Wind-triggered events of phytoplankton downward flux in the Northeast Water Polynya

Wind-triggered events of phytoplankton downward flux in the

Northeast Water Polynya

S. Pesant a,*, L. Legendre b, M. Gosselin c, E. Bauerfeind d, G. Budeus e

aDepartment of Fisheries and Oceans, 200 Kent Street, 12th floor, Ottawa, Ontario, Canada K1A 0E6bStation Zoologique, BP 28, 06234 Villefranche-sur-Mer Cedex, France

cDepartement d’Oceanographie, Universite du Quebec a Rimouski, 310 Allee des Ursulines, Rimouski, Quebec, Canada G5L 3A1dInstitut fur Ostseeforschung, Biol. Meererskunde, Seetstrasse 15, D 18119 Rostock-Warnemunde, GermanyeAlfred-Wegener-Institut fur Polar- und Meeresforschung, Postfach 120161, D-27515 Bremerhaven, Germany

Received 1 November 1999; accepted 21 August 2001

Abstract

Phytoplankton carbon fluxes were studied in the Northeast Water (NEW) Polynya, off the eastern coast of Greenland (79� to81�N, 6� to 17�W), during summer 1993. The downward flux of organic particles was determined during 54 days using a

sediment trap moored at a fixed location, below the pycnocline (130 m). The hypothesis of the present study is that wind events

were ultimately responsible for the events of diatoms downward flux recorded in the trap. Wind conditions can influence the

vertical transport of phytoplankton by affecting (1) the environmental conditions (e.g. hydrostatic pressure, nutrient

concentrations, and irradiance) encountered by phytoplankton during their vertical excursion, and (2) the aggregation and

disaggregation of phytoplankton flocs. The first mechanism affects the physiological regulation of buoyancy, whereas the

second one affects the size and shape of settling particles. Using field data (wind velocity, density profiles and phytoplankton

abundance), we assessed the potential aggregation and the vertical excursion of phytoplankton in surface waters. The results

show that, upstream from the trap, wind and hydrodynamic conditions were sometimes favourable to the downward export of

phytoplankton. Lag-correlation between time series of wind and phytoplankton downward flux shows that flux events lagged

wind events by ca. 16 days. Given that the average current velocity in the top 100 m was ca. 10 cm s� 1, a lag of 16 days

corresponded to a lateral transport of ca. 130 km, upstream from the sediment trap, where phytoplankton production was lower

than at the location of the trap. According to that scenario, 21% to 60% of primary production was exported to depth during

wind events. If we had assumed instead a tight spatial coupling between the material collected in the trap and the relatively high

phytoplankton production at the location of the trap, we would have concluded that < 7% of primary production was exported

to depth. The difference between the two scenarios has great implications for the fate of phytoplankton. Our results stress the

importance of investigating the spatial coupling between surface and trap data before assessing the pathways of phytoplankton

carbon cycling. D 2002 Elsevier Science B.V. All rights reserved.

Keywords: Wind; Phytoplankton; Sinking; Aggregation; Trap; SETCOL; Polynya; Arctic

0924-7963/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved.

PII: S0924-7963 (01 )00065 -3

* Corresponding author. Centre for Water Research, University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, Australia.

E-mail addresses: [email protected], [email protected] (S. Pesant).

www.elsevier.com/locate/jmarsys

Journal of Marine Systems 31 (2002) 261–278

1. Introduction

Phytoplankton in the Northeast Water (NEW) Pol-

ynya was studied during summer 1993. Diatoms

dominated the large-sized fraction of phytoplankton

(Pesant et al., 1996) so that, in the present paper, this

size fraction will be referred to as simply diatoms.

During our investigation, diatoms were characterised

by high production rates and relatively low-standing

stocks (Smith, 1995; Smith et al., 1995; Pesant et al.,

1996). The imbalance between these two variables in

the euphotic zone indicates potentially high losses of

diatom biomass to other compartments of the pelagic

food web (e.g. grazing, viral attack), to regions out-

side the study area (i.e. lateral flux), and/or to deeper

waters and ultimately to the benthic food web and

sediments (i.e. downward flux). Pesant et al. (1998)

estimated that large-sized dinoflagellates, copepods

and appendicularians potentially grazed < 60% of

diatom production, so that an important fraction of

that production was potentially exported downwards

and/or laterally.

