BNL-108210-2015-JA

PUBLICATIONS
Journal of Geophysical Research: Atmospheres
RESEARCH ARTICLE Observed drag coefcients in high winds in the near
10.1002/2015JD023172
offshore of the South China Sea
Special Section: Xueyan Bi1,2,3, Zhiqiu Gao1, Yangang Liu3, Feng Liu4, Qingtao Song5,6, Jian Huang2, Huijun Huang2,
Fast Physics in Climate Models: Weikang Mao2, and Chunxia Liu2
Parameterization, Evaluation
and Observation 1
State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry, Institute of Atmospheric
Physics, Chinese Academy of Sciences, Beijing, China, 2Guangzhou Institute of Tropical and Marine Meteorology, China
Key Points: Meteorological Administration/Guangdong Provincial Key Laboratory of Regional Numerical Weather Prediction,
Momentum uxes in high winds in Guangdong, China, 3Brookhaven National Laboratory, Upton, New York, USA, 4Climate and Atmospheric Science Section,
shallow water were measured
Division of Illinois State Water Survey, Prairie Research Institute, University of Illinois at Urban-Champaign, Champaign,
Impact of depth of water on the
air-sea momentum ux Illinois, USA, 5National Satellite Ocean Application Service, Beijing, China, 6Key Laboratory of Space Ocean Remote Sensing
Momentum uxes were estimated by and Application, SOA, Beijing, China
three methods and were compared

Abstract This paper investigates the relationships between friction velocity, 10 m drag coefcient, and
10 m wind speed using data collected at two offshore observation towers (one over the sea and the other
Correspondence to:
Z. Gao, on an island) from seven typhoon episodes in the South China Sea from 2008 to 2014. The two towers
[email protected] were placed in areas with different water depths along a shore-normal line. The depth of water at the tower
over the sea averages about 15 m, and the depth of water near the island is about 10 m. The observed
Citation: maximum 10 min average wind speed at a height of 10 m is about 32 m s
1. Momentum uxes derived
Bi, X., Z. Gao, Y. Liu, F. Liu, Q. Song, from three methods (eddy covariance, inertial dissipation, and ux prole) are compared. The momentum
J. Huang, H. Huang, W. Mao, and C. Liu uxes derived from the ux prole method are larger (smaller) over the sea (on the island) than those from
(2015), Observed drag coefcients in
high winds in the near offshore of the the other two methods. The relationship between the 10 m drag coefcient and the 10 m wind speed is
South China Sea, J. Geophys. Res. Atmos., examined by use of the data obtained by the eddy covariance method. The drag coefcient rst decreases
120, doi:10.1002/2015JD023172. with increasing 10 m wind speed when the wind speeds are 510 m s
1, then increases and reaches a peak
value of 0.002 around a wind speed of 18 m s
1. The drag coefcient decreases with increasing 10 m wind
Received 28 JAN 2015
Accepted 10 JUN 2015 speed when 10 m wind speeds are 1827 m s
1. A comparison of the measurements from the two towers
Accepted article online 19 JUN 2015 shows that the 10 m drag coefcient from the tower in 10 m water depth is about 40% larger than that from
the tower in 15 m water depth when the 10 m wind speed is less than 10 m s
1. Above this, the difference
in the 10 m drag coefcients of the two towers disappears.

1. Introduction
The dependence of the drag coefcient (CD) on wind speed under tropical cyclone conditions is critically impor-
tant for understanding and modeling storm intensity [Rogers et al., 2013; Soloviev et al., 2014]. On the basis of a
theoretical study of the energy balance of a typhoon system, Emanuel [1995] argued that storm intensity
depends on the ratio of the enthalpy coefcient (Ck) to CD and the ratio lies in the range of 0.751.5 in the high
wind region of intense storms. The Ck has little relation to wind speed [Jeong et al., 2012], that means CD does
not continue to increase at higher wind speeds. The idea of sea surface drag saturation at high wind speed has
attracted the attention of several research communities and has been conrmed by subsequent eld observa-
tions [Powell et al., 2003; Black et al., 2007; French et al., 2007; Jarosz et al., 2007; Holthuijsen et al., 2012], labora-
tory experiments [Alamaro et al., 2002; Donelan et al., 2004; Troitskaya et al., 2012], modeling studies [Moon et al.,
2004; Bye and Jenkins, 2006; Zweers et al., 2010], and theoretical studies [Emanuel, 2003; Makin, 2005]. These stu-
dies show that when the saturation 10 m drag coefcient (CD10) falls between 0.002 and 0.0025, the correspond-
ing u10 is approximately 3033 m s
1 [Powell et al., 2003; Donelan et al., 2004; French et al., 2007; Jarosz et al.,
2007; Holthuijsen et al., 2012]. However, the relationship between CD10 and u10 in strong storms has been exam-
ined mainly over deep water. Limited eld experiments were conducted over shallow water under high wind
speeds, and details of the relationship between CD10 and u10 remain unresolved [Anctil and Donelan, 1996;
Zachry et al., 2013a; Liu et al., 2012]. Drag coefcient parameterizations for deep water are commonly applied
2015. American Geophysical Union.
to shallow water. Additional and improved eld measurements are essential to gaining a better understanding
All Rights Reserved. of CD behavior under high wind conditions over shallow water.

BI ET AL. DRAG COEFFICIENTS IN HIGH WINDS OVER SEA 1

 Journal of Geophysical Research: Atmospheres 10.1002/2015JD023172

Seven typhoon cases (Hagupit, Chanthu,

Nockten, Nesat, Kaitak, Rammasun, and
Kalmaegi) were captured from 2008
to 2014 at two offshore instrumented
towers (Figure 1). With observations from
two buoys anchored 6.7 and 81.9 km
from the towers, Zhao et al. [2015] esti-
mated the relative water depth (0.9) near
the towers during typhoons, which is
much less than the threshold (3) for
shallow-water conditions. The relative
water depth is equals to n h, where n
is the wave number and h is the water
depth. The main objective of this paper
is to examine the relationships between
the friction velocity (u*), CD10 , and u10
Figure 1. An illustration of the platforms and locations of the tower over
over shallow water under high wind
water (blue upward pointing triangle) and the tower on Zhizai Island
(red pentagram). The bathymetric contours (m) are also shown. conditions by using these unique
observations.

