You are on page 1of 12

JOURNAL

OF GEOPHYSICAL

RESEARCH, VOL. 94, NO. C5, PAGES 6273-6284, MAY

15, 1989

Similarity Scalingof Viscousand Thermal


Dissipationin a ConvectingSurfaceBoundaryLayer
C. P. LOMBARDO1 AND M. C. GaEGG
Applied Physics Laboratory and Schoolof Oceanography,College o] Ocean and Fishery Sciences
University o] Washington, Seattle

By continuously deploying a turbulence profiler in the upper ocean, we observed the rate of
viscousdissipation of turbulent kinetic energy, e, and the rate of diffusive smoothing of turbulent

temperaturefluctuations,X, during elevendiurnal cyclesof the surfaceboundary layer (SBL).


Even though we restricted our analysisto times when the oceanlost buoyancyat a nearly constant
rate, we observed a wide range of conditions, including dominance of the turbulent production
by the wind stress.Throughout, e was normalized very well by the sum of the similarity scalings
for turbulence produced by wind stress and by convection. Scaling of X was less successfuland
applied only when turbulent production was dominated either by wind stress or by convection,
and then only within part of the SBL.

1.

INTRODUCTION

f52
pedz 0.01
El0,where
pisthedensity
inkgm-3 ande

As an unexpected bonus while looking for turbulent


patches in the upper thermocline, we observed eleven cycles in the diurnal growth and decay of the surface bound-

ary layer (SBL) (Figure1). By analogywith the planetary


boundarylayer (PBL) of the atmosphere,
we take the SBL
as the near-surface zone directly affected by the wind stress
and buoyancy flux applied at the surface. By profiling continuously, we sampled a wide range of turbulent regimes in
the SBL. Becauseturbulence is a major factor in the dynamics of the boundary layer, our immediate goal is to determine
how well similarity scalingdescribesprofiles of e, the rate of
viscous dissipation of turbulent kinetic energy, and X, the
rate of diffusive smoothing of turbulent temperature fluctuations. As the first step, we restricted ourselves to those
times during the night when the ocean lost buoyancy to the
atmosphere at a steady rate. Our ultimate goal, however, is
to use these unique observationsto increase our understanding of the SBL, and thereby improve our abihty to model
it.

Oceanographershavelearned many things about the SBL,


but they still lack even an approximate energy budget and
are as uncertain about the disposition of the energy as they
are about its sources. For example, Richman and Garrett

[1977]estimatedthe rate of energyenteringthe oceanfrom


the wind as (0.02- 0.1)Et0, whereEt0 is the wind work

10m abovethe surface,


in unitsof W m-2. Relatively
little of this energy goes into deepeningthe SBL. Typically,
increasesin the potential energy as the layer thickens ac-

count for only (0.001- 0.002)El0 [Denman and Miyake,


1973]. By contrast, viscositydissipates10 times as much
of the wind work. For instance, in analyzing their observa-

tionsof 20 m thick SBL, Oakeyan Elliott [1982]estimated

hasunitsofW kg- 1. In viewofthedisparity


between
theenergiesgoing into deepeningand into dissipation, oceanographers will not closethe energy budget or realistically model
the SBL until they accurately parameterize the turbulence.
Although for 20 years atmosphericscientistshave parameterized turbulence in the PBL with similarity scahng, the
difficulty of making the measurements has impeded a corresponding developmentfor the SBL. Seawater is much harder
on sensorsthan air. Also, in contrast to thicknesses of 1 km
in the atmosphere, the oceanic boundary layer usually extends only tens of meters, and it has the added comphcation

of breaking waves. Therefore, only recently did Oakey and

Elliott [1982] make the first direct comparisonsof e measurements in the SBL with Et0, and as yet only a handful

of other studieshave been done. Dillon et al. [1981] inferred the dissipation rate from the spectral shape of temperature microstructure measured close to the surface of a

lakeandfoundthate(z),,z-t, thestructure
predicted
by
wind stressscaling. Shay and Gregg[1984, 1986] reported
the first evidencefor convectivescaling,using e profilesfrom

two oceansites. Brubaker[1987]then demonstratedconvective scaling of X at shallow depths in a lake. Taken together
these studiesprove that similarity scahngapplies in the SBL
at some times and places. However, no attempt at X scaling in the ocean has been reported, nor do we know the
limitations on when and where different parts of similarity
scaling apply. Our purpose here is to examine these questions systematicallyfor the nighttime phase of the diurnal
cycle.
As necessarybackground, in section 2 we summarize the
similarity scalingof e and X in the PBL. In section3 we describe the what and where of our observations in the SBL,
followed in section 4 by comparisons of e and X with similarity scaling. We summarize and briefly discussthe implications

of our results

in section

5.

1Nowat NavalPostgraduate
School,
Monterey,
California.
Copyright 1989 by the American Geophysical Union.

2.

SIMILARITY

SCALING

APPLIED

TO THE PBL

In most cases,the PBL developsvertically, being set in


motion and controlled by the wind stress and heat flux at

Paper number 88JC04308.

0148-0227/89/88JC-04308505.00
6273

6274

LOMBARDO

AND GREGG:

SIMILARITY

SCALING

DURING

NIGHTTIME

CONVECTION

-8

0.1

0.2

. 0.3
Q- 0.4.

