JOURNAL
OF GEOPHYSICAL
RESEARCH, VOL. 103, NO. C5, PAGES 10,289-10,311, MAY 15, 1998
Upper ocean heat and salt balances in response to a
westerly wind burst in the western equatorial Pacific
during TOGA COARE
Ming Feng,• PeterHacker,andRogerLukas
Schoolof Ocean and Earth Scienceand Technology,University of Hawaii, Honolulu
Abstract. Two volume control methods are used to analyze the upper ocean heat
and salt balancesin responseto a westerlywind burst event in the westernequatorial
Pacific during the Tropical Ocean Global AtmosphereCoupled Ocean-Atmosphere
ResponseExperiment. One method usesa fixed-thicknesssurfacelayer, and the
other usesan isopycnaldepth as the lower boundary. Horizontal advectionterms in
the budgetcalculationsare estimatedusingthe R/V Wecomarepeathydrographic
survey data within a 133 km x 133 km region. In both methods,the upper ocean
heatbudgetis balanced
within 10 W m-2 of the surfaceair-seaflux observations
during a 19-day time period, which coversthe December 1992 westerly wind burst
and a low-wind recovery period in early January 1993. The standard error in
the estimationof heatadvection
is 11 W m-2. The salt budgetyieldsa rain rate
estimate of 15.4 mm d -1 with an error bar of 4 mmd -1. This estimate is within
20% of the optical rain gaugemeasurements.The advectionterms are important
in both the heat and salt balances.
Meridional
advection
dominates
over zonal
and vertical advection, acting to decreasetemperature and increasesalinity in the
surface layer. From the isopycnalboundary method, the diapycnal turbulent flux
transports
a meanheatfluxof 17W m-2 intothe thermocline.
Diapycnaladvection
is almost equally important, so that the total heat flux into the thermocline is
estimated
to be morethan30W m-2 duringthestudytimeperiod.Bothtermsare
also important in the salt budget.
1.
During the IOP, strong intraseasonal oscillations
Introduction
The western equatorial Pacific warm pool, where
sea surfacetemperature exceeds29øC, is the warmest
open ocean water in the world. It plays an important
role in the ocean-atmospherecoupling during the E1
(ISO) occurredwithin the IFA. The ISO, or 30-60oscillation, is characterizedby an eastwardpropagationof a
large-scale
regionof anomalous
deepconvection
[Chen
et al., 1996]. The anomalousconvection
showsmultiscale structures in time and space. The active phase
Nino SouthernOscillationevents[WebsterandLukas, of the ISO is associatedwith intermittent, strongsur1992].Extensiveupperoceanand air-seaflux observa- face westerly wind anomaliesknown as WWBs, which
tions wereconductedin the intensiveflux array (IFA) typicallylast for lessthan I week[Weller and Andercentered at 2øS, 156øE in the warm pool during the son, 1996;Harrison and Vecchi,1997].The WWB has
Tropical OceanGlobal Atmosphere(TOGA) Coupled a meridional scale of a few hundred kilometers and a
Ocean-Atmosphere
Response
Experiment(COARE) in- zonal scaleof more than 1000 km, can generatestrong
tensiveobservingperiod (IOP) from November1992 equatorial waves,and can strongly affect the air-sea in-
through February 1993. The goal of theseobservations teraction processesin the tropics.
was to quantify the air-sea interaction processes,esThis paper utilizes the IOP observationsto address
pecially those associatedwith the westerly wind burst two issues;one is to determine the important oceanic
(WWB) phenomenon
in the warmpoolregion[Webster processesthat affect the upper ocean heat and salt baland Lukas,1992].
ancesin responseto a WWB during December 1992 to
early January 1993, and the other is to usethe heat and
• Alsoat Institute of Oceanology,ChineseAcademyof Sci- salt balancesin the upper ocean to evaluate the indeences,Qingdao, People'sRepublic of China.
pendent estimates of heat and freshwater fluxes across
the air-sea interface.
Copyright1998 by the AmericanGeophysicalUnion.
Paper number 97JC03286.
It has been difficult to improve closure of the surface energy budget in the warm pool to better than 60-
0148-0227
/ 98/ 97JC-03286$09.00
80 W m-2 beforeCOAREobservations
[Godfreyand
10,289
10,290 FENGET AL.: HEAT AND SALTBALANCESIN RESPONSE
TO A WESTERLYWIND BURST
Lindstrom,1989]. The freshwaterbudgetin this re- Niiler, 1983; Cronin and McPhaden, 1997; Godfrey et
gion is also a challengedue to the large net freshwater al., personalcommunication,
1996].The fixed-thickness
flux [Donguy,1987;LukasandLindstrom,1991].Using method is easy to develop,and we can recognizewhich
COARE bulk formulae[Fairallet al., 1996],Wellerand advection terms play important roles in the budget.
Anderson[1996]calculatethe flux termsusingsurface However, this method is very sensitiveto the vertical
meteorologyobservations
from a centralmooring(de- motion of the thermocline. The benefit of the isopycployed by Woods Hole OceanographicInstitute, here- nal boundary method is that we can more closelyrelate
after calledthe WHOI mooring). On the basisof in- the mean temperature in the layer to the sea surface
tercomparisonswith nearby ships, they find that the temperature[Stevenson
and Niiler, 1983],and we can
uncertainty in the net heat flux can be reduced to less eliminate the vertical advection effects of low-mode in-
than 10 W m-2 in meanscomputedoverone to sev-
ternal
waves.
eral weeks. However, there are still discrepanciesin the
This paper is organized as follows: in section 2, we
meanrain rate estimatesfromdifferentplatforms[God- introduce the data sets used in the present study; in
section 3, we introduce the formulae for the heat and
frey et al., 1997].
Previousupper oceanbudgetstudiesfor the IOP sug- salt budgetsin the two volume control methodsand the
gestedthat one-dimensional
(l-D) physicssometimes calculationmethods for different terms; in section4, we
is not enough to describethe upper ocean responsesto give the calculation results;and in section5, we discuss
a WWB [Smythet al., 1996b;Andersonet al., 1996; the results, followed by a short summary in section6.
Croninand McPhaden,1997]. Calculations
of the 3-D
heat and salt budgets at the center of the IFA are the
subjectsof this paper and a paper by Antonissenet al.
2.
Data
Figure 1 gives the positionsof the IOP observations.
(personalcommunication,
1997).
During the IOP, the researchvessel(R/V) Wecoma The WHOI mooring was located at 1ø45'S, 156øE at
conducteda repeated butterfly pattern surveycentered a water depth of 1744 m, the R/V Moana Wavewas
at 1.8ø S, 156.1øE with a spatial extent of 133 km in within 10 km of the WHOI mooring,andthe R/V Vickboth zonal and meridionaldirections(Figure 1). Each erswasat 2øS, 156.2øE, all nearthe centerof the R/V
Wecomabutterfly survey.
The WHOI mooring provided continuousmeteorological
and oceanographicmeasurementsfor the durathe track [Huyer et al., 1997]. In this paper, two voltion
of
the IOP. Sensorson its surface buoy sampled
ume control methods are used to analyze the 3-D upwind
speed
and direction, relative humidity, air temper oceanheat and salt budgetsfrom the R/V Wecoma
perature,
barometric
pressure,incoming shortwaveradata over the butterfly survey domain. One is a fixed
diation,
incoming
longwave
radiation, sea surfacetemthickness surface layer analysis, and the other uses an
perature,
and
current
speed
and direction at 5 m depth
isopycnaldepth to bound the volume[Stevenson
and
circuit of the butterfly took about 1.5 days. Temperature, salinity, and velocity profileswere measuredalong
0ø
!
i
i
!
i
i
i
I ßWI-IOI
ß Vickers
[Weller and Anderson,1996].The COARE version2.5
bulk formulae[Fairall et al., 1996]havebeenusedto
calculate the latent and sensibleheat fluxes, and true
sea surface temperature or skin temperature was used
to estimatethe outgoingradiation[Weller and Ander-
son,1996]. We haveapplieda 7 W m-2 decrease
of
the net longwaveradiation as suggestedby Bradley and
Weller [1997]. The rain rate data are from the optical rain gaugemeasurements
on the RfV Moana Wave;
also, the sensibleheat flux due to rainfall has been cal-
culatedfrom the R/V Moana Wavemeasurements
by
Fairall et al. [1996]. The flux data usedhereinare
hourly averages.
The R/V Wecomaobservations
weremadeon three
cruises.We use the cruise2 data for the presentstudy
(duringDecember20, 1992,to January8, 1993[Huyer
et al., 1997]). ContinuousacousticDoppler current
Figure 1. Intensive flux array region map. Solid areas are islands. The bold line is the standardbutterfly
pattern of the R/V Wecoma.The circleis the position
of the WHOI buoyand the R/V Moana Wave,and the
solidtriangle is the positionof the R/V Vickers.The
light line is the time integrationof the 5-m WHOI current from December20, 1992, to January 10, 1993. The
spacingbetweenthe shadedtriangles is I day.
profiler velocity measurementswere made from 18 to
about 300 m. For our analysis, we use gridded data
vertically averaged to 10-m-depth bins. The Seasoar
conductivity-temperature-depthtemperature and salinity measurements were gridded from the surface to
about 250 dbar with a 2 dbar spacing. The horizontal resolution is about 1.5 km, with data gaps near
FENG ET AL'
HEAT AND SALT BALANCES IN RESPONSE TO A WESTERLY WIND BURST
10,291
the surface and 250 dbar on some sections because of
Qo- Rs(O)+ FT(O)is the net surfaceheatflux,
variable Seasoarhandling. For our analysis,the velocity, temperature, salinity and potential density data are
averagedhourly. Only the data that fall in the butterfly region are used. The Seasoardata are accurateto
+0.01øC in temperature, +0.01 practical salinity unit
is the penetrating solar radiation at h, So is the surface
salinity, P and E are the precipitation and evapora-
(psu)in salinity,and +1 dbarin pressure
[Huyeret al.,
1997]. The hourly velocitydata have an accuracyof
+1 cm s-1.