2. Experimental Setup
The data were collected at two offshore towers xed along the shore-normal line during seven typhoon
events from 2008 to 2014. Figure 1 shows a map of the two towers and the bathymetric contours around
the towers. One tower (21260 24N, 111230 26E; hereafter sea tower) is on an integrated observation plat-
form for marine meteorology over the sea at Bohe, Maoming. It is operated by the Institute of Tropical and
Marine Meteorology, China Meteorology Administration [Huang and Chan, 2011]. It stands 6.5 km offshore
in water depth of 15 m in the South China Sea. The observation platform is about 11 m above the mean
sea level (msl). Its upper part is a 25 m high steel tower. Wind, temperature, and humidity sensors are
mounted on 2 m booms on ve different planes (13.4 m, 16.4 m, 20.0 m, 23.4 m, and 31.3 m above msl) with
a sampling frequency of 1 min. The models of these sensors are RM Young/05106 and HMP45C. Gill
Windmaster Pro ultrasonic anemometers are installed on 2 m booms at 27.3 m and 35.1 m above msl with
a sampling frequency of 10 Hz. A schematic view of the instruments deployed on this tower can be found
in Zhao et al. [2013]. The other tower (hereinafter island tower), operated by the Guangdong Climate
Centre, is on the Zhizai Island (21270 3.12N, 111220 28E), about 2.0 km from the sea tower. The island has
an above-water area of approximately 90 m 40 m and is covered by sand and sparse weeds. The tower is
10 m above msl [Wang et al., 2013]. The NRG #40 wind sensors are mounted on 2.5 m booms on six different
planes (10.0 m, 20.0 m, 40.0 m, 60.0 m, 80.0 m, and 100.0 m above msl) with a sampling frequency of 1 min. Gill
Windmaster Pro ultrasonic anemometers with a sampling frequency of 10 Hz were placed on the 2.5 m
booms at a height of 60 m during Typhoon Hagupit and at a height of 40 m during Typhoon Chanthu.
The instruments were placed on the east side of the towers facing the seaward direction to minimize ow
distortion. The small island Dazhuzhou lies southwest of the towers. The instruments and measurements
used in this work are listed in Table 1.

3. Characteristics of the Typhoons

Figure 2 shows the typhoon tracks relative to the location of the towers. Strong Typhoon (41.550.9 m s
1)
Hagupit moved over the towers on 24 September 2008. The other six typhoons passed by south of the
towers. Measurements from the sonic anemometers were collected for ve typhoons, and the time series
of observed wind speed (WS), wind direction (WD), u10 , and distance of the typhoon center to the towers
are shown in Figure 3. The red color indicates the portion of observations left after data quality control
(detailed below). It can be seen from Figure 3 that the observations from the sonic anemometers are fairly
consistent with those from the wind sensors. Typhoon Hagupit was the most intense storm observed from
the towers, although Typhoon Chanthu had the highest u10 of 32 m s
1 based on the quality-controlled
data. Figure 3 illustrates that from 4:30 to 6:00 A.M. on 24 September 2008 during Hagupit, the observed

BI ET AL. DRAG COEFFICIENTS IN HIGH WINDS OVER SEA 2

 Journal of Geophysical Research: Atmospheres 10.1002/2015JD023172

u10 on the towers decreased more than 30 m s
1, and the

u10(m s )
Maximum

1
Sensors: RM Young/05106 HMP45C
Anemometers and Temperature

26.3
23.2
27.0
27.8
19.6
27.1
wind direction varied by nearly 180, indicating that the

center of Typhoon Hagupit passed almost directly over
the tower with widespread typhoon force winds. Its gale

31.3, 23.4, 20.0, 16.4, 13.4

radius reached 500 km, and its storm radius reached

31.3, 23.4, 20.0, 16.4

31.3, 23.4, 20.0, 16.4
31.3, 23.4, 20.0, 16.4
200 km. The minimum wind speed is 17.2 m s
1 for a gale

31.3, 20.0, 16.4

31.3, 20.0, 16.4
On the Platform Over Water

Heights (m)
Instrument
and 24.5 m s
1 for a storm. The maximum 10 min average
wind speed was ~47 m s
1 at a height of 60 m on the

island tower. According to the ofcial record for the storm
(http://www.typhoon.gov.cn/), it reached strong typhoon
intensity as the eye moved over the towers during
the early morning hours of 24 September. Typhoon
u10 (m s )

1
Maximum
Chanthus typical characteristics were rapid intensica-
Gill Windmaster Pro
Sonic Anemometer:

18.7
25.7

23.4

tion near the coast, widespread typhoon force winds, and
rainfall. The towers were in the right front quadrant of
Typhoon Chanthu when the maximum wind speed was
Instrument
Height (m)

observed. The minimum distance between the towers and

35.1
35.1

27.3
a

the center of Typhoon Chanthu was about 60 km. The maxi-

mum observed wind speed was 36.4 m s
1 at a height of
40 m on the island tower.
u10 (m s )

1
Maximum

44.1

34.3
32.6
Anemometers: NRG #40

4. Methodology
The drag coefcient is dened as
100,80,60,40,20,10
80,60,40,20,10
On the Tower on the Island

100,40,20,10

u2
Instrument
Height (m)

CD ; (1)
u2

where u is the horizontal wind speed at the instruments

height [Garratt, 1992]. To allow for comparison of
CD obtained from measurements at different heights
u10 (m s )

1
Maximum

and for the convenience of parameterization, the 10 m

Gill Windmaster Pro
Sonic Anemometer:

32.1
23.8

drag coefcient under the neutral stability condition is

estimated by
Table 1. Description of the Observations During the Seven Typhoon Events

u2
Instrument
Height (m)

C D10 : (2)
u210
60
40

A logarithmic surface layer has been widely used [Dyer,

Pressure
Central

1974; Powell et al., 2003; Liu et al., 2013], although the

(hPa)

940
970
980
950
968
910
960

validity of the traditional logarithmic surface layer in

the inner core of typhoons and its use in the inference
Distance to the

of the drag coefcient at high wind speeds have been

Towers (km)
The Closest

questioned [Smith and Montgomery, 2014]. We have

220
195

160
194
65

83
0

checked our data, and the results show that most cases
follow the logarithmic wind prole. Zhao et al. [2015]
plotted the mean wind proles during Typhoons
1619 Aug 2012
2324 Sep 2008

2930 Sep 2011

1517 Sep 2014

2123 Jul 2010
2930 Jul 2011

1819 Jul 2014

Hagupit and Chanthu. Thus, we used the logarithmic

: no data are available.

wind prole to convert u at different measuring heights

Period
Time

to u10 in order to eliminate the inuence of different

height on CD:
u
u10 u
 lnz=10; (3)
Rammasun

k
Kalmaegi
Typhoon

Chanthu
Nockten
Hagupit

Kaitak
Nesat

where k = 0.40 is the von Krmn constant and z is the

a

measurement height.