0.5

0.6

0.7

13

14

15

16

17

18
October

,,,

19

,,

20

, ?,

21

22

23

1986

Fig. 1. Diurnal cyclesduring the PatchesExperiment (PATCHEX). Each day the oceanlost heat and buoyancy,
starting several hours before sunset and continuing until a few hours after sunrise. These lossesare shown by
the shaded portions of the surface heat and buoyancy fluxes in the top panel. In response, the SBL slowly

deepened(lowerpanel). The solid


line marks
D, the middle of the entrainmentzone, and the lightest shading
7
1
shows10-s W kg-1 < e < 10- W kg- . The shadingincreasesby decades,so that the darkestshadeis e >
10- Wkg -1 .

the surface. Because of this control from the bottom, much

of the average and turbulent structure within the boundary layer is determined by only four variables: the heat flux

at thesurface,
jq0,in unitsofWm-2; thewindstress
at

spheric scientists distinguish within the P BL a sequenceof


three sublayers: the surface layer, the mixed layer, and the

entrainmentzone [Panofskyand Dutton, 1984]. Similarity


scaling is applied differently in each sublayer.

the surface, r, in pascals; the distance above the surface,

z, in meters; and the innate buoyancyof heated air, g/T,

The Surface Layer

in unitsof ms-2K-1. (Hereg is the gravitational


acceleration.) In the atmosphericliterature, the wind stressis

usually
statedin termsofthefrictionvelocity,
u. ----/,

andtheheatfluxisexpressed
asQ = Jq/pcp,
where
Cpis

thespecific
heatat constant
pressure,
in unitsof Jkg- 1K- 1

ThenJ = (g/T)Qisthebuoyancy
flux,inunitsofWkg-1
The use of these four variables to normalize boundary layer

measurementsis known as similarity scaling [Monin and


Obukhov,1954]. Similarityscalingworkswell whenthe PBL
is controlled vertically by horizontally uniform fluxes at the
surface.

The application
of similarityscalinghaschanged
a atmospheric scientists have learned more about the internal
structure of the PBL. Like the boundary layer in the up-

When the convectingPBL is strongly established,only its


bottom 10% is directly affected by the surface. Within this
surfacelayer, the height, z, establishesthe energy-containing
scale of the turbulence. Because only three of the four similarity variables are independent, a secondlength scale can
be formed to give the distance above the surface where the
wind stress and buoyancy are equally effective at producing turbulence. Known as the Monin-Obukhov length, this
scale is defined

as

-u.3

L---J2

(1)

where von Krm.n's constant, = 0.4, is included by tradi-

per ocean,the PBL waxesand wanesin a diurnalcycle. tion,asistheminus


sign,which
makes
L negative
whenJ2is
Over land, convectionusually beginssoon after sunrise,and
by early afternoon it pushes the PBL up to about l km.
The thickness, D , of the PBL is defined by the first inversion, where temperature reverts to its normal increase with

height. Once D growsbeyonda few hundredmeters, atmo-

positive. L is the fundamental length scale used to separate


the two asymptoticregimes-where z/- L 1, wind stress
dominatesthe production of turbulence;where z/- L 1,
buoyancy controls production. During strong convection,
-L _ 10m. Hence, in midafternoon the wind-stressregime

LOMBARDO

AND GREGG:

SIMILARITY

SCALING

DURING

NIGHTTIME

CONVECTION

6275

TABLE 1. Characteristic
SimilarityScalesfor theThree
Sublayersof thePlanetaw Bounda Layer
SurfaceLayer

Characteristic

Range
Length

Wind Stress

FreeConvection

0 <-z <<-L

-L <<-z << D

Mixed Layer

-L <<-z < D
z,D

31/2

u. j

w.

- %pu.

%pu

_%pw.

Ef=j0

g =j0

Es= --

u.T.2

ufTf2

w.0.2

Terminologydiffers slightly from atmosphericusage. The range specifiedis meant to be general.reflecting


the uncertaintyin atmosphericliterature.

is usually confined to the lower 10% of the surface layer,

=- W,

accounting
foronlythebottom
!% ofthePBL.
The scaling procedure is to nondimensionalizeboundary
layer parameters using the similarity variablqs, e.g., dimen-

sionlesstemperatureis T/T., whereTable I showsT. to be


the ratio of Q and u.. More complex parameters are constructed from simpler ones. As an example, in Table I we
see that Xs is formed from u., T., and z. When properly

nondimensionalized,
parametersare then functionsof z/L
that can be determined only by observation. For instance,

after observinge between0 < z/Cotd, 1971]fitted e/es with

L < 2.5 Wyngaardand

z2/3)3/2
--- (1+0.sll
In the free convectionregime, i.e., where z/-

L >> 1,

columnof Table 1. With onelessvariable,z/L is no longer

(s)

Typically,w, 2ms-x givingTrot 8min for D =


1000m.

As in the free convection sublayer of the surface layer, e

is approximately constant with height. However, X falls off


with distance from the surface, a result obtained indirectly
by matching the scalingsin the indistinct transition between
the free-convection sublayer and the mixed layer,

X___
_Xml
XXrnl
X (Z)
XJ'
Xl _Xml
' 4/3

(2)

u. is dropped becauseit is no longer important, leading to


the similarity parameters shown in the "Free Convection"

[4

(4)

(Here,
Xml/X.i
= (z/D)4/:3isobtained
byinserting
thesimilarity variables.)In the free-convection
sublayer,X/X.t'is
constant. Therefore,

a dimensionlesslength scale. As a result, dimensionlesspa-

X = a(3)-4/3

rametersare constantwith height[Wyngaardet al., 1971],

Xrn/

e.g.,clef-- elJi5
= const
andxlx! -- const.Theuniformity of e with height is one of the distinguishing features of
convection and carries up into the mixed layer.
The Mixed Layer

Above the surface layer, potential temperature and velocity are nearly constant, demonstrating that most of the
PBL is well mixed. Thus, it is not surprisingthat the energycontaining scalesin the mixed layer are limited only by D.
Consequently,similarity scaling in the mixed layer is like
that in the free convectionregime of the surfacelayer, but
with D in place of z, as shown in the rightmost column in

where a is a constant determined from observations. This

scaling often apphes from the free-convection sublayer to


deep within the mixed layer, and sometimeseven to the top
of the layer: temperature fluctuations carried down from the
entrainment zone can be larger than those coming up from
the surface.

The Entrainment

Zone

When the PBL is growing upward, the inversion rises by


entrMning air from aloft. The thicknessand character of the
entrainment zone vary greatly, depending on subsidenceof
Table 1. (Notwithstanding,
someparametersstill vary with air above the PBL, as well as on the surface fluxes, but active
z.) Sincethe verticalvelocities,w., dominatethe motions, entrainment typically extends over 0.8 D _< z _< 1.2 D. Althe time scale of the mixed layer is
though some scaling has been successfulin the entrainment

6276

LOMBARDO

GItEGG-

SIMILARITY

SCALING DURING NIGHTTIME

CONVECTION

differencing adjacent points. The vertical resolution, however, is somewhat larger than 10 m.

