Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Ocean Dynamics (2015) 65:375–384

DOI 10.1007/s10236-015-0811-4

Swell and the drag coefficient


Henry Potter

Received: 31 October 2014 / Accepted: 22 January 2015 / Published online: 21 February 2015
# Springer-Verlag Berlin Heidelberg 2015

Abstract Simultaneous measurements of waves and turbu- momentum via the momentum flux ! τ . In part, the mo-
lent fluxes were collected from a moored surface buoy in the mentum flux impacts the climate by driving currents,
Philippine Sea. Waves were partitioned into their wind sea and creating waves, triggering ocean mixing, influencing
swell components, and the ratio of swell to wind sea energy whitecap coverage, forcing aerosol production, and alter-
was used to assign a swell index. The 10-m neutral drag co- ing atmospheric and oceanic stability. The exchange of momen-
efficient was calculated using the eddy correlation method. tum at high wind speed is also important in tropical cyclones.
Four hundred hours of data were processed in 30 minute runs Being a crucial component of the climate system, the momen-
for wind speeds 8.5 to 16.5 m s−1 when the peak wave direc- tum flux is a key parameter used in atmospheric, oceanic, and
tion was within 90° of the wind direction and included obser- wave models, and with recent advances in coupled atmosphere–
vations during mixed seas, swell dominant, and wind sea ocean–wave modeling (e.g., Chen et al. 2013), this is even more
dominant conditions. The data were analyzed to explore the important.
influence of swell on the drag coefficient. It was found that Direct measurement of ! τ requires sampling turbulent
when compared to periods of equal wind speed, the drag co- fluctuations of the horizontal downwind u and cross-
efficient was reduced up to 37 % when swell energy was twice wind v components of the velocity and correlating them
that of the wind sea energy. It is believed that this reduction with the vertical component w. In stationary and ho-
was due to a decrease in the turbulent flux around the swell mogenous conditions, ! τ is assumed to be constant
frequency, suggesting that the swell diminishes the surface within the surface flux layer and above the viscous
aerodynamic roughness. sublayer and is calculated from:
h   i
Keywords Dragcoefficient . Waves . Swell . Momentumflux !
τ ¼ρ −u0 w0 bi þ −v0 w0 bj ¼ ρu2* : ð1Þ

Here ρ is air density, bi and bj are unit vectors along and per-
pendicular to the mean wind direction, overbars represent time
1 Introduction average (O 30 min), primes denote fluctuating components
0  
u ¼ v0 ¼ w0 ¼ 0, and u* is the friction velocity. Given the
The marine atmospheric boundary layer (MABL) is a dynam- difficulty of measuring fluxes at sea, many applications, espe-
ic region where the ocean is strongly coupled with the atmo- cially those that require extensive spatial coverage, use a bulk
sphere. Within the MABL, energy is exchanged across the air- formula to determine the momentum flux, which is conven-
sea interface through the vertical transport of horizontal tionally written as:

jτ j ¼ ρC Dz U 2z ð2Þ
Responsible Editor: Jörg-Olaf Wolff
H. Potter (*)
where U z and C Dz are the mean wind speed and drag
Naval Research Laboratory, 4555 Overlook Ave., SW,
Washington, DC, 20375 USA coefficient, respectively, at height z above the surface.
e-mail: henry.potter.ctr@nrl.navy.mil Monin and Obukhov (1954) established a similarity theory
376 Ocean Dynamics (2015) 65:375–384

which relates the gradient of U to the u* through a universal


dimensionless gradient function φu:
∂U u
¼ φ ðζ Þ: ð3Þ
∂z κz u

Here, κ≈0.4 is the Von Kármán constant and ζ=z/L is the


Monin–Obukhov stability parameter. L is the Obukhov length
(Obukhov 1946) which accounts for the effect of buoyant
production of turbulence on the wind profile:

u3 θ
L¼− * v  ð4Þ
κg w0 θ0 v

where g=9.8 m s−2 is acceleration due to gravity, θv is the


Fig. 1 Example of wave partitioning from 1-D wave spectrum taken
mean virtual temperature, and w0 θ0 v is the virtual temperature from a period of U10N = 8.9 m s−1. The vertical dashed line is the
flux. Integrating Eq. (3), the velocity profile becomes: frequency f1 =g/(2πU101.2) which partitions ES (swell energy) from EW
    (wind wave energy)
u z
U z −U 0 ¼ ln −ψu ðζ Þ ð5Þ
κ z0 wave speed at the peak frequency, to account for the speed of
the roughness elements in relation to the wind. This was first
where U0 is mean wind speed at the surface (typically set to 0), supported by Donelan (1982) whereby yielding an expected
z0 is the surface roughness length for momentum (usually drag coefficient relationship of CDN =f(U10N, Cp /u*). Howev-
defined relative to zero velocity), and ψu(ζ) is the integrated er, attempts to quantify the influence of wave age on surface
form of φu(ζ). In neutral conditions, where buoyant forcing roughness are often hampered by self-correlation issues which
becomes negligible, |z/L| goes to 0 whereby the last term in arise because the roughness length is calculated from the fric-
Eq. (5) becomes zero and the classic logarithmic mean wind tion velocity via Eq. (5), as such both the roughness length and
profile is formed. The expression for the 10 m neutral drag the wave age are determined using u* and can result in spuri-
coefficient can then be written as CD10N =u2*/U210N which can ous correlation. Furthermore, although wave age dependence
be expressed as a unique function of the roughness length: reasonably well describes the wind stress under pure wind
seas, it has been shown that swell obscures the relationship
κ2
CD10N ¼  2 ð6Þ between roughness and wave age (e.g., Drennan et al. 2005)
ln 10
z0 limiting its applicability under conditions when the ocean has
a significant swell component. This was demonstrated by
Physically, z0 is a measure of the upwind terrain roughness Smedman et al. (2003) who showed that as the energy of
experienced by the wind due to the surface roughness waves moving faster than the wind increases relative to the
elements and viscosity. Theoretically, z0 is the height where energy of waves generated by the wind, the dependence of the
U goes to 0. Using dimensional reasoning, Charnock (1955) drag coefficient on wave age becomes weaker.
proposed an empirical expression which accounts for ampli- The influence of swell on stress depends on the wind speed
fied roughness length from wave growth due to surface stress and swell properties and on the angle between the wind and
as wind speed increases: z0 =αu2*/g: wheree α is the Charnock swell (e.g., Pan et al. 2005). When swell propagates along the
parameter. Charnock took α to be a constant, thus this expres- wind direction, it tends to decrease surface drag (e.g., Drennan
sion for z0 can be substituted into Eq. (6) and CD10N can be et al. 1999; García-Nava et al. 2009) because swell is thought
expressed as a function of U10N alone: CD10N =f(U10N). Typ- to modify the wind sea roughness associated with younger
ical values of α range from 0.012 to 0.035 (Wu 1980). waves whereby limiting the frictional coupling between the
Several field campaigns have found that the drag coeffi- atmosphere and ocean (Donelan and Dobson 2001). When
cient increases linearly with wind speed, e.g., HEXOS (Smith opposing the wind, swell has also been found to increase the
et al. 1992); SWADE (Donelan et al. 1997); AWE (Drennan surface drag (e.g., Donelan et al. 1997). Through its impact on
and Shay 2006). However, due to the scatter inherent in these surface roughness, swell is, in part, liable for variability in the
datasets (e.g., Fig. 1 of Drennan, 2005a) no single parame- drag coefficient that cannot be attributed to wind speed. Being
terization sufficiently models all the data. Kitaigorodskii and able to quantify the influence of swell’s impact on surface drag
Volkov (1965) suggested that α should be a function of wave is essential for improving modeling efforts that rely on param-
related parameter such as the wave age Cp /u*, where Cp is the eterization of momentum exchange across the air–sea
Ocean Dynamics (2015) 65:375–384 377

interface such as the forecasting of waves, currents, and trop- EASI to operate as a single point triplet surface follower wave
ical storms which have direct impact on spill combat, ship- buoy. More information about EASI as a wave buoy can be
ping, recreation, fisheries, and search and rescue efforts, found in Collins III et al. (2014a, b) and further information
amongst others. Despite this, lack of observational data means about the ITOP experiment can be found in D’Asaro et al.
substantial gaps in our ability to quantify the momentum flux (2011) and Drennan et al. (2014).
exist and underscore an inherent need for more MABL mo- Data were processed in runs of 30 min. Each run was
mentu m flux measurements in order to improve analyzed for spikes in u, v, and w. Spikes were generally rare
parameterizations. and those which were both isolated and greater than four
In this study, concurrent measurements of direct fluxes and standard deviations from the run mean were interpolated
waves from a moored surface buoy are presented in order to through. Any runs for which more than 0.5 % of values fell
quantify the damping by swell of the small-scale wind-driven outside four standard deviations were discarded, accounting
waves which support the surface roughness. Data collected for the removal of 17 runs. Data quality of individual runs
during the Impact of Typhoons on the Ocean in the Pacific was assessed by inspecting both the linear cumulative sum-
(ITOP) experiment are analyzed over wind speeds of 8.5 to mations of the covariance, and the cumulative integrals of
16.5 m s−1 and include measurements made in three typhoons cospectrum for the downwind and crosswind stress (see
and with significant wave heights between 1 and 8.5 m. The Fig. 3 of Potter et al. 2015); 72 runs were removed in this
diversity of sea states recorded during ITOP makes this data manner. Following quality control, the wind vector was ro-
set ideally suited to examine the interaction between the tated so that u pointed into the mean wind direction and
waves and momentum flux. The impact of swell is assessed u ¼ w ¼ 0. Mean wind speeds were raised to their 10 m
by comparing CD10N for different swell indices which are neutral equivalent values following Monin and Obukhov
determined from the ratio of swell to wind sea. Results show (1954). Drennan et al. (2014) investigated the effect of
that the presence of swell dampens the surface roughness EASI’s motion on wind speed due to changes in anemome-
resulting in a reduction in CD10N up to 37 %. This result helps ter height caused by platform tilt. They found that reduction
explain some of the variability in the drag coefficient and has in mean anemometer height was at most 1 % and had a
important implications to future modeling efforts which cur- negligible effect on measured wind speed.
rently parameterize the drag coefficient as a unique function of
wind speed.
This paper is laid out as follows: in Section 2, the ITOP 3 The swell index
data is introduced; Section 3 contains a description of the
swell index; in Section 4, results are presented; Section 5 is Wave energy was separated into the swell ES and wind sea EW
reserved for the discussion; and summary is in Section 6. components using 1-D wave partitioning (e.g., Smedman
et al. 2003):
Z f1
2 ITOP data
ES ¼ S ð f Þd f ð7Þ
0
Data were collected from an extreme air–sea interaction
(EASI) buoy (Drennan et al. 2014) between August and De- Z ∞
cember 2010 as part of the ITOP experiment. EASI was de- EW ¼ S ð f Þd f ð8Þ
f1
ployed by the R/V Roger Revelle in the Philippine Sea around
127°E 21°N, 750 km east of Taiwan, in approximately
Where
5,500 m of water depth where it was anchored to the seabed
via ~7,000 m of line attached to a ~3,100-kg cast iron anchor. g
f1 ¼ ð9Þ
EASI is an adaptation of the Navy Oceanographic Meteoro- 2π1:2U 10
logical Automatic Device and was fully instrumented to mea-
sure air-sea fluxes of momentum, heat, and mass, as well as Here, S(f) is the 1-D wave spectrum, f is frequency, and f1 is
mean meteorological and oceanographic parameters. Momen- the separation frequency. This partitioning is based on the CP /
tum fluxes were recorded using a Gill R2A sonic anemome- U10 definition of wave age for which CP /U10 ≥1.2 corresponds
ters at 20 Hz and stored to a below deck data acquisition to swell conditions (Pierson and Moskowitz 1964). Applying
system. The sonic was mounted on the rear mast 5.45 m above this to the equation for phase speed at the peak frequency CP =
mean sea level and positioned to minimize the effects of flow g(2πf1)−1 yields the swell definition g(2πU10 f1)−1 ≥1.2 from
distortion. EASI’s motion was recorded by two full motion which Eq. (9) is formed.
packages. The buoy motion was used to motion correct the The definition of f1 is slightly different from the one pre-
sonic data following Anctil et al. (1994) and also enabled sented by (Smedman et al. 2003). In the original definition,
378 Ocean Dynamics (2015) 65:375–384

