The Labrador Sea Deep Convection Experiment

Abstract

In the autumn of 1996 the field component of an experiment designed to observe water mass transformation began in the Labrador Sea. Intense observations of ocean convection were taken in the following two winters. The purpose of the experiment was, by a combination of meteorological and oceanographic field observations, laboratory studies, theory, and modeling, to improve understanding of the convective process in the ocean and its representation in models. The dataset that has been gathered far exceeds previous efforts to observe the convective process anywhere in the ocean, both in its scope and range of techniques deployed. Combined with a comprehensive set of meteorological and air-sea flux measurements, it is giving unprecedented insights into the dynamics and thermodynamics of a closely coupled, semienclosed system known to have direct influence on the processes that control global climate.

Full text

2033
Bulletin of the American Meteorological Society
*The Lab Sea Group:
J. Marshall, Massachusetts Institute of Technology, Cambridge,
Massachusetts.
F. Dobson, Bedford Institute of Oceanography, Dartmouth, Nova
Scotia, Canada.
K. Moore, University of Toronto, Toronto, Canada.
P. Rhines, University of Washington, Seattle, Washington.
M. Visbeck, Lamont-Doherty Earth Observatory, Palisades, New
York.
E. d’Asaro, Department of Meteorology, University of Washing-
ton, Seattle, Washington.
K. Bumke, Department of Meteorology, Institut für Meereskunde,
University of Kiel, Kiel, Germany.
S. Chang, Naval Research Laboratory, Monterey, California.
R. Davis, Scripps Institution of Oceanography, San Diego, Cali-
fornia.
K. Fischer, Environmental Research Institute of Michigan, Ann
Arbor, Michigan.
R. Garwood, Naval Postgraduate School, Monterey, California.
P. Guest, Naval Postgraduate School, Monterey, California.
R. Harcourt, Naval Postgraduate School, Monterey, California.
C. Herbaut, Massachusetts Institute of Technology, Cambridge,
Massachusetts.
T. Holt, Naval Research Laboratory, Monterey, California.
J. Lazier, Bedford Institute of Oceanography, Dartmouth, Nova
Scotia, Canada.
S. Legg, Woods Hole Oceanographic Institution, Woods Hole,
Massachusetts.
J. McWilliams, University of California, Los Angeles, Los An-
geles, California.
R. Pickart, Woods Hole Oceanographic Institution, Woods Hole,
Massachusetts.
M. Prater, University of Rhode Island, Kingston, Rhode Island.
I. Renfrew, University of Toronto, Toronto, Canada.
F. Schott, Department of Meteorology, Institut für Meereskunde,
University of Kiel, Kiel, Germany.
U. Send, Department of Meteorology, Institut für Meereskunde,
University of Kiel, Kiel, Germany.
W. Smethie, Lamont-Doherty Earth Observatory, Palisades, New
York.
Corresponding author address: Dr. John Marshall, Bldg. 54-1256,
Department of Earth, Atmosphere, and Planetary Studies, Mas-
sachusetts Institute of Technology, Cambridge, MA 02139.
E-mail: marshall@gulf.mit.edu
In final form 24 July 1998.
©1998 American Meteorological Society
The Labrador Sea Deep
Convection Experiment
The Lab Sea Group*
ABSTRACT
In the autumn of 1996 the field component of an experiment designed to observe water mass transformation began
in the Labrador Sea. Intense observations of ocean convection were taken in the following two winters. The purpose of
the experiment was, by a combination of meteorological and oceanographic field observations, laboratory studies, theory,
and modeling, to improve understanding of the convective process in the ocean and its representation in models. The
dataset that has been gathered far exceeds previous efforts to observe the convective process anywhere in the ocean,
both in its scope and range of techniques deployed. Combined with a comprehensive set of meteorological and air–sea
flux measurements, it is giving unprecedented insights into the dynamics and thermodynamics of a closely coupled,
semienclosed system known to have direct influence on the processes that control global climate.
1. Introduction
a. Meteorology and oceanography of the Labrador
Sea
The northwest corner of the Atlantic Ocean (the
Labrador Sea sketched in Fig. 1) is a region of power-
ful physical forces, extremes of wind and cold, incur-
sions of icebergs and sea ice, great contrasts in buoy-
ancy of air and seawater, and a region of great biological
activity. Intense air–sea interaction occurs here with
strong upward heat flux at the sea surface. The proxim-
ity of the region to the principal North Atlantic storm
track of the atmosphere results in a strong modulation
of air–sea interaction by passing extratropical cyclones.

2034
Vol. 79, No. 10, October 1998
The response of the Labrador Sea involves a fun-
damental fluid dynamical process: buoyancy-driven
convection on a rapidly rotating planet. Heat loss from
the ocean is induced by cyclonic atmospheric circula-
tion over the North Atlantic in winter, which advects
cold, dry arctic air over the relatively warm (~2°C)
waters of the Labrador Sea. Peak heat losses in winter
can reach many hundreds of watts per square meter
(G. Moore et al. 1998, manuscript submitted to J. Cli-
mate, hereafter MAH; I. Renfrew and G. Moore 1998,
manuscript submitted to Mon. Wea. Rev.) and the re-
sultant buoyancy loss causes the surface waters of the
ocean to sink. But because the fluid is
stiffened by the earth’s rotation, sinking
of the cooled water compresses “Taylor
columns,” generating strong horizontal
circulation. The heat lost from the ocean
is taken up by the atmosphere, which also
responds in a convective manner. But be-
cause the timescale of response in the at-
mosphere is so much shorter, here rotation
is not an important constraint on the mo-
tion. As a result, the convection that occurs
over the Labrador Sea has a very differ-
ent manifestation from that which occurs
in it, often being organized in a quasi-lin-
ear manner that results in roll clouds that
are a ubiquitous feature in satellite images
of the region (e.g., see Fig. 2). However,
atmospheric convection is tied to the to-
pography of the basin and thus affects the
pattern of surface buoyancy fluxes; it is
very much part of the coupled problem.
The important convection, climate,
and circulation of the Labrador Sea, to
be described further below, encouraged
us to follow the historic lead of earlier
Canadian initiatives and develop a mul-
ticomponent program of observations
and modeling. Prompted by simulations
of rotating convection on the computer
(e.g., Jones and Marshall 1993) and in
the laboratory (e.g., Maxworthy and
Narimousa 1994) and by the establish-
ment of a National Oceanic and Atmo-
spheric Association (NOAA)-funded
time series mooring in the central Labra-
dor Sea, the U.S. Office of Naval Re-
search formed the Accelerated Research
Initiative on Oceanic Deep Convection.
The deep convection experiment,
whose field program began in the autumn of 1996, has
as its primary focus the oceanic convective process and
its interaction with geostrophic and basin-scale eddies
and circulation. But its proximate goals have grown
to be major efforts in themselves: the investigation of
the atmospheric, synoptic, and mesoscale dynamics
that result in intense air–sea interaction in the region;
the coupled dynamics of the deep convection process
in the atmosphere and ocean; the communication of
newly convected waters of the Labrador Sea with the
World Ocean; and the relation between convection and
decadal climate variability.
FIG. 1. Schematic showing the cyclonic circulation and preconditioning of the
Labrador Sea. The typical depth of the
σ
= 27.6 isopycnal in the early winter is con-
toured in meters. The warm circulation branches of the North Atlantic Current and
Irminger sea water (ISW) and the near-surface, cold, and fresh East/West
Greenland and Labrador Currents are also indicated. (From Marshall and Shott 1998.)

2035
Bulletin of the American Meteorological Society
The experiment took place,
quite fortuitously, in the larger
context of the Frontal and Atlan-
tic Storm-track Experiment
(FASTEX) and the validation
program for the National Aero-
nautics and Space Administra-
tion (NASA) scatterometer on
the Advanced Earth Observing
Satellite. The FASTEX goals
were to investigate, with a new
range of forecast models, the de-
velopment and evolution of low-
pressure systems over the North
Atlantic Ocean (see Joly et al.
1997). We also benefited by the
large-scale oceanographic de-
scription provided by interna-
tional efforts organized by the
World Ocean Circulation Ex-
periment (WOCE) in the Atlan-
tic Ocean. The context provided
by these related experiments
will provide a much clearer and
more complete picture of the
synoptic meteorology and ambi-
ent oceanographic conditions
that occurred in the Labrador
Sea and its environs.
Advanced and newly con-
ceived technologies abound in
the Labrador Sea experiment;
beyond classic hydrographic
sections and moorings measuring velocity, salinity,
and temperature, we deployed drifting and profiling
floats; three-dimensional nearly Lagrangian drifters
that can follow the convective process vertically as
well as horizontally; acoustic tomography and verti-
cal echo sounding aimed at long-baseline temperature,
salinity, and currents; newly designed conductivity–
temperature–depth profiler (CTD) moorings, and
moored and lowered acoustic Doppler current
profilers; shipboard air–sea flux instrumentation; wave
radar systems; airborne and satellite passive micro-
wave and scatterometer systems; and synthetic aper-
ture radars. In the second season of field work,
1997–98, autonomous underwater vehicles were also
deployed to map fine structure in the boundary layer.
The datasets that have been and are being gathered far
exceed previous efforts to observe the convective pro-
cess anywhere in the ocean, both in the scope and range
of techniques deployed. We are now in a position to
test new theoretical ideas, the fidelity of ocean gen-
eral circulation models, parameterizations of convec-
tive mixing, and to explore new and exciting scientific
territory.
Here we give an overview of the important scien-
tific issues that are being addressed by the experiment
and provide a preliminary description of results from
the field work. The paper is set out as follows. After a
statement of the major aims of the experiment, we dis-
cuss the circulation and climatological context in
which it is being carried out and some of the theoreti-
cal and modeling issues that motivate it in section 2.
The planning of the multifaceted experiment is dis-
cussed in section 3 and some of our preliminary find-
ings in section 4. Finally, conclusions and the future
outlook are presented in section 5.
FIG. 2. Infrared advanced very high resolution radiometer image from the NOAA-14 po-
lar orbiter at 1141 UTC on 7 February 1997 showing an extratropical cyclone over the North
Atlantic. The location of the low pressure center is indicated by the L. The bright and there-
fore high clouds to the north of the low pressure center are associated with the system’s
warm sector. The less bright and therefore shallow clouds to the east of the low, organized
into quasi-two-dimensional bands, are associated with the northwesterly flow that results in
the advection of cold and dry polar air over the Labrador Sea.