Atmospheric Forcing
The winds were weak, with speeds usually less than

10ms-x andEx0risingonlyonceto 1 W m-2. RobPinkel


had sampleda full suite of meteorologicalsensorson FLIP at
four times per second,and graciouslygave us a copy of the
data for processing.From the 16 min summariesin Figure 3,
13

14

15

16

17

18

October

19

20

21

22

23

we see that the calms of October

13-16

and October

20-22

were separated on October 17-19 by mild winds, which pro-

1986

Fig. 2. Except for brief interruptions to relocate the ship, we


operated AMP continuously, first alongside FLIP, then following
the RINO float, and finally by FLIP again. The rate of profiling was somewhat less in the second half because the MSP was
operating more frequently.

zone [Wyngaardand LeMone, 1980], it is not well coupled


to mixed layer scaling, and we do not apply it in this paper.

ducedEl0 = (0.2- 0.6)Wm-2. OnOctober


23,a passing
frontbrieflyraisedE0 to lWm -2. Eventhen,the time
scale of wind fluctuations,
El0

Ts= 10El0
/Otl[s]

(6)

was at least several hours, allowing wind-produced turbulence in the SBL to remain in equilibrium with the surface.
In addition to having a diurnal cycle, the surfaceheat flux,

j0,alsovaried
overseveral
days,
modulated
bychanges
in
the Mr-sea temperature contrast and in the winds. The sea-

3.
The

OBSERVATIONS

AND BACKGROUND

Observations

In October 1986, as our contribution to PATCHEX, we

took the R/V Thompson


to 34oN, 127oW, locatedin the
outer reaches of the California

Current

and the site of the

Mixed-LayerDynamicsExperiment(MILDEX) [Paduanet
al., 1988]. We started profilingnext to the Floating Instrument Platform (FLIP), which was anchored,then followed
a midwaterfloat (calledRINO for Richardson
number),and
finishedby FLIP again(Figure2). By the time we returned,
the RINO

float had drifted

surfacetemperaturestayednear 18.5C, fluctuatingdaily


by several tenths of a degree. The air was usually several degreescooler, but warmed both times the winds increased, particularly while the front went by. Falling air
temperatures on the 17th and 18th, combined with the
moderate winds, produced the largest nighttime heat losses,

j0= 200- 300Wm-2. Bycomparison,


onthe23rdthe

heatlosswasonly100W m-2 because


the Mr wasnearly
the same temperature as the sea. As a result of these fluc-

tuations,
thecumulative
heatloss,
H(t) = f pjqO
dr,var-

iedbetween
4-10MJm-2. Spreaduniformlyovertheupper
50m, the maximumH(t) wouldchangethe temperatureby

19 km from FLIP.

For our primary turbulence observations, we took Mmost 700 profiles with the Advanced Microstructure Profiler

(AMP), eachgoingto 3 MPa (300m). We averaged2.4 profilesper hour (Figure 2). Althoughthis rate is adequatefor

1.o

studying the diurnal cycle, it is three or four times lessthan


we could have achievedby observing only the SBL.

o.t3

Sensors on the AMP resolved the dissipation spectra of


centimeter-scale velocity and temperature fluctuations, al-

0.2

19
:t
Sea
Surface

lowing us to computee and X by integratingspectra[Shay


and Gregg,1986]. Veused standard definitions,i.e., ----

7.5'(Ou/Ox3)
2 andX ----6nT(OTt/Ox3)
2, wherex3 is vertical distance, is the kinematic viscosity, and nT is the
thermal diffusivity. Both definitions assume isotropy, accounting for the multiplier of 7.5 for e and a factor of 3 in the

X definition. (The additional factor of 2 for X is standard


for oceanographicturbulencemeasurements.)To remove
any contamination from the initial AMP wobble or from
the Thompson's wake, we avoided all data above 0.05 MPa
and some above

o ,

-12

-o

-24 I
13

0.1 MPa.

14

15

17

18
October

In addition to the AMP, we launched and recovered the

Fig. 3.

19

20

21

22

23

1986

Two calm intervals were interrupted by mild wincls and

Multi-Scale Profiler (MSP) and operatedthe Thompson's followed by a sharp front (top panel). The sea surface tem150 kHz Doppler acoustic profiler, made by RD Instruments.

perature was steady, except for small diurnal fluctuations, un-

The

til it increased
about0.5Cafterthe frontpassed(middlepanel).

MSP

data

confirmed

the

turbulence

levels measured

with AMP, Mthough they were less densely sampled. As


background for this study, the Doppler measurements provide nearly continuousshear profiles through the SBL. They
were digitized every 4m, and we calculated shear by first-

Throughout, the air remained cooler than the sea, contributing to


the positive heat flux each night. However, the average nighttime

heat lossvaried by about a factor of 3 (bottom panel), causing


the cumulative integral of the heat flux, H, to fluctuate every few
days between net waxming and cooling.

LOMBAPO

AND GREGG:

Sunset
-

I--

-2

10

CONVECTION

6277

J= (g/p)
[(a/cp)Jq
+(/5s/(1
- s)e)jeq], where t isthe

coefficient ot thermal expansion, /3 is the coefficient of


line contraction, s is the salinity in concentration units,

(00Q

isthelatent
heat,andJ isthelatentheatflux.Each
night,

15hours
18hours

SCALING DURING NIGHTTIME

Sunrise

OIDX

SIMILARITY

convection
beganI-

2i hoursbeforesunsetand continued

untilI -3 hoursaftersunrise
(Figure4), whenthesun
finally overcamethe evaporative and longwave fluxes. Dur,

12

14

16

18

20

ingthese15-18hours,Y wasnearlysteady
from(1 - 3) x
10-?Wkg-1 (Figure1) forat least13hours.Forexample,
during
thenightofthe17th,J = (1.3+0.3)x 10-? Wkg-z,

..,

22

GMT / hours

Fig.4. The timeswhenJ changed


signareplotteds circles and the fluctuationstook at least an hour (Figure 5). Conand the times of sunset and sunrise as crosses. Convection began
sequently,the buoyancytime scale[ Wgmgaard,
1973]
when the buoyancy flux went positive, several hours before sunset,

andcontinued
until J wentnegative,
I to 3 hoursaftersunrise.