Additionally, microstructure measurements were
madefrom the CHAMELEON profiler[Mournet al.,
1995]on the R/V Moana Wave.Temperature,salinity,
potential density, Brunt-Vaisala frequency,and turbu-
tion rates, respectively,and FT[-h and Fs[_n are the
turbulent heat and salt fluxesat h, respectively.
3.2.
Isopycnal Boundary
Method
The isopycnalselectedfor this purposehas a depth
h just below the mixed layer during the WWB, so that
it can best representthe active surfacelayer and minimize the influence of variations in the thermocline.
The
equations(1) and (2) are rewrittenin the senseof ver-
tical mean temperature and salinity over the layer betweenthe seasurfaceand the isopycnaldepthfollowing
computed using these data. The vertical resolution is
Stevenson
and Niiler [1983]. In this format, the adia4 m. The hourly mean data were calculatedby Smyth
batic variationsdue to the vertical displacements
of the
et al. [1996a].Also,solarradiationpenetrationprofiles thermocline are excludedfrom the calculation, and it
havebeendaily averagedfor the R/V Vickersmeasureis easy to integrate with time. The rewritten equations
ments[Siegelet al., 1995]. Thesedata are for every are
lent kineticenergydissipationrate (e) profileshavebeen
meter
from
I to 100 m.
OT
3.
Methods
I
Ot -- pcph
(-Qo
-•-•S]-h
-•-FT]_n)
- h
In this section, we first introduce the formulae for
the heat and salt budgets of the two volume control
methods. Then we give the calculation methods for
different
T - T_n
•
- •. •T- • •.
h
(5)
T•u•dz,
terms.
os
3.1.
We
Fixed
Thickness
ot = n((-So(PE)+Zsl-)
- S-
Method
Assumingno horizontalmixing, the temperatureand
- •' vS- • V'
salinity tendency equationsare written as
OT
OT
) ORs
pocp•/+u.vT+W•z z - Oz
os
os
Ot+ u. X75
+ wOz
OZ '
OFs
Oz
•
h
S•u•dz,
h
We
(6)
(1)
where
• - • fohTdz,T' = T- T,S- • fonSdz,
S'=
S- •, anda - • fonudz,andu' - u - a. •, •, and
(2)
H are the mean temperature, salinity, and horizontal
velocity,respectively;T •, S•, and u • are the deviations
from the vertical mean valuesin the water column;Qo,
Rs, Fr, So, P, E, and Fs are the same as in the fixed
The variablePois the meanwater density,% is the spe- depthanalysis;TI-n, $l-n and wl_n are the tempercificheat, u = (u, v) is the horizontalvelocity,w is the ature, salinity, and vertical velocityat h, respectively.
vertical velocity, Rs is the penetrating solar radiation, Note that h is a function of time and horizontalspace.
and Fr and Fs are the vertical turbulent heat and salt The valuew, - dh/dt + wl_n is the diapycnalvelocity,
whered/dr = o/at + uO/Ox+ vO/Oy.
fluxes,respectively.
The last two termsin (5) and (6) are the horizontal
Vertically integrating the above heat and salt tendencyequationsfrom a fixed depth h to the seasurface, advection terms. The first term is the vertical mean
horizontal velocity acting on the horizontal gradients
the followingforms are obtained:
of verticalmeantemperature(salinity),and the second
term is the convergence
of heat (salt) due to a stratified shearflow. Someauthorshaveneglectedthe second
:• poCp
OT =
•-dz
-
-
h
poCp
u.vT+W•zz dz
Q0 + ZZsl- + Frl-,
(3)
advectionterm, becausethey do not have crediblemea-
surements
to estimatethis term [Stevenson
andNiiler,
1983; Cronin and McPhaden,1997]. In this study,we
estimateboth termsbasedon the R/V Wecomadata.
3.3.
dz
= _
-
h
u. vS+W•z
z
So(P- E) + Fsl-.
Evaluation
of Different
Terms
dz
(4)
3.3.1. Temporal and spatial gradients. In the
two volume controlmethods,it is necessaryto estimate
10,292
FENG ET AL.: HEAT AND SALT BALANCES IN RESPONSE TO A WESTERLY WIND BURST
the mean values and the temporal and spatial gradients of the variablessuchas T, 5', p, and u in a certain
layer
andT, 5',•, h,føaT•u•dz,
andfoa$•u•dz
' etc.
On the basisof the R/V Wecomasurveypattern, the
abovevariablesovertwo completebutterflycircuits(approximately3 days)are fit with a meanvalueand linear
trends in time and horizontal space. Note that the approximately 3-day fit period representsa compromise:
short enough so that near-inertial period changescan
be resolved and long enough to obtain two measure-
of horizontalvelocity and the other usesisopycnaldisplacements.
In the first method,
w-w0- •
o
• Ou
+ Ov
(9)
The variable z0 = 0, the sea surface,is the reference
depth of the integration,and w0 is the verticalvelocity
at z0 and is assumedto be zerooverthe 3-day timescale.
The horizontal gradientsof the velocitiesare calculated
ments in time at each spatial location for evaluation of with the linear fit. The divergencebetween0-20 m is
the temporal tendency and to suppressthe tidal noise. assumed
to be constant.A tidal removalprocess[Feng
Thus we evaluate
et al., 1998]is appliedbeforethe velocitygradientcalculation. This processdoesnot noticeablychangethe
d(x(t),y(t), t) - d + dxx-Fdyy-Fdtt
-F d'(x(t),y(t), t).
(7)
Variabled - d(x(t), y(t), t) is the variableto be fit, x
and y are the longitudeand latitude, respectively,relative to the crossoverpoint, t is the time relative to
the centertime of the interval,d, d•, dy, and dt are
constants,and d• is the residual. In this sense,the d
representsthe mean value of d at the crossoverpoint,
d•.and dy representthe meanzonalandmeridionalgradients, and dt is mean rate of temporal change. The
linear fit results but can reduce the variances in the
residual and the error estimations.
The other method assumes adiabatic motion in the
density equation, that is,
Op
Op
Op
Op
W•zz- at U•x
x- vo--•
'
(10)
Variable p is the water density;the gradientsof p are
calculated from the linear fit. This method is used be-
low the mixed layer depth, which is definedas equiva-
lentto a 0.167kgm-s densitychange
fromthe surface
vertical gradients of T and S in the vertical advection
[Sprintalland Tomzak,1992].In the mixedlayer,the
term are calculatedas the vertical gradientsof T and $.
vertical
velocityis assumedto be linearlydecreasing
to
Moving the near 3-day window forward, the time series
zero at the sea surface.
of all variablesare obtained.BootstrapanalysistElton
There is qualitative agreement between the two
and Tibshirani,1986]is usedto estimatethe errorsin
(rather noisy)verticalvelocityestimates.In practice,
the mean value and gradients(AppendixA). Considwe averagethe two estimatesweightedby their staneringa peak velocityof 50 cms-• in the surfacelayer dard errors.
during WWB, the advection distance is about 65 km
(smallerthan the surveydomain)in 1.5 days. Thus we
can evaluate the advection effects for both heat and salt
balancesfrom the R/V Wecomadata.
3.3.2.
Turbulent
fluxes.
The turbulent
heat and
3.3.4. Diapycnal velocity.
There are three meth-
odsto calculatethe diapycnalvelocityacrossan isopycnal depth from the available data.
1. Use the definition, that is,
salt fluxes at a certain depth are calculated as
dh
w•-•+w]_•.
F• - -pc•K•Oz
Fs- -Ksas
Oz' (s) Variableh is the isopycnaldepth; dh/dt =
(11)
Oh/Or+
u]_•Oh/Ox+ v[_•Oh/Oy, whichcan be calculatedby
where the thermal and haline diffusivitiesare set equal
usingthe linearfit of the R/V Wecoma
data,andw[_•
to the densitydiffusivity,
KT -- Ks - Kp - Fe/N•. is the vertical velocityfrom the divergenceintegration.
The variable e is the turbulent kinetic energydissipation
rate calculated
from the microstructure
measurements
From the Bootstrap analysis,the errors in the estima-
tion of eitherterm on the right-handsideof (11) are big
[Smythet al., 1996b].N is the buoyancy
frequency,
and
enoughto mask the signals.Also, the diapycnalvelocF is the mixingefficiency
[Osborn,1980;Mourn,1990]. ity sometimesevolveswith the diurnal processes,
which
On the basisof laboratoryobservations,
Osborn[1980] are not resolvedin the 3-day linear fit. Therefore this
used 0.15 as an upper bound of the flux Richardson
method is not usedin this study.
number yielding an upper bound of F equal to 0.2.
2. Usethe mixed layer turbulentkineticenergyequa-
Mourn[1990]pointedout that F hasa variationrange
of 0.1-0.4 by comparingthe dissipationmeasurements
and the temperature variance measurements. Because
a F - 0.2 is still the choiceof many authors[Gregg,
1987; $myth et al., 1996b],we usea constantF - 0.2
tion [Niiler andKraus,1977]:
h
weAbh(1
- sRi•
•)- 2muS.
+ •((1+n)Bo2
2
(1-•)lB01)
+((n-•) +(n+•)•-•")•zs(o),
(12)
throughout this study.