BI ET AL. DRAG COEFFICIENTS IN HIGH WINDS OVER SEA 3

 Journal of Geophysical Research: Atmospheres 10.1002/2015JD023172

Figure 2. Map of the typhoon tracks along with the locations of the two observational towers. The different dashed lines
denote the different typhoon categories. The black circles mark the distances to the towers of 50 km, 100 km, 150 km,

1
1
1
200 km, and 250 km. Super strong typhoon is larger than 51 m s . Typhoon is from 32.7 m s to 41.4 m s , and strong

1
1
tropical storm is from 24.5 m s to 32.6 m s . Note that the locations of the towers on the island (blue pentagon) and
over the water (red upward pointing triangle) largely overlap each other due to their close proximity.

In this study we used three different methods to estimate u*: eddy covariance, inertial dissipation, and ux
prole methods. We also used the Coriolis correction of friction velocity given by Donelan [1990] in the analysis.
To meet the neutral stability condition, the measurements with
0.1 < z/L < 0.1 were chosen (L is the Obukhov
length). For the measurements from the eddy covariance and the inertial dissipation methods, L is computed by
u3 T 0
L
0
; (4a)
gkw 0 v
0
where g is the gravitational acceleration, w 0 v is the ux of virtual potential temperature at height z, and T 0 is
the mean air temperature at height z. For the measurements using the ux prole method, L is estimated
from the gradient Richardson number (Ri) based on an empirical relationship proposed by Arya [1982]:
z
Ri Ri < 0
L
; (4b)
z Ri
Ri > 0
L 1
5Ri
where Ri is the average Riz. Every Riz is calculated from observations in which two levels by
" #  z 2
T ln z21
Ri z gT 1 d  z1 z2 ; (5)
z1 z2 2 u
p
where z z1 z 2 ; T is the mean temperature at height z; T T 2
T 1 , u u2
u1 and T 1 and T 2 are the
temperatures at heights z1 and z2, respectively; and d = 0.0098 K m
1 is the dry adiabatic lapse rate.
4.1. Eddy Covariance Method
In the eddy covariance method, friction velocity is directly calculated from the measurements of turbulent
velocity uctuations:
 2  2 1 =4 ;
u*s u0 w 0 v 0 w 0 (6)

where u0 , v0 , and w0 are the turbulent uctuations of the three wind components. The over bar indicates
Reynolds averaging, and u* s is the friction velocity without the Coriolis correction.

BI ET AL. DRAG COEFFICIENTS IN HIGH WINDS OVER SEA 4

 Journal of Geophysical Research: Atmospheres 10.1002/2015JD023172

Figure 3. Temporal changes of the raw 10 min wind speed (WS), wind direction (WD) obtained from the sonic anemometers
and the standard wind gauge, 10 m wind speed (u10) obtained by using a logarithmic wind prole (i.e., equation (3)), and
the distance of the typhoon centers to the towers (data obtained from the website http://www.typhoon.gov.cn/). The red
color represents the portion of observations from sonic anemometers left after quality control.

BI ET AL. DRAG COEFFICIENTS IN HIGH WINDS OVER SEA 5

 Journal of Geophysical Research: Atmospheres 10.1002/2015JD023172

Figure 3. (continued)

4.2. Inertial Dissipation Method

Assuming that the dissipation of kinetic energy is balanced by the production of turbulence by the shear ow
in the surface layer at the same level [Sjblom and Smedman, 2004], u* s is estimated by
u3*s kz= ; (7)
where is the rate of dissipation of turbulent kinetic energy and is the dimensionless dissipation function
of stability and equals one under neutral conditions.
The dissipation rate is further related to the power spectral density Su(f) of the streamline direction wind
speed via [Panofsky and Dutton, 1984]
h  i3=2
f 5=3 Su f = u u2=3 ; (8)

where the Kolmogorovs constant u = 0.52 and f is the frequency.

4.3. Flux Prole Method
The ux prole method is based on the logarithmic wind prole relationship:
u
u *s lnz=z0 ; (9)
k
where z0 is the sea surface roughness length and u* s/k is the slope of the least squares tting of u and ln(z).

BI ET AL. DRAG COEFFICIENTS IN HIGH WINDS OVER SEA 6

 Journal of Geophysical Research: Atmospheres 10.1002/2015JD023172

4.4. Coriolis Correction

Strictly speaking, a constant stress layer
occurs only at the equator (no Coriolis
acceleration) in stationary and homoge-
nous conditions; thus, the use of u* s as
the surface friction velocity u* introduces
a systematic underestimation of u*.
Donelan [1990] introduced equation (10)
to correct this so-called Coriolis bias:
 
0 f c z
u u2*s 1 ; (10)
u*s
Figure 4. Ratio of the friction velocity obtained with the Coriolis correction
(u*) to that without the Coriolis correction (u* s) as a function of 10 m wind
speed (u10). The data are obtained from the eddy covariance method. The where 0 12 for neutral conditions, and
symbols and bars represent the median values and interquartile ranges, fc = 1.454 10- 4 sin(la) s
1 is the Coriolis
respectively. parameter, where la is the latitude. This
Coriolis correction is applied in our analy-
sis. We also examine it as a function of u10.
Figure 4 shows that the ratio of u* to u* s decreases with increasing u10. For u10 = 1015 m s
1, u* s underestimates
u* by approximately 3%10%. For u10 > 15 m s
1, u* s underestimates u* by about 2%.