U10 was multiplied by a factor of cosθ, where θ is the differ- that the waves were generally moving faster than the wind and
ence between the wind and wave angle at the peak of the wave conditions were SD. Runs of δ3 taper off as U10N increases and
spectrum. This term originates from the wave age parameter do not exist above ~14 m s−1. Runs for δ2 (black) fill the
CP /(U10cosθ), which was motivated by the need to account for intermediate wind speed range between ~6 and 16 m s−1 as
the wind component in the wave propagation direction. How- the relative amount of swell to wind sea energy is decreased.
ever, swell effects on the MABL are thought to be similar for Cases with δ1 (green) are prevalent at moderate to high wind
both swell traveling 90° relative to the wind direction and speeds (above ~8 m s−1) when the waves largely moved
wind following swell (Högström et al. 2009; Smedman et al. slower than the wind and conditions were WSD.
2009). Thus, inclusion of the cosine term in the wave age
definition may lead to erroneous classification of sea state,
as discussed by Högström et al. (2011). The inclusion of the 4 Results
factor 1.2 in the denominator of Eq. (9) also differs from the
original definition of f1 which separated the spectrum into To focus on the relationship between waves and the momen-
waves moving faster and slower than the wind. Here instead, tum flux, a region of interest from 8.5 to 16.5 m s−1 was
the partitioning variation used by Sahlée et al. (2012) was isolated for closer examination. Within this range, all swell
applied and the 1.2 term was included in the denominator such indices are well represented and the drag coefficient increases
that ES represented the swell part of the spectrum and EW linearly with wind speed. Above this range, only δ1 exists and
represented the waves influenced by the local wind. A typical below this range there is increased scatter in the drag coeffi-
1-D wave partition is shown in Fig. 1. The swell ratio was cient which may be attributed to gustiness, which causes de-
determined from ES /EW (denoted δ) and used to quantify the parture from Monin-Obukhov similarity theory, and smooth
relative amount of swell energy for each run. Runs were flow, in which viscosity helps support surface stress (see
indexed according to their swell ratio into one of three groups: Drennan (2005a) for a discussion of these parameters). These
ES /EW <0.5 (δ1), 0.5≤ES /EW ≤2 (δ2), or ES /EW >2 (δ3), where considerations are beyond the scope of this paper.
higher values relate to greater swell energy relative to wind In Fig. 3, friction velocity is shown as a function of wind
sea. Note that δ1 represents wind sea dominant (WSD) condi- wave energy. The data are separated in EW bins of 0.1 m2.
tions because the wind sea energy is at least double the swell Within each bin, the data are plotted according to their respec-
energy, and δ3 represents swell dominant (SD) conditions be- tive swell indices and shown with 90 % confidence intervals
cause the swell is at least twice as energetic as the wind sea. around the friction velocity. Higher friction velocities occur
To satisfy sea state classification established by Donelan during periods of greater wind sea energy, and u* decreases
et al. (1985), which states that only one component of wave within each bin as the relative amount of swell increases even
energy is designated wind sea, analysis was restricted to wave though the amount of wind sea energy remains approximately
spectra with, at most, a single dominant peak in the wind wave equal. These results show that surface roughness elements are
frequency range. Data was further restricted to conditions associated with wind waves and that the friction velocity is
when the mean wind was traveling within 90° of the peak reduced in the presence of swell. At its lowest value, u* ≈
(dominant frequency) waves, in order to prevent waves mov- 0.3 m s−1 which is consistent with Foreman and Emeis
ing faster than, but against, the wind being erroneously classed (2010) who found that for U10 ≥8 m s−1 the sea surface tran-
as wind waves. This classification of wind waves is more sitions to become aerodynamically rough and boasts a friction
liberal than was proposed by Donelan et al. (1985) who de- velocity of ~0.27 m s−1. The lack of separation between swell
fined them as waves traveling within 45° of the wind. How- indices at low EW suggests that swell may have limited impact
ever, in this study, data are grouped as either θ≤45° or 45° during this transition regime. Combining results from Figs. 2
<θ≤90° and processed separately to explore if swell’s effects and 3, it can be inferred that the drag coefficient is both a
on the surface roughness is similar for swell traveling function of the wind speed and the characteristics of
within 90° relative to the wind direction and for wind the wave field.
following swell. In Fig. 4, CD10N is plotted as a function of U10N for≤45 (a)
The result of partitioning and indexing the waves is and 45°<θ≤90° (b). Plots are delineated by swell index each
displayed in Fig. 2 where CD10N vs. U10N is plotted for runs with a linear best fit. The drag coefficient increases with wind
up to 20 m s−1 and θ≤90°. Coefficient CD10N increases ap- speed for each swell index while higher swell indices appear
proximately linearly with U10N, but there is scatter throughout to have reduced drag coefficient.
depicting approximately two-to-threefold variability in CD10N In Fig. 5, the regression lines from Fig. 4 are shown but
which cannot be attributed to wind speed. The plot contains 1, now include 90 % confidence intervals. The drag coefficient is
364 values for which the number of runs N within each swell seen to be significantly influenced by the presence of swell for
index are as follows: δ1: N=345, δ2: N=495, and δ3: N=524. θ≤45° (Fig. 5a). This assertion can be made when comparing
At lower wind speeds the swell index was high (blue) showing δ1 to δ3 (98 % confidence) or δ2 to δ3 (90 % confidence), but
Ocean Dynamics (2015) 65:375–384 379