2036
Vol. 79, No. 10, October 1998
b. Aims of the experiment
Many of the details of the water mass transforma-
tion process in the ocean remain largely unknown be-
cause they are difficult to observe and model. The
overarching goal of the Labrador Sea Convection Ex-
periment (LSCE) is then to improve our understand-
ing of the convective process in the ocean, and hence
the fidelity of its parametric representation in large-
scale ocean models, through a combination of meteo-
rological and oceanographic field observations,
laboratory studies, theory, and modeling.
The water mass transformation process in the
ocean is inherently complicated, involving air–sea
interaction and the interplay of a hierarchy of oceanic
scales: convective plumes (on scales of order 1 km)
that act to homogenize properties to form a “mixed
patch,” eddies that orchestrate “lateral exchange” be-
tween the mixed patch and the ambient fluid through
advective processes (on a scale of a few tens of kilo-
meters), and the large-scale circulation itself (over
hundreds of kilometers) involving the ocean gyre and
boundary currents. The scales of the key phenomenon
are represented schematically in Fig. 3.
Our modeling and observational strategies were de-
signed to address each of the scales and its interaction
in the context of the prevailing meteorological forc-
ing that drives the whole process in the depths of win-
ter. The key oceanographic and meteorological
objectives are outlined below.
1) OCEAN
The objectives at each of the oceanographic scales
are as follows.
•Plume scale (100 m–1 km): Determine the char-
acteristic scales, properties, and integral fluxes of
a population of convective plumes and how they
depend on the atmospheric forcing and their local
environment.
•Eddy scale (5 km–100 km): Understand how the
convective process is related to and organized by
its large-scale environment and the relative impor-
tance of balanced (geostrophic) versus unbalanced
(nonhydrostatic) processes in the flux of heat and
salt both laterally and vertically.
•Gyre scale (50–1000 km): Determine the large-
scale factors that control the volume and tempera-
ture/salinity (T/S) properties of the convectively
created water masses and how they are subse-
quently accommodated into the general circulation
of the ocean; describe the mean and seasonal varia-
tion in the circulation of the Labrador sea.
2) ATMOSPHERE
The aims of the meteorological component of the
experiment are to
•understand the physics of the atmospheric pro-
cesses in the Labrador Sea that force oceanic mix-
ing and deep convection;
•collect a set of high quality in situ surface fluxes
of heat, fresh water, radiation, and momentum in
conditions representative of those in which deep
convection occurs;
•use the in situ measurements to “test” remotely
sensed products; and
•use the in situ measurements to assess the ability
of atmospheric numerical models to correctly rep-
resent the air–sea interaction that occurs in the re-
gion and, where needed, improve the boundary
layer parameterization in these models so as to bet-
ter represent the interaction.
2. Background
a. The Labrador Sea
The weak density stratification of the Labrador Sea
is broken down each wintertime, recently to depths
greater than 2000 m, making it one of the most extreme
ocean convection sites in the World Ocean. A lens-
shaped water mass (Labrador sea water, or LSW) of
FIG. 3. Scales of phenomena involved in deep convection: the
mixed patch on the preconditioned scale created by convective
plumes and geostrophic eddies that orchestrate the exchange of
fluid and properties between the mixed patch and the stratified
fluid associated with the peripheral boundary current.

2037
Bulletin of the American Meteorological Society
dimension 500 km × 700 km × 2 km deep has devel-
oped in response to this wintertime air–sea heat flux
(Fig. 4). It is weakly stratified, with a temperature near
2.8°C, a salinity of 34.83 ppt, and a potential density
of 27.78 kg m−3. Deep convective cooling to the atmo-
sphere competes with the buoyant, low-salinity near-
surface waters nearby and with the warmth of the
subtropical waters just beneath the surface. The net
effect on the oceanic general circulation is to transport
salt and heat poleward in the surface layers and low-
salinity, cooled waters southward to the rest of the
World Ocean at depths between 1 and 2 km, produc-
ing fresh new deep water on a quasi-continuous ba-
sis, with all the climate implications of such a
production. The Labrador Sea is also an important
component of the “thermohaline circulation,” the glo-
bal meridional-overturning circulation that is respon-
sible for roughly half of the poleward heat transport
demanded by the atmosphere–ocean system.
Figure 5 shows the average mean sea level pres-
sure, 10-m wind, and total heat flux fields for all win-
ter months during the period from 1968 to 1997, as
determined by the National Centers for Environmen-
tal Prediction–National Center for Atmospheric Re-
search (NCEP–NCAR) Reanalysis Project (Kalnay
et al. 1996). Note in this context a winter is defined
as the months of December, January, February, and
March. One can see that in winter the North Atlantic
is under the influence of the Icelandic low and the
Azores high. Thus one would expect cyclonic flow
over the North Atlantic associated with the movement
of synoptic-scale weather systems along a track from
the eastern seaboard of North America to Iceland. This
“mean” cyclonic circulation over
the North Atlantic results in the ad-
vection of cold and dry arctic air over
the relatively warm waters (~2°C)
of the Labrador Sea, resulting in a
large transfer of heat from the ocean
to the atmosphere as shown in Fig. 5.
In the center of the Labrador Sea, the
average winter heat loss exceeds
300 W m−2, a value that is of the same
order of magnitude as that which oc-
curs in the temperate Sargasso Sea to
the east of the Gulf Stream. Although
averaging tends to blur spatial gra-
dients, the highest heat loss in the
Labrador Sea occurs in an elliptical
region some 150 km wide situated
along the Labrador coast just off the
sea–ice edge where peak values can exceed
1000 W m−2.
The surface waters of the Labrador Sea are suffi-
ciently warm, except near the sea–ice margin, that
their contraction under cooling can cause convective
overturning. The delicate balance of the cold and fresh
water from continental runoff and sea–ice melt, and
the inflow of the warm and salty water of the Irminger
Current (see Fig. 1) maintains the temperature and
salinity of the surface waters at elevated values.
In addition to the large uncertainty that exists with
regard to the spatial and temporal variability in the air–
sea flux in the region, the equally important freshwa-
ter cycle has received much less attention. The
precipitation-induced supply of buoyancy to the sur-
face waters of the Labrador Sea can have a direct im-
FIG. 4. Autumn hydrographic section of potential temperature
(October 1996) along AR7 (marked in Fig. 1) showing the lens-
shaped bolus of Labrador Sea water extending down to about
2 km, formed by convection in previous winters (courtesy of
A. Clarke and J. Lazier).
FIG. 5. Average mean sea level pressure (contour, mb), 10-m wind (vector, m s−1),
and total heat flux (color scale, W m−2) fields from the NCEP–NCAR Reanalysis
over all winter months (December, January, February, March) during the period
1968–97.

2038
Vol. 79, No. 10, October 1998
pact on the convective process in the ocean (MAH).
Supply through river runoff, freshwater release in ice
melt as well as advection by ocean currents are all
important in the buoyancy budget. Direct measure-
ments show significant (> 1 Sverdrup) inflow from the
Arctic through the Davis Strait, as well as input via the
East Greenland Current. The freshwater runoff from
Canada is also very great, evident for example in the
high tritium concentration of surface waters (tritium that
is locked up in continental ground water and ice is rela-
tively undiluted, compared with ocean-borne tritium).
THE CONVECTIVE PROCESS
Observations suggest that there are certain recur-
ring features and conditions that predispose a region
to deep-reaching convection and that are common to
all known sites of deep convection—the Mediterra-
nean, Greenland, and Labrador seas (e.g., see Marshall
and Schott 1998). First, there must be strong atmo-
spheric forcing due to thermal and/or haline surface
fluxes. Thus open ocean regions adjacent to bound-
aries are favored, where cold and dry winds from land
or ice surfaces blow over water inducing large sensible
and latent heat and moisture fluxes. Second, the
stratification beneath the surface mixed layer must be
weak, made weak perhaps by previous convection.
And third, the weakly stratified underlying waters
must be brought up toward the surface so that they can
be readily and directly exposed to intense surface forc-
ing. This latter condition is favored by cyclonic cir-
culation associated with density surfaces that “dome
up” to the surface. All these conditions are readily sat-
isfied in the Labrador Sea (see Fig. 1).
Since the classic MEDOC experiment in the Medi-
terranean (MEDOC Group 1969) three phases of
ocean convection have been identified (sketched sche-
matically in Fig. 6) and provide a useful context
to consider the convective process in the Labrador Sea
(see Clarke and Gascard 1983): “preconditioning”
on the large scale (of order 100 km), “deep convec-
tion” occurring in localized, intense plumes (on scales
of order 1 km), and lateral exchange between the
convection site and its surroundings. The last two
phases are not necessarily sequential and often occur
concurrently.
During preconditioning (Fig. 6, panel I), the gyre-
scale cyclonic circulation and buoyancy forcing typi-
cal of the convection site predispose it to overturn.
Subsequent cooling events may then initiate deep con-
vection in which a substantial part of the fluid column
may overturn in numerous plumes (Fig. 6, panel II)
that distribute the dense surface water in the vertical.
The plumes are thought to have a horizontal scale of
the order of their lateral scale, ~1 km, with vertical ve-
locities of up to 10 cm s−1. They mix properties over
the preconditioned site, forming a homogeneous deep
mixed patch ranging in scale from several tens to per-
haps many hundreds of kilometers in diameter. With
the cessation of strong forcing, the predominantly ver-
tical heat transfer due to convection gives way to hori-
zontal transfer associated with eddying on geostrophic
scales. The mixed fluid disperses under the influence
of gravity and rotation, spreading out at its neutrally
buoyant level, leading, on a timescale of weeks, to the
disintegration of the mixed patch and reoccupation of
the convection site by the stratified fluid of the periph-
ery (Fig. 6, panel III).
The above conceptual idealization provides a use-
ful ordering of our ideas when thinking about the
convective process in the Labrador Sea, albeit modi-
fied by geographical detail and particularly the prox-
imity of boundaries and boundary currents (see Fig. 3),
which can provide an effective conduit for convected
fluid away from its formation region.
b. The climatic context
1) THE NORTH ATLANTIC OSCILLATION
The northwest Atlantic is an important center of
action for global climate, in part because of the huge
upward heat flux at the sea surface in winter and in part
due to the sympathetic arrangement of orography both
locally (the Greenland Plateau) and globally (in par-
ticular, the Rocky Mountains). The North Atlantic Os-
cillation (NAO) measures the strength of the cyclonic
circulation and climate variability over the region (van
Loon and Rogers 1978; Rogers 1990; Hurrell 1995).
The positive phase of the NAO occurs when the Ice-
landic low is anomalously deep and the Azores high
is anomalously shallow. When the NAO is high there
is greater cyclonic activity and hence a stronger mean
cyclonic flow over the North Atlantic with an enhanced
circulation of cold air out of the Canadian Arctic. The
opposite occurs in the negative phase. Generally, then,
one might expect higher oceanic heat loss from the
Labrador Sea during the positive phase of the oscilla-
tion and lower heat loss during the negative phase.
2) VARIABILITY IN DEEP CONVECTION
Orchestration of Labrador Sea and Greenland Sea
deep convection by the NAO is described by Dickson
et al. (1996). The ocean has an immediate, shallow
response to atmospheric variability but also has a