+0.06C.

lence in the SBL to equilibrate.

the heat balance.

Oceanic Background

The buoyancy
flux, J, reached
its nighttimevalue

Shearwaslow acrossthe pressurerangeof the SBL (Figure 6). But, howlow mustit be to not competewith surface

within about an hour after becoming positive and had


only minor excursions until about a half hour before going negative in the morning. The positive sign resulted
from the net heat flux at night, plus a small contribution from the effective salinity flux due to evaporation, i.e.,

2L '

'

'

'

..

'

'

. : ,'.., - q:

()

exceeded 5 hours, providing ample opportunity for turbu-

Since we observed a net increase of 0.08C in

the temperature of the nighttime SBL, weak advection also


affected

s I N
To=-ioso/ot

forcing in producing turbulence? Our only recourse is to


compare the PATCHEX data with observations above the
equatorial undercurrent. During Tropic Heat at the start

'

'

.....:..:..
.......... :...::...
.....

'

I
....

'

'

'

'

'

'

'

....

-,.--"":
:"'"'"'?:
'":'"
'-"'
'i.'"-"
'""'" ::'::
.....:. :.i;i?::ii,.;::i:d!:.,..,:
..........
17Oct1986

.. 4 ..

.............

. :..`...:;:.;.!*...::..:::i.;:..::::::>.i.a.*..`.::.;..........:
:.:.x:...:.:.:.:.:.:.:.::..`.x.g:.:.m:.;:.....>:.:.<.:::.:>.::z.:*:.:..*..<..:.:.`..;:.:.:.:..:.:.:`

II

II

II -Sunset

Sunrise-
I
I

I I

::::::::::::::::::::::::::::::::::::::::::
" ..'.:!i::i!!i::::::iiiii:!ii::i::i:!i
....
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
.................

1.
0.4

0.7 ,I ,I,.,,,,
0

.:::::::
:::::
:::::::
::::
::::::::::::::::::::::::::::
:;:::::::::::
:;:
:;:
:;:;:;::
:::::::::::::::::::::::::::::
::;:::::::;::
:::::::;:::::::::..
:::::::::::::::.
.::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::

-::::
:::::::::
::.::
::::::::::::::::::::::::
::::
::::
::::
::::::
::::::
:::::
:::::::::::::::::::::::::::
::::::::::::::::::::::::::::::::::::::
.::::::::
::::::::::::::::::::::::
:::.

12

16

2o

GMT / hours

Fig. 5. The six stagesof the diurnal cycleduring a typical day (refer alsoto Table 2). We analyzedprofilesfrom

stageII, entrainment,
andstageIII, equilibrium.
Duringbothstages,
d wasrelativelysteadyand-L (shown
by
solid squares)was lessthan 10m. The shadingof e is the sameas in Figure 1.

6278

LOMBARDO AND GREGG: SIMILARITY SCALING DURING NIGHTTIME

CONVECTION

0.1

0.2

0.3

0.4

0.5

0.6

0.7

19
October

2o

21

22

23

1986

Fig. 6. Two-houraveragesof shear2, contouredby decades,with D superimposed.


No diurnal shearcycleis
evident,but shear in the pressurerangeof the diurnal SBL is lowerthan in the thermocllne.

of the nightly deepening, the near-surfacezone was weakly

that intrusions usually accompany the daytime restratifica-

stratified,withsheare 3 x 10-4 s-e ande about10times tion of diurnal boundary layers and are not simply a pecuthe similarity scaling. As the deepeningproceeded,density

washomogenized,
sheare droppedto 5 x 10-5, and e
decreasedto the similarity level (H. Peters et al., Meridional variability of turbulence through the equatorial undercurrent, submitted to Journal of Geophysical Research,

liarity of the PATCHEX site. Their characteristics,though,


depend on the local horizontal temperature and sahnity gradients. Like the intrusions found in the Bahamas, those during PATCHEX were weak and apparently did not affect the
convection.

Below the seasonal thermochne, internal waves and dissi1989). In contrast,


duringPATCHEXshear
e wasalternatelygreaterthan andlessthanI x 10-5 s-e, averaging pation were unremarkable and had the magnitudes that we

about a decade less than in the undercurrent. Furthermore,


the absenceof a diurnal shear cycle leads us to conclude that
turbulent production by shear was not important.
Typical of the location, an irregular salinity minimum

containingthermohalineintrusionslay near 0.6 MPa (60 m),


the maximum depth of the SBL (Figure 7). Somedays, we
found the intrusions in the deeper part of the SBL, contributing to the restratification. Because Shay and Gregg

believe
typifythebackground.
UsingtheMSPdata,sheare
calculated over Az = 10 m was close to the levels predicted

by Garrett and Munk [1975][Greggand Sanford,1988].The


shear appeared completely random, and we found no trace
of shear from persistent near-inertial motions. Dissipation

rateswerelowanddecreased
withdepthase ~ N e, aspredicted by interaction calculations for background internal

waves[Greggand Sanford,1988].

[1986] observedsimilar featuresin the Bahamas,we believe


4.

SBL EVOLUTION

AND SIMILARITY

SCALING

0.0

Evolution o the SBL

Although
thechanging
signof J2 forces
thediurnalcycle, the state of the boundary layer is not simply convecting

Nig

or nonconvecting.
It alsodepends
onwhether
J2 issteady

0.3

or changing and whether the thickness of the layer, D, is


growing, shrinking, or constant. In Figure 5 and Table 2,
we distinguish six stages: I-IV were convecting, and V-V!
were restratifying under a negative buoyancy flux. We restricted this study to stagesII, entrainment, and III, equilib-

0.4

0.5
:Da)

0'61.' i. ' .432.94 33.;423.73 23.760 2


O/ C

103salinity

(70/ kgm'3

102N
/ rads'

40.0 0.6 -1.2


102S
/ s'

Fig. 7. Typical day and night profiles. During the night (heavy
lines), the SBL usually descendedto about 0.SMPa, where it
encountered a large increase in stratification. During the day

(thin lines),restratification
producedN 0.003s-1 overthe
pressure range of the nighttime SBL. We found the shear to be
the same day and night. These profiles were taken October 18.

rium,because
onlythenwasJ2positive
andsteady,
making
Tb >> Tml. Stage II, however,may differ from stage III; during entrainment some of the energy released at the surface
by convectiongoesinto the growing potential energy of the
deepeninglayer, and somegoesinto internal wavesradiating
from the oscillating entrainment zone.
The entraining layer advanced slowly. For instance, on
October 17 it took 10 hours to grow 40m. As the layer

LOMBARDO AND GREGG:

SIMILARITY

TABLE 2.