5. Data Processing and Quality Control

5.1. Sonic Anemometer Observations Preprocessing
Sonic anemometer measurements are crucial to both the eddy covariance method and the inertial dissipa-
tion method, as indicated in equations (6) and (7), respectively. It is necessary to apply several corrections
to raw sonic anemometer data to produce uctuations and means [Foken et al., 2005; Lee et al., 2005; Oh
et al., 2010a]. Eddy covariance data processing has been widely investigated in previous studies [Lee et al.,
2005] and consists of several steps: spike detection, coordinate rotation, and detrending to separate the
turbulent signals from the mean ow [Lee et al., 2005]. Each step of the process can be approached in various
ways, and selection of specic approaches depends upon both site and weather conditions [Vickers and
Mahrt, 1997; Mauder et al., 2007, 2008]. In this work, the raw eddy covariance data were processed by the
following steps:
1. Spike detection and removal
Spike detection is similar to that of Hojstrup [1993]. A point is considered a spike and is discarded if it falls out-
side the prescribed limits. The prescribed limits are 60 m s
1 for the horizontal wind speed, 10 m s
1 for the
vertical wind speed, and 050C for air temperature. A point is also considered a spike if it is greater than 6
standard deviations of the difference between consecutive data points of a 10 min record. The data are
not treated as spikes if four or more are detected consecutively. The discarded spikes are replaced by linearly
interpolated values. A record is eliminated if the number of spikes is greater than 1% of the total number of
data points in a 10 min segment.
2. Averaging period
The choice of the averaging period is considered a crucial factor in turbulent ux calculation using the eddy cov-
ariance method. The Ogive method is used to determine an optimal averaging period at any site [Lee et al., 2005]:
f0
Oguw f 0 Couw f df ; (11)

where Couw(f) is the cospectrum of uw at frequency f. The Ogive curve shows the cumulative contribution of
eddies of increasing period (decreasing frequency) to the total transport and reaches a constant after the
frequency falls below a certain value. The time period corresponding to this frequency represents the
minimum averaging time necessary to capture signicant ux-carrying eddies. Figure 5 shows the Ogive of
uw at different wind speeds during Typhoon Hagupit at a height of 60 m from the island tower. It is clear
that the Ogive curve approaches a constant at an averaging period of 10 min, which we chose as the
averaging period in our calculation of turbulent uxes.

BI ET AL. DRAG COEFFICIENTS IN HIGH WINDS OVER SEA 7

 Journal of Geophysical Research: Atmospheres 10.1002/2015JD023172

3. Coordinate rotation (tilt correction)

and detrending
The coordinate system of the sonic
measurements is transformed into the
mean-ow streamlines to eliminate
instrument tilt errors and cross con-
tamination among components of the
turbulent ux vector [Finnigan, 2004;
Oh et al., 2010b]. We applied the
double rotation approach [Kaimal and
Finnigan, 1994] in this work (it is intro-
Figure 5. Ogives from the consecutive 30 min average spectrum of uw
measured during Hagupit over the island at 60 m height (the solid line duced in detail in the Appendix A).

1
is for a wind speed of 10.5 m s ; the dashed line is for a wind speed of The detrending method uses a block

1
22.4 m s ). The 10 min averaging time applied in this study is marked by time average recommended by Lee
the vertical line. et al. [2005] to separate means and
turbulent components. The dominant
wind direction at the towers is easterly. In order to minimize the inuence of ow distortion by the
platform, the data observed on the sea tower with wind directions between 240 and 300 are removed.
Considering the inuence of ow distortion by Dazhuzhou Island, we also removed the data observed
on the tower on Zhizai Island with wind directions between 180 and 300.

5.2. Footprint Analysis

The footprint analysis method is usually applied to identify the region that inuences the ux measurements.
Schmid [1994] recommends the Eulerian analytic scalar ux source area model (FSAM) for those uxes
obtained from the eddy covariance method, and the Eulerian analytic scalar concentration source area model
(SAM) for those obtained from the ux prole, the Bowen ratio, or other indirect methods, which rely on
observations of mean concentrations. FSAM and SAM are both based on the inverted plume assumption
and require the parametersu, z, z0, and z/L, and the intensity of crosswind turbulence (uv ) as inputs.
Figure 6 shows the 50% and 90% of the surface source areas of uxes obtained from the eddy covariance method
on the island for wind speeds of 33.0 m s
1 at a height of 40 m during Typhoon Chanthu. The symbol xm repre-
sents the maximum source weight location. Figure 6 shows clearly that 90% of the source area upwind distances
from the sensors are between 500 m and 3000 m when the wind speed is 33 m s
1 at a height of 40 m. As stated in
section 2, Zhizai Island is 4.5 km from the shore and has an above water area of approximately 90 m 40 m. It is
clear that the minimum distance of the surface 90% source area from the sensors is about 500 m, larger than
the 90 m of the Zhizai Island area. The maximum distance of the surface 90% source area from the sensors is about
3.0 km, less than the 4.5 km from the island to the shore. So the uxes obtained by the eddy covariance method on
the island should not be inuenced by either the island or the shore.
Because there were no standard tem-
perature sensors on the island tower,
the z/L cannot be obtained from the ux
prole method and the corresponding
footprint cannot be analyzed. However,
Schmid [1997] indicates that measure-
ments obtained from the ux prole
method are inuenced by the local sur-
face patch. That is to say, measurements
obtained from the ux prole method
on the island are inuenced by the
island. This may be a key reason for the
Figure 6. The 50% and 90% ux source areas by the eddy covariance

1 differences between the uxes obtained
method on the island tower for winds of 33.0 m s at a height of 40 m
during Typhoon Chanthu. The symbol xm indicates the maximum source from the eddy covariance and the ux
location (upwind distance of the surface element with the maximum prole methods on the island (shown in
inuence on the sensor). section 6.1).

BI ET AL. DRAG COEFFICIENTS IN HIGH WINDS OVER SEA 8

 Journal of Geophysical Research: Atmospheres 10.1002/2015JD023172

Figure 7. The values of the friction velocity (u*) derived from the eddy correlation method (EC) (magenta), the inertial
dissipation method (ID) (blue), and the ux prole method (FP) (red) as a function of 10 m wind speed (u10) during different
typhoon events. The solid (hollow) symbols denote the observations from the sea (island) tower. The symbols and bars

1
represent the median values and interquartile ranges, respectively. The u10 bin size is 5 m s . The results given by Powell et al.
[2003] are also shown for comparison (black pentagon).

6. Results and Discussion

6.1. Comparison Between Different Methods
The momentum uxes can be estimated using the eddy covariance method, the inertial dissipation method,
or the ux prole method, each of which has advantages and disadvantages. The eddy covariance method,
the direct method for calculating the uxes using the high-frequency data, is sensitive to platform motion
and sensor tilt and suffers from sampling uncertainty. The inertial dissipation and ux prole methods are
insensitive to platform motion and distortion of airow but rely on the validity of similarity theory. The eddy
covariance and the inertial dissipation methods use some of the same data, but while the eddy covariance
method directly estimates u*, the inertial dissipation method equates production and dissipation turbulent
kinetic energy and derives the latter from the spectrum of the streamwise velocity component. Over land
in moderate wind conditions, the eddy covariance method is generally thought to be superior.
Figure 7 (Figure 8) compares the values of u* (CD10) obtained by the three methods as a function of u10 for
each typhoon case and for all cases combined. It is evident that u* and CD10 obtained from the eddy covar-
iance and the inertial dissipation methods on both towers have similar patterns. However, the results derived
from the ux prole method show that both u* and CD10 are larger (smaller) on the sea (island) tower than
those from the other two methods.