2039
Bulletin of the American Meteorological Society
longer response through the advection of
shallow salinity anomalies (together with
a response that can extend out to millen-
nia in the deep ocean).
Observations of SST in hostile re-
gions like the wintertime Labrador Sea
are very sparse and are difficult to inter-
pret because of the large (~6°C) annual
cycle of SST. At about 100-m depth,
however, the annual cycle is down to
~1.5°C and decadal variability stands out
(e.g., Levitus et al. 1994; Reverdin et al.
1997). At 1000 m the annual cycle is
0.2°C, decadal variability is muted, and
10–100-yr variations dominate. The
great volume of LSW makes it a useful
stable reservoir for climate analysis.
Over the past 100 yr, LSW has moved in
a great counterclockwise loop in the po-
tential-temperature/salinity diagram
(Fig. 7). In the past decade or so the sys-
tem has returned to a high 70-yr extreme
NAO index, wonderfully deep convec-
tion (to a depth of 2200 m in 1992), and
a Labrador Sea resembling that in the
first few decades of the century. Mean-
while, as suggested by the seesaw be-
tween the Greenland high and Icelandic
low, convection on the other side of
Greenland, in the Greenland Sea, has
been weak since the early 1980s.
c. Theory and modeling
Laboratory and numerical studies of
oceanic convection have been central to
the planning of the field experiment and, particularly
when used in concert with and scaled for comparison
with the observations, have led to advances in our
understanding of the general problem of convection
in a rotating stratified fluid. Marshall and Schott
(1998) review the key ideas and contributions in the
context of the observations, models, and theory.
Two aspects make ocean convection interesting
from a fundamental point of view. First, the timescales
of the convective process in the ocean are sufficiently
long that it may be modified by the earth’s rotation.
Second, the convective and geostrophic scales are not
very disparate in the ocean and so the water mass
transformation process involves a fascinating inter-
play between convection and baroclinic instability
(the interaction between phases II and III in Fig. 6).
This lack of a scale separation in the ocean should be
contrasted with the atmosphere (e.g., see Fig. 2) where
the convective scale (the “rolls” clearly evident in IR
image) have a much smaller scale than that of the syn-
optic system in which they are embedded. This dif-
ference in the parameter range of atmospheric and
oceanographic convection can be usefully expressed
in terms of the size of a “natural Rossby number” that
is small in the ocean but large in the atmosphere (see
Jones and Marshall 1993; Maxworthy and Narimousa
1994). Moreover, in the ocean large horizontal buoy-
ancy gradients on the edge of the convection patch
support strong horizontal currents in thermal-wind
balance with them—the “rim current” (see Fig. 3). If
the patch has a lateral scale greater than the radius of
deformation, then instability theory tells us that it must
FIG. 6. A schematic diagram of the three phases of open-ocean deep convec-
tion: (I) preconditioning, (II) deep convection, and (III) lateral exchange and
spreading. Buoyancy flux through the sea surface is represented by curly arrows
and the underlying stratification/outcrops by continuous lines. A boundary cur-
rent runs around the periphery. Fluid overturned and mixed by convection is
shaded. (From Marshall and Schott 1998.)

2040
Vol. 79, No. 10, October 1998
break up into deformation-radius-scale fragments.
These edge effects play a dominant role in the dynam-
ics and thermodynamics of the mixed patch, orches-
trating the exchange of fluid and buoyancy to and from
it (e.g., see Legg and Marshall 1993; Visbeck et al.
1996). Many of the issues can be beautifully illustrated
in a simple laboratory experiment in which a disc
of colored ice is gently floated on the surface of a ro-
tating tank of water. The ice melts, drawing its latent
heat of fusion from the water, thus inducing rotation-
ally modified plumes that penetrate downward, mix-
ing the water up from below. If the dish is rotating
rapidly enough, the body of convectively modified
water breaks up into eddies—conical structures that
parcel up the convected fluid and carry it away to the
periphery.
One of the central goals of the Labrador Sea project
is to learn how to parametrically represent convection
in limited-area and large-scale ocean models.
Although hydrostatic adjustment remains the primary
method of representing deep convection in ocean gen-
eral circulation models, nonhydrostatic simulation
of unsteady deep convection by Jones and Marshall
(1993) has demonstrated the need to include vertical
acceleration by buoyancy in the momentum budget
on scales of kilometers and smaller. The budget of
turbulent kinetic energy in deep convection, previ-
ously viewed simply as “turbulence,” is now known
to have interesting new physics that had never before
been considered in mixed layer models: planetary ro-
tation can influence the evolution of the mixing layer
in the ocean; nonhydrostatic and 2Ω cos(lat) terms can
also play an important role, as can “thermobaric”
effects (the increase in the thermal expansion coeffi-
cient with pressure; see Garwood 1991; Garwood et al.
1994).
A hierarchy of different models and resolutions has
been used to study the interplay of scales sketched in
Fig. 3 and the role of hitherto-neglected dynamics and
physics. Studies focusing on the plume scale are typi-
cally carried out over domains only a few kilometers
wide and are capable of resolving the full nonhydro-
static turbulent plume dynamics, as in Fig. 8. The
subplume scales may either be explicitly resolved at
Reynolds numbers somewhat less than the ocean
(Julien et al. 1996) or accounted for by subgrid-scale
turbulence models [as in the large-eddy simulations
(LES) of Harcourt et al. (1997); Fig. 8]. Investigation
of the interaction between the geostrophic scale and
the plume scale requires resolution down to the plume
scale but over domains several tens of kilometers
FIG. 7. Salinity (ppt) and potential temperature properties of
Labrador Sea water from 1928 to 1995 (years marked). The wa-
ter mass has moved around a counterclockwise path in this
oceanographic phase space (P. Rhines, personal communication).
The tilted dotted lines are isopleths of constant potential density
(kg m−3), referred to 2 km.
FIG. 8. Snapshot of surface T (K), u, and v fields, in LES
high-resolution simulation of ocean convection induced by a
400 W m−2 heat loss and 20 m s−1 winds applied at the surface.
Currents reach speeds of a few tens of centimeters per second.
The domain is 6 km by 6 km in the horizontal and 2 km deep. (See
Harcourt et al. 1997.)

2041
Bulletin of the American Meteorological Society
across (Jones and Marshall 1993; Legg et al. 1997).
Investigations of larger-scale dynamics, including the
interaction between the basin-scale circulation and the
water mass formation require parameterization of con-
vection to account for turbulent transfer by unresolved
plumes. Comparison of the results between these dif-
ferent model formulations at the overlapping scales
and in the context of field data allows the validity of
assumptions and approximations to be tested.
In addition to improving
our understanding of the physi-
cal processes, modeling studies
also help us to evaluate and in-
terpret the response of different
measurement systems to the
turbulence, plumes, and geo-
strophic eddies. Simulations of
Lagrangian and isobaric drifters,
and comparison of their mea-
surement records with Eulerian
measurements of a field of con-
vective turbulence (Harcourt
et al. 1997), suggest that each of
these three types of measure-
ment systems has a particular
advantage and that a mixture of
all three is useful. Indeed, as de-
scribed in section 3, such a mix
of observing strategies is being
employed in the experiment.
Other applications of modeling
to the interpretation of observa-
tions include the simulation of
synthetic aperture radar images of
the surface fields associated with
convection (Fischer et al. 1998).
3. Planning the field
experiment
a. Elements of the
observational strategy
1) OCEAN
The measurement of deep
oceanic convection presents a
formidable challenge due to
the wide range of space and
timescales involved and the in-
termittency of the process in
space and time. The experimen-
tal challenge was to measure both the large and small
spatial scales over both short and long timescales in a
remote location under very adverse conditions. A va-
riety of experimental techniques was used (see Table
1 and Fig. 9). Measurements were carried out over a
2-yr period (summer 1996–summer 1998) in order to
resolve the seasonal cycle and observe the convective
process twice. The field work for the Convection Ac-
celerated Research Initiative was embedded within an
Aug 1996 T/S, O, N F. Schott, IfM Kiel, Germany
Oct–Nov 1996 T/S, O, N A. Clarke, BIO, Canada
Feb–Mar 1997 T/S, O, N CFC R. Pickart, WHOI, US
May–Jun 1997 T/S, O, N J. Lazier, BIO, Canada
Aug 1997 T/S, O, N F. Schott, IfM Kiel, Germany
TABLE 1. Elements of the oceanographic observational program.
Notes:
T/S = temperature/salinty; O = oxygen; N = nutrients; CFC = chloroflourocarbons; Ts =
surface temperature; pa = atmospheric pressure; (u, v) = horizontal velocity; w = vertical
velocity.
BIO = Bedford Institute of Oceanography; IfM Kiel = Dept. of Meteorology, Institut für
Meereskunde; IOS = Institute of Ocean Sciences; SIO = Scripps Institute of Oceanogra-
phy; URI = University of Rhode Island; UW = Dept. of Meteorology, University of
Washington; WHOI = Woods Hole Oceanographic Institution.
Hydrography
Period Variables PI/Lab
3D Lagrangian float u, v, w, TE. D’Asaro, UW
PALACE/VCM T/S, u, v, wR. Davis, SIO
Profiling RAFOS T, u, v, wM. Prater, URI
PALACE/VCM T/S, u, v, wB. Owens, WHOI
PALACE T, u, vF. Schott, IfM Kiel, Germany
Surface drifter Ts, pa, u, vP. Niiler, SIO
PALACE T/S, u, vR. Schmidt, WHOI
Moorings
WOTAN rainfall D. Farmer, IOS, Canada
Seacat/Aanderaa u, v, wP. Rhines, UW, USA;
J. Lazier, BIO, Canada
Seacat/Aanderaa/ADCP u, v, wF. Schott, IfM Kiel, Germany
Tomography T, vel U. Send, IfM Kiel, Germany
Sound sources Rafos tracking M. Prater, URI
Instrument Variables PI/Lab
Floats and drifters
Instrument Variables PI/Lab

