early spring oceanic heat fluxes and mixing observed from drift stations north of svalbard*

Early Spring Oceanic Heat Fluxes and Mixing Observed from Drift StationsNorth of Svalbard*

ANDERS SIREVAAG

Geophysical Institute, University of Bergen, and Bjerknes Centre for Climate Research, Bergen,

and University Centre in Svalbard, Longyearbyen, Norway

ILKER FER

Geophysical Institute, University of Bergen, and Bjerknes Centre for Climate Research, Bergen, Norway

(Manuscript received 24 October 2008, in final form 16 June 2009)

ABSTRACT

From several drifting ice stations north of Svalbard, Norway, observations were made in early spring of the

ocean turbulent characteristics in the upper 150 m using a microstructure profiler and close to the under-ice

surface using eddy correlation instrumentation. The dataset is used to obtain average heat fluxes at the ice–

water interface, in the mixed layer, across the main pycnocline, as well over different water masses in the

region. The results are contrasted with proximity to the branches of the warm and saline Atlantic water

current, the West Spitsbergen Current (WSC), which is the main oceanic heat and salinity source both to the

region and to the Arctic Ocean. Hydrographic properties show that the surface water mass modification is

typically due to atmospheric cooling with relatively less influence of ice melting. Surface heat fluxes of O(100)

W m22 are found within the branches of the WSC and over shelf areas with elevated levels of mixing due to

strong tides. Away from the shelves and WSC, however, ocean-to-ice turbulent heat fluxes are typical of the

central Arctic. Deeper in the water column, entrainment from below together with equally important hori-

zontal advection and diffusion increase the heat content of the mixed layer and contribute to the heat flux

maximum in the upper layers. The results in this study emphasize the importance of mixing along the

boundaries, over shelves, and topography for the cooling of the Atlantic water layer in the Arctic in general,

and for the regional heat budget, hence the ice cover and cooling of the WSC north of Svalbard, in particular.

1. Introduction

At high latitudes, air–sea interaction is affected by the

presence of sea ice, which partly insulates the upper

ocean from solar radiation and wind stress. A significant

source of oceanic heat flux is thus the upward turbulent

mixing and entrainment of heat from warmer layers un-

derlying the cold well-mixed layer below the ice (Aagaard

et al. 1987). The West Spitsbergen Current (WSC) is the

northernmost extension of the Norwegian Atlantic Cur-

rent and carries warm and saline Atlantic water (AW).

North of Svalbard, Norway, where several drift stations

reported here are occupied, is the site where the WSC

enters the Arctic. In this area, the exchange of heat

between the ocean and the ice/atmosphere plays an

important role in modifying the core of the AW and

thickness of the ice cover (Untersteiner 1988).

Fram Strait, the passage between Greenland and

Spitsbergen, Norway, is a site of dramatic exchange of

mass, heat, and salt. On the eastern side, AW is carried

northward by the WSC and on the western side, the East

Greenland Current brings colder and less saline water

southward. The WSC transports a major fraction of the

heat and salt supplied to the Arctic Ocean (Rudels et al.

2000; Saloranta and Haugan 2001). Primarily confined to

the upper continental slope along Svalbard, the WSC

bifurcates into two branches around 798N where it meets

the Yermak Plateau (YP; Fig. 1) (Bourke et al. 1988;

Manley 1995). The Svalbard branch continues eastward,

inshore of the plateau and along the upper part of the

slope between the 400- and 500-m isobaths. The Yermak

* Bjerknes Centre for Climate Research Publication Number

A232.

Corresponding author address: Anders Sirevaag, Bjerknes Cen-

tre for Climate Research, Allegt. 70, 5007 Bergen, Norway.

E-mail: [email protected]

VOLUME 39 J O U R N A L O F P H Y S I C A L O C E A N O G R A P H Y DECEMBER 2009

DOI: 10.1175/2009JPO4172.1

� 2009 American Meteorological Society 3049

branch flows along the offshore western flank of the

YP following the 1000-m isobath. While a fraction of the

Yermak branch rejoins the Svalbard branch and enters

the Arctic Ocean, another fraction detaches north of

808N and turns westward, contributing to the recirculat-

ing Atlantic water (Bourke et al. 1988). Roughly, 20%

and 35% of the Atlantic water in the region is carried

by the Yermak and Svalbard branches, respectively,

whereas about 45% recirculates (Manley 1995). Fur-

thermore, the Yermak branch loses significant heat be-

cause of intense tidal mixing at the YP (Padman and

Dillon 1991; Padman et al. 1992).

The path of the WSC branches includes areas of in-

tense air–sea ice interaction where the warm core sup-

plies heat to the mixed layer, leading to surface heat

fluxes of O(100) W m22 (Aagaard et al. 1987; Dewey

et al. 1999). As the WSC progresses into the Arctic, it

loses its heat and surface fluxes are subsequently lower:

0–37 W m22 reported at 328E (Wettlaufer et al. 1990).

The objective of this study is to quantify the oceanic

heat flux toward the ice west and north of Spitsbergen at

varying proximity to the warm AW core and over

varying topography. During several drifting experi-

ments performed in early spring (within 1 April–1 May)

in the area 78.88–828N, 08–15.48E (Fig. 1), the turbulent

heat fluxes were either measured directly in the mixed

layer by an eddy correlation method or inferred from

dissipation measurements from a microstructure pro-

filer in the upper 150 m. The individual drifts were lo-

cated within, close to, or away from the branches of the

WSC and due to their short durations provide only

a snapshot of the early spring conditions in this area. The

measurements are from separate years, however at the

same time of year, and span the period from 2003 to 2007

(Table 1).

An overview of the measurement locations along with

the instruments and the data reduction details is pre-

sented in section 2. In section 3, a general description of

each drift is given together with salient observations.

Oceanic turbulent fluxes are presented and discussed in

section 4, followed by a summary and concluding re-

marks in section 5.

FIG. 1. Map of the study area showing the locations and drift trajectories of the six individual drifts

using TIC instrumentation (squares) and MSS instrumentation (circles) plotted over the bathymetry

around Svalbard. Bathymetry is from the General Bathymetric Chart of the Oceans (GEBCO) dataset

(IOC, IHO, and BODC 2003) and isobaths are in meters. Wide arrows indicate the WSC, the Svalbard

branch, the Yermak branch, and the recirculation branch. The thinner arrow indicates recirculation

across the Yermak Plateau (Bourke et al. 1988; Saloranta and Haugan 2001).

3050 J O U R N A L O F P H Y S I C A L O C E A N O G R A P H Y VOLUME 39

2. Measurements and methods

a. Drift stations and sampling

Geographically, the drift stations are located in the

Fram Strait, over the eastern flank of the Yermak Pla-

teau and on the shelf north of Svalbard, all occupied

within the period 1 April–1 May of a given year. Upper

ocean turbulence measurements were conducted using

two separate systems: (i) direct eddy correlation mea-

surements with a turbulence instrument cluster (TIC)

and (ii) dissipation measurements in the upper 150 m

using a microstructure profiler (MSS). The TIC is

equipped with an acoustic 5-MHz Sontek/YSI Acoustic

Doppler Velocimeter, an SBE04 standard conductivity

sensor, an SBE03 fast-response temperature sensor, and

an SBE07 microconductivity sensor, all manufactured

by SeaBird Electronics. All TIC sensors sample at the

same vertical level, providing three-dimensional veloc-

ity, temperature, and salinity at 2 Hz, resolving turbu-

lence well into the inertial subrange (McPhee 2002; Fer

et al. 2004; McPhee 2008). MSS, on the other hand, is

a loosely tethered free-fall profiler equipped with precision

CTD sensors and a suite of turbulence sensors including

two shear probes, a fast-response thermistor Fasttip probe

(FP07), and a microconductivity sensor (Fer 2006; Fer and

Sundfjord 2007). All MSS profiles were collected from the

research vessels moored to the ice floe, whereas TIC de-

ployments were made from hydroholes in the ice, more

than two ship lengths away from the vessels. The MSS

profiler was deployed using a manual winch equipped with

200 m of cable in total, limiting the total profiling depth.

A location map of the drifts, together with a sketch of

the WSC, is shown in Fig. 1. In the Fram Strait, two drift

stations (FSD) were occupied in deep water in 2005

(FSD1 and FSD2) by R/V Lance. In total, 51 and 26 MSS

profiles were collected for about 13- and 10-h durations

of FSD1 and FSD2, respectively. The Yermak Plateau

Drift (YPD) was conducted in 2003 during the ARK

XIX/1 cruise of R/V Polarstern. The sampling included

two TIC systems, deployed 4 m apart vertically at 1 and

5 m below the ice over a duration of 6 days. On the shelf

north of Svalbard, the Whaler’s Bay Drift (WBD) in

2003 (ARK XIX/1) was on the Svalbard branch of the

WSC, whereas the two drifts on the Norwegian Bank

(NBD1 in 2005 and NBD2 in 2007) were over shallow

and colder water, both occupied by R/V Lance. Sampling

consisted of TIC measurements at 1 and 5 m at WBD, at

1 m at NBD2, and MSS profiles at NBD1. A summary of

the sampling at each drift is tabulated in Table 1.