3.3.3. Vertical velocity. Two methodsare used
to calculatethe vertical velocity: one usesdivergence whereAb = -g(•-
P]-a)/Po, • is the meanpotential
FENG ET AL.: HEAT AND SALT BALANCES IN RESPONSE TO A WESTERLY WIND BURST
10,293
densityabovethe isopycnaldepth,p[-a is the potential solarradiation(short-wave),
,•--1 is 22.1 m according
density at the isopycnaldepth, Ri• is the bulk Richard- to the double exponential transmissionprofile observed
sonnumber, u, is the friction velocityof the wind stress, during COARE IOP [Siegelet al., 1995; Cronin and
B0 is the surfacebuoyancyflux, rRs(O) is 38% of the
McPhaden,1997], and m,n, and s are the efficiency
IIIIIIIIIIIIIIIIIIIIt
-10
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
/
I IIIIIIIIIIIIIIIIIII
lO b
0 -
-' ^
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
!
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
600
300
o
0
-300
•
.
3O
E 2O-
20
22
December
24
26
28
1992
30
I
3
5
7
9
January 1993
Figure 2. Time seriesof wind speedand air-seafluxes:(a) eastwardwind speed,(b) northward
wind speed,(c) total wind speed,(d) net surfaceheat flux, and (e) rain rate. The bold linesin
Figures2a- 2d are 3-day movingaverages.
10,294
FENG ET AL.: HEAT AND SALT BALANCES IN RESPONSE TO A WESTERLY
WIND BURST
factors. All the terms in the equation can be calcu- 4. Results
lated; however, there are uncertainties in the efficiency 4.1. Overview
factors, which we can estimate by comparison with the
next method.
The time period for our study coversthe December
WWB
period and the beginning of the January low3. Usethe densityconservation
equation[McDougall
and You,1990]:
Op
_0( OpaRs(z)
) ' (13)
weoz- Oz Kp•z-F poC•
where t• is the thermal expansion coefficient. The first
term on the right-hand side is due to the vertical gradient of the turbulent flux, and the secondterm is due
to the vertical gradient of the penetrating solar radiation (Godfrey et al., personalcommunication,1996).
In Appendix B, we give a brief discussionabout the
equivalency between the density conservation method
and the definitionof the diapycnalvelocity,(11); also,
we give the rationale for including both the turbulent
flux and diapycnal advection in the budget calculation.
In our study, we use the microstructure data for the
calculation of diapycnal velocity. The secondterm on
the right-hand side of (13) is two magnitudessmaller
than the first term for the isopycnal depth we select.
Neglecting the secondterm on the right-hand side and
0e
from the microstructure
T
S
a
b
periodis 32 W m-2.
The rainfall is very largeduringthis time period(average18.8mmd-1). Heavyrainfalloccursin the first
WWB peak becausethe deep convectionis right on top
of the IFA. During the secondWWB peak, the deep
east and the rain rate is small
This also results in less heat loss in the sec-
ond WWB peak due to larger solar radiation than in
the first WWB peak. The large rain rate on January
3-5 may be due to the westwardpropagatingnear 2-day
data.
U,V
W
I
i
c
(after January4). The significantheat lossperiodsare
roughly coincidentwith the WWB peaks. The average
heat lossacrossthe air-sea interface during this study
in the IFA.
we= N• Oz'
which we evaluate
ber 25, and between December 31 and January 2. The
timescaleof one WWB peak is a few days. The gap between the first and secondWWB peaks is so small that
from the near 3-day analysis, they are describedas one
peak. It is referred to as the first WWB peak. The last
WWB peak is referred to as the secondWWB peak.
During the WWB period, the ocean is losingheat to
the atmosphereuntil the beginningof the JLW period
convection moves farther
usingthe expression
for Kp, (13) becomes
F
windperiod(JLW) [WellerandAnderson,1996].There
are three WWB peaks during this time period (Figure 2), with peak wind speedon December22, Decem-
T
T
S
S
d
20
./
80
100
,
26
28
C
34
35
psu
-0.4
0
m/s
0.4
-2
0
-0.5
2
10-5 rn/s
10-5 C/m
0
0.5
10-5 C/m
-0.2
0
0.2
10-5 psu/m
-0120 012
10-5 psu/m
Figure 3. Cruisemeanupperoceanpropertiesat the R/V Wecomacrossover
point: (a) temperature, (b) salinity,(c) eastward(solid)and northwardvelocities(dashed),(d) verticalvelocity,
(e) zonal and (f) meridionaltemperaturegradients,and (g) zonal and (h) meridionalsalinity
gradients The shadedareasare the standarderrorsusingbootstrapanalysis The errorsin the
mean temperature, salinity, and horizontal velocitiesare lessthan 0.05øC, 0.01 practical salinity
unit and I cms-1, respectively.
FENG ET AL.- HEAT AND SALT BALANCES IN RESPONSE TO A WESTERLY
WIND BURST
oscillation near the end of the WWB period or due to
10,295
salinity are dominated by the meridional gradientsin
the mixed layer with increasingtemperature and deFrom the temporal mean profilesof the temperature creasing
salinitytowardthe north (Figure3). The mean
and salinity (Figure 3), the surfacelayer is well mixed meridional gradients of temperature and salinity bedown to about 60 m. Both temperature and salinity low the mixed layer depth havesignsconsistentwith a
gradients are important in stratifying the layers below geostrophicallybalancedmean current. The zonal gra60 m. The mean zonal current (u) is towardthe east dientsin both the temperature and salinity are smaller
(• 30 cms-•) in response
to the WWB in the surface than the meridional gradients.
layer. The subsurfacecurrent below 70 m is toward
Figure 4 showsthe temporal evolution of the u and
the west representingthe remnant of the South Equa- v componentsof the velocity and the temperature and
torial Current (SEC), whichis sandwiched
betweenthe salinity at the crossoverpoint above 100 m from the lineastwardflowingsurfacecurrentand the EquatorialUn- ear fit. The low-passfiltereddata (removingthe diurnal
dercurrent[Eldinet al., 1994;Huyer et al., 1997].The and semidiurnaltides)from the 5 m WHOI currentmemeridionalcomponent(v) of the meancurrentis much ter is used to interpolate the velocity from 20 m to the
smaller and has a northward tendency in the mixed sea surface. The u componentis eastwardin the mixed
layer(• 10cms-•). Themeanverticalvelocityisdown- layer during all the time in responseto the WWB. Both
ward during this time period, with a peak velocity of the surface eastward flow and subsurface westward flow
morethan -2 x 10-5 ms-• at 50 m. Temperatureand accelerateduring the two WWB peaks. The acceler-
localizedconvection
[Chenet al., 1996].
U (era/s)
o
50
lOO
V (cm/s)
lOO
Temperature
(C)
o
lOO
Salinity (psu)
i•:•4•-•i23•.:2•i•:•
•:•:::•:•i;.•.:.-•:•
•i• •2•:•.:•::•i•::•::•..•i•i•:::•
:'•;
............................................................
lOO
12/20
I
I
I
12/25
12/30
1/4
119
Figure 4. Temporal evolution of eastwardand northward velocities,temperature, and salin-
ity at the R/V Wecomacrossover
point. The negativevelocities(westwardand southward),
temperature above29øC and salinity below 34.2 psu are shaded.
10,296
FENG ET AL.: HEAT AND SALT BALANCES IN RESPONSE TO A WESTERLY WIND BURST
ation of the subsurfaceflow may be due to pressure similar vertical displacements.Despite the heavy raingradientandinertialmotions[Smythet al., 1996a].The fall during the first WWB peak, the surfacesalinity
surfacelayer v componentis northward mostof the time slightly increasesexcept during the JLW when shallow
and has a northward accelerationa few days after the fresh water pools developin the top few meters.
eastward accelerationof the u component,which is due
During most of the time, there is a northwardtemper-
to inertial motions[Smythet al., 1996a].Thereare indicationsof vertical propagationfrom the structure of
the v component, also seen in the moored observations
(S. Anderson,personalcommunication,
1997). Thus it
seemspossiblethat inertial motions are of importance
in the upper ocean heat and salt budgets becauseof
their effectson shear and mixing.
The mixed layer temperature decreasesin response
to the WWB (Figure4). During the JLW, the top few
meters are warmed up, and the isotherms below 50 m
shoal so that there is a trend of increasingstratifica-
ature gradientin the mixedlayer (Figure5). Significant
gradientsare found during December25-26, 1992, and
after January 3, 1993. The southwardsalinity gradient
has maxima during similar time periods in the mixed
layer (Figure 6). In the mixed layer, the zonaltemperature and salinity gradientsare generally smaller than
the meridional. Below the mixed layer depth, the internal wave signalsintroducesignificanterrorsin the fixed
depth gradientcalculations(Figure3). Thus it is more
difficult
to calculate
heat and salt advection
below the
mixed layer in the fixed thicknessanalysis;an isopycnal
tion in the surfacelayer. Smythet al. [1996b]argued depth may be more useful as a lower boundary.
that the turbulence generatedby the strong shear between the surface wind-driven
current
and subsurface
SEC might also contribute to this stratification process. The 28øC depth is almost coincident with the
4.2.
Heat
and
Salt
Balances
Above
a Fixed
Depth
Resultsfrom the cruisemean analysesare shownin
as = 22 kgm-s depth(alsothe 34.5psudepth)in the Figure 7 for heat and Figure 8 for salt. In Figure 7,
R/V Wecomadata, whichis alsothe casefromthe R/V the whole water columnabove100 m is losingheat at
MoanaWavedata [Smythet al., 1996b].Generally,
the a rate near 2 W m-s. Above60 m, the meridionalad28øC depth deepensgradually from 70 m to more than vection is a big contributor to the heat loss, and the
80 m in responseto the WWB, and it shoalsto 70 m zonal and vertical advection terms are much smaller.
after the WWB. Note that the lower boundary of the Accordingto (1), the verticalgradientof the penetratsurfacelayer eastwardflow in the u componentalsohas ing solar radiation flux is subtracted from the sum of
Zonal temperature
Meridional
-
'=
,,
....