2042
Vol. 79, No. 10, October 1998
array of autonomous floats deployed as part of the
WOCE Atlantic Circulation and Climate Experiment,
which mapped out the large-scale circulation.
The three-dimensional structure of the Labrador
Sea gyre was revealed by numerous hydrographic sec-
tions across the basin during four cruises in 1996–97.
These three-dimensional data were complemented by
acoustic tomography, measuring average properties
between several moored acoustic transceivers (see
Send et al. 1995). In addition, the WOCE AR-7 hy-
drographic section (see Fig. 9) was heavily instru-
mented with moorings measuring velocity, salinity,
and temperature throughout the water column at high
temporal resolution.
Finally, freely drifting surface
drifters and subsurface floats of
various designs provided an eco-
nomical way to measure a wide
range of space and timescales. A
large number (> 150) of such de-
vices have been deployed (see
Fig. 9). All measure horizontal ve-
locity and temperature at their lo-
cation; some measured vertical
velocity, a clear signature of con-
vection. Others periodically pro-
filed the T/S structure of the water
column, moving up to the surface
and relaying data back to base be-
fore dropping down to their refer-
ence level again. A broad range of
spatial scales was sampled by de-
ploying the instruments in both
small- and gyre-scale arrays.
2) ATMOSPHERE
The fluxes of heat, moisture,
and radiation across the air–sea in-
terface are the primary agents re-
sponsible for the densification of
the surface waters that triggers
convection. The objective of the
meteorological component of the
experiment was to document the
surface fluxes and to understand
the mechanisms responsible for
them. Our measurements benefited
from the larger-scale context pro-
vided by the FASTEX experiment
(Joly et al. 1997).
Historically, there have been no
direct measurements of the fluxes of heat, momentum,
and moisture across the air–sea interface in the Labra-
dor Sea—only bulk estimates. As a result, the confi-
dence that one can have in model-derived estimates
of these fluxes is reduced. It was therefore decided to
collect in situ data during a winter oceanographic
cruise of the R/V Knorr to test flux estimates of nu-
merical models and observe the detailed evolution of
the atmospheric boundary layer. Measurements of the
fluxes of incoming solar radiation, momentum, sen-
sible and latent heat, water vapor, and precipitation
were carried out by a consortium of groups. To im-
prove our understanding of the structure and evolu-
tion of weather systems over the region and to
FIG. 9. Map showing elements of the oceanographic field program. The hydrographic
stations occupied by the Knorr during February and March 1997 are indicated, as are
the floats deployed from Knorr (colored symbols; see key). Also shown are the posi-
tion of the moorings (open circles), the Central Float Deployment (CFD) region and
the tomographic array. Many other floats were deployed in cruises both before and after
that of the Knorr. The Ks indicate the position of Kiel moorings, the Bs are boundary
current moorings, Ws are inverted echo sounders, and SSs are RAFOS sound sources.
Investigators involved in the field program are indicated in Table 1.

2043
Bulletin of the American Meteorological Society
facilitate model initialization and verification efforts,
radiosonde launches were also made from the Knorr.
Table 2 outlines the wide variety of activities associ-
ated with the science program on the Knorr. In tan-
dem with the ship program, several aircraft missions
were flown to document the synoptic-scale environ-
ment over the region and to make measurements of
the mesoscale structures that modulate air–sea inter-
action in the region.
There was also a remote sensing component to the
experiment. Images were collected at prespecified
times and locations from the Canadian Radar Satel-
lite Synthetic Aperture Radar (SAR) and data were
collected on an ad hoc basis from the European Sat-
ellite Agency European Remote-sensing Satellite-2
(ERS-2) SAR, scatterometer, and radar altimeter. Two
well-equipped aircraft (a CV-580 from the Canada
Centre for Remote Sensing and a P3 from NASA/
Goddard Space Flight Center) flew over the ship on
separate 4-day missions collecting sea surface and
mean meteorological profile data using a wide vari-
ety of instruments.
b. Anticipating where convection will happen
During the planning stages of the convection ex-
periment we began to recognize that decadal climate
variability reviewed in section 2b and the vagaries of
the NAO might interfere with our goal to observe deep
convection in the Labrador Sea. The winter of 1993
had resulted in an extremely active convection season
penetrating down to more than 2200-m depth, but in
subsequent winters convection never reached those
depths again and in fact the winter of 1995–96 resulted
in convection that was probably no deeper than
1000 m.
With the drop in the NAO index during the win-
ter of 1995–96 we feared a second weak winter. It was
clear that we had to make sure that the instruments
designed to observe deep mixing were deployed in an
optimum way. Luckily we were in a position to de-
ploy deep mixed layer floats during the early part of
the Knorr winter survey, just a few weeks prior to the
expected deepest mixing. But where, precisely, should
we put them?
Starting in October 1996 we began predicting sev-
eral convection scenarios using both a high-resolution
numerical model (Marshall et al. 1997) and real-time
data coming in from some 30 profiling Autonomous
Lagrangian Circulation Explorer (PALACE) floats,
whose position in February 1997 is indicated in
Fig. 10. The temperature and salinity profiles provided
by these floats were used to compute the stratification
and obtain estimates of how much buoyancy was held
in the water column. This buoyancy content was com-
pared to the expected buoyancy loss between the time
of “forecasting” and the end of the winter season in
late March, enabling us to predict how deep convec-
tion might reach assuming it to be a one-dimensional
(1D) process. The expected depth to which the mixed
layer would reach in a moderate winter (defined as one
in which there was only 80% of a typical wintertime
buoyancy loss) assuming nonpenetrative deepening,
is indicated in Fig. 10. As reviewed in section 4a, early
in the winter heat losses were relatively weak raising
some concern, but our predictions suggested that we
would still have a reasonable chance of observing
mixing down to a depth of 800–1000 m. Nevertheless,
it was decided to withhold half of the floats to be able
to spread the risk over two convection seasons; the
remainder were deployed in the second phase of the
field work, in January 1998. By early January 1997 it
became clear that an interesting pattern of stratifica-
tion was revealed by the PALACE floats (see Fig. 10):
the sea was strongly stratified to the north and to the
south, but a band of weak stratification existed in
the middle, some 3° in width centered at 56°N.
Combining this information with the climatological
distribution of oceanic heat loss that shows a maxi-
mum just eastward of the Labrador coast (see Fig. 4)
we identified an area of weakest stratification and
expected maximum heat loss (the “yellow” region in
Fig. 10). This “prediction” also found support from
the high-resolution numerical model of the Labrador
Sea developed by C. Herbaut and J. Marshall (1998,
personal communication) (see Fig. 11).
Based on the above considerations, a few days
before the Knorr left Halifax, Nova Scotia, Canada,
it was decided to deploy the floats in a central float
deployment (CFD) area [the box marked in Fig. 9,
roughly coinciding with the position of the convec-
tion patch observed by Clarke and Gascard (1983);
also see the shaded patch in Fig. 1]. During early Feb-
ruary the float array was deployed. From October until
the middle of February our predictions of mixed layer
depth, based on the assumption of a moderate winter,
remained unchanged at 800–1000 m. However, to our
great delight and as described in section 4b, by the end
of the Knorr cruise (12 March) the mixed layer had
reached a maximum depth of 1500 m in one spot
50 km southward of our best-guess estimate. The cy-
clonic circulation pattern over the North Atlantic had
intensified during February, resulting in heat losses

2044
Vol. 79, No. 10, October 1998
Rawinsonde, 6–12 day U,D,Td, Ta Guest/NPGS Guest/NPGS, IfM Kiel
Solar pyranometer SW + LW incoming Guest/NPGS Guest/NPGS, IfM Kiel
Laser ceilometer CeilingHt, CloudHt White/ETL Guest/NPGS
3 GHz precipitation radar VertVel Liq. H2O Costa/ETL Dobson/BIO
Ship rain gauge P Uhlig/Kiel Uhlig/Kiel
Disdrometer P, P′Grossklaus/Kiel Uhlig/Kiel
Disdrometer P, P′Hare/ETL Dobson/BIO
IR cloud temp Tcloudbot Uhlig/Kiel Uhlig/Kiel
IR sea temp Ta Guest/NPGS Guest/NPGS
“Sea snake” Ta Hare/ETL Dobson/BIO
Intake temp Ta IMET/WHOI Pickart/WHOI
Pl. res. therm. Ta Hare/ETL Dobson/BIO
Thermistor Ta IMET/WHOI Pickart/WHOI
Fast thermometer Ta, Ta′Bumke/Kiel Uhlig/Kiel
Thermistor Ta, Ta′Anderson/BIO Anderson/BIO
Thermistor Ta, Ta′Anderson/BIO Anderson/BIO
Hygrometer RH IMET/WHOI Pickart/WHOI
Dewflinger e′Aktaturk/UW Anderson/BIO
Hygrometer h′, CO2′Hare/ETL Anderson/BIO
Wind monitor U, D, u′, d′Anderson/BIO Anderson/BIO
Gill sonic anem. U, D, Ta, u′, v′, w′, ta′Anderson/BIO Anderson/BIO
Anemometer U, D, Ta, u′, v′, w′, ta′Bumke/Kiel Bumke/Kiel
Motion package ai, TiltiUhlig/Kiel Uhlig/Kiel
Gyrocom-pass Ship’s course IMET/WHOI Pickart/WHOI
Gyrocom-pass Ship’s course Hare/ETL Dobson/BIO
Doppler log Ship’s speed IMET/WHOI Pickart/WHOI
GPS Ship’s position IMET/WHOI Pickart /WHOI
Marine radar Wave field Trizna/NRL Dobson/BIO
Wave height gauge Wave height Dobson/BIO Dobson/BIO
Directional buoy Wave dirn. spec Dobson/BIO Dobson/BIO
TABLE 2. Meteorological and air–sea flux measurements on R/V Knorr.
Instrument Variable(s) Owner/Lab PI
Caps = Mean quantities; primes = fluctuations about the means; U, u = wind speed; D, d = wind direction; v, w = wind components
crosswinds, vertical; Ta = air temperature; Ts = sea surface temperature; h = absolute humidity; CO2 = CO2 content; a =
acceleration; P = precipitation; e = water vapor; IMET = improved meteorological measurement system; Pl. res. therm. = platinum
resistance thermometer.
ETL = Environmental Technology Laboratory, NOAA; NPGS = Naval Postgraduate School.