Stage

Entrainment

IV. Decay

6279

Stagesof the Diumal Cycle

Duration,hours

jopositive
andincreasing

1%-2 , T=2

Jff steady
andD increasing

4-16, T=9

joandD steady

0-12,T=7

jodecreasing
andD steady

T=2

1TI. Equilibrium

jonegative
andD decreasing

V. Suppression
VI.

CONVECTION

Characteristic

I. Initiation

II.

SCALING DURING NIGHTTIME

jonegative
andSBLstratifying

Stratification

T=3

Refer to Figure5 for an illustration.

creased with the dropping air temperature, plunging the

deepened: D increasedfrom 13 to about 45 m, w. increased

from11to 20mms-!, and-L shrankfrom13to 8 m. Con- Monin-Obukhov length well below the boundary layer. This
sequently,D/-L increasedfrom lessthan 2 to over 15, and
Tml lengthened from 15 to 45 min. As a result, the pro-

one night let us observe the vertical structure of a deep,


stress-driven layer and accounted for all but one of the hours

duction of turbulence changedso slowly from wind stressto


convection that Tml was always much less than the 5 hours
estimated for Tb.
While the front passed over us on October 23, strong
winds kept the SBL in the turbulent regime throughout the

having D/-L
Figure 9.

night (Figure 8). AlthoughstageII beganwith D/-L

< 1 in the summary of turbulent regimesin

Scaling

Averaged throughout the boundary layer, dissipation


rates tend toward cs and ml for the asymptotic forcing
regimes. To compare all the data on the same criteria, we

< 1,

after a few hours the winds rose and the buoyancy flux de-

23 oct 1986
j

I-- Sunset

Sunrise -

-4

I,

'

' I'

'

'

'

'

'

0.3

....
.....

0.7

, I,

II

':-::i}i{{{i{!i?::{i/::::::.:.:.:.....
.....
.:--:::!:-:::--:-.
..............

===

'

,.

?.;iiii;i,1i,.;i{i'ii,liliiiii,iiii
................................
............................
?,:';!;!'{i',!:ii!,,!??:::
i:'i
}!'ii!iiii!111!11ii:;'i!!11iii!i:ii':!
i!!i';
{!!!i{if?'

0.6

I
I
I

i'""""

" I 8=10
-8Wkg
-1

'

, i

, , . I

, , , ! ,,'

---

I
I
I-

, , ,
12

, , ,
16

, ,
20

GMT / hours
ig. 8.
the SBL.

Ding

the ght of October 23, a reduced buoyancy fl,

rising winds, &ove -- below the be of

6280

LOMBARDO

AND GREGG:

SIMILAR/TY

SCALING DURING NIGHTTIME

-L/m
0

10

20

30

40

50

60

70

80

90

D/-L

< 1

0 -1

100

10

' ' ' ' ....i I' .......


i
i

10 -1

10 0
i
i

0.58

126
20

/Jb

/s

100
0.0

10

CONVECTION

30
0.5

60

.................

' ...............

' ......... ' ......... ' .........

Fig. 9. Hourly averagesof D/-L reveal a wide range of conditions. The llne to the right is D/-L = 1, and that to the left is
D/-L = 10. Most ratios duster near 10, indicative of strongconvection in the bulk of the SBL. A few, however,have D/-L < 1;

1.0

all of these, except one, occurred on October 23.


D/-L

plotted
11(D
- 5)f/i, dzand11(D
- 5)f/iJ2dzversusD/-L

(Figure 10). (We beganthe integrationat 5 m

to avoid contamination by the questionable data close to

the surface.) Ratios normalizedby es clusternear 1 when


D/-L < 1 but are between10 and 100 when D/-L > 10.

Conversely,
thosenormalized
by J2 approach
0.6forlarge
D/-L

and rise to 1-10 when D/-

10 3

.....

< 1

D/-L

> 10

1.5

Fig. 11. When D/-L < 1, normalizeddissipationprofilesfollow


wind-stressscalingthroughoutthe boundarylayer (left panel).

When D/-L

> 10, profilesobeyconvective


scaling(rightpanel).

Shading marks the 95% confidenceestimates calculated with the

bootstrapprocedure[Elton and Gong, 1983].

L < 1. Therefore,windstressscalingworkswell when D/- L < 1, and convection


scalingworkswell when D/- L > 10.

To examine the vertical structure during the wind-stress


x
x

10 2

xx

x o

'xX

that similarity scaling normalizes the vertical decay, with


no distinction between entrainment and equihbrium stages,

(ko
o
x

Cb'----, 10 1
I

10 0

10-1

101

.....
x

ox

x
x

i I , [

x x

xo

x 0.6 o

___

',

' .,% x

X
o(
;'.....
x ,, x x xXx
o 7o
%
xX,.o2.
'b---o
.....x............
.,_.x
o
x x o x
o x o ooox x ox-o Xo
......
o o
xo
x
o

XOoOo

o 0%
0
o

10 2

D/-L

Fig. 10. Verticalaverages


of /s and /J.

Eachpoint is

Wind-stress scal-

ing (top panel) is a good normalizationwhen D/-L

< 1, but

greatly underestimates the dissipation when convection domi-

nates, D/-L

> 10, the normalizeddissipation

rates fall off gradually with depth and scatter much lessthan
is very close to that previously found in both atmosphere

1 1

an average of all drops taken in an hour.