BI ET AL. DRAG COEFFICIENTS IN HIGH WINDS OVER SEA 9

 Journal of Geophysical Research: Atmospheres 10.1002/2015JD023172

Figure 8. The same as Figure 7 except for the drag coefcient (CD10) as a function of 10 m wind speed (u10).

Figure 9a further compares the simultaneous u* derived from the eddy covariance and inertial dissipation
methods. The correlation coefcient is 0.93, and 88.1% of the data points concentrate between lines y = (1 + 30 %)
x and y = (1
30 %)x. Figure 9b compares u* from the eddy covariance and ux prole methods. The correlation
coefcient is 0.78, and only 28.5% of the data points concentrate between lines y = (1 + 30 %)x and y = (1
30 %)x.
It is clear that the inertial dissipation method better correlates with the eddy covariance method than with the ux
prole method. Thus, only the measurements obtained from the eddy covariance method are analyzed in the
remaining analysis.
The different results from the ux prole method compared to the other two methods can be attributed to (1)
the island inuences the uxes obtained from the ux prole method on the island (as discussed in
section 5.2), (2) the ux prole method used the data collected at two levels which are inuenced by different
fetches, and (3) the inherent errors in slow response anemometers signicantly inuence the accuracy of the
ux prole method that uses the wind speed difference between two levels to calculate the ux. The ux
prole method has been used for many years for turbulent ux calculation. It is valid for homogeneous
surfaces such as open ocean surfaces [Powell et al., 2003].

6.2. Impact of Height on Momentum Flux and Drag Coefcient

Because there were no simultaneous measurements at different heights, we compared the median values of
u* and CD10 binned according to a 5 m s
1 of u10 interval for each typhoon case. Figure 10 shows the values of

BI ET AL. DRAG COEFFICIENTS IN HIGH WINDS OVER SEA 10

 Journal of Geophysical Research: Atmospheres 10.1002/2015JD023172

Figure 9. The friction velocity obtained using (a) the inertial dissipation method (u* ID) and (b) the ux prole method
(u* FP) compared to those obtained by using the eddy correlation method (u* EC). The dashed lines indicate the 30%
deviations from the solid 1:1 line, and r is a correlation coefcient. The different colors denote different ranges of 10 m wind
speed (u10).

u* and CD10 as functions of u10 for each typhoon case. The results indicate that neither relationship varies
systematically with height. This suggests the validity of the constant stress layer hypothesis in this work.

6.3. Impact of Water Depth on the Air-Sea Momentum Flux

As shown in Figure 1, the two towers are placed in areas with different water depths along a shore-normal
line. The water depth at the sea tower averages about 15 m [Huang and Chan, 2011], and the water depth
near Zhizai Island is about 10 m. Donelan et al. [2012] showed that water depth (<30 m) has a strong effect
on CD10 during strong winds. The observations at the towers provide an opportunity to investigate the
impact of water depth on momentum ux in shallow water. Figure 10 shows the values of u* and CD10 as
functions of u10 for each typhoon case. The magenta (blue) denotes the observations from the island (sea)
tower. Below 10 m s
1, CD10 is about 0.0018 on the island tower and is about 0.00125 on the sea tower.
The values of CD10 in 10 m water depth are about 40% larger than those in 15 m. This difference of 40% is
larger than measurement uncertainties, which were about 1015% for neutral conditions [Kessomkiat et al.,
2013]. Anctil and Donelan [1996] observed that the value of CD10 is about 0.0028 in 4 m water depth and about
0.0018 in 12 m depth when u10 is 14 m s
1. Limited previous eld campaigns and laboratory measurements
found that CD10 is higher over shallow water than over deep water at wind speeds below hurricane force
because of the markedly different wave conditions [Anctil and Donelan, 1996; Zachry et al., 2013a, 2013b].
Our measurements support this conclusion, but only when u10 < 10 m s
1. Beyond 10 m s
1, the difference
in CD10 between the two towers disappears. Unfortunately, no simultaneous wave data were available near

BI ET AL. DRAG COEFFICIENTS IN HIGH WINDS OVER SEA 11

 Journal of Geophysical Research: Atmospheres 10.1002/2015JD023172

Figure 10. The values of (a) the friction velocity (u* EC) and (b) the 10 m neutral drag coefcient (CD10) derived from the
eddy correlation (EC) method as a function of 10 m wind speed (u10) during different typhoon events. The symbols and

1
bars represent the median values and interquartile ranges, respectively. The u10 bin size is 5 m s . The solid magenta
(blue) symbols denote the observations from the tower over the island (water).

the observation platform for further analysis. This reinforces the conclusion evident in Figure 6 that the uxes
obtained from the eddy covariance method on the island are solely inuenced by the sea surface not by the
island or shore. Therefore, we combine the observations from the two towers and perform a composite analysis.

6.4. Characteristics of Friction Velocity and Drag Coefcient

Figure 11 shows that the values of u* and CD10 yielded from the eddy covariance method as a function of
u10 , with the data derived from the ve typhoon events combined. The median values and interquartile
ranges binned according to 5 m s
1 intervals of u10 are shown; also shown are the sample numbers in
every bin. For the sake of clarity, the median values are also shown in 1 m s
1 bins. Figure 11a shows that
u* increases with increasing u10 when u10 = 5
35 m s
1. The values of u* are less than those reported by
Powell et al. [2003] for u10 = 25
35 m s
1. Figure 11b shows that CD10 initially decreases with increasing
u10 for u10 = 5
10 m s
1, then increases and reaches a peak at u10 = 18 m s
1 before decreasing again,
and then levels off beyond 27 m s
1. When u10 = 10
18 m s
1, CD10 increases with u10 in much the same
manner as reported in Large and Pond [1981]. However, below 10 m s
1 and above 18 m s
1, our measure-
ments show a quite different pattern from those previously reported [Powell et al., 2003; Donelan et al.,
2004; Jarosz et al., 2007]. Our measurements are larger for u10 = 5
10 m s
1 but are smaller for
u10 > 18 m s
1. The values for the saturation CD10 of 0.002 are similar to those of Powell et al. [2003]
and French et al. [2007], but the saturation u10 of 18 m s
1 is much less than the 3035 m s
1 reported
by others.