2045
Bulletin of the American Meteorological Society
greatly exceeding climatological values (see section
4a), which induced deeper-than-expected convection.
4. Preliminary results from the 1996–97
experiment
a. The atmospheric forcing during the winter of
1997
1) SYNOPTIC CONDITIONS DURING WINTER 1997
The winter of 1997 provided an excellent oppor-
tunity to document the variability in the air–sea inter-
action that exists within a given winter season. Figure
12 shows the average mean sea level pressure, 10-m
wind, and total heat flux fields for January 1997
(Fig. 12a) and February 1997 (Fig. 12b) as determined
from the NCEP–NCAR Reanalysis project. A com-
parison between the two months and the winter
climatology (Fig. 5) shows that a dramatic
transition occurred in the flow regime
over the North Atlantic and that this re-
sulted in a significant change in the mag-
nitude of the surface cooling in the
Labrador Sea region. January 1997 was
a month in which the circulation over the
North Atlantic was significantly differ-
ent from the winter climatology; the
presence of a blocking high over
Europe and anomalously strong high
pressure over Greenland resulted in a
significant westward shift in the center
of cyclonic activity. It is interesting to
note that even with this weakening of the
cyclonic flow over the North Atlantic,
the mean total heat loss in the center of
the Labrador Sea was, at 260 W m−2,
larger than the climatological winter
mean. In contrast, February 1997 was a
month in which the circulation pattern
over the North Atlantic was significantly
stronger than average. As one might ex-
pect in such a flow configuration, the
mean heat loss in the center of the La-
brador Sea, some 420 W m−2, was sig-
nificantly above the climatological
winter mean (see Table 3). These ex-
tremely high oceanic heat losses contrib-
uted to making what might otherwise
have been a lackluster winter, from the
perspective of forcing deep convection,
into a “good” winter.
2) VARIABILITY IN HEAT AND MOISTURE FLUXES
DURING THE WINTER OF 1997
Although a period of one month is a convenient
period of time over which to average, the influence and
phasing of individual events tends to be lost by such
averaging. Another and perhaps more illuminating
view of the variability of air–sea interaction in the La-
brador Sea during the winter of 1997 is depicted in
Fig. 13. This figure shows the 6-hourly values of mean
sea level pressure at a location in the center of the cli-
matological Icelandic Low (62°N, 30°W; Fig. 13a) and
total heat flux at the Bravo site (56°N, 51°W; Fig. 13b)
as determined from the NCEP–NCAR Reanalysis. It
is clear that the North Atlantic flow regime in Decem-
ber 1996 and early January 1997 was significantly
different from that in the latter part of the winter. The
generally high sea level pressure and weak cyclonic
activity in the early part of the winter can be classi-
FIG. 10. Position of PALACE floats (black discs; courtesy of R. Davis) at the
beginning of February 1997. Based on the measured buoyancy content in the water
column at these positions, climatological estimates of the buoyancy loss from the
surface, estimates of the expected maximum depth of the mixed layer are indi-
cated in meters (courtesy of M. Visbeck) assuming a moderate winter (defined as
80% of the climatology), and a 1D, nonpenetrative process. The region of weak-
est stratification is marked. Its overlap with the region of maximum buoyancy loss
to the atmosphere led to our prediction of the position of deepest mixed layer depth
(marked yellow). This, together with the numerical model results shown in Fig. 11,
guided our choice of position of the CFD area.

2046
Vol. 79, No. 10, October 1998
fied, according to Vautard (1990), as being associated
with either a “blocking” or “Greenland anticyclone”
regime, while the low sea level pressure and strong
cyclonic activity during the latter part of the winter
belongs to a more “zonal” flow regime. The first two
weeks of January were unusual in that the pressure near
Iceland underwent a monotonic decrease. This change
in flow regime had a dramatic impact on the surface
cooling that was taking place in the Labrador Sea.
From December 1996 to 15 January 1997 the average
total heat flux at the Bravo site was only 150 W m−2.
Indeed, the lowest heat fluxes occurred during the first
two weeks of January. Over the flowing six-week pe-
riod ending on 28 February 1997, however, it was in
excess of 420 W m−2 with peak fluxes greater than
1000 W m−2.1 There is significant high-frequency
variability in the magnitude of the total heat flux, a
signature of the passage of North Atlantic cyclones.
This can be seen from the high degree of anticorrela-
tion that exists between the two time series (see
Fig. 13). Furthermore, the data explain why the sur-
face cooling during January was above the climato-
logical mean even though the monthly mean sea level
pressure field indicated a weaker than average cyclonic
flow over the North Atlantic. The more vigorous cy-
clonic circulation that developed after the middle of
the month and the concomitant elevated heat fluxes led
to a large monthly mean heat flux for January.
With a few exceptions, the time series of precipi-
tation is similar to that of the total heat flux. The high-
est precipitation rates, in excess of 2 mm h−1, occurred
in late January and February. A careful examination
of the phasing of the precipitation and total heat flux
maxima in the Labrador Sea region indicates that the
former typically leads the latter by approximately one
day. This is a consequence of the structure of a typi-
cal extratropical cyclone, such as the one shown in
Fig. 2, which results in a separation of the region of
high precipitation from that of high heat flux.
Table 3 presents monthly mean values of the total
heat flux from NCEP–NCAR, precipitation rate, and
net radiative flux at the historic location of the Ocean
Weather Station Bravo (56°N, 51°W) for the months
of December 1996, and January, February, and March
1997. December 1996 was a month in which the
fluxes of heat and momentum were anomalously low,
when compared to the 30-yr mean for the month of
December. In contrast, January and February 1997
were months in which these fluxes exceeded the 30-yr
means. When considered as a whole, the winter of
FIG. 11. Simulation of water mass transformation in a high-
resolution model of the Labrador Sea developed by C. Herbaut
and J. Marshall (1998, personal communication): (a) hydrographic
section of temperature across the model’s AR-7 section, (b) cur-
rents at a depth of 100 m showing the boundary current and ed-
dies, and (c) mixed layer depth in March 1992 obtained by driving
the model with NCEP winds and fluxes starting in summer 1991.
The position of the region of deepest mixed layer is roughly co-
incident with that indicated in Fig. 10.
1It is important to note that there is a clear discrepancy between
analyzed fluxes and the bulk estimates from the Knorr; com-
pare Fig. 14 with Fig. 13, for example. Are they real or the result
of differences in sampling or of reporting?

2047
Bulletin of the American Meteorological Society
1997 was one in which the
fluxes of heat, moisture, and net
radiation exceeded the 30-yr
means by approximately one
standard deviation and so was
highly conducive to convective
activity in the Labrador Sea
region.
3) AIRCRAFT AND REMOTE
SENSING MISSIONS
The meteorological aircraft
studies consisted of three mis-
sions over the Labrador Sea.
Flight planning took into ac-
count the position of the Knorr
and meteorological conditions
reported from the ship, and on
one occasion sampling took
place in the vicinity of the ship.
The flights were completed dur-
ing February 1997 with research
aircraft from the 53d Weather
Reconnaissance Squadron of
the U. S. Air Force. The C-130
aircraft used were equipped with
dropsonde systems and were
able to record state parameters
at flight level. To assist in flight
planning, as described in
Renfrew et al. (1998), the Naval
Research Laboratory ran a spe-
cial version of their Coupled
Ocean–Atmosphere Mesoscale
Prediction System (COAMPS) (see Hodur 1997) re-
gional atmospheric forecast model over a domain that
included the Labrador Sea twice daily out to 36 h. The
model data, including all important surface fields such
as the heat, radiative, and moisture fluxes, were made
available in real time to scientists in the field, greatly
helping the planning process. These data were also
used to generate daily synoptic weather summaries and
forecasts for use by the Knorr in scheduling opera-
tional and scientific ship activities.
Two remote sensing aircraft programs were car-
ried out over the Knorr during the experiment. On
22–24 February a Convair 580 from the Canada Cen-
tre for Remote Sensing equipped with a C-band syn-
thetic aperture radar flew a series of four missions,
imaging the ship and the waters around her to deter-
mine the effects of polarization on the ability of the
instrument to image the sea surface. On 3–9 March a
P-3 aircraft from NASA flew four ocean wave imag-
ing and passive microwave missions over the ship.
Both missions carried out intensive surface meteoro-
logical and wave measurements in concert with the
shipboard measurement program. In addition, ERS-2
SAR images of the ocean surface were obtained and
attempts made to identify the surface signature of deep
ocean convection by using hydrodynamic models of
convection and a sensor model for SAR (see Fischer
et al. 1998). Identification of a convective surface sig-
nature could allow unique information about the spa-
tial characteristics of the convective process to be
determined such as the extent of convecting region,
the size of individual convective plumes and eddies,
the ratio of converging to diverging area, and the sur-
face current strain. Initial results, though speculative
FIG. 12. Average mean sea level pressure (contour, mb), 10-m wind (vector, m s−1) and
total heat flux (shaded, W m−2) fields from the NCEP–NCAR Reanalysis for (a) the month
of January 1997 and (b) the month of February 1997.

2048
Vol. 79, No. 10, October 1998
in nature, showed several scenes that had an appear-
ance that mimicked model results.
4) IN SITU MEASUREMENTS FROM THE KNORR
The R/V Knorr was in the Labrador Sea proper
from 7 February to 12 March 1997. During that pe-
riod the atmospheric conditions were surprisingly con-
sistent and in the “zonal regime”. Most of the
low-pressure systems that approached from the south
and west tracked to the south of the ship, and for vir-
tually the entire cruise the wind direction was west-
erly to northerly. Upper-air winds were generally from
the west. We were in a “cold air outbreak” regime (see
Figs. 2 and 12b).
At this early stage in the analysis what is really
available to us, besides real-time uncalibrated esti-
mates of the fluxes from our on-line logging systems,
are bulk flux estimates and the evidence of our senses.
They tell us that the entire period was dominated by a
flow of cold dry air from the northwest, intense trans-
fers of heat and water vapor to the atmosphere, ever-
present precipitation in the form of snow squalls, and
rapid deepening of the oceanic mixed layer. Table 4
indicates the ranges and types of meteorological con-
ditions encountered during the cruise.
Figure 14 is a plot of preliminary time series of air-
sea fluxes as deduced from heat and momentum trans-
fers encountered during the cruise of the Knorr. (The
bulk meteorological parameters on which they are
based have not been fully corrected, although the
Total heat flux (W m−2) 117 258 422 227 271
(30-yr mean and (189 ± 64) (223 ± 91) (247 ± 98) (198 ± 88) (223 ± 59)
standard deviation)
Precipitation rate (mm h−1) 0.114 0.247 0.175 0.192 0.182
(30-yr mean and (0.167 ± 0.043) (0.162 ± 0.050) (0.137 ± 0.046) (0.114 ± 0.039) (0.147 ± 0.028)
standard deviation)
Net radiative flux (W m−2) 53.2 50.8 44.8 −20 32
(30-yr mean and (57 ± 7) (55 ± 10) (25 + 10) (−25 ± 7) (28 ± 5)
standard deviation)
TABLE 3. Fluxes and precipitation from the NCEP–NCAR Reanalysis.
Dec Jan Feb Mar Winter of
1996 1997 1997 1997 1997
Mean values of the total heat flux (W m−2), precipitation rate (mm hr−1) and net radiative flux (W m−2) at 56°N,51°W for the months
of December 1996 and Jan–Mar 1997 as diagnosed from the NCEP–NCAR Reanalysis. Also shown are the means for the winter
of 1997 (1 Dec–31 Mar 1996 inclusive). The 30-yr means for the various time periods as well as the standard deviation from the
means are also shown.
FIG. 13. (a) Six-hourly valves of mean sea level pressure (mb)
at 62°N, 30°W and (b) total heat flux (W m−2) at 56°N, 51°W for
the period 1 December 1996–1 April 1997 as determined from
the NCEP–NCAR Reanalysis.