Turning
nowto theconvection
regime
andaveraging
/J2
for all times when D/-L

duringthe wind-stress
regime(Figure 11, right). The trend

10
-1 .... 10,Io

althoughthe scatteris large. (The shadingshowsthe 95%


confidencelimits from the bootstrap estimator [E/ron and
Gong,1983])The normalizeddissipationrate beginsdropping off near 0.SD, which is often the top of the entrainment
zone. Plottingthe wind-stress
comparison
versusz/-L (not
shown)revealsthe same uniformity with depth over the
range 0.15 < z/-L < 0.8. They average1.76, comparable
to atmosphericobservations.For example,evaluating(2)
for z/-L = 0.5 givese/es = 1.5. Therefore,when the wind
stressdominates the turbulence, the PATCHEX dissipation
rates match both the vertical decay and the magnitude predicted by similarity scaling.

Q ,---,t.r) 10 0

o.... Entraining
x .... Equilibrium

regime,we averaged/ for all timeswhenD/-L < 1 (Figure 11, left panel). The uniformitywith depth clearlyshows

and ocean(Figure12), althoughthe verticalaverageof 0.58


is slightlysmaller.Returningto Figure10 (bottom),wefind

thate/J2issignificantly
lower
during
entrainment
thanduringequihbrium.
Separately
averaging
e/J2forstageII and
stageIII revealsthat the bias is statistically significantin the

upper and lowerparts of the SBL (Figure 13). Acrossthe

> 10. Convectivescaling(bottom panel) behaves entire layer, the net differencesare large: vertical averages

in the opposite manner. Dashed and dotted lines show averages

duringentrainment(circles)and duringequihbrium(crosses).Al-

of /J2are0.66q-0.13and0.44q-0.06duringequilibrium
and entrainment,respectively.(The confidencehmits are
95%.) The sizeof the contrastsuggeststhat the variability

though data from the entrainment and equilibrium stages show


no differencesfor wind-stressscaling, the data during equilibrium
are consistently higher that those during entrainment for convec-

ofvertical
averages
of/J2 among
observations
at different

tive scaling(bottom).

times and places may result from varying ratios of equilib-

LOMBARDO

AND GPEGG: SIMILARITY

SCALING DURING

/ Jb0
iO-a

lO-t

NIGHTTIME

/ Jb0
lO o

iO-a

lO-t

..

CONVECTION

/ Jb0
lO o

::

lO-a

iO-

lO o

:;

0.4---6
/Jb=0.58
i ,2
L-/Jb=0.72

::

. /Jb
=0.64+

o. 6 -

:: '?'?:: -

. o ....................
7":"=?
....r .......................... ......'"r...............................

..=.
Fig. 12. The verge of

HEX '

PATCHEX pro

ing

when D/-L

5281

' +

................

Atmosphere '

> 10 (left) is si

o obseNa[lom dg

cold-

k oubre over w-core ring [a ad rcgg, 1986](de) d [o [mosphericmeemen[s [en in


Minneso [I{aima a., 1976]d nearAschch in he U[ed Kgom [Oaghy and Pamr, 1979](right).
However, [he veric verge fod ding PATCHEX was slx
bs, de
d gh[, in, cagegS codence

smer.

ShrUg,

on ghe left, d

horizon

rium and entrainment stages in the data ensemblesin the


averages. For instance, both stages were included by Shay

in between,i.e., when 1 < D/-L

and Gregg[1986]. Having relativelylower dissipationrates

tists make empiricalfits to e/e. We reasonedthat both

during entrainment is consistent with some of the energy


releed by convectiongoing into potential energy and some
going into internal waves.
So far, we have matched data during the two asymptotic
regimeswith similarity scaling,but have not dealt with those

wind stress and convection are significant sources in this

10-1

10 0

midrange.
Consequently,
theirsum,1.76e-F0.58
J, should
be a better normalization,as indeedit is (Figure 14, left).
Since it fits the intermediate data so well, we applied it to
all our observations,with results that are almost as good

101
,

.0

,11

< 10. As noted in section

2, when working in part of this range, atmospheric scien-

10
s_,
-1
/(0.58Jlb
+1'76Ss)
10 10-1
s_,/(0.58Jlb
o+1.76Ss)
10
,

i , i

''''l
I

,-4- 0.84

Entraining I

r'
J- 0.87

Equilibrium
_

0.5

0.5

'""ii
:]

1.0

.-"i!i
........

D/-L> 10
D/-L>O
1.5

Fig. 14. Scaling with 1.76s+0.58/J worksverywellwhen


Fig. 13. WhenD/-L > 10, /J dnringequilibrium
is larger 1 < D/-L < 10. Using this snm for all profLies(D/-L > O)
than it is during entrainment. The vertical average during equiworksaswenasusing0.58/J aloneworkswhenD/-L > 10,
librim

is 0.65, compared to 0.44 during entrainment.

as seen by comparing with Figure 11.

6282

LOMBARDO AND GREGG: SIMILARITY

SCALING DURING NIGHTTIME

10 5

twice as large as our definition, we multiplied his value for


a by 2 to put it on the same basis as ours. Similarly, atmospheric scientists usually define X without the factor of

10 4
N
o

103

o
o

'----,u')10 2 -

101
10 0

x
x

x
o

o o

o x
o
x

o ooXoo
xo

Xx

%xxo xi
- xx ox oOo%
oxxo
xffo x
x x

o x o

--

.....

10

'

'

oXx x

I
'

o
o (

Xo

''1

o.... Entrainlug
x .... Equilibrium

6 _

8o

- x

-4/3

o (1

x %ooto o

2 -

-----F--?.-?-
<-.--.
- __x
xo
-o
'%
x
....
xx......
o---o----o o
'x x
-.... x -xX -

-2 --

o o

-6

( 0'o-'-- ' .... x....

oo x x

-4-

)"'"'

ox

o
i

i i

'1')o

scaling regime for X extended as deep into the SBL as found


previouslyin the atmosphereand in a lake. We, however,
did not approach as close to the surface as was done in the
lake and atmosphere.

