurricane prone and nonhur-
zo = 0.03 m兲 gust wind speed. Thus, for Exposure D, the ratio is ricane prone regions differ.
equal to 冑Kz where Kz is evaluated at z = 10 m 共32 ft兲. As indi- The approach taken to develop return periods associated with
cated in Fig. 6 the wind speed ratios produced by the new wind different importance factors 共occupancy categories兲 began with
field model are approximately consistent with the ratio associated the premise that the nominal wind load computed using the meth-
with Exposure D for all wind speeds, particularly for larger values ods given in ASCE 7-05 and earlier, when multiplied by the wind
of RMW, whereas for large wind speeds, the wind speed ratio load factor, was representative of an “ultimate” load. Further-
associated with the older model, using a drag coefficient that does more, it was assumed that the variability of the wind speed domi-
not have an upper limit, the wind speed ratio approaches that of nates the calculation of the wind load factor. The assumption that
Exposure C in ASCE 7. the variability of the wind speed dominates the wind load is an
approximation, and depends on the assumptions of the uncertainty
in the various components of the wind loading chain 共Davenport
1983兲. The wind speed uncertainty is combined with uncertainties
Return Periods for Design with a Wind Load Factor in pressure coefficient, exposure factor, gust factor and modeling
of 1.0 error as sums of squares, and because the uncertainty effect of the
wind speed is doubled since it appears in the wind load equation
In the development of the design wind speed map used in ASCE as a pressure, the net result is the wind speed uncertainty contrib-
7-98 the Wind Load Task Committee 共WLTC兲 revisited the hur- utes in the range of 70–80% of the total uncertainty.
ricane importance factor that had been in use in the U.S. Stan- The “ultimate” strength wind load, WT, is given as
dards since 1982. The primary reasons for revisiting the hurricane
importance factor, IH, was the recognition that the importance WT = CF共V50Iv兲21.6 共9兲
factor likely varied with location along the coast and using a
constant value of 1.05 was not appropriate. Note that the hurri- where CF = building/component/structure specific coefficient that
cane importance factor is not the same as the importance factor includes the effects building height, building geometry, terrain,
applied to the design of Category I, III, and IV buildings, but is a gust factor, etc., as computed using the procedures outlined in
factor that takes into account the fact that the slope of the wind ASCE 7, V50 is the 50-year return period design wind speed, 1.6

620 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2010

J. Struct. Eng. 2010.136:613-625.


Downloaded from ascelibrary.org by Universidad Politecnica De Valencia on 06/09/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 11. Twenty-five-year return period peak gust wind speeds at a height of 10 m in open terrain

is the wind load factor, and Iv is the importance factor applied to VT/V50 = 关0.36 + 0.1 In共12T兲兴 = 冑1.6 共12兲
the wind speed.
In order to estimate the magnitude of the hurricane importance and from Eq. 共12兲, the return period T associated with the ulti-
factor, IH, the committee required that the annual probability of mate wind speed in the nonhurricane prone portion of the United
exceeding the ultimate wind load in the hurricane and nonhurri- States is
cane regions of the United States should be the same. Recalling
that the nominal design wind speed in the nonhurricane regions of T = 0.002 28 exp共10冑1.6兲 = 709-years 共13兲
the United States is associated with a return period of 50-years,
the WLTC sought to determine the return period associated with Therefore, Vdesign = V709 / 冑1.6⬇ V700 / 冑1.6.
the wind speed producing the ultimate load in a representative The importance factor used in ASCE 7-05 and earlier for the
nonhurricane prone region. As defined in ASCE 7-98 through computation of wind loads for the design of Categories III and IV
ASCE 7-05, over most of the nonhurricane prone coastline of the structures is defined so that the nominal 50-year return period
United States, the ratio of the wind speed for any return period, T, nonhurricane wind speed is increased to be representative of a
to the 50-year return period wind speed can be computed from 100-year return period value. The importance factor is applied to
共Peterka and Shahid 1998兲 the design of all Categories III and IV buildings whether or not
they are located in a hurricane prone region. Following the ap-
proach used above to estimate the resulting ultimate load return
VT/V50 = 关0.36 + 0.1 In共12T兲兴 共10兲
period associated with the 50-year design wind speed we find that
for the 100-year return period case
where VT = T-year return period wind speed. In the nonhurricane
prone regions of the United States, the ultimate wind load, WT,
occurs when T = 0.002 28 exp关10共V100/V50兲冑1.6兴 共14兲
where for V100 / V50 computed from Eq. 共12兲 we find that T
WT = CFVT2 = CF1.6V250 共11兲 = 1 , 697 years. In the development of Eq. 共14兲, the term
共V100 / V50兲冑1.6 replaces the 冑1.6 used in Eq. 共13兲, effectively re-
thus sulting in a higher load factor for Categories III and IV structures