The relative importance of downward vs. lateral

export has significant implications for seabed com-

munities. In areas where downward export dominates

over lateral export (i.e. a tight bentho–pelagic cou-

pling), seabed communities depend in part on local

pelagic production. In contrast, in areas where lateral

export dominates over downward export (i.e. a weak

bentho-pelagic coupling), seabed communities depend

in part on pelagic production upstream.

The downward flux of particulate organic carbon

in the NEW Polynya was investigated during 1992–

1993 by Bauerfeind et al. (1997), using sediment traps

at ca. 300 m. The seasonal pattern of the downward

flux was characterised by an important peak in

autumn and a moderate one in spring. The summer

flux was temporally variable, with relatively small

peaks occurring episodically. The autumn and spring

peaks corresponded to changes in ice conditions

(Bauerfeind et al., 1997; Ramseier et al., 1997), i.e.

progressively decreasing values during spring, and

rapidly increasing values during autumn. The autumn

peak was likely triggered by the formation of new ice,

which results in the formation of superdense water

and, thus, in a downward movement of particle-rich

surface waters (Gawarkiewicz and Chapman, 1995).

In contrast, the spring flux was likely triggered by the

melting of snow and ice (Ramseier et al., 1997),

which released ice algae into the euphotic zone and

ultimately to depth (Bauerfeind et al., 1997). These

two mechanisms, however, could not have triggered

the episodic downward pulses during summer because

the air–sea heat transfer at the time was positive

(Schneider and Budeus, 1997), which did not allow

the formation of superdense water, and the diatom

species, which dominated in the summer trap samples,

occurred in the water not in the ice (Hellum von

Quillfeldt, 1997). The present paper explores other

mechanisms that could have triggered episodic down-

ward fluxes during summer.

According to Stokes’ law (V=(2gr2)(d1� d2)/9l,where V is settling velocity, g is gravity, r is particle

radius, d1 is particle density, d2 is water density and lis water viscosity), the settling velocity of phytoplank-

ton is proportional to the square of particle diameter

so that aggregation should increase settling. There is,

however, increasing evidence that the settling velocity

of aggregates is not solely related to their sizes

(Diercks and Asper, 1997; Waite et al., 1997) and is

also influenced by the density and shape of cells or

aggregates. The density of actively growing cells can

be controlled metabolically (Steele and Yentsch,

1960; Bienfang et al., 1983) and can, thus, be influ-

enced by the physiological condition of the cells and

the environmental conditions, e.g. nutrient concentra-

tions and light. Beyond a given size of aggregates,

however, it is likely that physiological control is no

longer possible.

When nutrients and/or irradiance are limiting pri-

mary production, phytoplankton can lose their ability

to control their buoyancy, thus increasing their settling

velocity (Bienfang et al., 1983; Waite et al., 1992a,

1997). In the NEW ice-free waters, the nutricline and

the euphotic zone were generally shallower than the

surface pycnocline (Pesant et al., 1996; Kattner and

Budeus, 1997). Thus, nutrient and irradiance condi-

tions in the NEW generally allowed for physiological

control of buoyancy within the surface mixed layer.

Episodic wind events, which deepen the mixed layer,

can modify these conditions. There are some exam-

ples of the influence of wind events on the redistrib-

ution of nutrients over depth (Hitchcock et al., 1987;

Kiørboe and Nielsen, 1990) and the deep excursion of

phytoplankton (Yamazaki and Kamykowski, 1991;

Franks and Marra, 1994; Webster and Hutchinson,

S. Pesant et al. / Journal of Marine Systems 31 (2002) 261–278262

1994). The deep excursion of cells can reduce the

average irradiance experienced by phytoplankton and

disable the physiological control of buoyancy. Wind-

induced turbulence can also influence the aggregation

and disaggregation of phytoplankton and form large-

sized flocs that settle rapidly (Kiørboe, 1993; All-

dredge and Jackson, 1995), at least when buoyancy

control is not possible.

The hypothesis of the present paper is that wind

events (e.g. storms) were ultimately responsible for

the events of phytoplankton downward flux recorded

in a shallow (130 m) sediment trap, deployed in the

NEW during summer 1993. First, we present theoret-

ical considerations of the influence of wind on the

seasonal stratification and the aggregation of phyto-

plankton; second, we use environmental and biolog-

ical data to predict the location of areas in the NEW

where particles are expected to be exported down-

wards; and, finally, we cross-correlate time series of

wind and particulate carbon flux recorded in the trap.