.
gradient (CLIO0 km)
gradient
,
/
'
/
•.............
\
--Z:• .... _
/'
,
:.:
t
•
'
:--:-_-,:
'-'-,
//
',
I
[
•
.
5 I
•
,
---
5-•-'-----
,, .
.
'-._2
- ....
,
.' ,W'
,;•/
;'//
.
I
II
100
12120
12/2•
121•0
114
119
Figure 5. Temporal evolutionof zonaland meridionaltemperaturegradientsfrom the linear fit.
Positivegradientsare toward the north and east. Negativegradientsare shaded.
FENG ET AL.: HEAT AND SALT BALANCES IN RESPONSE TO A WESTERLY
WIND BURST
10,297
Zonal salinity gradient (psu/100 km)
Meridional
gradient
i•1 ( •
.• ß•
I
•
•i•ii¾iiiii:iii::•:ii•!i:•:::•!ii•:i:•:.'•i•!::•
.........................
•:.:a:::c-:::::;i:i;;;;;;;;;'•"""•!:•*:':•:•.•*•e;:•:-'::•:'•
......................
so
:::?;a ':'*•
100
"-
' '" '
'
I
'
'
'
I
12125
12/20
'
'
'
'
12/30
1
'
'
'
'
1/4
1/9
Figure 6. Temporalevolutionof zonal and meridionalsalinitygradientsfrom the linear fit.
Positivegradientsare towardthe north and east. Negativegradientsare shaded.
pCpTt
i
!
i
pc p uT x
i
i
pc p vT y
i
i
pc p wT z
!
R
sum
Residual
I
.
ß
d
a
g
2O
ß
• 40
r-•
ß
60
..
:
ß
8O
;.
...?/
?•)..
..........
...i•
.......
lOO
':...•L:.
!
-5
o
5
5
-5
o
5
-5
0
5
-lO
o
lO
-10
0
10
Figure 7. Cruisemean of heat tendencyequation(1): (a) temporal rate of heat content,
(b) zonal,(c) meridionaland (d) verticaladvections
of heat,(e) the summation
of the abovefour
terms,(f) the penetratingsolarradiation,and (g) the residualbetweenFigures7e and 7f, which
is -dFT(z)/dz.
10,298
FENG ET AL.: HEAT AND SALT BALANCES IN RESPONSE TO A WESTERLY WIND BURST
St
0
i
vS
uSx
i
i
i
i
wS
y
i
•
i
z
sum
i
i
i
.
b
c
d
e
20
•- 40
I•
60
8o
lOO
_
-0.02
0.02
-0.•2 0.02 -0.02
0.02
psu/day
-0.02
0.02
-0.02 0.02
Figure 8. Cruise mean of salt tendencyequation(2): (a) temporalrate of salinity change,
(b) zonal,(c) meridionaland (d) verticaladvectionof salt, and (e) the summationof the above
four terms,whichis also-dFs(z)/dz.
Heat flux
a
Salt flux
:•
f
2O
#
2"-'-'-'-
1
I
I/
I
I
%
4O
%
%
6O
:;i:i8O
100
•.:'.-•.• '•
,
.....
' ........
'...............
-1 O0
0
W/m2
7......
1 O0
0
1
psum/day
0
5o
percent
Figure 9. Cruise mean turbulent fluxes: (a) FT(Z) derivedfrom the residualof the advection terms (bold solid line) and that derivedfrom the microstructuremeasurementassuming
F - 0.2 (bold dashedline) and F - 0.1 and 0.4 (light dashedlines). The shadedarea is the
standarderror of the advectiontermsintegratedfrom seasurface.(b) The sameas Figure9a but
for Fs(z). (c) The percentageof hourly microstructuredata availablefor the calculationwith
N • _> Nmi
• n -- 3 x 10-6 s-• .
lOO
FENG ET AL.' HEAT AND SALT BALANCES IN RESPONSE TO A WESTERLY WIND BURST
the temporal and advection terms. The residual, or
the remaining term, then includes the vertical gradi-
ent of the verticalturbulentflux, -dFT(z)/dz, and the
errors(Figure 7). The R/V Vickersobservations
have
10,299
boundary. This selection provides a compromisebetween minimizing errors in the advection calculations
and maximizing the information of the mixed layer.
During the 19-day time period (from day 355 to the
missedsomeof the energyspectrumnear the seasurface center time of the last 3-day fit), the heat contentin
(S. Anderson,personalcommunication,
1997),sothere the upper50 m decreases
by 150MJ m-2 (equivalent
is a secondarymaximum near about 6 m (Figure 7). to a flux of 92 W m-• for 19 days). Overthis time peThis effect is not adjusted, becauseit does not affect riod, only about one third of this heat lossis due to the
the results below 6 m.
Knowing the surface turbulent heat flux FT(0)
air-seaflux, 52 MJ m-2 (32 W m-•) (Figure10). We
see that the total
advection
term also contributes
to the
(184W m-2) fromthe WHOI andR/V MoanaWave coolingof the surfacelayer and has a cumulativeeffect
of 70 MJ m-2 (43 W m-2), whichis largerthanthe air-
data, the residualof Figure 7 is integratedfrom the surfacedownward(equation(3)). In this way,the vertical
profile of the turbulent heat flux, F•(z), is calculated
(Figure9a). The F•(z) derivedfrom the residualof the
advectioncalculationand the F•(z) calculatedfrom (8)
radiation term has a cumulative effect of 13 MJm -•
usingthe microstructure
data agreewithin 10 W m-2
bulentflux, the residualis nearly20 MJ m-•.
between 40 and 70 m when selectingF - 0.2 (Figure 9a). Thusa fixed-thickness
analysisof the heat bud-
Comparing the residual with the turbulent flux at
50 m, the largest discrepancyoccurs during the first
WWB peak, with the ocean losing more heat than
get is not sensitiveto the lower depth selectionwithin
this depth range in the cruise mean sense. There is
significant bias between the two flux estimates below
sea flux during this time period. The penetrating solar
(8 Wm-2). In Figure10a,the turbulentflux term at
50 m is included
which that
in the residual.
Without
can be mixed downward.
vertical
tur-
This could be due
to the underestimates of the turbulent flux, the sea sur-
70 m.
face heat loss, or the advectivecooling. Near the end
of this time period, the situationreverses(Figure 10b,
ing cruise2 (Figure8). As in the heat balancecase,the alsoseeFigure 13). As a cruisemean,the residualcomzonal and vertical advectionterms are small in affecting pares well with the turbulent flux at 50 m, and the heat
the surfacelayer. The meridionaladvectionbringsmore budgetis balanced
to 2 W m-• (Figure10b). The stansalt to the IFA mixed layer than the salinity change. dard error in the heat advection terms for the whole
This difference,the residualin Figure 8, has to be bal- time periodis 11 W m-2, whichis comparable
with eranced by the vertical gradient of the turbulent fluxes rors in other terms (Table 1). However,the errorsin
accordingto (2). Assumingthat the turbulent flux the advectionterms may havebeenoverestimated(ApFs is equal to the result from microstructuredata at pendix A).
50 m, the residual is integrated to obtain Fs(z). In
The combined precipitation and evaporation has a
this way, the predicted turbulent flux at the sea sur- cumulativeeffect of 9.3 psum (272 mm of freshwater;
faceis 0.38psumd-• (Figure9b), whichis equivalent Figure 11a), tendingto decreasethe salt contentin the
to 11.1 mmd-• precipitationminusevaporation
rate. upper 50 m; however, the salt content in this layer acUsing the WHOI evaporationdata, the predictedaver- tually increasesslightlyby 4.7 psum (-137 mm of fresh
ageprecipitation
is 15.4mmd-• duringourstudytime water). Most of the increaseoccursduring the first
period. This precipitation number is consistent with WWB peak. The advection terms play a major role in
the R/V Wecomaopticalrain gaugemeasurements;
the balancing this differenceand account for a changeof
R/V Moana Waverain rate is about 20% higherthan 7.6 psum (-222 mm). However,the residual,6.4 psum
this prediction, which may be partly due to spatial vari- (-187 mm), is still very large.
ations[Bradleyand Weller,1997].The F$(z) compares The turbulent flux has a cumulative effectof 4.2 psum
well with that derived from the microstructure
data in
(-123 mm) and doesnot accountfor all the aboveresidthe depth rangeof 40-70 m (Figure9b). The biasbelow ual. The imbalance is about 20% of the rain rate estima70 m is less than in the heat flux.
tion from the R/V Moana Wave,as shownbefore.The
The results from the microstructure data seem to unerror in the advection estimation can barely accountfor
derestimate the turbulent fluxes between 20 and 40 m
this imbalance(Figure 11b). Thus this imbalancecan
2 n - 3 )<10-6 s--2, N 2 be due to an overestimationof the rain rate by 20% in
(Figure9). WhenN 2 • Zmi
is assumed
to be equalto Nmi
• n in calculatingthe tur- the R/V Moana Wavedata or due to the spatialvariaThe salinity in the mixed layer slightly increasesdur-
bulent fluxesfrom the microstructuredata [$myth et
al., 1996b]. Note that only about half of the hourly
microstructure
data haveN • _>Nmi
• n above50 m (Figure 9c), so that the estimatefrom the microstructure
data may be biased(W. Smyth, personalcommunication, 1996). The 10 m microstructuredata are contaminated by the ship wake.
In the following temporal evolution analysis of the
heat and salt balances, we pick h-50 m as the lower
tionsof the rain rate [Bradleyand Weller,1997]or due
to error in the vertical
turbulent
flux estimate.