Fig. 2 CD10N as a function of


U10N for runs up to 20 m s−1 and
θ≤90°. The points are color coded
according to the swell index as
follows: δ1 (green), δ2 (black),
and δ3 (blue). Note that some
points are plotted behind others

no distinction can be made between measurements for which 5 Discussion


0<δ<2. Results show that for wind following swell, when the
amount of swell energy is at least twice the energy in the wind 5.1 Momentum flux
sea (i.e., conditions are SD), a distinctive reduction in the drag
coefficient occurs compared to conditions of equal wind speed In neutral conditions, CDN is thought to have a one-to-one
but higher relative wind sea energy. An increase in swell index relationship with the roughness length z0 (Eq. 6), which is a
from δ2 to δ3 was found to reduce CD10N up to 16 %, measure of the ocean roughness elements. This roughness is
while an increase from δ1 to δ3, i.e., transitioning from supported by the wind sea waves which themselves grow
WSD to SD conditions, is associated with a reduction under direct forcing from the wind whereby increasing the
of CD10N ranging from 5 to 37 %. This reduction range surface stress. As such, wind speed is directly related to am-
of CD10N was effectively the same throughout the entire plified roughness and an increased drag coefficient. However,
wind speed range. it has been shown here that, for wind speeds from 8.5 to
For 45°<θ≤90° (Fig. 5b), the regression value of CD10N 16.5 m s−1 and θ≤45°, an increase in the ratio of swell to wind
was reduced by 19 % at 8.5 m s−1 and 26 % at U10N of sea energy is also associated with reduction of the drag coef-
14 m s−1 between δ1 and δ3. However, the decrease was only ficient insomuch as an increase in the amount of swell energy
found to be significant at 60 % confidence. These results may relative to wind sea energy is linked to reduced surface rough-
be attributed to the small sample size of δ3 (N=14, of which ness. This relationship was also explored for 45°<θ≤90° for
only three data points were above 10 m s−1) rather than the which results suggest a similar mechanism exists but could not
lack of a relationship, but results remain inconclusive. No be confirmed due to a small sample size. The reduction of
difference was found in CD10N when comparing δ2 to either CD10N reached maximum when SD conditions were compared
δ1 or δ3. to WSD conditions, for which a reduction in the drag coeffi-
cient between 5 and 37 % was found.
The bearing of swell on momentum flux is visualized in
Fig. 6 (top row) where 3 h of u-w cospectra are plotted for three
periods during ITOP. The mean of the six runs in each plot is
shown by a black line and the red dashed lines are universal
curves from Miyake et al. (1970) dimensionalized by mean
values of u*, U10N, and z. Concurrent mean wave spectra are
shown in Fig. 6 (bottom row) where f1 from Eq. (9) is indicated
by a gray dashed line. For each period, (a), (b), and (c), the angle
between wind and waves at peak frequency was narrow, but
each had unique wave spectra, wind speed, wave age, wave
height and swell ratio, which are summarized in Table 1. During
period (a), conditions were swell dominant (δ=2.5, U10N /CP =
0.54); during period (b), conditions were developing with mixed
seas (δ=0.5, U10N /CP =0.78); and in period (c), the wind sea was
dominant (δ=0.02, U10N /CP =1.3).
The mean cospectrum in Fig. 6(a) is similar to the universal
Fig. 3 Friction velocity u* is presented as a function of wind sea energy
EW. Data are divided into EW bins of 0.1 m2 and further separated within curve of Miyake et al. (1970); however, there is a dip in the
each bin according to their swell indices cospectrum around the frequency of the swell system and a
380 Ocean Dynamics (2015) 65:375–384