2049
Bulletin of the American Meteorological Society
winds have been transformed from rela-
tive to the ship into absolute.) We have
made no attempt to separate the data into
periods when the ship was still or mov-
ing, so the time series is a mix of time and
space. The large heat fluxes to the atmo-
sphere are a reflection of the environment
the ship was working in, but note that
they are somewhat smaller than those im-
plied by the models (see Fig. 13). They
are largest in magnitude at times when
the ship approached the ice pack on the
Labrador coast.
It is interesting to use the average
of the total heat flux to compute the
cooling effect of the fluxes themselves.
Assuming that a layer of ocean 1000 m
deep was cooled over the course of the
Knorr’s stay in the Labrador Sea, the
average flux of 360 W m−2 would have
cooled that water by about 0.24°C. This
is a representative calculation only, be-
ing a combination of the fluxes at differ-
ent locations in the Labrador Sea at
different times. Nevertheless, the heat loss as mea-
sured by the CTD was very close to what would be
expected based on the ship measurements.
Estimates of mean evaporation E and precipitation
P for the period are incomplete at time of writing
and hence inaccurate. Based on the bulk latent heat
flux from Table 4, E is about 5 mm day−1; based
on visual estimates of snow accumulation rates on
deck, P is 5–15 mm day−1, in both cases of liquid
water. We intend to improve both, using eddy corre-
lation latent heat fluxes for E and disdrometer mea-
surements for P.
Heat was lost over the entire sea, but by different
mechanisms in different parts of the area. Near the
western boundary, the precipitation was at a minimum
but the measured upward fluxes of heat and water va-
por were the largest. The snow squalls occupied more
of the total surface area as we progressed eastward, so
that from one-third of the way from the west to the east
side it snowed more or less continuously. Whereas the
net moisture transfer was upward on the western side,
it was downward over most of the remainder of the
area. This downward flux of moisture as precipitation
was necessarily accompanied by an upward transfer
of heat, since the water that received the snow had to
give up the latent heat of fusion as well as some sen-
sible heat. Thus there was upward transfer of heat over
the entire sea, but the flux of buoyancy was upward
on the western side and downward to the east. Using
a bulk surface heat flux method (Smith 1988), the av-
erage sensible heat flux during the time the ship was
in the Labrador Sea proper was about 160 W m−2. The
average latent heat flux was about 150 W m−2.
The spatial distribution of observed heat and mois-
ture transfer suggests that at least during cold air out-
breaks (northwesterly winds), the Labrador Sea forces
a strong coupled response in the atmosphere and that
the air–sea transfers ensure that the entire atmosphere–
ocean system can be treated (and modeled) as a single
coupled process. However, when the wind is from the
south (a rare occurrence over the duration of the cruise)
there is a ready supply of warm advective moisture for
precipitation.
In addition to surface atmospheric measurements,
many upper-air rawinsonde profiles were performed
from the Knorr, the only rawinsonde measurements
ever undertaken in the central part of the Labrador Sea.
The upper-air profiles demonstrated that deep convec-
tion existed in the atmosphere as well as in the ocean.
Figure 15 shows three examples: the deepest atmo-
sphere boundary layer (ABL) observed during the
cruise (0500 UTC 9 February), a more typical bound-
ary layer in the same location a few hours later
(1400 UTC 9 February), and a relatively shallow ABL
FIG. 14. Time series of surface fluxes and wind stress as deduced from con-
tinuous shipboard measurements on the Knorr as it moved around the Labrador
Sea taking hydrographic sections and deploying floats (courtesy of F. Dobson and
P. Guest). Investigators involved in the meteorological measurements taken from
the Knorr are indicated in Table 2.

2050
Vol. 79, No. 10, October 1998
just downwind of the ice edge (22 February). In gen-
eral, the ABL tended to be shallowest just off the up-
wind ice edge (as in the 22 February case) and it
deepened rapidly offshore. However, considerable
temporal variability (e.g., see the two 9 February pro-
files plotted in Fig. 15) prevented a detailed assess-
ment of spatial variations in ABL characteristics from
the ship data alone. One of the significant effects of
the deep ABLs was that the temperatures and humidi-
ties required considerable time to adjust to the sea
surface conditions due to the large quantity of air that
had to be modified. Evidently, in cold air outbreaks
over the Labrador Sea, the temperatures and humidi-
ties remain low and large turbulent sensible and la-
tent heat fluxes extend all the way across the Labrador
Sea.
b. The response of the ocean
1) WINTER CRUISE: 7 FEB–12
MARCH 1997
(i) Hydrography
The planned cruise track was de-
signed to cover a broad area of the La-
brador basin, with emphasis on the
western sector (see Fig. 9). After repeat-
ing the along-basin section of the fall
cruise (October–November 1996) it
was evident that the developing region
of deepest convection was of limited
lateral extent. Cross-basin sections were
thus placed more closely together, and
“dog-legs” devised in order to achieve
a broad coverage (Fig. 9). We ensured
that detailed boundary crossings were
made on both sides of the basin (on the
Labrador side these were limited by the
ice pack) so that the role of the warm–
salty Irminger water on both the con-
vection and restratification process
could be addressed. The central cross-
basin section, where the deepest mixed
layers were observed, was repeated a
second time in order to shed light on the
relative influence of vertical versus
horizontal advection. Finally, we per-
formed two detailed fine resolution
CTD surveys, the second of which cap-
tured the deepest convection of the ex-
periment (1500 m).
There were numerous surprises that
will require further analysis using the
combined resources of the convection
experiment. The rapidity at which the oceanic mixed
layer deepened during the cruise was impressive. One
of the CTD stations on the central cross-basin section
was repeated three times during the 5-week period. A
simple 1D mixed layer model (Pickart and Smethie
1998) required a sustained forcing of 1000 W m−2 to
produce the observed mixed layer depths of the sec-
ond and third occupations. This buoyancy loss is sig-
nificantly greater than that observed, suggesting that
the mixed fluid must have been advected from else-
where to the point at which it was observed. Another
surprise was how close convection occurred to the
western boundary. In fact, the deepest observed mixed
layers were near the 3100 m isobath, close to the core
of the deep western boundary current with important
ramifications for the spreading process. One view is
Air pressure (mb) 973.1 1017.3 994.9 10.3
Wind speed (m s−1) 0.4 24.7 11.7 3.9
Wind direction 291
Air temp (°C) −17.11 4.15 −7.84 3.80
Sea sfce temp (°C) −1.47 5.72 2.89 5.72
Relative humidity (%) 42.0 95.8 68.7 9.0
Wind stress (Pa) 0.0009 1.274 0.274 0.180
Sens heat flux −15.6 374.2 163.5 79.9
Latent heat flux −1.8 305.1 147.1 54.8
Net SW rad flux −524.3 0.0 −33.1 61.7
Net LW rad flux 11.0 174.0 84.3 24.7
Net rad flux −422.1 165.7 51.1 62.9
Total heat flux −255.3 732.7 361.7 160.8
TABLE 4. Range of meteorological conditions encountered during Knorr cruise.
Min Max Mean Std dev
Notes:
1. These statistics are based on 5-min averages of the 15-s samples of the Knorr’s
IMET system (sensor height 23 m) and samples from the U.S. Naval Postgraduate
School’s radiation measurement system. They cover only the period from 0000
UTC on 8 Feb 1997 until 24 UTC on Mar 13, when the Knorr was actually in the
experimental area.
2. All heat fluxes are in W m−2, and positive upward (out of the ocean).

2051
Bulletin of the American Meteorological Society
that convection occurs in the center of
the western cyclonic gyre (e.g.,
Clarke and Gascard 1983), then
slowly spreads outward, predomi-
nantly by the action of baroclinic in-
stability (Legg and Marshall 1993;
Visbeck et al. 1996). Instead our ob-
servations suggest that some of the
convected water may directly pen-
etrate the western boundary current,
thereby providing a swift escape route
to southern latitudes. The increased
surface salinity near the boundary, in-
fluenced by the Irminger water, may
also play a role in the deep mixed lay-
ers observed there.
Throughout, the convection was
observed to be accompanied by defor-
mation-scale (~10 km) geostrophic
eddies and baroclinic instability.
Toward the end of the cruise, particu-
larly during the repeat of the central
cross-basin section (see Fig. 9), we
surmise that measurements were
conducted during or shortly after
active convection. The spatial vari-
ability in the CTD casts was remark-
able. Intrusions were prevalent, and
often the downcast trace would differ
significantly from that of the upcast.
Our second CTD tow-yo survey cap-
tured the capping-over of a convective
patch, complete with numerous such
intrusions. Some of this rich structure
may be due to the proximity of con-
vection to the boundary, where strong
contrasts exist between resident water
masses. Finally, we found that the re-
gion of deep convection was rather
confined laterally, as revealed by the
distribution of observed mixed layer
depth (Fig. 16). This map is aliased
in time with, for example, a ridge of
shallow mixed layer depth corre-
sponding to the along-basin section,
which was occupied early in the
cruise. Eventually this map will be ad-
justed for synopticity using all the
available time series information. The
relative importance of atmospheric
forcing versus oceanic precondition-
FIG. 16. Mixed layer depth deduced from the Knorr hydrography, an average over
the period 7 February–12 March 1997. The deepest convection is localized to the
west, where a mixed layer of depth 1100 m is observed (courtesy of R. Pickart).
FIG. 15. Potential temperature vs height from the 1997 Knorr Labrador Sea cruise
based on rawinsonde (weather balloon) profiles. Regions of near-constancy of po-
tential temperature with height are indicative of convection in the atmospheric bound-
ary layers. The potential temperature of 0500 UTC 9 February increases slightly
between 1200 and 4500 m (following a “pseudo-adiabat”) due to latent heat release
within a cloud, but is indicative of a well-mixed layer of depth ~4 km forming the
ABL. The ABL depths on 1400 UTC 9 Feb and 2300 UTC 22 Feb are 1250 and 980 m,
respectively (courtesy of P. Guest).