Ancillary data include atmospheric measurements

and navigation data from the vessels. In 2003 and 2007,

TIC measurements were supplemented by shipboard

CTD profiles. As a part of field work exercise, Univer-

sity Centre in Svalbard (UNIS) students conducted

various sampling during the R/V Lance 2005 cruise.

Here we use data from ice cores and ship ADCP, col-

lected at FSD1 and FSD2 (F. Nilsen 2005, personal

communication).

b. Data reduction

TIC time series are segmented into 15-min intervals and

velocity data are rotated into a streamline coordinate

system. The chosen averaging period of 15 min for the

turbulence statistics preserves the covariance in the tur-

bulent eddies (typically of several minutes in the under-ice

layer) while excluding contamination from variability at

longer time scales. To calculate turbulent fluxes, deviatory

velocity (u9, y9, w9), temperature T9, and salinity S9 are

obtained by removing the linear trend fitted to the actual

time series at each 15-min interval. Fluxes are obtained by

zero-lag covariances assuming eddies advected past the

TABLE 1. Details of the ice drift stations.

Drift WBD YPD FSD1 FSD2 NBD1 NBD2

Platform R/V Polarstern R/V Polarstern R/V Lance R/V Lance R/V Lance R/V Lance

Year 2003 2003 2005 2005 2005 2007

Start time (doy)a 91.51 100.81 118.53 120.46 121.95 113.73

Lat, start 8N 80.429 81.789 78.780 79.888 80.244 80.347

Lon, start 8E 12.817 9.491 0.531 3.603 15.402 14.920

Drift length (km)b 10.05 38.20 11.97 17.54 4.53 27.55

Duration (h) 18.0 157.1 14.1 9.5 8.7 25.5

No. of profilesc 2 4 54 26 29 6

Drift speed (m s21) 0.16 0.07 0.24 0.52 0.06 0.30

Dml (m) 19 70 21 6 5 10

Dp (m) 28 83 51 19 14 27

a Day of year with 1200 UTC 1 Jan 5 1.5.b Cumulative length derived from navigation data.c Number of CTD profiles for 2003 and 2007, and number of MSS profiles for 2005.

DECEMBER 2009 S I R E V A A G A N D F E R 3051

sensors over the averaging time are representative of the

ensemble of instantaneous turbulent fields (Taylor’s fro-

zen turbulence hypothesis). Turbulent heat flux is calcu-

lated as FH 5 rCPhw9T9i, in units of watts per meters

squared, where r is the seawater density and CP is the

specific heat of seawater. In the following, deviatory tur-

bulent quantities are denoted by primes and angle

brackets denote averaging in time or space. The Reynolds

stress per unit mass is t 5 hu9y9i 1 ihy9w9i, expressed in

complex notation, where i 5 (21)1/2, and the local friction

speed is u*5 jtj1/2. The term ‘‘local’’ indicates fluxes

obtained at the measurement depth and must be dis-

tinguished from the theoretically calculated ice–ocean

interface fluxes, which are denoted with subscript 0.

MSS microstructure profile data are processed as de-

scribed elsewhere (Fer 2006; Sundfjord et al. 2007). Ini-

tial 1024-Hz sampling is averaged over four data points

(to 256 Hz) to reduce noise. The dissipation rate of tur-

bulent kinetic energy (TKE) per unit mass « in units of

watts per kilogram is calculated using the isotropic re-

lation « 5 7.5nhu92z i, where n is the viscosity of seawater,

hu92z i is the shear variance of the horizontal small-scale

velocity, and angle brackets denote appropriate averag-

ing. The instrument fall speed (typically ;0.7 m s21) is

used to convert from frequency domain to vertical

wavenumber domain using Taylor’s hypothesis, and the

shear variance is obtained by iteratively integrating

the reliably resolved portion of the shear wavenumber

spectrum of half-overlapping 1-s segments. Narrowband

noise peaks induced by the probe guard cage are above

the wavenumber range chosen for the analysis. Dissi-

pation data in the upper 8 m are unreliable owing to

disturbances from the ship and the initial adjustment to

free fall. The profiles of precision CTD and « are pro-

duced as 10- and 50-cm vertical averages, respectively.

The diapycnal eddy diffusivity of mass is calculated

assuming a balance between the production of TKE,

its dissipation, and the buoyancy flux as Kr 5 G«N22

(Osborn 1980), using the flux coefficient (the ratio of

work done against gravity to energy dissipated by fric-

tion) G 5 0.2. The choice of G is a major uncertainty;

however, it follows common practice and allows for di-

rect comparison with other studies. Both observational

evidence (Moum 1996) and direct numerical simulation

results (Smyth et al. 2001) show significant variability.

According to a parameterization that accounts for the

evolution over the lifetime of an overturn (Smyth et al.

2001), to employ G 5 0.2 will underestimate the median

Kr by a factor of 2, whereas according to the studies by

St. Laurent and Schmitt (1999) and Arneborg (2002) it will

overestimate Kr by about a factor of 2 in shear-induced

and patchy turbulence. The buoyancy frequency, N 5

[2g/r(›r/›z)]1/2, is approximated using Thorpe-ordered

su profiles (Thorpe 1977), derived from the precision

sensors. The density gradient is obtained as the slope of

the linear regression of depth on su, over 4-m-length

moving segments (i.e., typically over 40 data points).

The background temperature and salinity gradients are

derived similarly from the precision temperature and

salinity records as the slope of the linear regression of

depth on T and S. As inferred from the Osborn model

Kr is ill defined in the absence of stratification. We

therefore excluded the segments where the temperature

gradient or the density gradient was not significantly

different from zero (identified as twice the uncertainty of

the slope inferred from the regression described above).

An independent estimate of eddy diffusivity for heat

is obtained using KT 5 3kTCx (Osborn and Cox 1972),

using the temperature microstructure sampled by the

pre-emphasized FP07 signal. Here, Cx 5 h(›T9/›z)2i/h›T/›zi2 is the Cox number, kT 5 1.4 3 1027 m2 s21 is

the molecular diffusivity for heat, h(›T9/›z)2i is the

small-scale temperature gradient variance, and h›T/›ziis the background temperature gradient. We estimate the

dissipation rate of thermal variance, x 5 2kTh3(›T9/›z)2i,by fitting the Batchelor’s form (Dillon and Caldwell 1980)

constrained by the measured « to the resolved wave-

number band of the temperature gradient spectrum over

4-m-length segments. The vertical wavenumber spec-

trum of the temperature gradient ›T9/›z is obtained as

STz(m) 5 WSTt( f), where STt( f) 5 v2ST( f), v 5 2pf,

W is the fall rate of the profiler, m 5 f/W is the cyclic

wavenumber, and the temperature spectrum ST is ob-

tained using the transfer function for the ideal amplifi-

cation of the circuit. The resolved wavenumber band

corresponds to 1–30 Hz [1.4–43 cycles per meter (cpm)

using the nominal fall rate of 0.7 m s21].

CTD profiles, either from the microstructure profiler

with 0.5-m vertical resolution or from the ship CTD

system with 1.0-m vertical resolution, are used to de-

termine the mixed layer depth (Dml) and the pycnocline

depth (Dp). The Dml is determined using the split-and-

merge algorithm of Thomson and Fine (2003), where

linear fits are made to individual segments of the density

profile and the range of the uppermost, vertical fit de-

fines the extent of the mixed layer; Dp is determined as

the depth of maximum N2 using 5-m smoothed buoy-

ancy frequency profiles.

c. Ice drift velocity and under-ice surfacefriction speed

Ice drift velocity is derived from the 5-min averages of

navigation data collected at high temporal sampling

(10 s in R/V Lance, 60 s in R/V Polarstern). For each

drift station, latitude–longitude pairs were converted to


a polar stereographic grid centered at the mean position

of the drift, with a positive y axis aligned with north. The

ice drift velocity vector is obtained as the first difference

of drift positions.