We also address the temporal and spatial coupling

between phytoplankton produced in surface waters

and cells collected in the trap, and we determine

which proportion of primary production was exported

downwards and collected in the sediment trap.

2. Materials and methods

2.1. Study site and sampling

The NEW Polynya is located at the southern limit

of the permanent Arctic ice pack, where ice-free

waters extend on the continental shelf of Greenland

(79� to 81�N) from May to October. Sampling was

conducted on board the R.V. ‘Polarstern’ from (i) 26

May to 18 June and (ii) 3 to 27 July 1993 (cruises

ARK IX/2 and 3). A moored sediment trap was

deployed (Mooring G; 80�27VN, 13�41VW; Fig. 1) at

a depth of 130 m during the whole sampling period.

Ice concentrations were monitored using a Special

Sensor Microwave/Imager (data provided by the

Microwave Group Ottawa-River) and hydrographic

conditions were determined using a CTD profiler

(results in Budeus et al., 1994; Budeus and Schneider,

1996). Wind conditions were measured onboard the

ship, 20 m above the sea surface (Koenig-Langlo and

Marx, 1997), and at the Henrik Krøyer Holmes

meteorological station (80�39VN, 13� 43VW; station

M in Fig. 1, located near the mooring position) 10 m

above the ground (data provided by the Danish

Meteorological Institute).

Water was collected using a rosette sampler equip-

ped with twelve 12-l Niskin bottles and a LI-COR 185

B underwater quantum PAR meter. Niskin bottles

were equipped with silicone- or Teflon-coated stain-

less steel springs. Water samples were collected at

seven photic depths (100%, 50%, 30%, 10%, 5%, 1%,

0.1% of surface irradiance) and at the depth of the

deep chlorophyll maximum (DCM) as determined

from in vivo fluorescence (Haardt instruments). Water

was immediately drawn from the bottles for various

phytoplankton measurements including chlorophyll a

(chl a), particulate organic carbon (POC), settling rate,

and microscopic analysis.

The sediment trap (Kiel type; 0.5 m2 opening) was

programmed for 18 successive collections from 3 June

to 24 July, i.e. each collection cup sampled a period of

ca. 3 days. The trap was equipped with a baffle of

aspect ratio 5:1 and the collection cups were filled

with filtered (0.2 mm) sea-water enriched with HgCl2to reach a final concentration of 0.07%. All samples

were split on board into eight aliquots using a rotating

splitter, immediately after recovery. Swimmers and

appendicularian houses were carefully removed from

all samples prior to the splitting, and were analysed

separately. One aliquot was kept in solution for

microscopic analysis and the others were used for

analysis of sediment matter, e.g. particulate silica

(PSiO2), POC, and particulate organic nitrogen

(PON).

2.2. Laboratory analyses

Diatoms from the trap and from the depth of the

DCM were preserved with acidic Lugol’s solution, for

identification, enumeration, and size measurements

under the inverted microscope (Lund et al., 1958).

The volume and plasma content of diatoms were

calculated from mean cell sizes for the genus, using

formulas given by Edler (1979), and the plasma

content was converted to organic carbon according

to Strathmann (1967). A distinction was made bet-

ween empty frustules and diatoms containing plasma

and chloroplasts so that only the latter counts were

used to calculate the diatom C flux.

S. Pesant et al. / Journal of Marine Systems 31 (2002) 261–278 263

Appendicularian and copepod faecal pellets col-

lected in the trap were enumerated under the dissect-

ing microscope. During identification, the widths of

pellets were measured; the lengths could not be

measured because most pellets were broken. The

volume of these copepod pellets was estimated from

a width to length relationship derived empirically

using measurements made on intact pellets (Bauer-

feind et al., 1997). The volume and organic carbon

content (method as for POC) of intact copepod pellets

collected in the trap were determined and used to

calculate an organic carbon to volume ratio, i.e. 0.058

mg C mm� 3. This value is in agreement with those

reported in the literature, e.g. Gonzalez et al. (1994), it

was used to determine the carbon flux corresponding

to the enumerated copepod pellets. Appendicularian

pellets were assigned, according to their sizes, to

large- ( > 500 mm) or small-sized ( < 500 mm) pellets.

The organic carbon content of intact appendicularian

pellets in the two size classes were determined

(method as for POC) on several trap samples. The

median values for the two size classes, i.e. 0.745 and

0.149 mg pellet � 1, were used to determine the carbon

flux corresponding to the enumerated appendicularian

pellets.