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2010 / 621

J. Struct. Eng. 2010.136:613-625.


Downloaded from ascelibrary.org by Universidad Politecnica De Valencia on 06/09/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 12. Fifty-year return period peak gust wind speeds at a height of 10 m in open terrain

equal to WLF共V100 / V50兲2. Thus for Category III and IV structures, speeds that are nominally constant over each of the four regions
a design wind speed of V1700 / 冑1.6 is appropriate. Similarly, the and then combining the wind speeds from hurricanes and nonhur-
25-year return period wind speed associated Category I buildings, ricanes using Eq. 共15兲 and then developing the contours, while an
equates to a 300-year return period wind speed with a load factor approximation, is an improvement over the subjective approach
of 1.0. used to define the location of the hurricane-nonhurricane line used
In order to develop design wind speed maps, hurricane and in the preparation of the final wind speed map given in ASCE
nonhurricane wind speeds were combined together as statistically 7-98 through ASCE 7-05.
independent events as Design wind speed maps to be used with a WLF of 1.0 are
given in Figs. 7–9 for return periods of 300, 700 and 1700 years
P共V ⬎ v兲 = 1 − P共VNH ⬍ v兲P共VH ⬍ v兲 共15兲
respectively. Figs. 10–13 present contour maps associated with
where P共VNH ⬍ v兲 = annual probability that the nonhurricane wind wind speeds with return periods of 10, 25, 50, and 100 years.
speed is less than v, P共VH ⬍ v兲 is the annual probability that the Wind speeds for Hawaii have been obtained from the Hurricane
hurricane wind speed is less than v, and P共V ⬎ v兲 is the annual Hazard mitigation grant program study reported in Applied Re-
probability that the wind speed is greater than v considering the search Associates, Inc. 共2001兲. Fig. 14 shows the hurricane and
contribution from either meteorological phenomena. Following nonhurricane wind speeds for Hawaii plotted versus return period,
the methodology described in Peterka and Shahid 共1998兲, the ex- taken from Applied Research Associates, Inc. 共2001兲. Note that
ceedance probabilities associated with the nonhurricane winds Fig. 14 is a also a good example demonstrating the combination
were computed using four different “minisuperstations” for hur- of hurricane and nonhurricane winds, where for Hawaii for return
ricane prone areas affecting the Western Gulf coast area 共Texas periods less than ⬃30 years, nonhurricane winds dominate the
and Louisiana兲, the mid-Gulf and Florida Peninsula, and the wind climate. Wind speeds for the Puerto Rico and the U.S. Vir-
Southeast and Northeast Atlantic coastlines. A superstation is de- gin Islands are given in Vickery and Wadhera 共2008a兲. The ty-
fined as a group of relatively closely spaced meteorological sta- phoon simulation study for Guam uses the hurricane simulation
tions whose yearly maxima are produced by different storms, and model described in Vickery and Twisdale 共1995兲 coupled with the
can therefore be combined to produce effective record lengths wind field model described in Vickery et al. 共2009a兲. Typhoon
longer than associated a single station. For example, two stations radii to maximum winds and the typhoon pressure profile param-
each with 30 years of data can be combined together to produce eter B are modeled as described in Vickery and Wadhera 共2008b兲.
one superstation with 60 years of data. Using nonhurricane wind The statistics of storm intensity, heading and translation speed

622 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2010

J. Struct. Eng. 2010.136:613-625.


Downloaded from ascelibrary.org by Universidad Politecnica De Valencia on 06/09/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 13. One-hundred-year return period peak gust wind speeds at a height of 10 m in open terrain

have been derived using information on storms passing within In the case of American Samoa no published studies are cur-
250 km of Guam using the Hong Kong Observatory typhoon rently available to support a change in the design wind speeds.
database encompassing the period 1961 through 2000. The wind speeds for American Samoa presented herein have been
In Figs. 7–9, the wind speeds given for Alaska have been derived by converting the 50-year values given in the prior
developed using the wind speeds given in ASCE 7-05, and ad- version of ASCE 7 共Versions 95 through 05兲 using the conversion
justed using the factors given in Table 1, which are derived from factors given in Table 2. Values given in bold in Table 2
the conversion factors for Alaska wind speeds given in ASCE are interpolated or extrapolated using the other values given in
7-05. Values given in bold in Table 1 are interpolated or extrapo- Table 2.
lated from the values given in Table C6-7 of ASCE 7-05.