2052
Vol. 79, No. 10, October 1998
ing in setting the relatively confined scale of the
mixed patch needs to be addressed.
(ii) Tracers
The chlorofluorocarbons (CFCs) CFC-11,
CFC-12, and CFC-113 were measured at 44 of the
hydrographic stations and provide a vivid record of
convective activity. CFCs enter the ocean at the sur-
face and become incorporated in LSW during the deep
convective process. In Fig. 17 the basin-scale distri-
bution of CFCs is shown along the AR7 section (see
Fig. 9 for position). The highest surface water CFC
concentrations, close to equilibrium with the atmo-
sphere, were observed at the margins of the Labrador
Sea where convection was not occurring. In the cen-
tral Labrador Sea, surface water concentrations were
lower and well below equilibrium with the atmosphere
as a result of deep convection transporting CFCs
downward faster than they entered the surface water
by gas exchange. LSW is readily identifiable in the
CFC distribution as two distinct homogeneous layers,
an upper layer extending from the surface to 500–
1000-m depth and a deeper layer extending from the
base of the upper layer to a depth of about 2200 m.
The upper layer is LSW that has just formed, deeper
on the western side of the section. The deeper layer
was formed during a previous winter when convec-
tion reached to ~2 km, probably in February and
March 1992, as discussed in section 2b.
2)FLOATS
(i)Profiling floats (PALACE) and
vertical current meters (VCM)
These floats periodically dive to
depths near 1600 m collecting profiles
of temperature and salinity. By adjust-
ing buoyancy, the float then moves to
a preprogrammed depth of either 1500
or 600 m where it follows currents for
periods of 10 or 20 days. It then rises
to the surface for 24 h during which
time profile data is relayed to Argos
satellites, which also locate the float.
After this surface period, the only
time when float position is known, the
instrument descends to collect another
profile and velocity observation.
Figure 18 shows as colored vec-
tors the trajectories of 36 WOCE
floats deployed between November
1996 and January 1997 and 8 similar
NOAA floats deployed in 1994 and
1995. These floats map out some ex-
pected and unexpected aspects of the
circulation and should be compared
with the schematic shown in Fig. 1.
They show the Greenland Current
as a strong boundary current off the
FIG. 18. The trajectories PALACE floats in the Labrador Sea. Colored vectors
indicate submerged displacements over 10–20 days of floats drifting between 600
and 1500 m deployed between 1994 and 1997. Black vectors indicate the submerged
displacement of VCM floats deployed near 400 m as part of the Convection ARI.
These are 4-day displacements (courtesy of R. Davis).
FIG. 17. Vertical distribution of CFC-11 (pmol kg−1) along-
line AR7 of the Knorr winter Labrador Sea cruise. The data
are preliminary, based on shipboard calculations (courtesy of
W. Smethie).

2053
Bulletin of the American Meteorological Society
Greenland coast; sustained flows up to 25 cm s−1 are
observed off the southwest coast. This flow generally
follows the bathymetry across the northern Labrador
Sea, forming the intermediate depth portion of the
southwestward flowing Labrador Current that even-
tually makes its way past the Grand Banks of New-
foundland. These boundary currents surround the
central Labrador Sea, where organized flows are weak
compared to the substantial synoptic-scale variabil-
ity. In most places there is no apparent pattern.
A signature of convection in the Labrador Sea is
the mixing of relatively fresh surface waters to in-
creasing depth as surface fluxes extract buoyancy
from the upper-water column. Figure 19 shows a time
series of potential temperature and salinity profiles
from a float deployed near 60.5°N, 57°W that moved
southeastward roughly paralleling isobaths (it is
marked purple in Fig. 18). As time progresses, cool
upper-layer temperatures erode the thermocline tem-
perature maximum until it disappears; simulta-
neously, freshwater is mixed to progressively greater
depths. This sequence suggests convection reached
about 1200 m by the end of March. Unfortunately,
temporal and spatial variability within the water col-
umn make it difficult to determine precisely how deep
the convective stirring reached.
In late January 1997, 31 profiling vertical current
meter (VCM) floats were deployed in
the CFD area marked in Fig. 9 where
deep convection was expected to be
most vigorous. Individual displace-
ments from 16 of these floats are
shown as black vectors in Fig. 18.
These instruments recorded vigorous
vertical velocities associated with
convection. Figure 20 shows a set of
three temperature profiles describing
a mixed layer that deepens by about
100 m over 10 days. The figure also
shows vertical velocity time series
from the two intervals between these
profiles. Vertical flows of 5 cm s−1 are
frequent and there is one example
of a plume with peak velocities of
10 cm s−1 and a duration of 6–8 h. The
general magnitude of vertical veloc-
ity is comparable to that observed by
Schott et al. (1996) from moored
acoustic Doppler current profilers
(ADCPs) during convective events in
the northern Mediterranean. The
event duration is much longer than was seen in the
moored results, consistent with the idea that plumes
with widths of order 500 m were advected past the
moorings while the VCM approximately follows the
plume’s horizontal motion.
(ii) Three-dimensional Lagrangian floats
The three-dimensional trajectories of water parcels
in deep convection were measured using a new type
of subsurface float, the “deep Lagrangian float”. These
combine high drag, a compressibility that is very close
to that of seawater, and nearly neutral buoyancy to
closely match the physical properties of seawater and
thus follow its motion in three dimensions. They are
tracked acoustically (RAFOS)2 in the horizontal and
by pressure in the vertical and measure the tempera-
ture of the water with submillidegree resolution. A to-
tal of 13 deep Lagrangian floats were deployed in the
Labrador Sea in February 1997 in the CFD area; 10
more will be deployed in January 1998.
Figure 21 shows some preliminary data. The float
is deployed, sinks to 1000 m below the layer of ac-
tive convection, and begins a 7-day autoballasting se-
quence. It then lightens itself slightly and rises into
FIG. 19. Time sequence of potential temperature and salinity profiles from a pro-
filing float. There is a profile every 10 days from 2 December 1996 to 25 April 1997
following the color sequence in the temperature plot from top to bottom. The strong
salinity-stabilized temperature inversions are typical. It appears that convection had
reached 1200 m by late March (courtesy of R. Davis).
2RAFOS (not an acronym) is a float that receives sound signals.

2054
Vol. 79, No. 10, October 1998
the convecting layer. For the next 25 days, the floats
cycle between the surface and 500–800 m, with ver-
tical velocities often exceeding 10 cm s−1. Two excur-
sions to 1000 m can be seen. The rms vertical velocity,
2.3 cm s−1, is approximately that expected for convec-
tion with a surface heat flux of 400 W m−2. The water
is cooler when going down than when coming up, as
expected for convection. Overall, the data indicate the
presence of a nearly continuously convecting layer,
deepening with time.
(iii) RAFOS floats
A total of 33 RAFOS floats were deployed from
October 1996 to May 1997 from three cruises of the
CSS Hudson (twice) and the R/V Knorr (see Fig. 9).
The nominal pressure surfaces were set to one of the
following: 150, 350, or 600 dbars. The mission lengths
extended from 1 to 10 months. The early floats clearly
demonstrated the cooling of the surface mixed layer
during October and November (Fig. 22) and imply a
surface heat flux estimate of 160 W m−2. Floats de-
ployed later measured the effects of convective events,
particularly strong downward velocities, and the as-
sociated turbulent heat fluxes. In addition, floats have
been deployed off Greenland to capture the eddy-
driven transport of Irminger water into the Labrador
Sea, and the resulting capping and advective-driven
restratification of the basin interior. As more data be-
come available, we anticipate mapping the spatial and
temporal distributions of convective activity in the
context of the geostrophic eddy and basin circulation
fields.
3) MOORINGS AND TOMOGRAPHY
In May 1994 a long-term site mooring was es-
tablished at 56.75°N, 52.5°W near the location of former
weather station Bravo (Lilly et al. 1998). The moor-
ing, together with the repeated hydrographic sampling
on CSS Hudson, had provided an extended record of
wintertime convection as a lead-in to and motivation
for the intensive experiment during 1996–98. It has
15 instruments and measures temperature, salinity,
three-component velocity, as well as passive noise.
The velocity field revealed by the mooring involves
strong synoptic-scale and mesoscale eddies with
nearly barotropic vertical structure, in which are em-
bedded finescale convective plumes. Both scales of
eddy are energized at the time of deep convection.
An array of moorings was deployed along the AR7
hydrographic line over the winter of 1996–97 (see
Table 1 and Fig. 9). It was designed to measure hori-
zontal currents by rotor current meters and ADCPs to
record variability in convection by temperature and
conductivity measurements, as well as vertical currents
from ADCPs and to determine the integral effects of
deep convection by acoustic tomography.
The Bravo mooring gives an Eulerian picture of the
deep convection process, sampling a 100-km diameter
region of ocean as eddies sweep fluid past it. It thus
complements the wider range of scales sampled by the
drifting floats. The potential temperature field for
October 1996–May 1997 (Fig. 23) shows the build up
FIG. 20. Upper panel shows a sequence of three temperature
profiles at 5-day intervals from a VCM float. Lower panels are
4-day time series of vertical velocity from the intervals between
the profiles taken near 380-m depth. Thick lines indicate directly
measured velocities and thin lines are a confirming measurement
from a simple model balancing vertical drag and float buoyancy.
Deepening of the quasi-mixed surface layer and vigorous verti-
cal motion indicate this was a period of active convection (cour-
tesy of R. Davis).

2055
Bulletin of the American Meteorological Society
of warm, saline, buoyant fluid about January, when
convective deepening arrives at the top instrument (96-
m depth). The buoyancy barrier of the upper few hun-
dred meters is slowly eroded, and as the
air–sea heat flux intensifies, the weaker
deep stratification is rapidly broken
through. The weak winter of 1995–96
left a large amount of heat and freshwater
in the upper ocean, but the much stron-
ger winter cooling of 1996–97 was able
to reach great depth. Convection reached
1500 m in waters passing the mooring,
though the average depth of the mixed
layer is closer to 1000 m.
The water column at the mooring
was homogeneous down to 1000 m for
only about two weeks in late March 1997,
suggesting that convection is indeed in-
homogeneous. Direct measurement of
convective plumes by the ADCP showed
vertical velocities exceeding 10 cm s−1.
The short duration of the strong down-
welling (~2–4 h) suggests that the width
of the features advecting past is between
200 and 800 m. Restratification of the
water column occurs deeply and quickly
after winter, the freshwater invading near
the surface and saline, warm water at
all depths below; it reflects the inhomo-
geneity of the convection, though the
scale of that inhomogeneity is not yet
certain.
Vertical velocities were also recorded
on other moorings. For example, the K1
ADCP (see Fig. 9) over the depth range
440–700 m shows plumes of 3–6 cm s−1
during 12–22 February and again during
3–9 March. While during the first period,
significant temperature fluctuations in
the water column remained above the
ADCP, homogeneity was fairly com-
plete during the second period. During
the Meteor cruise in summer 1997, the
array was redeployed, this time moving
the tomography transceiver from K3 to
K2, in order to concentrate on the area
that had shown the most convective ac-
tivity during the 1996–97 winter Knorr
cruise. Some of the lines covered with to-
mography were also designed to coin-
cide with the inverted echo sounder
sections, with the hope of combining transport esti-
mated from the latter with tomography heat content.
FIG. 21. Data from Lagrangian floats deployed in the Labrador Sea over a pe-
riod of ~1 month starting in mid-February 1997. Continuous convection to 800 m
with deeper convection to 1000 m is evident (courtesy of E. d’Asaro).
FIG. 22. Temperature profiles as measured by a RAFOS float. The first six
profiles were made at 3-day intervals, the last two profiles were at 5-day inter-
vals. The dots are individual temperature–pressure measurements while the thick
line is from a CTD cast taken 3 days prior to the first profile. The steady cooling
and gradual deepening of the mixed layer can be clearly seen (courtesy of M.
Prater).