Subsamples (200 ml) from the seven photic depths

and the depth of the DCM were filtered on 25 mm

Poretics polycarbonate membranes with 5 mm nominal

pore size for determination of diatom chl a, assuming

that these dominated phytoplankton in the >5 mm size

fraction. Filtration was under vacuum pressure < 100

mm Hg. Concentration of chl a was determined using

a Turner fluorometer (model 112), after 24-h extrac-

tion in 90% acetone at 5 �C without grinding (Parsons

Fig. 1. Map of the Northeast Water showing the study area and the locations of moorings A and B (current meters at 75-m depth), moorings F

and G (long-term sediment trap at 300 m depth and short-term sediment trap at 130 m depth), and the Henrik Krøyer Holmes meteorological

station (M; 80�39VN, 13�43VW). Arrows: general surface circulation; hatched areas: edges of the ice barriers. The study area includes two

shallow banks ( < 200 m), i.e. Ob Bank to the north and Belgica Bank to the south.

S. Pesant et al. / Journal of Marine Systems 31 (2002) 261–278264

et al., 1984). Subsamples from all depths (400 ml) and

from the trap were filtered on precombusted (500 �Cduring 5 h) GF/F filters for later POC and PON

determination using a Perkin-Elmer CHN analyser

(water samples), and a Heraeus Instrument (trap

samples). PSiO2 in the trap samples was determined

according to Bodungen et al. (1991).

2.3. Settling velocity (SETCOL) determination

Settling velocities of phytoplankton were estimated

using a large settling column (9 cm diameter and 60

cm height; Bienfang (1981)) filled with a subsample

(ca. 4000 ml) from the depth of the DCM. The

subsample was shaken vigorously before being

poured in the columns, so that the initial distribution

of particles was nearly homogenous. Aggregates and

long chains of phytoplankton were probably disrupted

during that process so that if phytoplankton aggre-

gated in the field, their settling velocity would be

underestimated in the present experiments. The set-

tling velocity determined here reflects the morphology

(e.g. size, shape, appendages) of mostly single cells

and small chains of phytoplankton and their physiol-

ogy (e.g. ionic pump, senescence). Ship movement,

particularly when breaking ice, can cause vibration in

the settling column and influence phytoplankton set-

tling. In order to minimize this potential bias, experi-

ments were run only at stations that involved

prolonged (4–12 h) deck- and ice-work, i.e. not

during steaming time. In order to maintain the column

in darkness and at a temperature close to that in situ, it

was kept inside an opaque sleeve with running sea-

water pumped from 8 m. Particles were allowed to

settle during 2–6 h, depending on the abundance of

phytoplankton. Being in the dark for < 6 h should not

affect the physiological control of settling (Waite et

al., 1992b). The top and bottom fractions in the

column were then collected, shaken vigorously to

break aggregates, and split in two aliquots of ca.

200 ml. One aliquot was filtered on a 25 mm What-

man GF/F filter and the other on a 25 mm Poretics

polycarbonate membrane with 5 mm nominal pore

size, to determine the chl a biomass of phytoplank-

ton in the total and large-sized fraction (>5 mm),

respectively (method as for the water samples). Con-

centration of chl a in the small-sized fraction ( < 5

mm) was calculated by subtracting the large-sized

value from the total concentration. Settling velocities

(day � 1) of total phytoplankton and the two size-

fractions were calculated using the formulae of Bien-

fang (1981).

2.4. Calculations

Alldredge et al. (1987) and Ruiz (1996) quantified

the residence time of particles in the surface mixed

layer by comparing the time for settling (h/w) to the

time for turbulent diffusion (h2/K; Mann and Lazier,

1991), which are characteristic of particle and envi-

ronmental conditions, respectively; h is the thickness

of the modelled layer (m); w is the settling velocity

(m s � 1); and K = 0.17W––

103tm

� 1 is a simplified

expression for the coefficient of vertical diffusivity

(m2 s� 1). The constant value 0.17 was calculated for

seawater by Mann and Lazier (1991) and has units

m � 1 s3. K varies as a function of wind velocity (W––

10;

m s� 1) and the time required for mixing the modelled

layer (tm; s). The latter parameter was assumed to be

7200 s for the NEW, which is a moderate value for

coastal and oceanic systems (Mann and Lazier, 1991).