As a cumulativeresult, the meridionaladvectiondom-
inatesthe total advection,103 MJ m-2 (63 W m-2)
in the heat balanceand 9.5 psum (-278 mm) in the
salt balance(Figure 12). The effectsof the meridional
advection terms are to decreasethe temperature and
increase the salinity in the surface layer after each of
the two WWB peaks. This is due to the northward
10,300
FENG ET AL.- HEAT AND SALT BALANCES IN RESPONSE TO A WESTERLY WIND BURST
0
i
i
i
i
i
i
i
,
i
i
i
i
i
i
i
i
i
i
Rs(50)
%
i
i
,
-20
; %•'
-40
''N'Advection
..... Res•ual
E -6o
Qo
-8o
-lOO
-12o
Temporal
-14o
-16o
40
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
i
I
I
•
]
22
•
=
24
[
[
26
'
]
28
•
[
30
]
=
1
.=
]
3
[
•
5
'
•.
7
[
20
-20
-40
20
December
1992
9
January 1993
Figure 10. (a) Cumulativeheat budgetin the 0-50 m layer. The advectionterm includesthe
zonal, meridional, and vertical advection. The residual includesthe vertical turbulent flux at
50 m and estimationerrors. (b) The comparison
betweenthe residualin Figure 10a with the
verticalturbulentfluxestimatefromthemicrostructure
data.Es (5 W m-2) andEa (11W m-2)
are the cumulative errors in the surfaceheat flux and advectionterms, respectively.
inertial current during these two time periods and is
consistentwith the observednorthward propagationof
frontlike featuresin both surfacetemperature and salin-
surface heat flux tends to heat the surface layer, advection still causesthe mean temperature in the water
ity [Huyer et al., 1997].Duringthe JLW, whenthe net
Generally, zonal advection showssome eddylike behavior, but the net effects are small compared with
columnto decrease(Figures10 and 12).
meridional
Table 1. Comparisonof Error Estimationsof Different
Terms.
Terms
Net surfaceheat flux [Weller and
Anderson,1996]
Turbulentheat flux at 50 m [Smyth
et al., 1996b]
Heat advection(presentstudy)
Errorestimations
givenin Wm -2.
Standard
< lO
lO
11
Error
advection.
The
zonal
salt advection
is as
important as the meridional advection until January 2
when the zonal salinity gradient changessign and the
current turns northeastwardin the surfacelayer. Vertical advection is unimportant until the end of the cruise
whenthe surfacelayeris stratified(Figure 12). The net
effects of vertical
advection
are to increase the surface
temperature and decreasethe surface salinity; this is
causedby the downward vertical velocity in this layer.
The instantaneousheat balance in the 0-50 m layer,
using the near 3-day analysis, is shown in Figure 13.
The temperature decreasesall the time. The vertical
flux, which includesthe net surfaceheat flux, the pen-
FENG ET AL.' HEAT AND SALT BALANCES IN RESPONSETO A WESTERLY WIND BURST
5
lO
i
i
'
i
i
i
i
i
i
i
i
i
i
i
i
i
i
i
i
10,$01
i
a
l::
-5
,
....
-lO
So(P-E
)
....
-15
15
N
Temporal
--
Advection
Residual
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
[I
I
I
I'
I
lO
l::
Residual
Ea
--5
I
20
I
22
December
I
I
24
I
I
26
I
I
28
1992
I
I
30
I
I
I
1
I
I
3
i
5
I,
I
7
I
9
January 1993
Figure 11. (a) Cumulativesalt budgetin the 0-50 m layer. The advectionterm includesthe
zonal, meridional, and vertical advection. The residual includes the vertical turbulent flux at
50 m and estimationerrors. (b) The comparison
betweenthe residualin Figure 11a with the
vertical turbulent flux estimate from the microstructure data. Ea is the cumulative error in the
advection
terms.
etrating solar radiation at 50 m and the vertical turbu-
is estimated only at one point, and the rain events are
lent flux at 50 m, is negativeduring most of the time sparse
in space[Shortet al., 1997].
except near the end of this time period. The advection
coolsthe surfacelayer duringthe two northwardadvec- 4.3. Diapycnal Velocity
tion time periods. The residualsare larger than in the
The entrainmentprocessof the mixed layer can be
cumulativeresultsduring some3-day periods,with im- seenfroma comparison
of the mixedlayerdepthandthe
balances
aslargeas90 W m-2. The largeerror,onthe as = 22 isopycnaldepth (Figure14a). Both depthsare
one hand, is causedby averagingovershort time peri- calculatedfrom the R/V Wecomasurveydata so that
ods, so that the errors in the different terms are more they contain both temporal and spatial variations. The
than doubledthe numberslisted in Table 1. However, mixedlayer depthis calculatedusinga densitygradient
the temporal variation term has standard error of about criterion
of 0.01kgm-4 [LukasandLindstrom,
1991].
20 W m-2, whichhasbeenavoided
by usingthe mean
Note that there are two occasions that both the mixed
value instead of integrating the temporal gradient in layer depth and the isopycnaldepth deepenwith the
the cumulativecalculations(AppendixA). The instan- occurrencesof the WWB peaks,one during December
taneous salt balance is not shown because the rain rate
22-25 and the other betweenDecember30 and January
10,302 FENG ET AL.' HEAT AND SALT BALANCESIN RESPONSETO A WESTERLY WIND BURST
0
i
i
•1
i
i
20
i
i
i
i
i
i
i
i
i
i
i
i
i,
I
zonal
-----•-
•
vertical
-2O
-4O
meridional
-6O
-80
-
-100
-
I
-120
10
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
ll
I
,,
i
I
i
i
i
i
i
!
i
i
i
i
i
i
i
i
i
i
i
/
/'
_
/'
/-
sum
•"meridional
("3
zonal
2-
0
vertical
-2
ß rl
-4
20
.I
22
December
I
I
24
I
I
.I
26
I
28
1992
I
. I
30
. I
I
I
1
I
I
3
I,
5
I
I
I
7
9
January 1993
Figure 12. Cumulative
effects
of the zonal,meridional,
andverticaladvection
in (a) heatand
(b) salt balancesin the upper50 m layer.
2. Convergence
in the surfacelayer (abovece = 22) WWB peak. This deepeningmay generate a pressure
may causethe deepeningof the isopycnaldepthduring gradient in the zonal directionand thus be related to
January 4-6 when the mixed layer shallows.
the acceleration of the subsurface SEC.
The isopycnaldepthdeepensto the north (tilts upward to the south) in responseto the northwardinertial currentduringDecember27 to JanuaryI and after
January6 (Figure 14b). Thesesouthwardtilts of the
isopycnaldepth may causethe meridionaltemperature
and salinity gradientsin the surfacelayer. The zonal
gradientof the isopycnaldepthis smallmostof the time,
but it deepenstoward the east in responseto the first
In the isopycnalboundary method, the ae-22 isopycnal is used as the lower boundary of the surfacelayer.
Becausethe layer above the a•-22 isopycnaldepth is
not completelywell mixed duringthe entiretime period,
both positiveand negativediapycnalvelocitiesmay affect the propertiesof the surfacelayer.
Generally,the vertical gradientsof the temperature,
salinity,and potential densityabovethe isopycnaldepth
FENG ET AL.- HEAT AND SALT BALANCES IN RESPONSE TO A WESTERLY WIND BURST
lOO
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
10,303
I
a
-lOO
-200
Temporalchange
Vertical
flux
Advection
Residual
I
-300
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
......
..:-...::..
.•.,,..•-,,1•,
•"'•'•'••••::-'"..'.',
:'-,•",.•i•'"':•
' '......•:'.-"
....
•F.:.
"•.':'
"'..:":.
.::
0
,,•-ø•
' .....
ii:....
....
......
""'•{i•*:•"
'""::•:'""::'"";'•::'•:;:':-:•::(:'
"•"
"'•::'
'::':'''
'"'......
:'"'"
':'"'f
::•:'•':"""'"•'"•
%::'•:"
'.....................
•::':'•
....:"'•'•
'
.
,. =============================.
...............
..•,•,
.....
•..::•i:i:..,.•.,-.....!;•:.
ß''"'"':'•"
'•':*':•"•
..............
::•..x::.........,•.:.
.........
....
,•.:,..::•:*:::•½.-:,•-'.•
....
•
•:,•.I'•i•:•.:,::.•!•
.......
....:
..
' ....
...:
.!. .
l• -100
_oo
- '- '- Zonal advection
----Meridional
advection
_-.:• V_ertic_al
ad.
vect!on
.
20
22
December
24
26
28
1992
30
1
3
5
7
9
January 1993
Figure 13. (a) Heat balancein the near 3-day analysis,and (b) the zonal, meridional,and
vertical advection effectsin the near 3-day intervals in the upper 50 m layer. The shaded areas
represent the standard errors in the estimates.
are smaller than those below the isopycnal. Becausediapycnal advection is an upwind process,positive we,
equivalentto the entrainment in the mixed layer analysis,is more efficientin modifyingthe surfacelayer than
Significantpositivewe occursduringthe WWB peaks,
from both the density conservationmethod and the entrainment calculation using the Niiler-Kraus formula
(Figure 16). The resultfrom the Niiler-Krausformula
is dominated by the diurnal cycle. After the last WWB
tweenthe vertical gradientsaboveand belowthe isopyc- peak, there is substantial negative we using the dennal asthe efficiencyratios betweennegativeand positive sity conservationmethod; this may be due to the rem-
negativewe (detrainment).Thuswe definethe ratio be-
we (detrainmentand entrainment)(Figure 15). Using nant of turbulencenear the as = 22 depth [Smythet
a linear fit, the detrainmentof the temperatureis 84% al., 1996b]. Usingn - 0.16,m - 0.24, and s - 0.24,
as efficient as the entrainment. The salinity efficiency the estimated cumulativeentrainmentusingthe NiilerKraus formula and the cumulativepositivewe usingthe
is 75%, and the densityefficiencyis 79%.