BI ET AL. DRAG COEFFICIENTS IN HIGH WINDS OVER SEA 12

 Journal of Geophysical Research: Atmospheres 10.1002/2015JD023172

Figure 11. Relationships of (a) the friction velocity (u* EC) and (b) the 10 m drag coefcient (CD10) (derived from the eddy
correlation method) as a function of 10 m wind speed (u10) for ve integrated typhoon events. The blue upward pointing

1
1
triangles and magenta solid circles represent the median values binned according to 1 m s and 5 m s of u10 interval,
respectively. The corresponding bars represent the interquartile ranges. Also shown are the data points or tting curves
reported in literature for comparison.

The discrepancies may have arisen from numerous factors related to the conditions of the sea, such as
water depth. Since depth-induced wave breaking becomes an important dissipation term in a sloping
bed surf zone as at our research sites [Holthuijsen et al., 2012; Richter and Sullivan, 2013], a shoaling effect
can lead to a lower saturation u10 .
In order to better demonstrate the inuence of wind-wave conditions on CD10, Figure 12 also shows CD10 as a
function of u10 in different wind directions: 090 is the right front quadrant of a typhoon and 90180 is the
right rear quadrant. Observations from Coupled Boundary Layer and Air-Sea Transfer reveal the distinctively
different wave spectra in the different storm quadrants in association with the sea state. However, the value
of CD10 is independent of the storm quadrant [Black et al., 2007]. Our measurements also show that there are
no obvious differences in CD10 between different quadrants of typhoons.

7. Conclusions
Measurements collected from two offshore towers in the South China Sea during seven typhoon cases are
examined. Analysis of the measurements from the two towers, placed in regions of 10 m and 15 m water
depths along a shore-normal line, supports the existing notion that CD10 is higher in shallower water when
wind speeds are below hurricane force but only for u10 < 10 m s
1. Beyond 10 m s
1, the difference in CD10
between the two towers disappears. The dependence of CD10 on u10 shows a different pattern when
u10 = 5
10 m s
1 and u10 = 18
27 m s
1: CD10 decreases with increasing u10 and levels off beyond

BI ET AL. DRAG COEFFICIENTS IN HIGH WINDS OVER SEA 13

 Journal of Geophysical Research: Atmospheres 10.1002/2015JD023172

27 m s
1. The value of saturation CD10

of 0.002 is similar to the results of
Powell et al. [2003] and French et al.
[2007], but saturation u10 = 18 m s
1 is
signicantly less than other observa-
tions of 3035 m s
1. The value of
CD10 is independent of the typhoon
quadrant. The interaction between
waves and the local bathymetry that
causes wave conditions to dramati-
cally change in shallow water under
Figure 12. Relationship of the 10 m drag coefcient (CD10) (derived from high winds may be responsible for
the eddy correlation method) as a function of 10 m wind speed (u10) in these results. Unfortunately, no simulta-
different wind directions for ve integrated typhoon events. The blue solid neous wave data were available near
downward pointing triangles (magenta solid circles) and bars represent the towers for comparison with those
the median values and interquartile ranges in wind directions of 090

1 in deep water.
(90180). The u10 bin size is 5 m s . Also shown for comparison are the
data points or tting curves reported in literature. Three commonly used methods for esti-
mating the momentum uxes (eddy
covariance, inertial dissipation, and ux
prole methods) are compared, and the comparison shows that the results obtained from the eddy covariance
and the inertial dissipation methods are in good agreement. The relationships between u*, CD10, and u10 do not
exhibit a systematic change with measurement height. This supports the existence of a constant stress layer.
The results derived from the ux prole method on the sea (island) tower are larger (smaller) than that from
the other two methods, suggesting that the island inuenced the uxes derived from the ux prole method
and thus caution is needed in using the ux prole method over a heterogeneous surface.
Although still limited, the results derived from the unique observations shed new light on the relationship
between CD10 and u10 under high winds in shallow water. More data and analysis are needed to conrm the
ndings. In this work, just the momentum uxes were analyzed. The importance of the heat and moisture
uxes and the lack of observations in high wind have been discussed in some studies [Richter and Stern,
2014; Soloviev et al., 2014]. Because our preliminary analysis shows the complexity of the measurements
of heat and moisture uxes, further research should be conducted. This is the next topic of our
future studies.

Appendix A: Coordinate Rotation (Double Rotation)

Double rotation is based on works by Tanner and Thurtell [1969], as proposed by Aubinet et al. [2000]. It is
related to period-by-period rotation, which can be accomplished for every data averaging period (10 min).
Two rotations are conducted. The rst rotation is around the z axis into the mean wind such that v 1 0:
u1 um cos v m sin
v 1
um sin v m cos ;
w1 wm
where um, vm, and wm are the observed wind speeds in three coordinate axes and tan
1 v m um . The
symbols um and v m are the mean wind speeds along the x and y axes; u1, v1, and w1 are the wind speeds
in the new coordinate system after the rst rotation; and u1 , v 1 , and w 1 are the mean values of u1, v1, and w1.
The second rotation is around the new y axis, so that w 2 0:
u2 u1 cos w 1 sin
v2 v1 ;
w 2
u1 sin w 1 cos

1
where tan w 1 u1 . The symbols u2, v2, and w2 are the wind speeds in the new coordinate system after
these two rotations. Having performed these rotations, the coordinate system of the sonic anemometer is
moved into alignment with the mean streamlines.