The balance between turbulent diffusion and settling

of particles (D dimension less) was expressed as the

ratio of the two times (settling to diffusion):

D ¼ h=w

h2=K¼ K=wh: ð1Þ

Hence, when D�1, settling is the dominant mecha-

nism of vertical transport and, when D�1, diffusion

governs the vertical movements of phytoplankton.

The energy needed to homogenise a previously

stratified layer is equivalent to the potential energy

(EP; J m� 2) of that layer. This value was calculated

according to Simpson et al. (1978):

EP ¼1

h

Z 0

�h

ðq � qÞgzdz, ð2Þ

where h = 50 m is the thickness of the modelled layer,

which included the seasonal pycnocline, but not the

permanent pycnocline in the NEW (located at ca. 100

m). The in situ density and average depth density (qand �, respectively) were determined from CTD depth

(z) profiles (dz = 1 m) and g is the gravitational

acceleration. In the case of weak or no stratification,

S. Pesant et al. / Journal of Marine Systems 31 (2002) 261–278 265

EP is low and, in the case of a strong stratification, EP

is high.

The wind energy available for mixing a previously

stratified layer (EW; J m� 2) was calculated using a

simple equation, from Simpson et al. (1978):

EW ¼ �k�AW�3

10Dt; ð3Þ

where d is the efficiency of wind mixing (0.023), k is

the surface drag coefficient multiplied by the ratio of

wind induced current to wind speed (6.4� 10� 5), qAis the density of air (1.29 kg m � 3), and W

––10 is the

average wind speed (m s� 1) 10 m above the sea level

for a period Dt (s). We used wind data recorded on

board the ship 20 m above the sea surface, assuming

that the values were comparable to those 10 m above

the sea surface. Eq. (3) is equivalent to the equation of

Denman and Miyake (1973).

In a given size fraction, the critical cell abundance

(ACR; m� 3) for which growth is balanced by losses due

to coagulation, i.e. transfer of cells to larger aggregates,

was calculated according to Jackson (1990):

ACR ¼ 0:048lðacÞ�1r�3 ð4Þ

where l = 0.5 day � 1 is the average maximum phyto-

plankton growth rate predicted by Eppley (1972) for

water temperatures from � 2 to 4 �C, which are those

encountered in the NEW. The other terms are stickiness

(a = 0.1; Kiørboe and Hansen, 1993), shear rate (c;s � 1) and cell radius (r; m). The shear rate was

calculated as in Jackson (1990):

c ¼ ð0:13E=mÞ0:5; ð5Þ

where e (m2 s� 3) is the turbulent energy dissipation

rate, and m = 10� 6 m2 s� 1 is the kinematic viscosity

for sea water. e was estimated from friction velocity (u;

m s� 1; calculated from wind velocity) for a given

latitude (u; rad) and water depth (z; m), using a

relationship from van Aken (1984):

E ¼ 53:5u2f exp½z=ð0:068u=f Þ; ð6Þ

u ¼ ðCDqa=qwÞ1=2W���

10; ð7Þ

where f = [1.46� 10� 4sinu] is the Coriolis parameter

(s � 1), CD = 1.25� 10 � 3 is the drag coefficient,

qa = 1.29 kg m� 3 is the density of air and qw = 1026

kg m � 3 is the density of sea water.

2.5. Time series analysis

In the present paper, values in the two time series

(wind velocity and particulate carbon flux) were

transformed into binary variables, whose status are

‘event’ and ‘no event’. The ‘events’ have observed

values >median value for the total sampling period,

and the ‘no events’ are the opposite. Thus, the trans-

formed time series do not reflect the intensity of wind

or flux, but only the presence or absence of events in

time. The analysis of the corresponding, but not

necessarily synchronised, variations in the two time

series was done by ‘lag-correlation’. In the case of

qualitative time series, the lag between two series can

be determined using two-way contingency tables

(Legendre and Legendre, 1998). Contingency tables

were computed and tested for 20 positive and negative

lags between wind and flux events, a lag being equal

to 1 day. For each of the 41 resulting tables (including

one table for lag = 0), the strength of the relationship

between the two time series was quantified by com-

puting the Pearson chi-square statistic (vP2). Hence,

the term ‘lag-correlation’ is used here in a general

sense, since the computed statistic was not Pearson’s

correlation, but Pearson’s chi-square. Because wind

velocity was measured before and after the sampling

period, the time series of wind velocity is not finite so

that all the tables contained the same number of

observations. This also implies that, in each of the

41 tables being tested, the data used to construct the

table differed from the data used in the other tables, by

one observation. Given that all tables were 2� 2

(df = 1) and using a = 0.05, vP2 > 3.84 rejects the

hypothesis of independence between the two time

series. However, to account for multiple testing, we

conducted a Bonferroni correction (Legendre and

Legendre, 1998). It consisted in progressively reduc-

ing a when testing the contingency tables from no lag

to a lag of 20 or � 20, i.e. aAiA = ai = 0/AiA, where i isthe lag and ai = 0 = 0.05. Hence, for a lag of 20 or � 20