10,304
FENG ET AL.: HEAT AND SALT BALANCES IN RESPONSE TO A WESTERLY WIND BURST
ß
ß
ß
ß
20 '
a
ß
ß
**
ß
•' 40'
ß •lo ill
r'• 60
00elll• Illo'
""'"''•'
'
ß
Oo
''
'le
ß
ß
,
ß .,• •.
ß
'
,
I•:.3d•
.
...
.
me ß ee o
•I m•*
.
ß
*lflee
ooze
,
80
lOO
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
ß
:..
•::
20
::
.:.
Zonal
-lO
20
22
December
24
26
28
30
1992
Ol
03
05
07
09
January 1993
Figure 14. (a) The hourlymixedlayerdepth (circles)and the c0 = 22 isopycnaldepth (solid
line) from the R/V Wecomadata. The bold line represents
the linear fit meanof the isopycnal
depth. (b) The eastward(dashed)and northward(solid)gradientsof the isopycnaldepth. Units
(m/degree)are per degreelongitudeand latitude respectively.The shadedareasrepresentthe
standard
errors in the estimation.
density conservationmethod are very close. When the
isopycnalis closeto the mixed layer depth, the entrainment velocity using the Niiler-Kraus formula can
pict the positivediapycnalvelocityby selectingthe
ciencyfactors. The diapycnal velocity determinedfrom
the microstructure data depicts the turbulent structure
near the isopycnaldepth; it equally representsthe positive and negative diapycnal velocities. Therefore, in
our calculations
of the heat and salt balances above an
isopycnaldepth, the diapycnal velocity calculatedfrom
the microstructure
4.4.
data is used.
Heat and Salt Balances Above ce:22
The mean temperatureabovethe c0 - 22 isopycnal
decreases
by 0.7øCduringour study time period(Fig-
ure 17a). This temperaturedecreaseis about equally
accountedfor by the net sea surfaceheat flux, the horizontal advection, and the vertical processesat the iso-
FENG ET AL.' HEAT AND SALT BALANCES IN RESPONSE TO A WESTERLY
10,305
pycnal surface which are represented by the residual
in Figure 17a. The penetrating solar radiation is very
Temperature
0.8
WIND BURST
ß
small.
From Figure 17b, the diapycnalturbulent flux occurs
only at the end of the time period when the mixed layer
0.6
is shallow
and there
is still
extensive
turbulence
near
the isopycnaldepth [Smythet al., 1996b]. Diapycnal
E
advection occurs during the two WWB peaks with a
cumulative effect almost as important as the diapycnal
turbulent flux. Consideringa mean isopycnaldepth of
76 m, the diapycnalflux has an effect of 17 W m-2
andthe diapycnaladvectionhasan effectof 13 W m-2
0.2
while the penetrating solar radiation only has an effect
0
0.2
0
0.4
0.6
0.8
of 2.4 W m-2, duringthe 19-daytime period.Adding
theseup, thereis a total of 32 W m-2 of heatentering
acrossthe ere - 22 isopycnal. The imbalance is near
C/m
0.06øC(about10 W m-•).
The temporal variation of mean salinity aboveere- 22
is only of the order of 0.1 psu, which is smaller than
the temperature in terms of its effect on the density
Salinity
(Figure 18a). The surfacefleshwaterinput tends to
decreasemean salinity by 0.12 psu during this time period; the advection terms tend to increase the mean
salinity by 0.13 psu. Thus the increaseof mean salin-
ity (0.087 psu) is almostcoincidentwith the residualin
ß
Figure 18a. The diapycnalturbulent flux and diapycnal
advectionare almostequallyimportant(Figure18b),as
-0.1
-0.2
-0.2
in the temperature case. The sum of these two terms
does not account for the above residual, and the imbalance is near 0.03 psu. Multiplied by the mean water depth of 76 m, this is comparable to the imbalance
from the fixed thicknessmethod. Thus the isopycnal
boundary method yields the same rain rate estimate,
-0.1
15.4 mmd-i, as in the fixed thicknessmethod. The
psu/m
standard
errors
in the rain
rate
estimations
from
the
oceanbudgetareequivalent
to 4 mmd-1 in bothmethods.
Density
5.
C.
Discussion
It has long been assumedthat the horizontal advec-
ß
-0.1
tion effect on the heat balance in the center of the warm
pool region is negligible. However, during the Decem-
ber 1992WWB event,the advection
effect(43 W m-2)
ß
ß
E -0.2
is more important than the air-sea flux in coolingthe
surfacelayer at the centerof the IFA from the IOP observations.A similarshort-termstrongadvectionevent
-0.3
-0.4
-0.4
is identifiednear the equator[Cronin and McPhaden,
1997]. Thus, for someshorttime periods,advectionis
-0.3
-0.2
-0.1
0
Figure 15. Comparisonof the vertical gradientsof the
important in the surfacelayer heat balance. In the total
advection, the meridional advection term is dominant
in both heat and salt balancesduring this 19-day period
(Figure 12). There are two meridionaladvectionevents
during this time period related to the two occurrences
(a) temperature,(b) salinity,and (c) densityabove(Y
axes)and below(X axes)the ao - 22 isopycnaldepth of WWB peaks. Thus the WWB-related eastward jet
from the R/V Wecomadata. The solid lines are the and inertial motionshavesignificantimpact on the heat
linear fits of the ratios.
and salt balances.
10,306
FENG ET AL' HEAT AND SALT BALANCES IN RESPONSE TO A WESTERLY WIND BURST
2O
10
E
-10
-2O
-3O
-40
20
I
10 -
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
b
-
I
-10
-2O
I
-
VV i[ /I'11'fl
'I
-3O
I
-40
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
20
We+
We-
-lO
-
/
Microstructure
Niiler-Kraus
-20
•
I
20
22
December
•
•
24
1992
•
•
26
I
I
28
I
I
30
I
01
03
05
07
09
January 1993
Figure 16. Diapycnalvelocitiesacrossthe as - 22 isopycnaldepthfrom the (a) microstructure
data and (b) Niiler-Krausformulaand (c) their temporalcumulativeeffects.The velocitiesare
separatedinto positive(we+) andnegative(we-) partsin the cumulativecalculations
(we- from
Niiler-Krausformulais not shown).
From the isopycnalboundary method, both diapyc-
nal turbulentflux and diapycnaladvection(verticalentrainment) are importantto the surfacelayerheat and
salt balancesabovethe as - 22 depth (Figures17 and
18). The total heat flux acrossthis isopycnalis more
than 30 W m-2. The WWB conditions
typicallyprevail for about 2 monthsof a year [Smythet al., 1996b].
During the rest of the year, the total heat flux across
as - 22 is weak[Wijesekeraand Gregg,1996].If weuse
2 W m-2 to represent
the restof the year,whichis the
upper limit of the heat flux at 100 m during the weak
wind cruiseI surveyof the IOP [Wijesekeraand Gregg,
1996],then we obtain an annualmeanheat flux across
the as - 22 depthof 7 W m-•. McPhadenandHayes
[1991]estimateda long-termmeanheatfluxof 5 W m-•
at 75 m in the warm pool regionbasedon the entrainment flux only due to wind work; this is comparable
with our estimate.
We have also analyzed the salt budget during the
othertwo cruisesof the R/V Wecoma
survey.The anal-
FENG ET AL' HEAT AND SALT BALANCES IN RESPONSE TO A WESTERLY WIND BURST
I
I
I
I
I
I
I
I
I
I
I
I
I
I
[
I
I
I
10,307
I
Rs(z:
• ••R..esidual'
'-. -...... ..... .
Advection
-0.2
QO
-0.4
dT
-0.6
-0.8
I
I
--
....
0.2
I
I
I
R•$i•Juai
Diapycnal turbulentflux
, . Diapycnalflux + advection
Ea
0
0
eme
ememelmememememelmem
e ß s elmem
-0.2
,
20
I
I
22
I,
I
24
December 1992
I
I
26
I
I
28
I
I
30
I
I
01
I
I
03
I
,
I
05
I
I
07
I
09
January 1993
Figure 17. (a) Cumulativemeantemperaturechangeabovethe ao = 22 isopyenaldepth. The
advectionterm is the sum of mean advectionand stratified shearflow advection.The diapycnal
turbulent flux and diapycnaladvectionare part of the residualterm. (b) Comparisonof the
residualwith the diapycnalturbulent flux, the sum of diapycnalturbulent flux, and diapycnal
advection acrossthe isopycnaldepth. E• is the cumulative error in advection terms.
ysisof cruise1 yields a mean rain rate of 2.2 mm d-•
to simply assumethat detrainment has no effect on the
duringa 15-daytime period (November15-30, 1992), surfacelayer.
and the analysisof cruise 3 yields a mean rain rate of
One other problemrelated to our budgetestimationis
4.6 mmd-• duringa 15-daytime period(January29 that the vertical flux terms are all from point measureto February12, 1993) [Feng,1997;Bradleyand Weller, ments. There is no problem for the surfaceheat flux,
1997]. Averagingover the three cruises,a mean rain according to the intercomparisonwork by Weller and
rateof 8 mmd-• isobtained.Considering
a meanevap- Anderson[1996]. However,the rain rate estimations
orationrate of 3.8 mmd-•, the net freshwater
flux from still needfurther intercomparison
[Godfreyet al., 1998;
our analysisis consistentwith the annual mean estima- Bradley and Weller, 1997]. Cruiseaverageradar rain
tion of Donguy[1987],1.5my-• (4.1mmd-•).
rate data showspatial variationsoverthe surveydomain
Becausethe layer above the eye- 22 depth is not of a factorof 4 [Shortet al., 1997].Suchvariationsmust
an isothermalor isohalinelayer, both positiveand neg- be taken into account when calculatingupper ocean
ative diapycnal velocities will affect the heat and salt freshwaterbudgets. The oceansurfaceprocesses
in rebalancesin the surfacelayer, while negativeadvection sponseto the WWB have spatial scalescomparableto
(detrainment)is lessefficientthan positiveadvection the surveydomainso that a point measurementmight
(entrainment). Thus it is only a crudeapproximation be insufficientto determinethe meanvertical (diapyc-
10,308
FENG ET AL.' HEAT AND SALT BALANCES IN RESPONSETO A WESTERLY WIND BURST
0.15
0.1
0.05
-0.05
-S(P-E)
-0.1
!