Table 1. Factors Used to Adjust 50-Year Return Period Wind Given in


ASCE 7 to Other Return Period Values for Alaska
Return period 共years兲 Conversion factor
5 0.81
10 0.87
25 0.94
50 1.00
100 1.06
200 1.12
300 1.15
500 1.18
700 1.21
1,700 1.29
Fig. 14. Wind speed versus return period in Hawaii 关adapted from
Note: Values in bold are interpolated or extrapolated, other values taken
Applied Research Associates Inc. 共2001兲兴 directly from Table C6-5 in ASCE 7 95.

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2010 / 623

J. Struct. Eng. 2010.136:613-625.


Table 2. Factors Used to Adjust 50-Year Return Period Wind Given in Table 3. Comparison of Design Wind Speeds with ASCE 7-05 Values for
ASCE 7 to Other Return Period Values for American Samoa Selected Locations
Return period 共years兲 Conversion factor V700 / 冑1.6
ASCE 7-05
10 0.84 Location Exposure C Exposure C Exposure D
25 0.93
Bar Harbor, Me. 97 95 103
50 1.00
Boston 106 103 112
100 冑1.15= 1.07
Hyannis, Mass. 117 112 122
300 1.18
New Port, R.I. 117 109 119
700 冑1.6= 1.26 Southampton, N.Y. 120 110 119
1,700 冑1.15⫻ 冑1.6= 1.36 Atlantic City, N.J. 114 102 111
Downloaded from ascelibrary.org by Universidad Politecnica De Valencia on 06/09/15. Copyright ASCE. For personal use only; all rights reserved.

Note: Values in bold are interpolated or extrapolated, other values are Wrightsville Beach, N.C. 132 119 129
computed using the square root of the load factor times the hurricane Folly Beach, S.C. 131 115 125
importance factor from ASCE 7 02 as shown.
Miami Beach, Fla. 145 136 148
Clearwater, Fla. 128 115 125
Comparison of ASCE 7-05 Design Wind Speeds to Panama City, Fla. 129 107 116
Wind Speeds Presented herein Biloxi, Miss. 138 129 140
Galveston, Tex. 131 119 129
Table 3 presents a comparison of the nominal design wind speeds Port Aransas, Tex. 134 117 127
presented here to those given in ASCE 7-05. For the purpose of Hawaii 105 103 112
comparing the wind speeds we use the 700-year wind speeds Guam 170 155 168
from Fig. 8 and divided by 冑1.6 so that we are comparing equiva-
lent design wind speeds. For coastal locations, we present both
the Exposure C wind speeds and the Exposure D wind speeds.
The Exposure D wind speeds are defined as the basic wind speed