Fig. 4 Drag coefficient as a


function of wind speed for θ≤45°
(a) and 45°<θ≤90° (b). Data are
separated according to their swell
indices as shown in Fig. 2. a N=
287 (green), 199 (black), and 76
(blue); b N=122 (green), 99
(black), and 14 (blue). Note that
some points are plotted behind
others

departure from the universal curve at frequencies below the universal curve at lower frequencies. In both Fig. 6a, b, the
swell. Despite reasonable run-to-run variability, the mean departure from the Miyake et al. (1970) curve at lower fre-
cospectrum in Fig. 6b also adheres well to the universal curve. quencies lies around the frequency range of the swell system.
There is some departure from the universal curve at frequen- This is consistent with previous results (cf. Figure 11 of
cies below the swell, but there is no systematic dip at the swell García-Nava et al. (2009)) which showed that the momentum
frequency. The mean cospectrum in Fig. 6c is a very close transfer is reduced, or even upward, at lower frequencies and
portrayal of the Miyake et al. (1970) curve and supports the frequencies coincident with swell. A sizable dip in the mean u-
work of Drennan et al. (1999) who also found copsectra fol- w cospectrum was also observed by Sahlée et al. (2012) which
low this curve under similar conditions. coincided with the swell frequency; but due to the run-
The departure of the u-w cospectra from the universal curve to-run variability, the authors claimed that the data did
visible in Fig. 6a was a reoccurring trait during wind- not systematically differ from the universal curve. Both
following swell conditions and an index=δ3, which resulted mean cospectra in Fig. 6a, b are reduced at lower fre-
from decreases in turbulent flux at the swell frequency. This quencies around the swell but in column (a) reduction
feature does not exist systematically in the cospectra in in cospectra occur systematically, are more pronounced,
Fig. 6b, but the mean curve does depart slightly from the and occasionally drop below zero whereby contributing

Fig. 5 Drag coefficient as a


function of wind speed for θ≤45°
(a) and 45°<θ≤90° (b). Linear
regressions for each swell index
are plotted as in Fig. 4, but 90 %
confidence intervals are also
included
Ocean Dynamics (2015) 65:375–384 381

Fig. 6 Top row: mean u-w


cospectra for periods (a), (b), and
(c) during ITOP. Individual runs
are in gray, the mean of the six
runs is in black, and the red
dashed lines are universal curves
from Miyake et al. (1970)
dimensionalized by mean values
of u*, U10N, and z. Bottom row
concurrent 3 h mean wave spectra
are with f1 from Eq. (9) indicated
by a gray dashed line.
Atmosphere and ocean conditions
for these periods are summarized
in Table 1

an upward momentum flux, causing a reduction in u*, 5.2 Wave partitioning


and subsequently decreasing in C D10N during swell
dominant conditions. While the mean cospectrum is re- In order to corroborate the wave partitioning method used in
duced at lower frequencies in Fig. 6b, it is above the this study, here wave energy is compared to results from pre-
universal curve at frequencies between ~0.1 and 2.5 Hz vious studies. Following the similarity theory (Kitaigorodskii
due to increased wind sea (because of EASI’s large 1962; 1973), it was proposed that the total energy of a wave
hull, a cut off exists at ~0.45 Hz and waves with great- system should be a unique function of the wind speed. Using
er frequency cannot be resolved). Because wind sea dimensional reasoning, the total energy Etot of a wave system
provides the necessary roughness elements, reduction can be expressed as:
of momentum transfer at lower frequencies due to swell
is seen to be offset in this case by increased energy εU 4re f
E tot ¼ ð10Þ
transfer at higher frequencies. In Fig. 6c, there is no g2
swell; consequently, there is no attenuation of the sur-
face roughness and no departure from the universal where Uref is the reference wind speed (set=U10N) and ε is the
curve, meaning CD is not reduced. proportionality coefficient for total energy. This equation de-
The drag coefficient has been shown to have an in- scribes the limits of fetch limited wave growth. Alves et al.
verse relationship with stability (e.g., Kara et al. 2005) (2003) determined ε = 3.64×10−3 by reevaluating experimen-
and could influence the results here. To explore this tal data of Moskowitz (1964) and Pierson and Moskowitz
possibility, atmospheric stability for each of the wave (1964). Kahma and Calkoen (1992) determined ε using di-
classes was calculated. Results showed condition were mensionless fetch X e ¼ gX =U 2 , where X is fetch in m, and
10
generally less unstable when the swell index was δ3 found ε = 9.25×10−7 X e 0.766 for stable and ε = 5.38×
compared to δ1 or δ2. However, mean z/L between each 10−7 X e 0.94 for unstable conditions. In Fig. 7, EW as defined
index varied by less than 0.04, and were not significant- by Eq. (8) and (9) are compared to results from these previous
ly different (95 % confidence). studies. The data for each swell index is fitted with a line of its
respective color. For comparison, also plotted are the curves
Table 1 Mean swell calculated from Alves et al. (2003; red) and Kahma and
ratio δ, wind speed U10N, Period (a) (b) (c)
Calkoen (1992) for stable (lower gray) and unstable (upper
significant wave height
HS, off-wind peak wave δ 2.5 0.5 0.02 gray) conditions. The Kahma and Calkoen curves in Fig. 7 were
angle θ, and inverse U10N 9.5 14.1 12.9 calculated using Xe = 8,000, which was shown by the authors to
wave age U10N /CP, for HS 2.3 3.6 2.4 be the point where fully developed conditions are approached.
periods (a), (b), and (c) In order to complete a fair comparison between the observed
plotted in Fig. 6 θ 17.5 14 36
U10N/CP 0.54 0.78 1.3 and parameterized values in Fig. 7, it was necessary to address
the fact that Eq. (10) determines the energy of the entire wave
382 Ocean Dynamics (2015) 65:375–384