At the TIC stations, direct measurements of friction

velocity u* in the boundary layer are available at 1 m

below the ice. However, u* can be different from the

under-ice surface (ice–ocean interface) friction velocity

u*0, and comparable measurements are not available for

MSS drifts. To be consistent between all drifts, we esti-

mate the surface friction velocity u*0 from a Rossby

similarity drag law (McPhee et al. 1999),

Ui

u*0

51

k[log(Ro*)�A� iB], (1)

where Ui is the ice velocity relative to the upper ocean

velocity, k 5 0.4 is the von Karman constant, Ro* 5 ju*0j/fz0, and A and B are similarity constants. For the TIC

stations, Ui is found from the measured velocity at 1 m,

whereas for MSS stations FSD1 and FSD2, Ui is found

by subtracting the ADCP velocity at 25 m from the ice

drift velocity. At NBD1, no current measurements are

available and u*0 is calculated from ice drift velocity

alone. For all stations, the surface roughness is set to

z0 5 2.0 3 1023 m, which is obtained from the WBD

data using the logarithmic law of the wall combined with

velocity and friction velocity measurements at 5 m be-

low the ice. The measurements from the deeper TIC are

more representative for the overall ice floe and ice in

WBD, because the instruments were deployed in an

area of relatively smooth first-year ice within the sur-

rounding, more heavily ridged multiyear ice and the

measurements at 1 m from the interface may only reflect

the local ice topography. The average value of z0 5 2.0 3

1023 m is slightly smaller than that of the multiyear floe

TABLE 2. Water mass definitions following Aagaard et al. (1985).

Water mass Acronym Temperature Salinity

Polar water PW ,08C ,34.4

Polar Intermediate

Water

PIW ,08C 34.4 , S , 34.7

Arctic Surface Water ASW .08C ,34.9

Atlantic water AW .28C .34.9

Lower Arctic

Intermediate Water

LAIW ,28C .34.9

FIG. 2. Vertical profiles of (a) temperature T and (b) salinity S together with (c) the corresponding T–S diagram,

observed at the indicated drifts. Profiles for FSD1 and FSD2 are averaged over all MSS casts. The average profile at

NBD1 is for the deeper side of the thermohaline front. YPD and WBD profiles are single casts collected before and

after, respectively, the TIC measurements and their T–S curves extend to the bottom (860 m for YPD and 178 m for

WBD). The water masses shown in (c) are after Aagaard et al. (1985) and the acronyms and the T–S characteristics

are summarized in Table 2. The freezing line (near 21.98C) and density anomaly at 0.2 kg m23 intervals are also

shown. The three diagonal yellow dashed lines indicate the trajectory along which water with (T, S) 5 (3.2, 35.1) is

cooled and freshened near the surface by ratios of atmospheric cooling to ice melt, R 5 0, 1, and 2. The ice is assumed

to have a salinity of 5 psu.


during the Surface Heat Budget of the Arctic Ocean

(SHEBA) campaign (McPhee 2002).

3. General description of drifts

a. Hydrography and water masses

The sampling is limited to the upper 150 m and deep

water masses are not observed. We therefore adopt a

relevant subset of the water mass categories defined in

Aagaard et al. (1985) and summarized in Table 2. Rep-

resentative temperature and salinity profiles and T–S

diagrams for each drift are shown in Fig. 2. FSD and

NBD1 profiles are time-averaged MSS profiles over the

duration of the drift, whereas YPD and WBD profiles

are single casts of ship CTD. No prominent cold halo-

cline water is observed, and the mixed layer depth varied

by one order of magnitude between a very shallow 6 m

at FSD2 and about 70 m at YPD. The warm core of the

WSC is composed of warm and saline AW. The Lower

Arctic Intermediate Water (LAIW) is of similar salinity

but colder, and at this location, can be a result of cooling

AW by heat loss to the atmosphere. Arctic Surface Water

(ASW), on the other hand, is both colder and fresher and

can be a result of cooling AW by a combination of heat

loss to the atmosphere and ice melting, which accounts

for the freshening (Boyd and D’Asaro 1994).

On average, profiles from FSD1 and FSD2 show a very

small fraction of LAIW (owing to the shallow sampling)

and we include LAIW in the definition of AW when

averaging properties over water masses (section 4d).

b. Fram Strait Drifts

Fram Strait Drift 1 was established west of the WSC

on the morning of 28 April 2005 (day 118). Ice thickness

FIG. 3. Observations during the Fram Strait Drift, FSD1. Time series of (a) the magnitude of the ice drift velocity

inferred from GPS fixes before each MSS cast, (b) depth distribution of dissipation rate « (color) and isotherms at 0.2-K

intervals (black), and (c) Kr (color) and isohalines at 0.05 intervals (black). Arrowheads mark the start time of each

MSS cast. Evolution of the mixed layer depth is shown by the dashed line.


was about 120 cm over a water depth of about 2500 m

near 798N, close to the prime meridian. The conductive

heat flux in the ice is calculated as FHi 5 2ki(›T/›z)i,

where ki is the thermal conductivity of sea ice and (›T/›z)i

is the temperature gradient in the ice. Temperature gra-

dient measured from an ice core yielded FHi 5 10 W m22

for FSD1. Microstructure profiling started at 1254 UTC,

and 54 casts were made at a typical sampling interval of

10–15 min.

The patch of open water on the side of the vessel

where the profiling was made froze throughout the drift.

Occasionally, profiling was interrupted because of ridging

and rafting, and after about 14 h the sampling was ter-

minated as the slack tether was unable to penetrate the

young thin ice, making it impossible to conduct free-fall

profiling. A malfunction of the automated logging of

the ship’s meteorological and navigation data went un-

noticed throughout the drift; however, positions at the

start of each cast and occasional descriptive meteoro-

logical information were hand-logged by the authors.

Air temperature was around 2138C throughout the

drift. The first couple of hours into the profiling, the wind

was calm with a speed of 2 m s21, then decreased prac-

tically to nil, and followed with a light breeze of about

3 m s21 from day 118.0. Despite the calm wind, the sta-

tion drifted gradually toward the northwest at an average

speed of 0.24 m s21 over warmer water (Fig. 3a).

Mean profiles of hydrography, dissipation rate of

TKE, and diapycnal eddy diffusivity Kr averaged over

all casts are shown in Fig. 4. On average, the mixed layer

was 21 m deep, with temperature 0.19 K above the

freezing point. Dissipation rate, in excess of 3 3 1027

W kg21 in the mixed layer, reduced to the noise level

(2 3 1028 W kg21) below about 60-m depth. Corre-

sponding eddy diffusivity Kr inferred from the Osborn

model decreases from O(1022) m2 s21 at the base of the

mixed layer to O(1024) m2 s21 below the pycnocline. The

time-averaged velocity profile derived from the ship

ADCP showed that the speed was about 15 cm s21 down

to 40 m, gradually decreased to about 8 cm s21, and

stayed nearly constant below 70 m (Fig. 5a). The angular

shear (change of current direction with depth) was sig-

nificant in the upper 70 m, and despite the strong stratifi-

cation, the 4-m gradient Richardson number [Ri 5 N2/Sh2,

where Sh2 5 (›u/›z)2 1 (›v/›z)2 is the shear squared] was

often below unity, suggesting that the enhanced « between

40 and 70 m (Fig. 4c) can be due to local shear instabilities.

Fram Strait Drift 2 was established on the morning of

30 April 2005 (day 120) on a relatively thin floe of about

30 cm. In contrast to the freezing conditions at FSD1,

repeated ice core measurements and mixed layer tem-

peratures indicated that the ice at FSD2 was melting

rapidly. The melting rate could not be determined with

confidence because of the thin ice floe and scarcity of ice

FIG. 4. Profiles averaged over the first 51 MSS casts during FDS1. (a) Temperature (black)

and salinity (gray), (b) potential density anomaly su (black) and buoyancy frequency N (gray),

(c) 2-m bin-averaged dissipation of TKE «, and (d) eddy diffusivity using Kr 5 0.2«N22. Vertical

gray line in (c) is the noise level for « (52 3 1028 W kg21). The gray trace in (d) is the corre-

sponding noise level for Kr using the mean N profile and ignoring segments with N , 1.5 cph.


cores. Conductive heat flux, averaged over two ice cores,

was 15 W m22. Air temperature gradually warmed from

2128 to 238C throughout the drift when wind averaged

6.4 m s21 (Figs. 6a,b). The mixed layer was shallow and

the upper layer temperatures gradually increased. The

time-averaged velocity profile derived from ship ADCP

was nearly unidirectional at 2408 in the upper 25–170 m

with the speed between 35 and 40 cm s21. Dissipation

rate and eddy diffusivity were significantly larger when

compared to FSD1, however they similarly reached the

noise level below about 60 m (Fig. 7). Compared to

FSD1, the stratification was weaker; nevertheless, owing

to the lack of shear, the 4-m Ri was one order of mag-

nitude larger (Fig. 5). Relatively enhanced mean dissi-

pation at around 50 m was largely a result of the event

that occurred toward the end of the drift, following the

pinching of the isotherms and isohalines (Figs. 6c,d).