2056
Vol. 79, No. 10, October 1998
5. Summary and future plans
The dataset that has been gathered in the Labrador
Sea far exceeds previous efforts to observe the con-
vective process anywhere in the World Ocean, both
in the scope and range of techniques employed and be-
cause of the presence of the contemporaneous meteo-
rological component. It is giving us unprecedented
insights into the meteorology and oceanography of the
Labrador Sea, a region known to have direct influence
on the processes that control global climate. The me-
teorology and oceanography are interesting in their
own right, but perhaps the most exciting aspect of the
experiment is the glimpse it is giving us of the inter-
play and interaction between the two fluids.
Our experiment was the first in which direct mea-
surements of the fluxes of heat, momentum, and mois-
ture were taken across the air–sea interface in the
Labrador Sea. Preliminary results suggest that model-
derived estimates of these fluxes are significantly
higher than our direct measurements.
The pronounced observed spatial pattern in the
precipitation (virtually none near the ice edge and in-
creasing with downwind distance from the edge) is
equivalent to salt being injected in to the ocean near
the edge and removed downwind. How important
is this pattern of salinity flux on deep ocean convec-
tion? Based on atmospheric data from
the ship and typical upper-ocean condi-
tions, we found that the direct effect of
heat fluxes on the ocean buoyancy was
about five times the salinity effects of
evaporation. Precipitation had a similar
but opposite (stabilizing) effect. Both the
pattern of heat fluxes (maximum near the
ice edge) and the salinity fluxes (maxi-
mum evaporation near ice edge, maxi-
mum precipitation downwind of ice
edge) would concentrate the destabiliz-
ing forcing of the ocean on the upwind
(here the western) side of the Labrador
Sea, as indicated schematically in Fig.
10. In addition, a potentially important
mechanism in the overall surface buoy-
ancy in the area, but one that we are un-
able to quantify or model, must be
surface meltwater from the Labrador ice
pack. We observed ice close to the pack
melting rapidly in the 3–4°C sea surface
temperatures.
Much is being revealed about the re-
sponse of the ocean by our combination of Eulerian
and Lagrangian measurements. Many questions are
being raised by the data. Are the geostrophic scales as
important in the vertical transfer of heat—balancing
a significant fraction of its loss from the surface—as
initial estimates from float data suggest? How do the
observed vertical velocity and temperature fluctuations
vary in space and time? Existing models of dynamics
on the plume scale predict how such properties are re-
lated to surface forcing and the general stratification.
A knowledge of these fields over the observational
area will allow both qualitative and quantitative tests
of these theories. Do theories explain the origin of the
observed intrusive T/S variability? This would seem
a critical indicator of the process by which convective
stirring eventually leads to mixing. Do theories pre-
dict the substantial difference in the timescale of ver-
tical velocity and temperature fluctuations? Since
floats are neither Eulerian nor truly Lagrangian, this
will likely require tracking model floats through the
density and velocity fields of plume-scale dynamical
models, as in Harcourt et al. (1997), in order to gener-
ate time series that can be compared to those observed.
The experiment has profound implications for the
representation of convection in large-scale models.
Parameterizations of convection widely used in large-
scale ocean models remain stubbornly one-dimen-
FIG. 23. The potential temperature (°C) field for October 1996–May 1997 at
the Bravo mooring (see Fig. 9) showing the build up of warm (and saline) buoy-
ant fluid until January, the subsequent deepening in excess of 1500 m, and then
restratification in May (courtesy of P. Rhines).

2057
Bulletin of the American Meteorological Society
sional and bear little relation to the phenomenology
we are observing in the field and in our process mod-
els of convection (both laboratory and numerical). The
detailed description of the water mass transformation
process that is emerging emphasizes the pervasiveness
of the quasi-horizontal stirring processes that preex-
ist in the ocean and are rapidly energized as convec-
tion proceeds. These lateral stirring processes may be
the ones that ultimately lead to mixing.
The second phase of the field experiment, during
January 1998, repeated a subset of the observations
described here and observed the mixed layer deepen-
ing down to 600 m or so. Moreover, new technology
in the form of self-propelled autonomous underwater
vehicles carried out fine-resolution surveys of the
convecting mixed layer akin to atmospheric bound-
ary layer aircraft flights. These, together with more
conventional observations, may give us a clearer view
of the structure of the convective elements themselves.
Acknowledgments. The LSCE was supported by ONR,
NOAA, and NSF in the United States, by DFG and BMBF in
Germany, and by DFO and the Panel for Energy R&D in Canada.
Access to the NCEP–NCAR Reanalysis products was provided
by the Climate Diagnostics Center of NOAA. ECMWF also made
available their analysis products to us and helped us ensure that
our in situ measurements got onto the GTS. We should like to
acknowledge, in particular, the unstinting support of Dr. Manuel
Fiadeiro of ONR, whose interest and enthusiasm for this project
made it possible.
References
Clarke, R. A., and J.-C. Gascard, 1983: The formation of Labra-
dor Sea water. Part 1: Large-scale processes. J. Phys.
Oceanogr., 13, 1764–1788.
Dickson, R., J. Lazier, J. Meinke, P. Rhines, and J. Swift, 1996:
Long-term coordinated changes in convective activity of the
North Atlantic. Progress in Oceanography, Vol. 38, Pergamon
Press, 241–295.
Fischer, K. W., S. Legg, W. H. Munk, R. A. Shuchman, R. W.
Garwood, and J. P. Palshook, 1998: Modeled Radar Surface
Signature of Deep Ocean Convection. IEEE Trans. Geosci.
Remote Sens., in press.
Garwood, R. W., Jr., 1991: Enhancements to deep turbulent en-
trainment. Deep Convection and Deep Water Formation in the
Oceans, P. C. Chu and J. C. Gascard, Eds., Elsevier Science,
197–213.
——, S. Isakari, and P. Gallacher, 1994: Thermobaric convection.
The Polar Oceans and Their Role in Shaping the Global En-
vironment, Geophys. Monogr., No. 85, Amer. Geophys. Union,
199–209.
Harcourt, R., L. Jiang, and R. Garwood, 1997: Numerical simula-
tion of drifter response to Labrador Sea convection. ONR
Naval Postgraduate Rep., 110 pp.
Hodur, R. M., 1997: The Naval Research Laboratory’s Coupled
Ocean–Atmosphere Mesoscale Prediction System (COAMPS).
Mon. Wea. Rev., 125, 1414–1430.
Hurrell, J. W., 1995: Decadal trends in the North Atlantic oscilla-
tion regional temperatures and precipitation. Science, 269,
676–679.
Joly, A., and Coauthors, 1997: The Fronts and Atlantic Storm
Track Experiment (FASTEX): Scientific objectives and
experimental design. Bull. Amer. Meteor. Soc., 78, 1917–
1940.
Jones, H., and J. Marshall, 1993: Convection with rotation in a
neutral ocean: A study of open-ocean deep convection. J. Phys.
Oceanogr., 23, 1009–1039.
Julien, K., S. Legg, J. McWilliams, and J. Werne, 1996: Penetra-
tive convection in rapidly rotating flows: Preliminary results
from numerical simulation. Dyn. Atmos. Oceans, 24, 237–
249.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year
Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.
Legg, S., and J. Marshall, 1993: A heton model of the spreading
phase of open-ocean deep convection. J. Phys. Oceanogr., 23,
1040–1056.
——, J. McWilliams, and J. Gao, 1998: Localization of deep ocean
convection by a mesoscale eddy. J. Phys. Oceanogr., 28, 944–
970.
Levitus, S., J. I. Antonov, and T. P. Boyer, 1994: Interannual vari-
ability of temperature at a depth of 125 meters in the North
Atlantic Ocean. Science, 266, 96–99.
Lilly, J., P. B. Rhines, M. Visbeck, R. Davis, J. R. N. Lazier, F.
Schott, and D. Farmer, 1998: Observing deep convection in
the Labrador Sea during winter 1994–95. J. Phys. Oceanogr.,
in press.
Marshall, J., and F. Schott, 1998: Open ocean deep convection:
Observations, theory and models. Rev. Geophys., in press.
——, C. Hill, L. Perelman, and A. Adcroft, 1997: Hydrostatic,
quasi-hydrostatic, and nonhydrostatic ocean modeling. J.
Geophys. Res., 102 (C3), 5733–5752.
Maxworthy, T., and S. Narimousa, 1994: Unsteady, turbulent. con-
vection into a homogeneous, rotating fluid, with oceanographic
applications. J. Phys. Oceanogr., 24, 865–887.
MEDOC Group, 1969: Observations of deep-water formation in
the northwestern Mediterranean Sea. Nature, 227, 1037–
1040.
Pickart, R. S., and W. M. Smethie Jr., 1998: Temporal evolution
of the Deep Western Boundary Current where it enters the
subtropical domain. Deep-Sea Res., in press.
Renfrew, I. A., G. W. K. Moore, T. Holt, and S. Chang, 1998:
Mesoscale forecasting during a field program: Meteorologi-
cal support of the Labrador Sea Deep Convection Experiment.
Bull. Amer. Meteor. Soc., in press.
Reverdin, G., D. Cayan, and Y. Kushnir, 1997: Decadal variabil-
ity of hydrography in the upper North Atlantic 1948–1990. J.
Geophys. Res., 102, 8505–8531.
Rogers, J. C., 1990: Patterns of low-frequency monthly sea level
pressure variability (1899–1986) and associated wave cyclone
frequencies. J. Climate, 3, 1364–1379.
Schott, F., M. Visbeck, U. Send, J. Fischer, L. Stramma, and
Y. Desaubies, 1996: Observations of deep convection in the
Gulf of Lions, northern Mediterranean, during the winter of
1991/92. J. Phys. Oceanogr., 26, 505–524.

2058
Vol. 79, No. 10, October 1998
Send, U., F. Schott, F. Gaillard, and Y. Desaubies, 1995: Observa-
tion of a deep convection regime with acoustic tomography.
J. Geophys. Res., 100 (C4), 6927–6941.
Smith, S. D., 1988: Coefficients for sea surface wind stress, heat
flux and wind profiles as a function of wind speed and tem-
perature. J. Geophys. Res., 93 (C12), 15 467–15 472.
van Loon, H., and J. C. Rogers, 1978: The seesaw in winter tem-
peratures between Greenland and Northern Europe. Part 1.
Mon. Wea. Rev., 106, 296–310.
Vautard, R., 1990: Multiple weather regimes over the North At-
lantic: Analysis of precursors and successors. Mon. Wea. Rev.,
118, 2056–2081.
Visbeck, M., J. Marshall, and H. Jones, 1996: Dynamics of iso-
lated convective regions in the ocean. J. Phys. Oceanogr., 26,
1721–1734.