00

xx x
ox x xx

2 we used. So we alsomultiphedthe reportedatmospheric


valuefor a by 2.) Overall, in termsof z/D, the convective

8 -

10-

o o

CONVECTION

x.... <.x
x

10

x-

x-

1a

D/-L

5. SUMMARY AND DISCUSSION

While profilingfor elevendays thrOUghthe surfacebound-

ary layer (SBL) of the ocean,we encountered


a wide range
of regimesproducingturbulence. In this paper, we focused

on thosetimes when the oceanlost buoyancyat a steady


rate to the atmosphere and asked how well and X in the
SBL were describedby similarity caling. The short answer
is: very well for and somewhat for X, provided that in
both caseswe considerturbulent production by wind stress

as well as by convection. Becausethe range of D/-L

is

much more restricted in the ocean than in the atmosphere,


in most situations both mechanismsof producingturbulence
must be consideredto obtain a complete description of the

Fig. 15. Vertical averagesof X/X scatter much more than similar comparisonsfor dissipation(comparewith Figure 10, top).
In addition, they fail to convergeat low D/-L during both entrainmentand equilibriumstages.Slopesof log(x/x.) converge full boundarylayer. (D is the depthof the SBL, and L is the
to -4/3 when D/-L > 10, but not at smallerD/-L. Profiles Monin-Obukhov
length.) Previouscomparisons
with simitaken during equilibriuxn approach -4/3 more closely than do larity scalingin the SBL dealt with only one of the asympthose taken during entrainment.

totic regimes.

More specifically,
the followingcasesare cnsidered:
1. When D/-L < 1, wind-stressscalingremovedthe

ofstress-produced
dissipation,
at leastfordepths
(Figure 14, right). Therefore,the sum of the two asymp- z- decay
totic regimes offers a good scaling for conditions between
the two regimesand an adequatescalingfor all PATCHEX
observationsdeeper than 5 m.

greater than 5 m, the shallowestwe considered. In this comparison,we find no differencebetweenentrainment and equi-

Scaling

Iog[z / Zs]

Wind-stress scaling is less successfulwith X. Vertical av-

5) D X/X &, t,l to convg

low

-1
0.0

(Figure 15, top). Nevertheless,averagingall profileswhen


D/-L < 1, we find that X/Xs is uniformnear 1 in the upper half of the boundary layer, in agreementwith similarity

scaling(Figure 16). The vertical averagesfail to converge,


owing to the large increasesnear the bottom of the bound-

arylayer.Although
mostoftheprofiles
weretaken during
equilibrium,we presumethat the deeprisein X/Xs waspro-

0.5

duced by temperature structure from the entrainment zone


and will investigate this further as part of a study of that
zone.

Scaling for the convectionregime also works in only part


of the boundary layer. Computing the average slope of

log(x/X) betweenz/D:

0.1 and z/D = 0.7, we find a

rough convergence
toward -4/3 when D/-L > 10 (Figure 15, bottom). AveragingX/Xrnt for all profileswhen
D/- L > 10 showsthat the expected-4/3 slopefits the
data well for 0.3 < z/D < 0.7 (Figure 17), endingnear the

1.0

D/-L<

depth where the entrainment zone is first encountered. By

comparison,Brubaker[1987]found the -4/3 slopebetween


1.5
0.04 < z/D < 0.7 in a 5 m thick SBL in a lake, and Guillemet
et al. [1983]reportedit between0.03 < z/D < 0.4 in the
Fig. 16. The averageof X/Xs for all profileswith D/-L < 1
atmosphere(both are includedin Figure 17). The scaling demonstrates
that wind-stress scaling works only in the upper

factor is a 4, compared to 2 in the lake and 1.1 in the

half of the SBL. The shadingmarks the 95% confidencelimits de-

atmosphere. (BecauseBrubaker[1987]definedXml to be

terminedwith the bootstrapprocedureElton and Gong, 1983].

LOMBARDO AND GREGG:

SIMILARITY

SCALING DURING NIGHTTIME

10-2

CONVECTION

6283

104

10-1

10-2

'

10-1

PATCHE (Z
/'D)-4/3

Lake

r 10'1

""
2(z/D)
'4/3

10-2

100

102

104 10-2

10

17,,
/ Y,,ml
Fig. 17. During PATCHEX whenD/-L

102

104

17,,
/ Y,,ml

> 10, X/Xrat followedthe expected-4/3 slopeonly for 0.3 < z/D < 0.7.

Observations
froma lake [Brubaker,1987]and the atmosphere
[Guillgrnetet al., 1983]had moreextensive-4/3
ranges. As explained in the text, we multiplied the reported amplitude coefficients,a, for the lake and atmosphere
by a factor of 2 to put them on the same basis as ours.

librium
stages.
Thevertical
average,
1/(D-5) e/es
dz=
1.76, is consistent with observations in the atmosphere

[Wyngaardand Cord, 1971]. We believethat the 1.76 multiplier representsthe additional contribution from the production of turbulence by convection.

2. When D/-

L > 10, mixed layer scalingnormalizes

, leaving only the gradual decreasewith depth found be-

fore.Thevertical
average
is1/(D- 5)ff 6/J)dz
= 0.58,
comparedto 0.72 [ShagIand Gre##, 1986] and 0.(34in the
atmosphere. However, during entrainment the average is
0.44 q-0.06 compared to 0.65 q-0.13 during equilibrium, sug-

Initially, the general successof similarity scaling in the


SBL surprisedus. During the two periods of elevated winds,
we observedthe elongatedsurfacestreaks marking the convergencelines of Lungmuir cells. Although we could not
relate them to individual AMP profiles, the streaks were
frequent. Viewing the Lungmuir cells as sourcesof turbulenceadditional to thoseconsideredby similarity staling, we
expected that would greatly exceedthe similarity levels in
much of the data. Although additional turbulence from the
Lastgmuircells may contribute to the scatter, evidently the

Lungmuircirculationis part of the large-scaleenergyof the

gesting
tousthatsome
oftheVariability
in /J results
from SBL, just as large-scalecoherentstructuresare an integral
differing ratios of entraining and equilibrium stages in the
data.

3. When 1 < D/-L

< 10, both wind stressand con-

vection are important and neither of the asymptotic scalings for fits the observations. However, normalizing by

componentof the PBL. Although we still expect to find some


differencesin dissipationrates in Lungmuircellsversuselsewhere, the net effect does not seem to invalidate similarity
scaling for .