Using a free-drift model (McPhee 1986; see also Boyd

and D’Asaro 1994) and representative values of the air–

ice (Ca) and ice–water (Cw) drag coefficients, relative ice

speed will be about 2.5% of the wind speed, Ui 5 UaraCa/

(rwCw), where Ua is the wind velocity and ra and rw are

the air and water densities, respectively. The actual ice

motion can be obtained by adding the average upper

ocean velocity. In FSD2, the residual between average

ice velocity inferred from the navigation (absolute ice

velocity) less the 2.5% of the mean wind speed yields a

0.35 m s21 average upper ocean velocity, which is con-

sistent with ship ADCP measurements. Also, applying

the same model and equating the ice–water stress and

the air–ice stress (calculated from absolute ice and wind

velocities) yields a friction velocity at the ice–ocean in-

terface that corresponds well in magnitude with that

derived from the Rossby similarity drag law [Eq. (1)].

c. Yermak Plateau Drift

The Yermak Plateau Drift, established on day 100 in

2003, had the largest duration with 6 days of sampling.

Camp was established on a large floe of 2–3-m thick,

multiyear ice; however, TIC instrumentation was deployed

through thinner ice (;0.5 m) at the edge of the larger

floe. The upper 70 m was well mixed in temperature and

salinity (Fig. 2, green curve). The mixed layer temper-

ature was the coldest and the temperature maximum

below the thermocline was the lowest of all stations. The

floe drifted at a slow pace of 7 cm s21, comparable to

NBD1 but about 1/4–1/3 the speed of FSD1 and NBD2.

The location is slightly southeast of the site where upward

FIG. 5. Time-averaged profiles of the absolute horizontal current vectors (toward north

pointing upward) in the upper 150 m, measured by the ship ADCP during (left) FSD1 and

(right) FSD2. Bins are 4-m averages. The 4-m first-differenced shear magnitude is used to-

gether with the 4-m averaged buoyancy frequency to calculate the finescale Richardson

number (lower axis in log-scale).


turbulent heat fluxes reaching 30 W m22, due to ener-

getic mixing events as a result of proximity to internal

wave sources, were reported by Padman and Dillon

(1991). It is also the same area where an automated buoy

measured surface heat fluxes of 22 W m22 over the el-

evated topography of Yermak Plateau, 50 days prior to

our deployment (McPhee et al. 2003).

d. Drifts on the North Spitsbergen shelf

Whaler’s Bay Drift, north of Svalbard, was located

where the continental slope branch of warm and rela-

tively salty AW transported by the West Spitsbergen

Current enters the Arctic. WBD was established on

1 April 2003 on an area of 1.1-m-thick first-year ice within

about twice as thick and more heavily ridged multiyear

ice. An ice core was extracted and ice temperatures

measured at 10-cm vertical resolution yielded a tem-

perature gradient corresponding to an upward conduc-

tive heat flux of about 41 W m22. The relatively high

temperature gradient reflects the low air temperature,

which averaged to 229.18C over the drift, while the

average wind was northeasterly at 9 m s21. Water depth

was about 180 m with a mixed layer depth of 19 m with

temperature 0.93 K above freezing.

Norwegian Bank Drift 1, established 1 May 2005, was

closer to land and drifted across a thermohaline front that

separated the well-mixed shallows of polar water (PW)

characteristics from the deep water of Arctic Surface

Water properties. Air temperature during NBD1 was

relatively warm, varying between seawater temperatures

and zero. The drift started over about 90-m-deep water

and crossed the thermohaline front halfway into the drift.

FIG. 6. Observations during the FSD2. Time series of (a) air temperature, (b) wind speed (black) and ice drift

velocity (red), (c) depth distribution of dissipation rate « (color) and isotherms at 0.2-K intervals (black), (d) Kr

(color) and isohalines at 0.05 intervals (black), and (e) depth from the ship’s echo sounder. Arrowheads mark the

start time of each MSS cast. Evolution of the mixed layer depth is shown by the dashed line.


The shallow side of the front was characterized by un-

usually high dissipation rates (Fig. 8), enhanced both in

the upper and the bottom layers, suggesting strong tidal

currents (section 3e) generating turbulence in the bound-

ary layer near the seafloor and under the ice. The T–S

profiles and stratification averaged on either side of

the front are contrasted in Fig. 9. Shallows are nearly

homogeneous and well mixed. The dissipation rate is at

least one order of magnitude larger on the shallow side of

the front (Fig. 9c) and decreases with distance from the

boundaries toward the midwater. Throughout NBD1,

measured dissipation rates are well above the noise level.

Norwegian Bank Drift 2 was established on 24 April

2007 in the same area as NBD1. Compared to NBD1,

the NBD2 drift was 5 times as fast and over a mixed

layer that was colder and twice as deep (Table 1). The

hydrography from the CTD casts shows a tongue of AW

below about 60 m and a relatively strong salinity gra-

dient at the front (Fig. 10). Average air temperature was

2138C and an average northwesterly wind of 9 m s21

accompanied the drift across the shallow topography of

the Norwegian Bank.

e. Tides

Tidal currents can be strong and influence the oceanic

turbulent fluxes at the Yermak Plateau (Padman and

Dillon 1991; Padman et al. 1992) and over the shelves

north of Svalbard (Sundfjord et al. 2007). We estimate

and contrast the tidal current ellipse parameters for each

drift using the 5-km-resolution Arctic Ocean Tidal In-

verse Model (AOTIM) (Padman and Erofeeva 2004).

Results predicted for the start location of each drift for

the two main constituents M2 and K1 are summarized in

Table 3. Because NBD1 extends into significantly shal-

low water, ellipse parameters for the end location of

NBD1 are also calculated. Generally, model water

depths interpolated to the drift positions agree with the

actual water depth. At the shallow end of NBD1, how-

ever, the model depth is 58 m, about double the actual

depth, and the modeled tidal currents shown in Table 3,

which scale inversely by the water depth, may therefore

underestimate the actual current by a factor of 2. Except

for FSD2, the tidal ellipses have clockwise rotation at all

drifts for both the diurnal and semidiurnal constituents

(negative semiminor axis length). Typically, the semi-

diurnal tide dominates, except from at YPD where the

dominant constituent is K1 and at FSD2 where both

constituents are of comparable magnitude. Tidal cur-

rents at WBD are significant whereas those toward the

end of NBD1 are the largest and will lead to significant

stirring both in the under-ice boundary layer and the

bottom boundary layer near the seabed.

4. Turbulent fluxes

a. Heat flux at the under-ice surface

The under-ice surface (ice–seawater interface) heat

flux can be estimated using the parameterization

FH0

5 rCpc

Hju*0jDT , (2)

FIG. 7. Same as in Fig. 4, but for FSD2 and averaged over 26 casts.


where cH is the turbulent exchange coefficient (Stanton

number), u*0 is the surface friction velocity, and DT is

the mixed layer temperature above freezing. The ex-

change coefficient has been shown to be nearly constant,

cH 5 0.0057, for the under-ice boundary layer (McPhee

et al. 2003). Although the local turbulent heat flux in the

under-ice boundary layer can be directly calculated from

the TIC data at 1 m below the ice (section 2b), the actual

surface flux can be different. The surface friction ve-

locity can be estimated using the Rossby similarity [Eq.

(1)] or it can be approximated using the local u* mea-

sured close to the ice undersurface (typically 1 m) when

available. To be consistent, we calculated u*0 via the

Rossby similarity for all drifts. When compared to 1-m

TIC measurements, ju*0j is 0.99, 0.30, and 1.11 times the

local ju*j for WBD, YPD, and NBD2, respectively. The

under-ice surface heat flux results are compared for all

drifts in Table 4. Also tabulated are the measured heat

fluxes at 1 m below the ice for the TIC drifts and mixed

layer averages for the MSS drifts. The largest inter-

facial heat flux occurred at FSD2, consistent with the

observations of rapid ice melting throughout the drift.

Significantly large upward FH0 was also observed at

WBD on the Svalbard branch of the WSC. Although

the drift speed during NBD2 was 5 times that during

NBD1, u* was only 1.4 times as large, and the mixed

layer temperature above freezing in NBD2 was 0.4

times that during NBD1. Consequently, according to

FIG. 8. Observations during NBD1. Time series of (a) air temperature, (b) wind speed (black) and ice drift velocity

(red), (c) depth distribution of dissipation rate « (color) and isotherms at 0.2-K intervals (black), and (d) Kr (color)

and isohalines at 0.05 intervals (black). Bottom topography (gray) is from the ship’s echo sounder. Occasionally the

profiler was run to the bottom. Arrowheads mark the start time of each MSS cast. Evolution of the mixed layer depth

is shown by the dashed line.


Eq. (2), the interfacial heat flux in NBD2 is only 60% of

that in NDB1 (Table 4). For NBD1, FH0 is calculated

assuming stagnant upper ocean, which, according to

the tidal estimates in section 3e, is probably not the

case.