Fig. 15. “Design” wind speed map defined as V700 / 冑 1.6

624 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2010

J. Struct. Eng. 2010.136:613-625.


Batts, M. E., Cordes, M. R., Russell, L. R., Shaver, J. R., and Simiu, E.
共1980兲. “Hurricane wind speeds in the United States.” Report No.
BSS-124, National Bureau of Standards, U.S. Dept. of Commerce,
Washington, D.C.
Black, P. G., et al. 共2007兲. “Air-sea exchange in hurricanes: Synthesis of
observations from coupled boundary layer air-sea transfer experi-
ment.” Bull. Am. Meteorol. Soc., 88, 357–374.
Davenport, A. G. 共1983兲. “The relationship of reliability to wind load-
ing.” J. Wind. Eng. Ind. Aerodyn., 13, 3–27.
DeMaria, M., and Kaplan, J. 共1999兲. “An updated statistical hurricane
intensity prediction scheme 共SHIPS兲 for the Atlantic and Eastern
North Pacific basins.” Weather Forecast., 14, 326–337.
Donelan, M. A., et al. 共2004兲. “On the limiting aerodynamic roughness in
Downloaded from ascelibrary.org by Universidad Politecnica De Valencia on 06/09/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 16. Variation of importance factor 共V1,700 / V700兲2 the ocean in very strong winds.” Geophys. Res. Lett., 31, L18306.
Garratt, J. R. 共1977兲. “Review of drag coefficients over oceans and con-
tinents.” Mon. Weather Rev., 105, 915–929.
multiplied by 冑Kz where Kz is computed at a height of 10 m. As Georgiou, P. N. 共1985兲. “Design windspeeds in tropical cyclone-prone
indicated in Table 3, there is a decrease in the Exposure C “de- regions.” Ph.D. thesis, Faculty of Engineering Science, Univ. of West-
sign” wind speeds at most locations; however, near the hurricane ern Ontario, London, Ont., Canada.
prone shoreline, because of the introduction of Exposure D, the Holland, G. J. 共1980兲. “An analytical model of the wind and pressure
design wind speeds are similar to those presented in ASCE 7-05. profiles in hurricanes.” Mon. Weather Rev., 108, 1212–1218.
Makin, V. K. 共2005兲. “A note on the drag of the sea surface at hurricane
Fig. 15 presents contour maps of V700 / 冑1.6 that can be used to winds.” Boundary-Layer Meteorol., 115, 169–176.
compare the wind speeds directly with those given in ASCE 7-05. Peterka, J. A., and Shahid, S. 共1998兲. “Design gust wind speeds in the
Fig. 16 presents the distribution of the importance factor, I, for United States.” J. Struct. Eng., 124, 207–214.
Categories III and IV buildings and structures, defined as Powell, M. D., et al. 共2005兲. “State of Florida hurricane loss projection
共V1,700 / V700兲2 computed at each of the 2,859 points used to de- model: Atmospheric science component.” J. Ind. Aerodyn, 93, 651–
velop the wind speed maps. The mean value of 共V1,700 / V700兲2 is 674.
1.15 varying from about 1.1 through to 1.24. Powell, M. D. 共1980兲. “Evaluations of diagnostic marine boundary-layer
models applied to hurricanes.” Mon. Weather Rev., 108, 757–766.
Powell, M. D., Vickery, P. J., and Reinhold, T. A. 共2003兲. “Reduced drag
Summary coefficients for high wind speeds in tropical cyclones.” Nature, 422,
279–283.
Vickery, P. J., Skerlj, P. F., Steckley, A. C., and Twisdale, L. A. 共2000a兲.
An overview of the approach used to develop ultimate wind speed
“Hurricane wind field model for use in hurricane simulations.” J.
maps for use in ASCE 7 has been presented. Reasons for differ- Struct. Eng., 126共10兲, 1203–1221.
ences between the hurricane wind speeds in the proposed map Vickery, P. J., Skerlj, P. F., and Twisdale, L. A., Jr. 共2000b兲. “Simulation
and the maps used in ASCE 7-98 through ASCE 7-05 are re- of hurricane risk in the U.S. using an empirical track model.” J.
viewed, including a detailed evaluation and discussion of the im- Struct. Eng., 126共10兲, 1222–1237.
pact of the Holland B parameter on predicted wind speeds. These Vickery, P. J., and Twisdale, L. A. 共1995兲. “Wind field and filling models
changes include recent findings on the surface roughness of open for hurricane wind speed predictions.” J. Struct. Eng., 121, 1700–
water and changes within the source hurricane simulation model. 1709.
The primary change in the wind speed maps from the previous Vickery, P. J., and Wadhera, D. 共2008a兲. “Development of design wind
version is the shift from a single map for an importance factor for speed maps for the Caribbean for application with the wind load pro-
buildings and other structures of 1.0 to three separate maps, one visions of ASCE 7.” ARA Rep. No. 18108-1, Pan American Health
Organization, Regional Office for The Americas World Health Orga-
for each category of occupancy, thus eliminating the need for
nization, Disaster Management Programme, 525 23rd Street NW,
importance factors that vary between hurricane and nonhurricane Washington, D.C.
regions. Vickery, P. J., and Wadhera, D. 共2008b兲. “Statistical models of the Hol-
land pressure profile parameter and radius to maximum winds of hur-
ricanes from flight level pressure and H*wind data.” J. Appl.
References Meteorol., 47, 2497–2517.
Vickery, P. J., Wadhera, D., Powell, M. D., and Chen, Y. 共2009a兲. “A
Applied Research Associates, Inc. 共2001兲. “Hazard mitigation study for hurricane boundary layer and wind field model for use in engineering
the Hawaii hurricane relief fund.” ARA Rep. No. 0476, prepared for applications.” J. Appl. Meteorol., 48, 381–405.
state of Hawaii, Raleigh, N.C. Vickery, P. J., Wadhera, D., Twisdale, L. A., Jr., and Lavelle, F. M.
ASCE. 共2005兲. “Minimum disign loads for building and other structures.” 共2009b兲. “United States hurricane wind speed risk and uncertainty.” J.
ASCE 7–05, New York. Struct. Eng., 135共3兲, 301–320.

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2010 / 625

J. Struct. Eng. 2010.136:613-625.

You might also like