Weareintrigued
byfindingthat /J is lessduringen1.76s+ 0.58J, thesumof thetwoasymptotic
scalings,trainment than during equilibrium. In a separate study,
fits the data very well when 1 < D/-L < 10 and fairly well K. Purvis(personal
communication,
1988)takes1- /Jo

for D/-L

> 1.

as the energy available for entrainment at the base of the


SBL. The PATCHEX result indicates that the energy balonly in the upper half of the SBL. Departures from the scalanceshiftsbetweenentrainmentand equilibriumstages,pering were equally frequent during entrainment and equilibhaps reflectinggreater energylossesduring entrainment to
rium stages.
internal wavesand the potential energyof the stratification.
5. When D/- L > 10, mixed layer scalingfor X showed If verified, this will need to be incorporated into models of

4. When D/-L

< 1, wind stre'Ss


scalingnormalizedX

theexpected
(z/D)-4/3 slopefor0.3 < z/D <'0.7. The

the SBL.

scaling constant, a -. 4, is twice that found by Brubaker

[1987]in a lake and 4 times that reportedfrom the atmosphere.

Acknowledgments. The Office of Naval Research supported


this work

under

contract

N00014-84-C-0111.

The

success of the

6284

LOBO

AND GREGG: SIMILARITY SCALING DURING NIGHTTIME CONVECTION

observatiohs was the result of careful work at sea and aahore by

the AMP crew: Jack Miller, Wayne Nodland, Steve Bayer, Dale
Hirt, Donna Sorensen,Hartrout Peters, joel Wesson,Pat McKe-

own,and GordyWelsh. The officersand crewof the R/V Thompson assistedthe observationsby patiently enduring the tedium of

a long sequenceof AMP stations. Dave Winkel, Keith Brainerd, Hartrout Peters, and Harvey Seim gaveus usefulcomments.
Contribution 1792 from the University of Washington School of
Oceanography.
REFERENCES

Brubaker, J. M., Similarity structurein the convectiveboundary


layer of a lake, Nature, 330, 742-745, 1987.

Oakey, N. S., and J. A. Elliott, Dissipation within the surface


mixed layer, J. Phys. Oceanogr., 12, 171-185, 1982.
Paduan, J. D., R. A. de Szoeke,and J. A. Richman, Balances of
heat and momentum at 33.5 N, 127 W in the upper ocean
during the Mixed-Layer Dynamics Experiment, J. Geophys.
Res., 93, 8147-8160, 1988.

Panofsky, H. A., and J. A. Dutton, Atmospheric Turbulence:


Models and Methods for Engineering Applications, 400 pp.
Wiley-Interscience, New York, 1984.
Richman, J., and C. Garrett, The transfer of energy and momentum by the wind to the surface mixed layer, J. Phys.
Oceanogr., 7, 876-881, 1977.
Shay, T. J., and M. C. Gregg, Turbulence in an oceanicconvecting
mixed layer, Nature, 31 O, 282-285, 1984.
Shay, T. J., and M. C. Gregg, Convectivelydriven turbulent mix-

Caughey,S. J., and S. G. Palmer, Someaspectsof turbulence


ing in the upper ocean, J. Phys. Oceanogr.,16, 1777-1798,
structure through the depth of the convectiveboundarylayer,
1986.
Q. J. R. MeteoroL Soc., 105, 811-827, 1979.
Wyngaard, J. C., On surfacelayer turbulence,in Workshopon

Denman, K. L., and M. Miyake, Upper layer modification at


Ocean Station Papa: Observationsand simulations, J. Phys.
Oceanogr., 3, 185-196, 1973.

Dillon, T. M., J. G. Richman, C. G. Hansen,and M.D. Pearson,


Near-surface turbulence measurements in a lake, Nature, 290,
390-392, 1981.

Efron,B., and (. Gong,A leisurelylook at the bootstrap,the


jadmife, and cross-validation, Am. Statist., 37, 36-48, 1983.
Garrett, C. J. R., and W. H. Munk, Space-timescalesof internal
waves:A progressreport, J. Geophys.Res., 80, 291-297, 1975.

Gregg, M. C., ahd T. B. Sanford,The dependenceof turbulent


dissipationon stratificationin a diffusivelystablethermocline,
J. Geophys. Res., 93, 12,381-12,392, 1988.
Guillemet, J. C., H. Isaka, and P. Mascart, Molecular dissipation
of turbulent fluctuations in the convective boundary layer, 1,

Height variationsof dissipationrates, Boundary Layer MeteoroL, 27, 141-162, 1983.

Kaimal, J. C., J. C. Wyngaard, D. A. Haugen, O. R. Cot,


Y. Izumi, S. J. Caughey,and C. J. Readings,Turbulencestructure in the convective boundary layer, J. Atmos.
2152-2169,

Micrometeorology, edited by D. A. Haugen, pp. 101-148,


American Meteorological Society, Boston, Mass., 1973.
Wyngaard, J. C., and O. R. Cot, The budgetsof turbulent kinetic energyand temperature variancein the atmosphericsurface layer, J. Atmos. $ci., 28, 190-201, 1971.

Wyngaard, J. C., and M. A. LeMone, Behaviorof the refractive


index structure parameter in the entraining convectiveboundary layer, J. Atmos. $ci., 37, 1573-1585, 1980.
Wyngaard, J. C., O. R. Cot, and Y. lzumi, Local free convection, similarity, and the budgets of shear stressand heat flux,
J. Atmos. Sci., 28, 1171-1182, 1971.

M. C. Gregg, Applied Physics Laboratory and School of


Oceanography,College of Ocean and Fishery Sciences,University of Washington, Seattle, WA 98105.
C. P. Lombardo, Department of Oceanography,Naval Postgraduate School,Monterey, CA 93943-5010.

$ci., 33,

1976.

Monin, A. S., and A.M. Obukhov, Basiclaws of turbulent mixing


in the groundlayer of the atmosphere,Tr. Geofiz. Inst. Akad.

Nauk $SSR, 2J (151), 163-187, 1954.

(ReceivedOctober4, 1988;
acceptedNovember7, 1988)

You might also like