For the drifts where the conductive heat flux was

calculated using the measured gradient in the ice (FSD1,

FSD2, and WBD), an estimate of melting rate is made

by assuming that any discrepancy between ocean heat

flux and conductive heat flux at the ice–ocean interface

is balanced by melting or freezing (e.g., Steele and

Morison 1993). From this method, melting rates of 0.3, 9,

and 5 cm day21 are estimated for FSD1, FSD2, and

WBD, respectively. Although freezing was observed in

open-water areas during FSD1, the ice in general was

slowly melting as the surface heat flux was larger than

the conductive heat flux in the ice. Accordingly, the salt

fluxes in the mixed layer were positive (but small), which

corresponds to an input of meltwater at the surface

(Table 5). For FSD2, the inferred melting rate of

9 cm day21 is comparable to the observed rapid melt-

ing. The upward (positive) salt fluxes in the mixed layer

were significantly larger than during FSD1, which

qualitatively supports significant melting. For WBD, the

melting rates are in line with the observed salt fluxes

reported in Sirevaag (2009).

b. Heat and salt flux in the mixed layer

The local turbulent heat flux in the under-ice bound-

ary layer is directly calculated from the TIC data as

FH 5 rCPhw9T9i (section 2b). At TIC stations, values

from the cluster at 1 m are averaged for WBD and

NBD2, and those from both levels (1 and 5 m) are

averaged for YPD to obtain representative heat flux

values in the mixed layer. Although WBD had another

cluster at 5 m below the ice, the mean heat flux was

more than double that measured at 1 m (620 and

259 W m22, respectively), with significantly large dif-

ferences after day 92.1 (Fig. 12d). Large values at 5 m

at WBD are likely the consequence of pressure ridge

keels and might be representative of variability in the

mixed layer for irregular under-ice topography. For

MSS stations, turbulent heat and salt fluxes ([1023r 3

salinity flux) are obtained using the eddy diffusivity for

heat from the Osborn–Cox model and the property

gradients (section 2b), both evaluated at 4-m vertical

segments:

FH

5�rCP

KT

dT

dz

� �(W m�2)

FS

5� r

1000K

S

dS

dz

� �(kg s�1 m�2).

(3)

FIG. 9. Average profiles for NBD1. Thin traces are the profiles averaged on the deep side of

the thermohaline front (casts 1–14) whereas the thick traces are on the shallow (well mixed)

side of the front (casts 19–29). (a) Temperature (black) and salinity (gray), (b) potential density

anomaly su (black) and buoyancy frequency N (gray), (c) 2-m bin-averaged dissipation of TKE

«, and (d) eddy diffusivity using 0.2«N22, ignoring segments with N , 1.5 cph.


In the calculations we assume turbulent flow leading to

eddy diffusivities KT ’ KS.

Mixed layer fluxes are obtained by averaging the

values for z , Dml for each cast. MSS-derived eddy

diffusivity is available below 8-m depth [i.e., center of

the first 4-m-long segment used in Eq. (3) is at 10 m],

hence we only retain values at casts where Dml . 10 m.

TIC measurements, on the other hand, are either within

the upper 5 m or 1 m below the ice. Therefore, the

values inferred from MSS (TIC) are biased toward the

lower (upper) part of the mixed layer.

A summary of the mixed layer turbulent heat flux as

measured by the TIC is given in Figs. 11–13 for YPD,

WBD, and NBD2, respectively. At YPD, the mixed layer

temperature had a pronounced diurnal signal (Fig. 11a),

consistent with the strong diurnal tides from AOTIM5

(Table 3) and previous observations of strong diurnal

tides in the region (Padman et al. 1992). Friction velocity

decreased from 1 to 5 m below the ice, suggesting a

mechanical source for TKE generation at the interface.

The average mixed layer heat flux of 2 W m22 was

comparable to the typical Arctic values and compared

favorably with the heat flux at the ice–water interface of

FH0 5 3 W m22 obtained from the parameterization in

Eq. (2). At the WBD, the variation in the measured

FH was large, both in time and between the two mea-

surement levels. The average value at 1 m of 259 W m22

was larger than, but comparable to, the interface value

of 203 W m22 (Table 4). The friction velocity was slightly

larger at 5 m, particularly in the late part of the drift,

suggesting TKE contribution from distant sources. At

NBD2, TIC measured through a gradually less saline

mixed layer and the front seen between the first two CTD

stations in Fig. 10 was absent in the TIC data (Fig. 13a).

Heat flux in the mixed layer increased from ;30 to over

80 W m22 as the floe drifted over warmer water, as much

as 0.45 K above freezing.

The mixed layer was only occasionally deeper than

10 m at FSD2 and NBD1, hence flux estimates in the

mixed layer using Eq. (3) with profiler data from these

sites are scarce and may not be representative of the

drift average. The flux estimate from FSD1, on the other

hand, is well resolved. Despite a mean upward heat flux

of 28 W m22, the open water at the surface was freezing

as a result of low air temperatures increasing the ocean/

air temperature difference and leading to heat loss to the

atmosphere that is greater than the oceanic heat flux.

Larger fluxes of heat at the surface compared to fluxes

across the pycnocline should reduce the total amount of

heat in the mixed layer. Heat content in the mixed layer

Hml is calculated from the temperature above freezing

and the mixed layer depth for each MSS profile, and the

rate of change of the heat content dHml/dt is found as

the slope of a line fitted to the evolution of Hml in time.

For FSD1 and FSD2, heat content in the mixed layer is

reduced at rates equivalent to heat fluxes of 2387 and

2496 W m22, respectively, where the negative fluxes

mean that heat is lost from the mixed layer. As in Steele

FIG. 10. Hydrographic conditions during NBD2 in 2007. Tem-

perature contours (color) at 0.1-K intervals and salinity (black) at

0.05 intervals are shown. Arrowheads at the top mark the SBE–

CTD times and the horizontal line shows the duration of the TIC

sampling. White dashed line is the mixed layer depth. Bathymetry

is from the ship’s echo sounder.

TABLE 3. Tidal current ellipse parameters derived at the drift start positions using the AOTIM5 model (Padman and Erofeeva 2004).

Those at the final position of the NBD1 are also shown.

WBD YPD FSD1 FSD2 NBD1 start NBD1 end

M2 Ma/mi (cm s21)* 8.7/22.1 3.9/20.7 3.2/20.4 4.3/0.4 8.4/20.1 12.6/21.5

Pha/inc (8)** 272/53 240/51 53/85 60/87 308/45 305/43

K1 Ma/mi (cm s21)* 4.8/21.4 8.2/25.1 2.0/20.1 4.1/1.6 2.4/20.4 2.9/20.8

Pha/inc (8)** 48/51 36/56 47/90 37/77 7/23 6/8

* Ma/mi are the semimajor/semiminor axis lengths.

** Pha/inc are the phase and the inclination.


and Morison (1993), we can estimate the contribution to

changes in mixed layer heat content from horizontal

advection and diffusion relative to the drift as the re-

sidual A 5 dHml/dt 2 FH0 1 FHpyc, where FHpyc is the

heat flux at the base of the mixed layer. Average values

of A for FSD1 and FSD2 are 2374 and 2282 W m22,

respectively. Values for FSD1 are biased as the ice

drifted past a weak thermohaline front in the mixed layer

and into gradually colder (and fresher) water until day

;118.7. The contribution from horizontal advection and

diffusion was below 22 kW m22 in this period, whereas

A 5 27 W m22 for the period after day 118.7. During the

first period, the horizontal advection and diffusion is far

more important than surface fluxes in determining the

mixed layer heat content. Significant contribution from

horizontal advection and diffusion was also inferred

northeast of Svalbard by Steele and Morison (1993). For

the latter period, A is of the same order as the surface

fluxes. In the case of FSD2, both the surface fluxes and

the advective fluxes are one order of magnitude larger

than the last period of FSD1, due to the proximity of the

WSC branch. However, both the advective effects and

vertical fluxes are of the same magnitude and seem to be

equally important in modifying the mixed layer heat

content, which decreased steadily throughout the drift.

c. Fluxes across the pycnocline

Time-averaged profiles of turbulent heat and salt

fluxes are shown in Fig. 14 for the profiling depth at each

drift. In this section, fluxes averaged between the base of

the mixed layer and the stratification maximum (Dml ,

z , DP) are presented. Average turbulent parameters «,

Kr, and KT, the buoyancy frequency, and the resulting

fluxes are summarized in Table 5. The stratification

varies between 3 and 5 cph, with the weakest at NBD1,

under the influence of strong mixing with exceptionally

large eddy diffusivities.