Högström et al. (2013). The authors reported that the wind


profile was characterized by a wind speed increase from the
surface to a height of about 7–8 m above which a layer of
virtually constant wind speed existed, indicating a departure
from the logarithmic wind profile predicted by similarity the-
ory. In order to explore the possibility that this feature existed
during ITOP and influenced the results of this study, two tests
of quality control were preformed. Firstly, wind speeds from
EASI measured at 5.45 m and within the region of logarithmic
wind profile observed by Högström et al. (2013) were com-
pared to those from R/V Roger Revelle measured at 17 m,
where wind speed would essentially be constant with height.
Should a deviation from similarity theory exist, U10N deter-
mined from each of these heights would not agree (with EASI
overestimating U10N and R/V Roger Revelle underestimating
Fig. 7 Wind wave energy EW as a function of U 210N. The dots are the it). However, when within 5 km of each other, the sensors
swell indices δ3 (blue), δ2 (black), and δ1 (green) as shown in Fig. 2; each
has a line of fit in its respective color. The red dashed line is determined
showed excellent agreement even during swell (r=0.92), sug-
from Alves et al. (2003) and the upper and lower gray lines show EW gesting any deviation from the classic wind profile was neg-
calculated from Kahma and Calkoen (1992) for stable and unstable ligible. A second sensitivity analysis was conducted by re-
conditions, respectively analyzing the ITOP data using U7N rather than U10N in order
to stay below the region of constant wind speed observed by
Högström et al. (2013). The impact on results was negligible
system rather for which the peak frequency satisfies f1 ≥ (not shown), confirming any deviation from a logarithmic
g(2π1.2U10)−1; whereas, observed values of EW only consider wind profile was nominal.
the wind sea component even when f1 ≥g(2π1.2U10)−1 is satis-
fied. In order to account for this, the results of Eq. (10) were 5.3 Wave age
multiplied by the ratio EW /Etot calculated from observed values
for 8.5 m s−1 ≤U10N ≤16.5 m s−1 and θ<90°. The mean value of Previous wave age studies have shown a link between swell
EW /Etot was found to be 0.66±0.02, which is consistent with and the drag coefficient (e.g., Drennan et al. 2003, 1997).
Sahlée et al. (2012) who, using the same technique, found However, more recently, Andreas et al. (2012) determined that
EW /Etot =0.73±0.02. friction velocity is almost independent of wave age, which
The curves in Fig. 7 for δ1, δ2, and δ3 are well predicted by suggests that a relationship is ambiguous. The authors also
the Alves et al. (2003) and unstable Kahma and Calkoen showed swell slightly increases rather than decreases surface
(1992) curves, lending credence to the wave energy drag over the range 8.5 m s−1 <U10N 16.5 m s−1, which con-
partitioning method applied to the ITOP data. The Alves flicts with the results of this study. For comparison, the data
et al. (2003) curve slightly overestimates observed EW and from this manuscript were used to determine if the results using
the Kahma and Calkoen (1992) unstable curve slightly under- ES/EW are reflected in a wave age analysis. To this end, the
estimates EW, but the differences are nominal suggesting that data were classified as either wind sea or swell using the wave
EW during ITOP was similar to previous field campaigns. Due age partitioning of Pierson and Moskowitz (1964), and plotted
to the unstable conditions that prevailed throughout the exper- as CD10N vs.U10N (Fig. 8). No evidence of decreased drag
iment, the significant underestimate of the stable condition coefficient was found for following or cross swell compared
curve of Kahma and Calkoen (1992) was not unexpected. to wind sea conditions. This is in relative agreement with
The slopes of δ1 and δ2 plotted in log-log both equaled 2, Andreas et al. (2012) and suggests the wave age parameter
verifying that EW is proportional to U 410N and that these data may have limited utility in drag coefficient analysis, while
adhere to the predictions of similarity theory. For δ3, the slope highlighting the applicability of using the swell ratio.
was 2.3 indicating a slight departure from similarity theory.
This will be addressed below. A similar plot of ES vs. U 210N 5.4 The Charnock parameter
showed no relation, as expected.
The log-log slope of 2.3 for δ3 in Fig. 7 suggests that a The results of this study highlight the importance of including
slight deviation from similarity theory may have existed in a wave related component in parameterizations of surface
swell dominated conditions during ITOP. Wind profile mea- roughness. Previous studies (e.g., Donelan et al. 1993) have
surements under similar conditions of swell traveling predom- used the wave age for this purpose in order to account for the
inantly in the wind direction have previously been studied by greater frictional coupling between the ocean and atmosphere
Ocean Dynamics (2015) 65:375–384 383