BI ET AL. DRAG COEFFICIENTS IN HIGH WINDS OVER SEA 14

 Journal of Geophysical Research: Atmospheres 10.1002/2015JD023172

Acknowledgments References
This work was supported by the National
Natural Science Foundation of China Alamaro, M., K. Emanuel, J. Colton, W. McGillis, and J. Edson (2002), Experimental investigation of air-sea transfer of momentum and enthalpy
under grants 40906023, 41205011, at high wind speed paper presented at Preprints, 25th Conf. on Hurricanes and Tropical Meteorology, Am. Meteorol. Soc., San Diego, Calif.
41475061, 41175013, 41076012, and Anctil, F., and M. A. Donelan (1996), Air-water momentum ux observations over Shoaling waves, J. Phys. Oceanogr., 26(7), 13441353,
41276019 and the National Program on doi:10.1175/1520-0485(1996)026<1344:amfoos>2.0.co;2.
Key Basic Research Project of China (973) Arya, S. P. (1982), Atmopheric boundary layers over homogeneous terrain, in Engineering Meteorology, edited by E. J. Plate, pp. 233267,
under grant 2011CB403501. This work is Elsevier, New York.
also supported by the U.S. DOE ESM via Aubinet, M., J. Moncrieff, R. C. Clement, M. Aubinet, J. Moncrieff, and R. C. Clement (2000), Estimates of the annual net carbon and water
the FASTER project (http://www.bnl.gov/ exchange of forests: The EUROFLUX methodology, Adv. Ecol. Res., 30, 113175.
esm/) and ASR program. The authors are Black, P. G., E. A. DAsaro, T. B. Sanford, W. M. Drennan, J. A. Zhang, J. R. French, P. P. Niiler, E. J. Terrill, and E. J. Walsh (2007), Airsea exchange
particularly grateful to D. Richter and two in hurricanes: Synthesis of observations from the Coupled Boundary Layer AirSea Transfer Experiment, Bull. Am. Meteorol. Soc., 88(3),
anonymous reviewers for their careful 357374, doi:10.1175/bams-88-3-357.
reviews and valuable comments, which Bye, J. A. T., and A. D. Jenkins (2006), Drag coefcient reduction at very high wind speeds, J. Geophys. Res., 111, C03024, doi:10.1029/2005JC003114.
led to substantial improvement of this Donelan, M. A. (1990), Airsea interaction. The Sea, in Ocean Engineering Science, vol. 9, edited by B. LeMhaut and D. M. Hanes, 250 pp.,
manuscript. We thank all the crew mem- John Wiley, New York.
bers who took part in the campaigns and Donelan, M. A., B. K. Haus, N. Reul, W. J. Plant, M. Stiassnie, H. C. Graber, O. B. Brown, and E. S. Saltzman (2004), On the limiting aerodynamic
collected the valuable data. The authors roughness of the ocean in very strong winds, Geophys. Res. Lett., 31, L18306, doi:10.1029/2004GL019460.
would like to also thank Zhiqun Shi and Donelan, M. A., M. Curcic, S. S. Chen, and A. K. Magnusson (2012), Modeling waves and wind stress, J. Geophys. Res., 117, C00J23, doi:10.1029/
Wenchao Chen of the Guangdong 2011JC007787.
Climate Centre for their help with the Dyer, A. J. (1974), A review of ux-prole relationships, Boundary-Layer Meteorol., 7(3), 363372, doi:10.1007/BF00240838.
data on the tower on Zhizai Island. We Emanuel, K. A. (1995), Sensitivity of tropical cyclones to surface exchange coefcients and a revised steady-state model incorporating eye
thank our colleague, Zhongkuo Zhao for a dynamics, J. Atmos. Sci., 52(22), 39693976, doi:10.1175/1520-0469(1995)052<3969:sotcts>2.0.co;2.
lot of useful discussion and help in data Emanuel, K. A. (2003), A similarity hypothesis for airsea exchange at extreme wind speeds, J. Atmos. Sci., 60(11), 14201428, doi:10.1175/
processing. The data and code (in MATLAB 1520-0469(2003)060<1420:ashfae>2.0.co;2.
formats) used in this paper can be Finnigan, J. J. (2004), A re-evaluation of long-term ux measurement techniques Part II: Coordinate systems, Boundary-Layer Meteorol., 113(1),
obtained from the rst author. 141, doi:10.1023/b:boun.0000037348.64252.45.
Foken, T., M. Gockede, M. Mauder, L. Mahrt, B. Amiro, and W. Munger (2005), Handbook of Micrometeorology: Post-Field Data Quality Control,
edited by X. Lee, W. Massman, and B. Law, pp. 181208, Springer, Netherlands, doi:10.1007/1-4020-2265-4_9.
French, J. R., W. M. Drennan, J. A. Zhang, and P. G. Black (2007), Turbulent uxes in the hurricane boundary layer. Part I: Momentum ux,
J. Atmos. Sci., 64(4), 10891102, doi:10.1175/jas3887.1.
Garratt, J. R. (1992), The Atmospheric Boundary Layer, pp. 4155, Cambridge Univ. Press, Cambridge.
Hojstrup, J. (1993), A statistical data screening procedure, Meas. Sci. Technol., 4(2), 153.
Holthuijsen, L. H., M. D. Powell, and J. D. Pietrzak (2012), Wind and waves in extreme hurricanes, J. Geophys. Res., 117, C09003, doi:10.1029/
2012JC007983.
Huang, J., and P.-W. Chan (2011), Progress of marine meteorological observation experiment at Maoming of south China, J. Trop. Meteorol.,
17(4), 418429, doi:10.3969/j.issn.1006-8775.2011.04.012.
Jarosz, E., D. A. Mitchell, D. W. Wang, and W. J. Teague (2007), Bottom-up determination of air-sea momentum exchange under a major
tropical cyclone, Science, 315(5819), 17071709, doi:10.1126/science.1136466.
Jeong, D., B. K. Haus, and M. A. Donelan (2012), Enthalpy transfer across the airwater interface in high winds including spray, J. Atmos. Sci.,
69, 27332748, doi:10.1175/JAS-D-11-0260.1.
Kaimal, J. C., and J. J. Finnigan (1994), Atmospheric Boundary Layer Flows: Their Structure and Measurement, 289 pp., Oxford Univ. Press, New York.
Kessomkiat, W., H.-J. H. Franssen, A. Graf, and H. Vereecken (2013), Estimating random errors of eddy covariance data: An extended two-tower
approach, Agric. For. Meteorol., 171172(0), 203219, doi:10.1016/j.agrformet.2012.11.019.
Large, W. G., and S. Pond (1981), Open ocean momentum ux measurements in moderate to strong winds, J. Phys. Oceanogr., 11(3), 324336,
doi:10.1175/1520-0485(1981)011<0324:oomfmi>2.0.co;2.
Lee, X., B. Law, and W. Massman (2005), Handbook of Micrometeorology: A Guide for Surface Flux Measurement and Analysis, Springer Science
+ Business Media, Inc., Dordrecht.
Liu, C., Q. Wan, F. Liao, and Z. Zhao (2012), Surface observations in the tropical cyclone environment over the South Chian Sea, J. Trop.
Meteorol., 18(2), 263274, doi:10.3969/j.issn.1006-8775.2012.02.015.
Liu, G., Y. Liu, and S. Endo (2013), Evaluation of surface ux parameterizations with long-term ARM observations, Mon. Weather Rev., 141(2),
773797, doi:10.1175/mwr-d-12-00095.1.
Makin, V. K. (2005), A note on the drag of the sea surface at hurricane winds, Boundary-Layer Meteorol., 115(1), 169176, doi:10.1007/
s10546-004-3647-x.
Mauder, M., S. Oncley, R. Vogt, T. Weidinger, L. Ribeiro, C. Bernhofer, T. Foken, W. Kohsiek, H. R. Bruin, and H. Liu (2007), The energy balance
experiment EBEX-2000. Part II: Intercomparison of eddy-covariance sensors and post-eld data processing methods, Boundary-Layer
Meteorol., 123(1), 2954, doi:10.1007/s10546-006-9139-4.
Mauder, M., T. Foken, R. Clement, J. Elbers, W. Eugster, T. Grnwald, B. Heusinkveld, and O. Kolle (2008), Quality control of CarboEurope ux
data? Part 2: Inter-comparison of eddy-covariance software, Biogeosciences, 5(2), 451462.
Moon, I.-J., I. Ginis, and T. Hara (2004), Effect of surface waves on airsea momentum exchange. Part II: Behavior of drag coefcient under
tropical cyclones, J. Atmos. Sci., 61(19), 23342348, doi:10.1175/1520-0469(2004)061<2334:eoswoa>2.0.co;2.
Oh, H.-M., K.-J. Ha, K.-Y. Heo, K.-E. Kim, S.-J. Park, J.-S. Shim, and L. Mahrt (2010a), On drag coefcient parameterization with post processed
direct uxes measurements over the ocean, Asia-Pac. J. Atmos. Sci., 46(4), 513523, doi:10.1007/s13143-010-0030-3.
Oh, H.-M., K.-E. Kim, K.-J. Ha, L. Mahrt, and J.-S. Shim (2010b), Quality control and tilt correction effects on the turbulent uxes observed at an
ocean platform, J. Appl. Meteorol. Climatol., 50(3), 700712, doi:10.1175/2010jamc2367.1.
Panofsky, H. A., and J. A. Dutton (1984), Atmospheric Turbulence: Models and Methods for Engineering Applications, Wiley, New York.
Powell, M. D., P. J. Vickery, and T. A. Reinhold (2003), Reduced drag coefcient for high wind speeds in tropical cyclones, Nature, 422(6929),
279283.
Richter, D. H., and D. P. Stern (2014), Evidence of spray-mediated air-sea enthalpy ux within tropical cyclones, Geophys. Res. Lett., 41,
29973003, doi:10.1002/2014GL059746.
Richter, D. H., and P. P. Sullivan (2013), Momentum transfer in a turbulent, particle-laden Couette ow, Phys. Fluids (1994-present), 25(5),
doi:10.1063/1.4804391.