The variability in the dissipation rate and the eddy

diffusivity at FSD1 is not significant, leading to a narrow

95% confidence interval on the parameters; however,

the standard deviation on resulting fluxes is comparable

to their mean, owing to the variability in the 4-m prop-

erty gradients. The mean values of Kr and KT for FSD1,

which was located a considerable distance away from

the core of the WSC, are one order of magnitude larger

than typical low-latitude open-ocean pycnocline obser-

vations (Gregg 1987). This increases by one order of

magnitude at FSD2. The dissipation rate is 3–5 times

larger at FSD2. The average current profile at FSD2

lacks mean shear (Fig. 5b). The pycnocline is shallower

and the boundary mixing penetrates to the pycnocline,

leading to entrainment of relative warm deeper water,

which also explains the slowly decreasing heat content

of the mixed layer. In contrast to FSD2, several layers of

stratification maxima are observed at FSD1 (cf. Figs. 4,

6). In these layers where densely spaced isotherms are

separated by relatively mixed layers, the dissipation rate

has to work against the finescale stratification.

d. Fluxes in water masses

Mean profiles of heat and salt fluxes at FSD1, FSD2,

and NBD1 (Fig. 14) show large vertical variability with

a layering structure. In all drifts the fluxes decrease with

distance from the surface in the upper ;20 m. On the

shallow side of the front at NBD1, the heat flux increases

toward the bottom whereas the salt flux remains small in

magnitude but changes sign. Here, horizontal (possibly

isopycnal) exchange with significantly warmer water of

comparable salinity from the deeper side of the front can

TABLE 4. Comparison of the average mixed layer oceanic

conditions and heat flux for all drifts.

WBD YPD FSD1 FSD2 NBD1 NBD2

Tml (8C) 20.56 21.81 21.68 20.30 21.03 21.49

Sml 34.53 34.16 34.19 34.54 34.49 34.37

DT (K) 0.93 0.05 0.19 1.58 0.86 0.37

FH (W m22) 259 2 28 206* 240* 69

ju*0j 31022 (m s21) 0.94 0.23 0.44 0.99 0.59 0.8

FH0 (W m22) 203 3 21 303 116 69

* Few data points.

TABLE 5. Average turbulent parameters in the mixed layer and in the layer between the base of the mixed layer and pycnocline. All

values are inferred from microstructure profiling. The buoyancy frequency, heat flux, and salt flux are given as mean values 6 one standard

deviation. Values of the dissipation rate and the eddy diffusivities are the maximum likelihood estimator and the 95% confidence limits

are given in brackets.

FSD1 FSD2 NBD1 FSD1 FSD2 NBD1

Layer z , Dml Dml , z , DP

N (cph) 1.6 6 1.0 3.6 6 1.3 0.7 6 1.0 4.8 6 0.6 4.4 6 1.4 3.0 6 1.0

« 31028 (W kg21) 23.6 [16 34] 598 [18 19713] 353 5.3 [4.8 5.9] 23.5 [15.1 36.5] 917.3 [290 2900]

Kr 31024 (m2 s21) — — — 3.9 [3 5] 26.1 [12.7 53.7] 3302 [430 25328]

KT 31024 (m2 s21) 80 [37 175] 61 [22 169] 276 3.1 [1.9 4.9] 33 [9 122] 151 [60 379]

FH (W m22) 28 6 44 206 6 29 240 8 6 8 89 6 103 301 6 267

FS 10263 (kg s21 m22) 1.86 6 5.4 15.9 6 8.2 30.9 0.44 6 0.4 4.2 6 5.8 33.1 6 36.1


lead to the observed flux profiles. To assess the turbulent

exchange between different water masses, we average the

fluxes over the extent of each water mass. All 4-m seg-

ments from the MSS casts with heat and salt fluxes in-

ferred from the Osborn–Cox model (section 2b) are used

to identify the water masses (Table 2) using the segment

mean temperature and salinity. Average values of strat-

ification, dissipation rate, inferred diffusivity, and turbu-

lent fluxes are given for the MSS drifts (Table 6). Figure 15

shows the average heat fluxes in different water masses

for the duration of the drifts, also including those in the

mixed layer for completeness. The uppermost water mass

[PW and Polar Intermediate Water (PIW)] typically ex-

tends below the mixed layer. FSD1 mixed layer heat

fluxes compare well with the Stanton number–based es-

timates for the surface fluxes, except for day ;118.63

when downward fluxes were observed when crossing

a thermal front in the mixed layer and toward the end of

the drift when significant convection led to enhanced

upward fluxes. In FSD2, there are only four casts that

resolve the mixed layer, hence the heat flux estimates

here are uncertain and are slightly smaller than the

Stanton number–based estimate. An increase in heat flux

toward the end of the drift is observed in all water masses,

corresponding to the mixing event at around day 120.81.

During the NBD1, heat fluxes within the PW, PIW, and

ASW water masses increase toward the end of the drift as

the floe approaches shallower topography. Generally for

all drifts below the mixed layer, layer-averaged heat

fluxes decrease with depth as a consequence of reduced

mixing and thermal diffusivity.

A summary of drift average heat fluxes and salt fluxes

within the three classes of water masses is given in Fig. 16.

All three drifts show a decreasing magnitude of heat

and salt fluxes with increasing depth, somewhat coun-

terintuitive to the general understanding that the main

source of heat and salt is the layers of Atlantic water.

This is mainly because the AW (and LAIW) has rela-

tively weak vertical temperature gradient when com-

pared to ASW, although the eddy diffusivity for heat in

these water masses is comparable (even larger KT in

AW at FSD1). Furthermore, the average value in the

AW is reduced because the temperature gradient re-

verses below the depth of the AW temperature maxi-

mum, and thus the direction of the heat flux is opposite

in the upper and lower parts of this water mass, reducing

the average (see, e.g., Fig. 14a). Heat is likely entrained

from the deeper AW into PW, as observed at FSD2;

however, advective heat flux is also significant (section

4b). The vertical variability of the heat fluxes within the

AW layer is significant at FSD2 (Fig. 14a). At the upper

side of the AW at about 40 m where there is a subsurface

FIG. 11. TIC measurements in the under-ice boundary layer at 1 m (solid) and 5 m (dashed)

below the ice during YPD in 2003. Ensembles of 15 min are hourly averaged and further

smoothed by a 3-h running-mean filter. (a) Temperature at both levels and elevation above

freezing point at 5 m (shaded area, right axis), (b) salinity, (c) local friction velocity, and (d) the

heat flux.


temperature maximum, the heat flux is in excess of

100 W m22, an order of magnitude larger than the layer-

averaged value. The nonzero convergence of the vertical

heat flux above the AW will heat up the overlying layer.

Water mass transformations in this region can be af-

fected by ice melting. Following Cokelet et al. (2008),

lines with a constant ratio of atmospheric cooling to ice–

ocean interface melting of ice with a salinity of 5 psu are

FIG. 12. Same as in Fig. 11, but for WBD on the northern shelf of Svalbard. Because the time

series is relatively short, all 15-min segments are shown using 1-h running mean.

FIG. 13. TIC measurements at 1 m below the ice during NBD2 in 2007. All 15-min segments

are shown using 1-h running mean. (a) Salinity (solid) and temperature (dashed), (b) elevation

above freezing (shaded area, left axis) and the local friction velocity (solid), and (c) the heat

flux.


plotted as yellow dashed lines in the T–S diagram in Fig. 2.

The ratio indicates that a water mass with initial prop-

erties T 5 3.28C, S 5 35.1 is modified by melting only

(R 5 0), by equal contributions of atmospheric cooling

and interface melting (R 5 1), or more by atmospheric

cooling (R 5 2). The limit R / ‘ (a vertical line on T–S

space) indicates modification by atmospheric cooling

alone. Generally, the upper water mass characteristics are

not produced along the melting line, but both the atmo-

spheric cooling and ice melting are important (approxi-

mately equally). This is a consequence of the proximity to

the marginal ice zone and the variable ice concentration.

From Fig. 2, the water column at FSD1 has a gentler slope

than FSD2, WBD, and NBD1. A slope with R , 1 in-

dicates that melting has played a more dominant role in

modifying the water column. FSD1 was situated west-

ward of the other drifts, where it is more likely to en-

counter water exported from the Arctic through the

Fram Strait, which has been modified by melting under

continuous ice cover. The FSD2, WBD, and NBD1 water

columns in the T–S diagram are along lines that indicate

that atmospheric cooling has made a larger contribution

to modification than melting, which is appropriate since

water masses are advected along the WSC and drifts are

within the marginal ice zone where air–sea–ice inter-

action is intense.

5. Summary and concluding remarks

In total, six drift stations north of Svalbard were oc-

cupied in early spring covering years 2003–07. Turbulent

characteristics were observed using a microstructure pro-

filer or eddy correlation instrumentation. The drift sta-

tions represent three different conditions: 1) within close

proximity to the West Spitsbergen Current (WSC) with

shallow and warm mixed layers and subsequent large heat

fluxes [Whaler’s Bay Drift (WBD) and Fram Strait Drift

2 (FSD2)]; 2) away from the WSC where the mixed layer is

deeper, colder, and displays relatively moderate vertical

heat fluxes [Yermak Plateau Drift (YPD) and Fram Strait

Drift 1 (FSD1)]; and 3) over the continental shelf in the

Svalbard branch where cooling and mixing of the WSC is

enhanced by the ice drift and tidal effects related to to-

pography [Norwegian Bank Drifts (NBD1 and NBD2)].