90°, resulting in 400 h of data. Waves were partitioned into


their swell ES and wind sea EW components using 1-D wave
spectra, and for each run the swell ratio EW /ES was deter-
mined. Runs were grouped according to swell index: ES /EW
<0.5 (denoted δ1), 0.5≤ES /EW ≤2 (δ2), or ES /EW >2 (δ3), where
higher values relate to an increase in the relative amount of
swell. Using the eddy correlation method, the 10 m neutral drag
coefficient CD10N was also determined.
The data were examined to explore the influence of waves
on the drag coefficient independent of variability due to wind
speed. For θ≤45°, a decrease in the drag coefficient was ob-
served with increasing swell index. When comparing δ1 to δ3,
Fig. 8 CD10N vs. U10N. Data are classified as either wind sea (orange), CD10N decreased up to 37 % (98 % confidence); for indices
following swell (gray), or cross swell (purple). The black dashed line is from δ2 to δ3, a decrease up to 16 % was observed (90 %
the linear best fit of all the data
confidence); these decreases were independent of wind speed.
No difference was found between δ2 and δ3. For 45°<θ≤90°,
associated with younger waves. However, quantifying the in-
a reduction in CD10N was only found when the swell index
fluence of the wave age on surface roughness is difficult be-
increased from δ1 to δ3, i.e., when comparing wind sea dom-
cause it is vulnerable to self correlation issues and has limited
inant to swell dominant conditions. The low confidence of this
applicability during swell. The use of the dimensionless swell
result (60 %) is thought to be a consequence of the small
index introduced here avoids these problems. Although results
sample size rather than lack of a relationship. The decreased
here were hampered by their own limitation of only being
drag coefficient in the presence of swell was linked to a re-
assessed during wind-following-swell, such conditions were
duction in the momentum flux at and around the swell fre-
prevalent for U10N between 8.5 and 16.5 m s−1, as such this
quency. This reduction was most prevalent during swell dom-
limitation proved minor.
inant conditions which featured a reduction in the u-w
Previous findings that the Charnock parameter α=gz0 /u2*
cospectra decreasing the momentum flux at and below the
(Charnock 1955) should not be treated as a constant but
swell frequency.
should depend on some measure of the ocean surface are
An evaluation of the Charnock parameter α for δ1, δ2, and
supported by the results of this study. The mean Charnock
δ3 returned values of 0.013, 0.012, and 0.005, respectively.
parameter for δ1, δ2, and δ3 was 0.013, 0.012, and 0.005,
For δ1 and δ2, α was consistent with previous studies, but α
respectively. No significant difference was found between δ1
for δ3 was appreciably lower, illustrating the reduced frictional
and δ2, but both were significantly different from δ3 (>90 %
coupling between wind and waves in swell dominant
confidence). Values of α for δ1 and δ2 are consistent with
conditions.
previous studies (e.g., Kraus and Businger 1994), α for δ3
was appreciably lower, illustrating the reduced frictional cou- Acknowledgments ITOP was funded by Office of Naval Research
pling between wind and waves in SD conditions. The swell under grant N0014-09-1-0392 with additional support from National Sci-
ratio can be incorporated into the Charnock expression, α= ence Foundation (OCE-0526442) for the development of the EASI buoy,
gz0 /u2* =f(ES /EW,θ), which translates to the drag coefficient as and Office of Naval Research (DURIP N00014-09-0818) for funding
construction of the second EASI buoy. I appreciate input and guidance
CDN = f(UN,ES /EW, θ). This outcome complements results from colleagues at the University of Miami that worked on the ITOP
from previous studies that link wave characteristics as well project, especially Will Drennan, Hans Graber, and Tripp Collins. I am
as wind speed to surface drag. Parameterization of CDN in also grateful to the captains and crew of the R/V Roger Revelle. Finally, I
terms of the ES /EW and UN will lead to improved results when acknowledge the support of the National Research Council for my Post-
doctoral Research Associate fellowship.
compared to using simple linear wind speed dependence and
should be incorporated into future modeling efforts.
References

6 Summary Alves JHGM, Branner ML, Young IR (2003) Revisiting the Pierson-
Moskowitz asymptotic limits for fully-developed wind waves. J
Direct measurements of momentum fluxes and waves were Phys Oceanogr 33:1301–1323
made from a floating platform over 87 days in the Philippine Anctil F, Donelan MA, Drennan WM, Graber HC (1994) Eddy-
correlation measurements of air-sea fluxes from a discus buoy. J
Sea in 2010. Data were processed in runs of 30 min. Analysis Atmos Ocean Technol 11(4):1144–1150
was restricted to runs with U10N from 8.5 to 16.5 m s−1 and Andreas EL, Mahrt L, Vickers D (2012) A new drag relation for aerody-
when the angle between peak frequency waves and wind θ≤ namically rough flow over the ocean. J Atmos Sci 69(8):2520–2537
384 Ocean Dynamics (2015) 65:375–384