BI ET AL. DRAG COEFFICIENTS IN HIGH WINDS OVER SEA 15

 Journal of Geophysical Research: Atmospheres 10.1002/2015JD023172

Rogers, R., et al. (2013), NOAAS hurricane intensity forecasting experiment: A progress report, Bull. Am. Meteorol. Soc., 94(6), 859882,
doi:10.1175/BAMS-D-12-00089.1.
Schmid, H. P. (1994), Source areas for scalars and scalar uxes, Boundary-Layer Meteorol., 67(3), 293318, doi:10.1007/bf00713146.
Schmid, H. P. (1997), Experimental design for ux measurements: Matching scales of observations and uxes, Agric. Forest Meteorol., 87(23),
179200, doi:10.1016/S0168-1923(97)00011-7.
Sjblom, A., and A.-S. Smedman (2004), Comparison between eddy-correlation and inertial dissipation methods in the marine atmospheric
surface layer, Boundary-Layer Meteorol., 110(2), 141164, doi:10.1023/A:1026006402060.
Smith, R. K., and M. T. Montgomery (2014), On the existence of the logarithmic surface layer in the inner core of hurricanes, Q. J. R. Meteorol.
Soc., 140(678), 7281, doi:10.1002/qj.2121.
Soloviev, A. V., R. Lukas, M. A. Donelan, B. K. Haus, and I. Ginis (2014), The air-sea interface and surface stress under tropical cyclones, Sci. Rep.,
4, doi:10.1038/srep05306.
Tanner, C. B., and G. W. Thurtell (1969), Anemoclinometer measurements of Reynolds stress and heat transport in the atmospheric surface
layer. ECOM 66-G22-F, ECOM United States Army Electronics Command, Research and Development.
Troitskaya, Y. I., D. A. Sergeev, A. A. Kandaurov, G. A. Baidakov, M. A. Vdovin, and V. I. Kazakov (2012), Laboratory and theoretical modeling of
air-sea momentum transfer under severe wind conditions, J. Geophys. Res., 117, C00J21, doi:10.1029/2011JC007778.
Vickers, D., and L. Mahrt (1997), Quality control and ux sampling problems for tower and aircraft data, J. Atmos. Oceanic Technol., 14(3),
512526, doi:10.1175/1520-0426(1997)014<0512:qcafsp>2.0.co;2.
Wang, B., L. Song, and W. Chen (2013), Drag coefcient during strong typhoons, Adv. Meteorol., 8, doi:10.1155/2013/650971.
Zachry, B. C., C. W. Letchford, D. Zuo, and A. B. Kennedy (2013a), Laboratory measurements of the drag coefcient over a xed shoaling
hurricane wave train, Wind Struct. Int. J., 16(2), 193211, doi:10.12989/was.2013.16.2.193.
Zachry, B. C., J. L. Schroeder, A. B. Kennedy, J. J. Westerink, C. W. Letchford, and M. E. Hope (2013b), A case study of nearshore drag coefcient
behavior during hurricane Ike (2008), J. Appl. Meteorol. Climatol., 52(9), 21392146, doi:10.1175/jamc-d-12-0321.1.
Zhao, Z., Z. Gao, D. Li, X. Bi, C. Liu, and F. Liao (2013), Scalar uxgradient relationships under unstable conditions over water in coastal
regions, Boundary-Layer Meteorol., 148(3), 495516, doi:10.1007/s10546-013-9829-7.
Zhao, Z.-K., C.-X. Liu, Q. Li, G.-F. Dai, Q.-T. Song, and W.-H. Lv (2015), Typhoon air-sea drag coefcient in coastal regions, J. Geophys. Res.
Oceans, 120, 716727, doi:10.1002/2014JC010283.
Zweers, N. C., V. K. Makin, J. W. de Vries, and G. Burgers (2010), A sea drag relation for hurricane wind speeds, Geophys. Res. Lett., 37, L21811,
doi:10.1029/2010GL045002.

BI ET AL. DRAG COEFFICIENTS IN HIGH WINDS OVER SEA 16