FIG. 14. (a),(c) Average profiles of turbulent heat flux FH and (b),(d) salinity flux FS calcu-

lated using 4-m segments of MSS profiles in FSD1 (black) and FSD2 (gray), and on the Nor-

wegian Bank for the deep side (black) and shallow side (gray) of the thermohaline front.


Measurements from drifts within or close to the WSC

show surface and mixed layer heat fluxes of 200–300

W m22 and estimated melting rates of 5–9 cm day21.

Heat fluxes are comparable to and provide support for

earlier studies that inferred vertical heat fluxes from heat

loss observed between subsequent cross sections of the

WSC (e.g., Aagaard et al. 1987; Saloranta and Haugan

2004; Cokelet et al. 2008). During FSD2, turbulence in

the upper layers is one order of magnitude larger than

that observed away from the WSC, and extends deep into

the pycnocline, entraining warmer water into the mixed

layer. Changes of the mixed layer heat content, however,

indicate that the contribution from horizontal advection

and diffusion is equally important in the heat balance

within the mixed layer.

Outside the WSC, turbulent fluxes in the mixed layer

are small. Over the northeastern flank of Yermak Pla-

teau where the mixed layer is near freezing point and ice

drift is slow, the average heat flux is 2 W m22 at YPD,

but varies significantly due to solar radiation and tides.

At FSD1, vertical heat fluxes and contributions from

horizontal advection and diffusion account equally for

changes in mixed layer heat content. In contrast to

FSD2, mixing at FSD1 is typically confined to the mixed

layer with localized enhanced mixing between 40 and

70 m due to strong velocity shear.

On the shelf area, observed elevated levels of mixing

are due to tidal and topographic effects and lead to

surface heat fluxes of O(100) W m22. As the shallows

are approached, an increasing fraction of the water

column is covered by bottom and upper boundary layers

and mixing by tides and topography gets increasingly

important.

Profiles of heat flux typically decrease with depth

when averaged in water mass classes. Average fluxes in

the Atlantic water (AW) and the Lower Arctic In-

termediate Water layers are considerably lower than the

surface heat fluxes (e.g., only 10% of the surface flux at

FSD1). This also shows that a considerable fraction of

the mixed layer heat is entrained from deeper layers or

advected into the area. Near the WSC the vertical var-

iability is large and heat flux on the crucial upper side of

the AW exceeds 100 W m22, one order of magnitude

larger than the layer-averaged value.

The slope of temperature against salinity properties

compared with that expected from a given ratio of

melting/atmospheric cooling shows that upper-layer

water masses in all our drifts north of Svalbard are

slightly more affected by atmospheric cooling than by

melting of ice. This is in line with Cokelet et al. (2008),

who show that the contribution from atmospheric

cooling decreases along the path of the WSC as the ice

cover gets more solid.TA

BL

E6

.Wa

ter

ma

ss–

av

era

ge

dp

rop

ert

ies.

Va

lue

sa

reth

ed

rift

av

era

ge

sfo

rT

an

dS

,av

era

ge

6o

ne

sta

nd

ard

de

via

tio

nfo

rd

ep

thz,t

hic

kn

ess

h,b

uo

ya

ncy

fre

qu

en

cyN

,an

dth

etu

rbu

len

t

flu

xe

sF

Ha

nd

FS,

an

dth

em

ax

imu

mli

ke

lih

oo

de

stim

ato

rw

ith

the

95

%co

nfi

de

nce

inte

rva

lin

bra

cke

tsfo

rth

ed

issi

pa

tio

nra

tea

nd

the

ed

dy

dif

fusi

vit

ies.

FS

D1

FS

D2

NB

D1

PW

an

d

PIW

AS

W

AW

an

d

LA

IW

PW

an

d

PIW

AS

W

AW

an

d

LA

IW

PW

an

d

PIW

AS

W

T(8

C)

21

.19

0.9

92

.51

20

.51

.03

3.0

32

0.7

30

.42

S3

4.3

03

4.7

53

5.0

33

4.5

33

4.7

63

5.0

73

4.5

63

4.6

9

z(m

)1

0.4

62

.25

4.5

64

.47

6.4

61

0.8

10

.06

0.0

15

.76

7.2

41

.16

11

10

.06

0.0

41

.06

7.9

h(m

)4

06

51

7.9

67

.74

2.2

61

7.4

9.5

63

.72

1.4

67

.28

3.5

62

42

6.6

61

1.3

29

.36

9.7

N(c

ph

)4

.16

0.3

4.9

60

.92

.06

0.2

3.9

61

.53

.56

0.6

1.5

60

.22

.56

0.7

2.1

60

.7

«3

10

28

(Wk

g21)

10

.9[8

.5–

14]

3.0

[2.7

–3.

4]

1.9

[1.9

–2]

85

.7[2

0.3

–36

2]

12

.4[8

.8–

17.4

]2

.3[2

.1–

2.5

]4

58

[21

7–9

64

]1

7.7

[9.9

–31

.6]

Kr

31

02

4(m

2s2

1)

12

.7[7

.9–

20]

1.1

[1.0

–1.

3]

3.8

6[3

.5–

.2]

26

5[2

2–

321

1]

8.9

[6.1

–13

]7

[6.1

–8]

15

94

[45

0–5

64

9]2

5.1

[14

.1–

44

.7]

KT

31

02

4(m

2s2

1)

14

[8–

24]

0.4

5[0

.3–

0.7

]3

.1[2

.3–

4.1

]1

46

[10

–21

12

]1

4.7

[6.1

–3

5]

7.2

[5.5

–9.

4]

12

7[7

8–

20

8]

23

.8[1

1.6

–4

9.1

]

FH

(Wm

22)

12

61

56

66

2.4

63

11

96

10

87

26

78

9.5

62

42

25

61

57

17

56

36

6

FS

10

26

3(k

gs2

1m

22)

0.7

76

1.8

0.5

46

1.5

0.2

36

0.1

67

.06

7.3

3.5

64

.60

.66

0.8

20

.86

17

.44

67


A quiescent Arctic interior with weak vertical diffu-

sion is crucial for the maintenance of the cold halocline

layer (Fer 2009) as well as the circulation of the AW

layer (Zhang and Steele 2007). In the absence of storm

events and eddies, oceanic heat fluxes in the deep basin

cannot contribute significantly to the ocean surface

heat budget. Krishfield and Perovich (2005), using ba-

sinwide drifting buoy observations, estimate an annual

oceanic heat flux of 3–4 W m22 to the Arctic pack ice,

with significantly larger values in summer. In contrast,

ice-tethered buoy observations in the Canada Basin

thermocline yield diffusive heat fluxes about one order

of magnitude less than the annual mean heat flux to the

ice (Timmermans et al. 2008). Although relatively el-

evated turbulence is observed over deep topography

(Rainville and Winsor 2008), most of the mixing and

removal of heat from the AW will occur along the

Arctic perimeters, and possibly dominate basin-averaged

heat flux. The warm and saline AW layer propagates

into the Arctic Ocean as a boundary current flowing

along the continental slope. Recently, Lenn et al. (2009)

reported that measured vertical mixing of the bound-

ary current on the east Siberian continental slope could

not account for the freshening and cooling of the AW,

suggesting the influence of shelf waters and lateral

mixing.

This study presents directly measured surface turbu-

lent fluxes in addition to turbulent fluxes and mixing

inferred from high-resolution turbulence profiling at

sites where similar datasets are scarce. The results show

that mixing and heat transfer within the branches of the

WSC and over the shelf areas close to the Fram Strait

are several orders of magnitude larger than those

reported by Lenn et al. (2009) farther upstream. Outside

the branches of the WSC and off the shelves, however,

we observe vertical mixing and oceanic heat fluxes

similar to values found in the central Arctic. The large

spatial variability and the lack of turbulent mixing in the

AW boundary current along the continental slope far-

ther upstream suggest that significant cooling of AW

occurs in local regions north of Svalbard with strong

currents and tidal forcing over topography. Diapycnal

FIG. 15. Times series of turbulent heat flux averaged over (first row) the mixed layer, (second

row) PW, (third row) Arctic Surface Water, and (fourth row) Atlantic water layers. Black

square markers correspond to data points derived for each MSS cast and the gray curves are

averages over 10 equally spaced bins over the duration of the drift. Values of heat flux to the ice

obtained using the Stanton number parameterization (gray squares with error bars) are com-

pared to the mixed layer fluxes.


mixing is important for the regional heat budget, hence

the ice cover and cooling of the WSC in the gateway to

the Arctic north of Svalbard. In contrast, mixing by

processes including lateral mixing, shelf–basin exchange,

and intrusions of cold shelf waters sinking downslope

may be more typical of the Arctic shelves. Although our

measurements are characterized by significant lateral

and temporal variability, they represent an important

contribution to the understanding of processes by which

the core of the WSC loses heat, and provide support for

previous hydrography-based estimates of vertical heat

fluxes.