days, a = 0.0025 and vP2 > 9.97 rejected the hypothesis

of independence between the two time series. Positive

lags corresponding to significant vP2 values mean that

S. Pesant et al. / Journal of Marine Systems 31 (2002) 261–278266

wind leads the flux, and negative lags mean the

opposite.

3. Results

3.1. Time series of wind conditions

From May to September 1993, wind velocities

measured on land at the Henrik Krøyer Holmes

meteorological station (Fig. 1; Station M) and those

measured on board ranged from ca. 1 to 15 m s� 1.

Winds measured onboard and on land were signifi-

cantly correlated ( p < 0.05) with a slope close to 1

(Fig. 2). Winds measured on land were used to

construct the time series because the ship was not

INSIDE the polynya between days 170 and 182 so

that it could not provide a complete time series. Winds

measured on board were used to calculate wind-

induced mechanisms Eqs. (1)–(3) at various locations

and times in the NEW. Maximum wind velocities

occurred during May and August. During the sam-

pling period (June and July), winds were < 10 m s� 1

with an average of 3.6 m s� 1. Daily averages showed

several short events (a few days) of higher-than-

average wind velocity but, in the present study, 3-

day averages are used to match the sampling fre-

quency of the sediment trap. The 3-day averages (Fig.

3a) show four events of higher-than-average wind

velocity, with two intense events around days 150

and 180. In the present paper, the four wind events

will be identified 1 to 4 as indicated on Fig. 3a.

During the first and third events, winds blew from

Fig. 2. Relationship between wind velocity measured on land, at the

Henrik Krøyer Holmes meteorological station, and on the ship from

26 May to 27 July 1993. The relationship is significant ( p< 0.05)

and the slope is close to 1.

Fig. 3. Time series of (a) wind velocity and (b and c) downward

carbon fluxes recorded in the sediment trap (mooring G, Fig. 1) from

26 May to 27 July 1993. (a) 3-Day running means computed using

data measured at 3-h intervals at the Henrik Krøyer Holmes

meteorological station. White horizontal lines: median wind velocity

during the sampling period (3.6 m s� 1). Four events of higher-than-

median wind velocity are identified (1 to 4). (b) White horizontal

lines: median diatom C flux during the sampling period (0.67 mg C

m� 2 day� 1). Three events of higher-than-median diatom C flux are

identified (1 to 3). No data were collected before day 154. (c) Faecal

pellet C flux: no data were collected before day 154 and on days 184

to 187. Note the different scales of the C fluxes in (b) and (c).

S. Pesant et al. / Journal of Marine Systems 31 (2002) 261–278 267

the Northwest, whereas South–westerly winds domi-

nated during the remainder of the time (Koenig-

Langlo and Marx, 1997).

3.2. Time series of downward carbon fluxes

The median flux values of particulate organic

carbon (POC), diatom C, non-diatom C, and faecal

pellet C for June and July together were 1.49, 0.67,

1.09 and 0.07 mg C m � 2 day � 1, respectively (Table

1). The time series of the diatom C flux was charac-

terised by four events of higher-than-median values

which together make up >50% of the summer flux

(Fig. 3b). The first event is not considered in the

present paper because the sediment trap did not

sample before day 154 so that it is not possible to

determine the duration and the intensity of that event.