-0.15
0.1
I
I
I
I
I
I
I
I
I
I
,
I
I
I
I
I
I
,,I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
0.05
0
Ea
-0.05
-
--
Residual
....
•
Diapycna!turbulentflux
Diapycnalflux + advection
.I
20
I
22
I
I
24
I
I
26
I
I
28
December 1992
I
I
30
I
,
,
01
I
I
I
03
I
05
I
I
07
I
,
09
January 1993
Figure 18. Sameas Figure 17 but for salinity.
nal) turbulentflux anddiapycnaladvectionoverthisre- horizontalgradientsof temperature and salinity on the
gion. Fortunately,the point measurements,
the WHOI timescaleof 3 days with the repeated survey time of
mooring,and R/V Moana Wave,are very closeto the 1.5 days. This survey pattern resolvesthe main variacrossoverpoint of the survey. Assuminglinear spatial tions related to the December 1992 WWB at the center
variations of the vertical processes,the point measure- of the IFA region.
Two methods are used to analyze the heat and salt
mentsshouldrepresentthe mean overthe wholeregion.
balances
in the near surface layer. One uses a fixed
The gradientRichardson
number(Ri) at the isopycnal
depth of ers= 22 doesnot have a significantmeridional thicknesscontrol volume from which the budgetscan
variation trend acrossthe survey domain, whereasthe be computeddirectly usingthe temperatureand salinity
meridional variation of Ri at 55 m seems to have a linear
tendency equations;the other usesan isopycnaldepth
trendoverthe domain(Figure19). The zonalvariations as the lower boundary of the control volume, in which
casethe mean temperature can be better related to the
of Ri are generally small.
seasurfacetemperature and the low-modeinternal wave
effectscan be effectively eliminated from the advection
6. Summary
In this paper,the R/V Wecoma
IFA surveydata are
term calculation.
In both cases,the surfacelayer heat budget can be
used to estimate the advection effects on the near sur-
balancedwithin 10 W m-2 overa 19-daytime period.
facelayer heat and salt balancesin responseto a WWB.
It turns out that the butterfly survey can resolvethe
ing the R/V Moana Waveprecipitationrate estimate
The surface layer salinity can be balanced by reduc-
FENG ET AL.' HEAT AND SALT BALANCES IN RESPONSE TO A WESTERLY WIND BURST
the ship is near the crossoverpoint to avoid misinterpreting the temporal and spatial gradientsdetermined
by the fitting. Each time interval has approximately
3 days, so that the contribution from variabilities on
shorter timescalescan not be fully resolved. Some intervals are excluded, such as near January 4 and after
January 9, when the data only cover part of the re-
103
102
101
• 0o
1
gion and would give a biasedestimateof the gradients.
10-1
The center of the time interval and the crossoverpoint
are used to represent the results. There are altogether
10-2
20 time intervalswe selectduring cruise2 of the R/V
103
102
101
10o
10-1
10-2
10,309
i
.
2.4S
i
i
,
,
2S
i
,
i
,,
1.6S
i
,
i
1.2S
Latitude
Wecoma survey, and the adjacent time intervals usually have overlappingtime. Thus the 20 time intervals
cover approximately 20 days. Note that in the cumulative calculation the integration is from December20,
1992, to the central time of the last time interval, which
is approximately 19 days.
In the temperature and salinity fits, the errorsin the
mean values are negligible. Thus, in the cumulative
property analyses,the temporal gradientsare not used.
Instead, we use the mean values themselves,such as
T and $, in order to avoid the cumulative errors of
the integrationsof the temporal gradients.In this way,
we can almost nullify the errors in these terms. The
Figure 19. GradientRichardson
number(Ri) calcu- error estimations of the cumulative advection terms related from the R/V Wecomadensityand velocitydata quire a decorrelation timescale, where 3 days is used.
as a functionof latitude at (a) 55 m and (b) the depth This timescale is selected because this is the minimum
of a0=22 kgm-3 depth.The solidlinesin thetwoplots
timescalethat the horizontalgradientscan be resolved
are Ri = 0.25.
and is also the approximate timescale of the WWB
by 20%. In other words, we can predict the rain rate
peaks.
In the fixed depth heat balance, from the bootstrap
fromthe oceanbudgetto be 15.4mmd-z duringthis analysis,the standarderrorsfor the estimatesof •(z),
time period. This number is closerto the optical rain 5(z) and w(z) are Au(z), Av(z), and Aw(z), and the
gaugemeasurements
on the R/V Wecoma
[Bradleyand errorsfor the temperaturegradientsare ATx(z) and
Weller,1997].Someof the difference
maybe dueto the ATy(z), then the errorsfor the threeadvectionterms
spatial variation of rain rate.
In both control volume methods, the advection terms
are importantin both the temperatureand salinitybalances,thoughthey have differentforms. Thus, evenin
the center of the warm pool region, the advectioneffectscannotbe generallyneglectedin orderto closethe
heat and salt budgets. The error in the heat advection
estimationis 11 W m-2 over this 19-daytime period,
which is comparablewith errorsin other terms, suchas
in the 3-day analysisare
zx
U dZ _<
h
ZX
u
OT
A
the surface heat flux and the vertical turbulent flux.
In the isopycnalboundarymethod,the estimateddiapycnalturbulent flux and diapycnaladvection(entrainment) are almost equally important. The total
heat flux across the a0 -
22 surface is more than
30 W m-2 duringthis studytime period.
Appendix
+ lulZX dz -
OT
A
z j <_ ZXw
Note that the errorsof OT/Oz are negligiblecompared
to thoseof w. Assumingthat •x, •y, and •z are independent,the error for the total advectionin the 3-day
analysisis
A- Linear Fit
In order to obtain the best estimates of the spatial
A
u+v
+w
+ )
and temporalgradientsfrom the R/V Wecomasurvey
data, the time interval used in each estimate includes
The residualof heat balance(Figure 13) tends to
two completecircuitsof the butterfly pattern. The be- have a decorrelationscaleof nearly 2 days. Also, we do
ginningand endof eachtime intervalarethe timeswhen not consider the decorrelation scale in the vertical inte-
10,310
FENG ET AL.: HEAT AND SALT BALANCES IN RESPONSE TO A WESTERLY WIND BURST
Radiation flux
gration of error. Assumingthat there are two degrees
of freedom in the vertical and a 2-day decorrelationin
0
F
0
F
0
time, we have overestimated the standard error in the
advection terms by nearly a factor of 2.
In the isopycnalboundary method, we first calculate
ßD
•
the mean valuesTa(xo,Yo),$a(XO,YO),and ua(x0,Y0)
by simply vertically integrating T(xo, yo,z),
$(xo, yo,z), and u(x0, y0,z) from seasurfaceto the isopycnaldepth at eachhourly data point (xo,Yo). Then
we determinethe deviationsT' (xo,yo,z), $' (xo,Yo,z),
and u•(x0,yo,z) by subtractingthe mean valuesfrom
T(xo, Yo,z), S(xo,yo,z), and u(x0, Y0,z). Finally, we
.. ::•'
,.
ß
.-•?
(a)
D';:
.
..
(b)
j
(c)
Temperature
T
T
T
vertically integrate T•u • and $•u • from sea surfaceto
theisopycnal
depth
at(x0,Y0)todetermine
f_o
hT*u•dz
andf_o
h$'u'dz,which
arealsofunctions
of (xo,yo).
The horizontal gradientsin the advectionterms are calculated from the near 3-day linear fit of above terms.
The bootstrap method is used to estimate their errors.
(d)
(e)
(f)
Figure
B1.
A sketch for the diapycnal velocity.
(a) The radiation profilenear depth D, which can be
separatedinto (b) a mean part with a valueof F and
It is assumed that the errors in the mean advection and
(c) a linear trend. (d) The originaltemperatureprostratifiedshearflow advectionin (5) and (6) are inde- file near depth D. (e) The temperaturechangedue to
pendent.
a mean part with a value of F. (f) The temperature
changedue to a linear tread. T is a selectedisotherm,
which deepensfrom D to D1 in Figure B lf becauseof
Appendix B- Diapycnal Velocity
positive diapycnal velocity.
Here we discussthe equivalencyof the two methods
to calculate the diapycnal velocity: the density conser-
vation method, (13), and the definition,(11). When
assuminga linear relation betweenp, T and $, we can
combine(1) and (2) yielding,
+u.
+wo-;-
ohso(
isotherm. Under the secondpart, the radiation flux at
depth D is zero. However,the flux convergence
causes
the temperatureto increasenear D (Figure Blf). The
isothermalsurfaceat depth D would deepenat a speed
.
w = (Op/Ot)/(Op/Oz)= [(aORs/Oz)/(poCp)]/(Op/Oz),
becauseof absorptionof solar radiation, so that we need
to relocate the isotherm T, which is at depth D1. We
call this entrainment or diapycnal advection. In this
Op Op
Op
process,water is in effectentrainedat the originaltemw• = 0-•+ u. VP+Wo•
perature of water between depths D and D•, which is
by replacingOh/Or with (Op/Ot)/(Op/Oz),etc. Com- lower than the averagetemperature above depth D.