Acknowledgments. Constructive comments and sug-

gestions of the two anonymous reviewers are appreciated.

This study partly received support from the Research

Council of Norway through the project CuNoS: Current

measurements north of Svalbard (178919/S30). We thank

the captains and the crews of R/V Polarstern and R/V

Lance, the oceanographic group of the ARK XIX/1

cruise, and the UNIS students of the R/V Lance cruises.

REFERENCES

Aagaard, K., J. H. Swift, and E. C. Carmack, 1985: Thermohaline

circulation in the Arctic Mediterranean Seas. J. Geophys. Res.,

90, 4833–4846.

——, A. Foldvik, and S. R. Hillman, 1987: The West Spitsbergen

Current—Disposition and water mass transformation. J. Geo-

phys. Res., 92, 3778–3784.

Arneborg, L., 2002: Mixing efficiencies in patchy turbulence.

J. Phys. Oceanogr., 32, 1496–1506.

Bourke, R. H., A. M. Weigel, and R. G. Paquette, 1988: The

westward turning branch of the West Spitsbergen Current.

J. Geophys. Res., 93, 14 065–14 077.

Boyd, T. J., and E. A. D’Asaro, 1994: Cooling of the West Spits-

bergen Current: Wintertime observations west of Svalbard.

J. Geophys. Res., 99, 22 597–22 618.

Cokelet, E. D., N. Tervalon, and J. G. Bellingham, 2008: Hydrogra-

phy of the West Spitsbergen Current, Svalbard branch: Autumn

2001. J. Geophys. Res., 113, C01006, doi:10.1029/2007JC004150.

FIG. 16. Summary of turbulent heat fluxes for (a) heat and (b) salinity averaged over water

masses in FSD1, FSD2, and NBD1. The values between (a) and (b) show the total length (km)

of data segments averaged.


Dewey, R., R. Muench, and J. Gunn, 1999: Mixing and vertical heat

flux estimates in the Arctic Eurasian Basin. J. Mar. Syst., 21,

199–205.

Dillon, T. M., and D. R. Caldwell, 1980: The Batchelor spectrum

and dissipation in the upper ocean. J. Geophys. Res., 85, 1910–

1916.

Fer, I., 2006: Scaling turbulent dissipation in an Arctic fjord. Deep-

Sea Res. II, 53, 77–95.

——, 2009: Weak vertical diffusion allows maintenance of cold

halocline in the central Arctic. Atmos. Oceanic Sci. Lett., 2 (3),

148–152.

——, and A. Sundfjord, 2007: Observations of upper ocean

boundary layer dynamics in the marginal ice zone. J. Geophys.

Res., 112, C04012, doi:10.1029/2005JC003428.

——, M. G. McPhee, and A. Sirevaag, 2004: Conditional statistics

of the Reynolds stress in the under-ice boundary layer. Geo-

phys. Res. Lett., 31, L15311, doi:10.1029/2004GL020475.

Gregg, M. C., 1987: Diapycnal mixing in the thermocline—A re-

view. J. Geophys. Res., 92, 5249–5286.

IOC, IHO, and BODC, 2003: Centenary Edition of the GEBCO

Digital Atlas. British Oceanographic Data Centre, CD-ROM.

Krishfield, R. A., and D. K. Perovich, 2005: Spatial and temporal

variability of oceanic heat flux to the Arctic ice pack. J. Geo-

phys. Res., 110, C07021, doi:10.1029/2004JC002293.

Lenn, Y. D., and Coauthors, 2009: Vertical mixing at intermediate

depths in the Arctic boundary current. Geophys. Res. Lett., 36,L05601, doi:10.1029/2008GL036792.

Manley, T. O., 1995: Branching of Atlantic Water within the

Greenland-Spitsbergen passage: An estimate of recirculation.

J. Geophys. Res., 100, 20 627–20 634.

McPhee, M. G., 1986: The upper ocean. Geophysics of Sea Ice,

N. Untersteiner, Ed., Plenum, 339–394.

——, 2002: Turbulent stress at the ice/ocean interface and bottom

surface hydraulic roughness during the SHEBA drift. J. Geo-

phys. Res., 107, 8037, doi:10.1029/2000JC000633.

——, 2008: Air-Ice-Ocean Interaction: Turbulent Boundary Layer

Exchange Processes. Springer, 215 pp.

——, C. Kottmeier, and J. H. Morison, 1999: Ocean heat flux in the

central Weddell Sea during winter. J. Phys. Oceanogr., 29,

1166–1179.

——, T. Kikuchi, J. H. Morison, and T. P. Stanton, 2003: Ocean-to-

ice heat flux at the North Pole environmental observatory.

Geophys. Res. Lett., 30, 2274, doi:10.1029/2003GL018580.

Moum, J. N., 1996: Efficiency of mixing in the main thermocline.

J. Geophys. Res., 101, 12 057–12 069.

Osborn, T. R., 1980: Estimates of the local rate of vertical diffusion

from dissipation measurements. J. Phys. Oceanogr., 10, 83–89.

——, and C. S. Cox, 1972: Oceanic fine structure. Geophys. Fluid

Dyn., 3, 321–345.

Padman, L., and T. M. Dillon, 1991: Turbulent mixing near the

Yermak Plateau during the coordinated Eastern Arctic Ex-

periment. J. Geophys. Res., 96, 4769–4782.

——, and S. Erofeeva, 2004: A barotropic inverse tidal model for

the Arctic Ocean. Geophys. Res. Lett., 31, L02303, doi:10.1029/

2003GL019003.

——, A. J. Plueddemann, R. D. Muench, and R. Pinkel, 1992:

Diurnal tides near the Yermak Plateau. J. Geophys. Res., 97,

12 639–12 652.

Rainville, L., and P. Winsor, 2008: Mixing across the Arctic

Ocean: Microstructure observations during the Beringia 2005

Expedition. Geophys. Res. Lett., 35, L08606, doi:10.1029/

2008GL033532.

Rudels, B., and Coauthors, 2000: Water mass distribution in Fram

Strait and over the Yermak Plateau in summer 1997. Ann.

Geophys., 18, 687–705.

Saloranta, T. M., and P. M. Haugan, 2001: Interannual variability

in the hydrography of Atlantic water northwest of Svalbard.

J. Geophys. Res., 106, 13 931–13 943.

——, and ——, 2004: Northward cooling and freshening of the warm

core of the West Spitsbergen Current. Polar Res., 23, 79–88.

Sirevaag, A., 2009: Turbulent exchange coefficients for the ice/

ocean interface in case of rapid melting. Geophys. Res. Lett.,

36, L04606, doi:10.1029/2008GL036587.

Smyth, W. D., J. N. Moum, and D. R. Caldwell, 2001: The efficiency

of mixing in turbulent patches: Inferences from direct simu-

lations and microstructure observations. J. Phys. Oceanogr.,

31, 1969–1992.

Steele, M., and J. H. Morison, 1993: Hydrography and vertical

fluxes of heat and salt northeast of Svalbard in autumn.

J. Geophys. Res., 98, 10 013–10 024.

St. Laurent, L. C., and R. W. Schmitt, 1999: The contribution of salt

fingers to vertical mixing in the North Atlantic Tracer Release

Experiment. J. Phys. Oceanogr., 29, 1404–1424.

Sundfjord, A., I. Fer, Y. Kasajima, and H. Svendsen, 2007: Ob-

servations of turbulent mixing and hydrography in the mar-

ginal ice zone of the Barents Sea. J. Geophys. Res., 112,C05008, doi:10.1029/2006JC003524.

Thomson, R. E., and I. V. Fine, 2003: Estimating mixed layer depth

from oceanic profile data. J. Atmos. Oceanic Technol., 20,319–329.

Thorpe, S. A., 1977: Turbulence and mixing in a Scottish loch.

Philos. Trans. Roy. Soc. London, A286, 125–181.

Timmermans, M.-L., J. Toole, R. A. Krishfield, and P. Winsor,

2008: Ice-tethered profiler observations of the double-diffusive

staircase in the Canada Basin thermocline. J. Geophys. Res.,

113, C00A02, doi:10.1029/2008JC004829.

Untersteiner, N., 1988: On the ice and heat-balance in Fram Strait.

J. Geophys. Res., 93, 527–531.

Wettlaufer, J. S., N. Untersteiner, and R. Colony, 1990: Estimating

oceanic heat flux from sea-ice thickness and temperature data.

Ann. Glaciol., 14, 315–318.

Zhang, J. L., and M. Steele, 2007: Effect of vertical mixing on the

Atlantic Water layer circulation in the Arctic Ocean. J. Geo-

phys. Res., 112, C04S04, doi:10.1029/2006JC003732.


early spring oceanic heat fluxes and mixing observed from drift stations north of svalbard*

Documents