Charnock H (1955) Wind stress on a water surface. Quart J Royal Högström U, Rutgersson A, Sahlée E, Smedman AS, Hristov TS,
Meteorol Soc 81:639–640 Drennan WM, Kahma KK (2013) Air–sea interaction features in
Chen SS, Zhao W, Donelan MA, Tolman HL (2013) Directional wind– the Baltic Sea and at a Pacific trade-wind site: an inter-comparison
wave coupling in fully coupled atmosphere–wave–ocean models: study. Bound-Layer Meteorol 147(1):139–163
results from CBLAST-hurricane. J Atmos Sci 70(10):3198–3215 Kahma KK, Calkoen CJ (1992) Reconciling discrepancies in the ob-
Collins CO III, Lund B, Waseda T, Graber HC (2014) On recording sea served growth of wind-generated waves. J Phys Oceanogr 22:
surface elevation with accelerometer buoys: lessons from ITOP 1389–1405
2010. Ocean Dyn 64(6):895–904 Kara AB, Hurlburt HE, Wallcraft AJ (2005) Stability-dependent ex-
Collins III CO, Lund B, Ramos R, Drennan WM, Graber HC (2014b) change coefficients for air-se fluxes. J Atmos Ocean J Atmos
Wave measurement inter-comparison and platform evaluation dur- Ocean Tech 22:1080–1094
ing the ITOP (2010) experiment. J Atmos Ocean Tech in press Kitaigorodskii SA (1962) Applications of the theory of similarity to the
D’Asaro EA, Black PG, Centurioni LR, Harr P, Jayne SR, Lin II, Lee analysis of wind-generated wave motion as a stochastic process. Izv
CM, Morzel J, Mrvaljevic RK, Niiler PP, Rainville L, Sanford TB, Akad Nauk SSR, Geophys Ser 1:105–117
Tang TY (2011) Typhoon-ocean interaction in the western North Kitaigorodskii, SA (1973) The Physics of Air-Sea Interaction, Isr. 859
Pacific: Part 1. Oceanography 24(4):24–31 Program for Sci Transl Jerus
Donelan MA (1982) The dependence of the aerodynamic drag coefficient
Kitaigorodskii SA, Volkov YA (1965) On the roughness parameter of the
on wave parameters. 1st intl. conf. on meteorol. & air-sea interac-
sea surface and the calculation of momentum flux in the near-water
tion of the coastal zone, Amer Meteorol Soc: Boston, pp. 381–387,
layer of the atmosphere. Izv, Atmos Ocean Phys 1:973–988
1982
Donelan MA, Dobson FW (2001) The influence of swell on the drag. Kraus EB, Businger JA (1994) Atmosphere–ocean interaction. Oxford
Wind stress over the ocean. Cambridge Uvi. Press, New York, pp University Press, New York, p 362
181–190 Miyake M, Stewart R, Burling RW (1970) Spectra and cospectra of
Donelan MA, Hamilton J, Hui WH (1985) Directional spectra of wind- turbulence over water. Quart J Royal Meteorol Soc 96(407):
generated waves. Phil Trans R Soc Lond A315:509–562 138–143
Donelan MA, Dobson FW, Smith SD, Anderson RJ (1993) On the de- Monin AS, Obukhov A (1954) Basic laws of turbulent mixing in the
pendence of sea surface roughness on wave development. J Phys surface layer of the atmosphere. Contrib Geophys Inst Acad Sci
Oceanogr 23:2143–2149 USSR 151:163–187
Donelan MA, Drennan WM, Katsaros KB (1997) The air-sea momentum Moskowitz L (1964) Estimates of the power spectrums for fully-
flux in conditions of wind sea and swell. J Phys Oceanogr 27(10): developed seas for wind speeds of 20 to 40 knots. J Geophys Res
2087–2099 69:5161–5179
Drennan WM (2005a) On parameterisations of air-sea fluxes. In: Obukhov AM (1946) ‘Turbulentnost’ v temperaturnoj–neodnorodnoj
Atmosphere–ocean Interactions, Vol. 2, WIT Press; pp 1–33 atmosfere (turbulence in an atmosphere with a Non-uniform tem-
Drennan WM, Shay LK (2006) On the variability of the fluxes of mo- perature). Trudy Inst Theor Geofiz AN SSSR 1:95–115
mentum and sensible heat. Bound-Layer Meteorol 119(1):81–107 Pan JD, Wang W, Hwang PA (2005) A study of wave effects on wind
Drennan WM, Kahma KK, Donelan MA (1999) On momentum flux and stress over the ocean in a fetch- limited case. J Geophys Res 110(C2)
velocity spectra over waves. Bound-Layer Meteorol 92:489–515 Pierson WJ, Moskowitz L (1964) A proposed spectral form for fully
Drennan WM, Graber HC, Hauser D, Quentin C (2003) On the wave age developed wind seas based on the similarity theory of S.A.
dependence of wind stress over pure wind seas. J Geophys Kitaigorodskii. J Geophys Res 69:5158
Res 108:C3 Potter H, Graber HC, Williams NJ, Collins CO III, Rafael RJ, Drennan
Drennan WM, Taylor PK, Yelland MJ (2005) Parameterizing the sea WM (2015) In situ measurements of momentum fluxes in typhoons.
surface roughness. J Phys Oceanogr 35:835–848 J Atmos Sci 72:104–118
Drennan WM, Graber HC, Collins CO III, Herrera A, Potter H, Ramos Sahlée E, Drennan WM, Potter H, Rebozo MA (2012) Waves and air‐sea
RJ, Williams NJ (2014) EASI: an air-sea interaction buoy for high fluxes from a drifting ASIS buoy during the Southern Ocean Gas
winds. J Atmos Ocean Technol 31:1397–1409 Exchange experiment. J Geophys Res 117(C8)
Foreman RJ, Emeis S (2010) Revisiting the definition of the drag coeffi-
Smedman, AS, Larsén XG, Högström U, Kahma KK, Pettersson H
cient in the marine atmospheric boundary layer. J Phys Oceanogr 40:
(2003) Effect of sea state on the momentum exchange over the sea
2325–2332
during neutral conditions. J Geophys Res: Oceans 108.C11, 3367
García-Nava, H, Ocampo‐Torres FJ, Osuna P, Donelan MA (2009) Wind
stress in the presence of swell under moderate to strong wind con- Smedman A, Högström U, Sahlée E, Drennan WM, Kahma KK,
ditions. J Geophys Res 114 (C12) Pettersson H, Zhang F (2009) Observational study of the marine
Högström U, Smedman A, Sahlée E, Drennan WM, Kahma KK, atmospheric boundary layer characteristics during swell. J Atmos
Pettersson H, Zhang F (2009) The atmospheric boundary layer dur- Sci 66:2747–2763
ing swell: a field study and interpretation of the turbulent kinetic Smith SD, Anderson RJ, Oost WA, Kraan C, Maat N, DeCosmo J,
energy budget for high wave ages. J Atmos Sci 66(9):2764–2779 Katsaros KB, Davidson KL, Bumke K, Hassee L (1992) Sea surface
Högström U, Smedman A, Semedo A, Rutgersson A (2011) Comments wind stress and drag coefficients: the HEXOS results. Bound-Layer
on ‘A global climatology of wind-wave interaction’ by K. E. Meteorol 60:109–142
Hanley, S. E. Belcher, P. P. Sullivan. J Phys Oceanogr 842(41): Wu J (1980) Wind stress coefficients over the sea surface near neutral
1811–1813 conditions: a revisit. J Phys Oceanogr 10:727–740

You might also like