The rate of increaseof mean temperature above the
biningthe abovetwo equations,(13) is obtained.
Now let us consider a simple case of the solar radia- depth of this isopycnalwill be reducedby the entraintion flux to illustrate how the diapycnal velocity affects ment process,as well as by direct heat loss out of the
the mean propertiesin the surfacelayer (S. Godfrey, bottom. The same argument appliesto the turbulent
Equation(11) can be rewrittedas
personalcommunication,1997). In this case,there is
flux.
no motion of any kind and no turbulence; the density is only determined from the temperature so that
Acknowledgments. We wouldlike to thank W. Smyth
an isotherm can be used to represent an isopycnal. for providing the R/V Moana Wave microstructuredata
Figure B la showsthe radiation flux near an isotherm and sharing his ideas on budget calculations. S. Anderson
depthD, wherethe temperatureequalsT (FigureBld). and R. Weller providedthe WHOI mooring data and ofduring this work. We alsothank
We assumethat the flux Rs(z) is linearly decreasing fered valuablesuggestions
R.
DeSzoeke,
E.
Firing,
S.
Godfrey,
C. Paulson,E. Antoniswith depth, and the flux at depth D is F. Because
sen, M. Cronin, and H. Wijesekerafor their helpful sugthis is a linear system, the radiation flux can be sepa- gestionsand discussions,and we thank the anonymousrerated into two parts: one part has a constantflux of F
viewersfor comments.The R/V Vickerssolar penetration
(FigureBlb); the other part containsthe linear trend data werekindly providedby D. Siegel,andthe R/V Moana
with zero flux at depth D (Figure Blc). As an effect Wave surfacedata were kindly provided by C. Fairall. We
of the first part, there is no temperature changenear thank both the Universityof Hawaii and the OregonState
University groupsfor the R/V Wecomadata processing.
depthD (FigureBle), but the radiationflux is nonzero. This work was fundedby the PhysicalOceanographyProThis heat loss needs to be included in estimating the gram under NSF grant OCE-9113948. This is SOEST conrate of change of depth-mean temperature above the tribution number 4543.
FENG ET AL.: HEAT AND SALT BALANCES IN RESPONSE TO A WESTERLY WIND BURST
References
Anderson, S.P., R.A. Weller, and R. Lukas, Surface buoyancy forcing and the mixed layer of the western Pacific
warm pool: Observationsand 1-d model results, J. Clim.,
9, 3056-3085, 1996.
Bradley, F., and R. Weller (Eds.), Fourth workshopof the
TOGA COARE air-seainteraction(flux) workinggroup,
report, 72 pp., Woods Hole Oceanogr. Inst., Woods Hole,
Mass., 1997.
Chen, S., R.A. Houze Jr., and B. Mapes, Multiscale variability of deep convectionin relation to large-scalecirculation
in TOGA COARE, J. Atmos. Sci., 53, 1380-1409, 1996.
Cronin, M.F., and M.J. McPhaden, The upper ocean heat
balance in the western equatorial Pacific warm pool dur-
10,$11
McDougall, T.J., and Y. You, Implications of the nonlinear equation of state for upwellingin the oceanfloor, J.
Geophys.Res., 95, 13,263-13,276,1990.
McPhaden, M.J., and S.P. Hayes, On the variability of
winds, sea surfacetemperature, and surfacelayer heat
content in the western equatorial Pacific, J. Geophys.
Res., 96, 3331-3342, 1991.
Mourn, J.N., The questfor Kp, J. Phys. Oceanogr.,20, 19801984, 1990.
Mourn, J.N., M.C. Gregg,R.-C. Lien, and M.-E. Carr, Comparisonof turbulent kinetic energydissipationrates from
two ocean microstructure profilers, J. Atmos. Oceanic
Technol., 12, 346-366, 1995.
Niiler, P.P., and E.B. Kraus, One-dimensionalmodelsof the
upper ocean, in Modelling and Prediction of the Upper
ing September-December1992, J. Geophys.Res., 102,
Layers of the Ocean,edited by E.B. Kraus, pp. 143-172,
8533-8553, 1997.
Pergamon, Tarrytown, N.Y., 1977.
Donguy,J., Recentadvancesin the knowledgeof the climatic Osborn, T.R., Estimates of the local rate of vertical diffusion
variations in the tropical Pacific ocean, Prog. Oceanogr.,
from dissipationmeasurements,J. Phys. Oceanogr.,10,
19, 49-85, 1987.
83-89, 1980.
Efron, B., and R.J. Tibshirani, Bootstrap methodsfor stanShort, D.A., P.A. Kucera, B.S. Ferrier, J.C. Gerlach, S.A.
dard errors, confidenceintervals and other measuresof
Rutledge and O.W. Thiele, Shipboardradar rainfall patstatistical accuracy,Star, Sci., 1, 54-77, 1986.
terns within the TOGA/COARE IFA, Bull. Am. MeteoEldin, G., T. Delcroix, C. Henin, K. Richards, Y. du Penrol. Soc., 78, 2817-2836, 1997.
hoar, J. Picaut, and P. Rual, Large-scalecurrent and ther- Siegel,D.A., J.C. Ohlmann,L. Washburn,R.R. Bidigare,
mohalinestructuresalong 156ø E during the COARE InC.T. Nosse,E. Fields,and Y. Zhou, Solarradiation,phytensive ObservationPeriod, Geophys.Res. Left., 21, 2681toplankton pigmentsand the radiant heatingof the equa2684, 1994.
torial Pacific warm pool, J. Geophys.Res., 100, 4885Fairall, C., E.F. Bradley, D.P. Rogers,J.B. Edson, and G.S.
4891, 1995.
Young, The TOGA COARE bulk flux algorithm, J. Geo- Smyth, W.D., D. Hebert, and J.N. Mourn, Local oceanrephys. Res., 101, 3747-3764, 1996.
sponseto a multiphasewesterly windburst, 1, Dynamic
Feng, M., Semidiurnaltide and heat/salt balance in the
response,J. Geophys.Res., 101, 22,495-22,512,1996a.
westernequatorialPacificduring TOGA COARE, Ph.D.
Smyth, W.D., D. Hebert, and J.N. Mourn, Local oceanrethesis, 143 pp., Inst. of Oceanol., Chin. Acad. of Sci.,
sponseto a multiphase westerly windburst, 2, Thermal
Qing Dao, 1997.
and freshwaterresponses,J. Geophys.Res., 101, 22,513Feng, M., M.A. Merrifield, R. Pinkel, P. Hacker and A.J.
22,533, 1996b.
Pluddemann, E. Firing, R. Lukas and C. Eriksen, SemidiSprintall,J., and M. Tomzak,Evidenceof the barrierlayer
urnal tides observedin the western equatorial Pacific durin the surfacelayer of the tropics, J. Geophys.Res., 97,
ing Tropical Ocean-Global AtmosphereCoupled Ocean7305-7316, 1992.
AtmosphereResponseExperiment, J. Geophys.Res., in Stevenson,
J.W., and P.P. Niiler, Upper oceanheat budget
press, 1998.
duringthe Hawaii-to-Tahitishuttleexperiment,J. Phys.
Godfrey, J.S., and E. Lindstrom, The heat budget of the
Oceanogr.,13, 1894-1907, 1983.
equatorialwesternPacificsurfacemixed layer, J. Geophys. Webster,P.J., and R. Lukas,TOGA COARE: The coupled
Res., 9•, 8007-8017, 1989.
ocean-atmosphereresponseexperiment, Bull. Am. MeteGodfrey, J.S., R.A. Houze Jr., R.H. Johnson,R. Lukss, J.orol. Soc., 73, 1377-1415, 1992.
L. Redelsperger,A. Sumi, and R. Weller, CoupledOcean- Weller, R.A., and S.P. Anderson,Surfacemeteorology
and
AtmosphereResponseExperiment(COARE): An interim
air-seafluxesin the westernequatorialPacificwarm pool
report, J. Geophys.Res., in press,1998.
duringthe TOGA CoupledOcean-Atmosphere
Response
Gregg, M.C., The coupled structures and fluxes in a deep
Experiment, J. Clim., 9, 1959-1990, 1996.
convectingmixed layer, in Dynamics of the OceanicSur- Wijesekera,H.W., and M.C. Gregg,Surfacelayer response
face Mixed Layer, Aha Hulikoa Proceedings,Honolulu,
to weak winds, westerly bursts, and rain squallsin the
HI, 1987, edited by P. Muller and D. Henderson,pp. 1westernPacific warm pool, J. Geophys.Res., 101, 977-
23, Univ. of Hawaii, Honolulu, 1987.
Harrison, D.E., and G.A. Vecchi, Westerly wind eventsin
the tropical Pacific, 1986-1995, J. Clim., 10 3131-3156,
997, 1996.
1997.
Huyer, A., P.M. Kosro, R. Lukas, and P. Hacker, Upper-
M. Feng, P. Hacker, and R. Lukas, School of Ocean
oceanthermohalinefieldsnear 2øS,156ø E duringTOGA- and Earth Scienceand Technology,University of Hawaii,
COARE, November1992 to February 1993, J. Geophys. Honolulu, HI 96822. (e-mail: [email protected];
[email protected];
[email protected];)
Res., 102, 12,749-12,784, 1997.
Lukas, R. and E. Lindstrom, The mixed layer of the western equatorial Pacific ocean, J. Geophys.Res., 96, suppl.,
3343-3358, 1991.
(ReceivedApril 8, 1997; revisedSeptember26, 1997;
acceptedOctober 15, 1997.)