The three other events of downward C flux will be

identified 1 to 3 as indicated on Fig. 3b. Particulate

material other than intact diatoms (Fig. 3b, 5) was

mostly empty diatom frustules, faecal pellets and

debris. Faecal pellets contributed very little to the

POC flux at any time (Fig. 3c; Table 1). During the

three events, the diatom C flux and the ratio of diatom

C flux to particulate silica (PSiO2) were significantly

higher than during the ‘no events’ (Table 1). There

were no significant differences between events and

Table 1

Particulate organic carbon (POC) fluxes and ratios of fluxes in the sediment trap (mooring G in Fig. 1)

Sampling period

(days of the year)

Median values of the downward fluxes

(mg C m� 2 day� 1)

Median values of flux

ratios (mol:mol)

POC Diatom Ca

(% POC)

Non-diatom C

(% POC)

Faecal pellet C

(% POC)

POC:PON Diatom

C:pSiO2a

Partly sampled event

(154–156)

1.30 0.86 (66) 0.44 (34) 0.23 (17) 9.63 4.81

No event (157–168) 0.93 0.20 (22) 0.73 (78) 0.07 (8) 9.22 1.60

Event #1 (169–178) 3.07 2.94 (96) 0.52 (4) 0.10 (3) 9.45 4.22

No event (179–183) 1.37 0.07 (5) 1.30 (95) 0.08 (6) 8.10 0.58

Event #2 (184–186) 3.51 2.34 (67) 1.17 (33) ND ND 7.71

No event (187–192) 1.85 0.46 (25) 1.39 (75) 0.13 (7) 8.87 1.53

Event #3 (193–201) 1.31 1.01 (77) 0.33 (23) 0.07 (6) 8.34 3.10

No event (202–204) 1.41 0.47 (33) 0.94 (67) 0.05 (4) 9.54 2.04

Total sampling period

(154–204)

1.49 0.67 (32) 1.09 (68) 0.07 (5) 9.31 2.46

The total sampling period was divided according to Fig. 3 into one partly sampled event and three fully sampled events of higher-than-median

values (total sampling period). Values are also reported for periods between events, identified as ‘no event’.a Median values for ‘events’ are significantly larger than for ‘no events’ (Mann–Whitney U-test; p-value < 0.05).

Fig. 4. ‘Lag-correlation’ performed between wind and the down-

ward diatom carbon flux. Negative lags mean that the wind leads the

diatom C flux, and positive lags mean the opposite. Positive values

of Pearson’s chi-square (vP2) indicate coincidence of events in the

two time series, and negative values indicate that the wind events

did not coincide with the flux events. Data points above the hatched

line for positive values and below the hatched line for negative

values indicate a significant relationship between the two series for

the corresponding lag. Shade: significant positive relationships are

found for lags of � 16 and � 17 days.

S. Pesant et al. / Journal of Marine Systems 31 (2002) 261–278268

‘no events’ for other fluxes (POC, non-diatom C, and

faecal pellet C) or for the ratio of POC: PON fluxes.

3.3. Lag-correlation between wind and downward

carbon fluxes

In Fig. 4, negative lags mean that the independent

time series (wind) leads the dependent time series

(downward diatom C flux), and positive lags mean

the opposite. Positive values of vP2 indicate coinci-

dence of events in the two time series, and negative

values indicate that events in the wind time series do

not coincide with events in the diatom C flux time

series. The only significant coincidences of events be-

tween wind and the diatom C flux (i.e. positive chi-

square values in Fig. 4) are for lags of �16 to �17

days. There is a significant lack of coincidence of wind

events with the diatom C flux for lags of �10 to �12

and + 16 days. There are nonsignificant positive peaks

for lags of ca. � 3 and + 10 days, and a nonsignificant

Fig. 5. Relationships (a) between the two predictor variables EW/EP and AFLD/ACR, (b, c) between BLmax/BLavg

and the two predictor variables, and

(d) between EW/EP and zm/zeu for BLmax/BLavg

< 3.0 only. Dashed lines are the thresholds (see Results) for BLmax/BLavg

(3.0), AFLD/ACR (0.06), and EW/

EP (0.75). Arrows in (a) indicate the potential effect of a wind event on the two predictor variables. The different symbols correspond to four cases:

aggregation alone (AFLD/ACR > 0.06, EW/EP < 0.75; +), wind mixing alone (AFLD/ACR < 0.06, EW/EP > 0.75; w), aggregation and wind mixing

combined (AFLD/ACR > 0.06, EW/EP > 0.75;5), and no aggregation and no wind mixing combined (AFLD/ACR < 0.06, EW/EP < 0.75; .). 6 (b and d

only): EW/EP > 0.75 but AFLD/ACR was not determined.

S. Pesant et al. / Journal of Marine Systems 31 (2002) 261–278 269

negative peak for a lag of + 5 days. Hence, the fre-

quency of the wind and diatom C flux events was ca. 12

days, and wind led the diatomC flux by ca. 16 days. No

significant lags were detected between wind and other

downward fluxes (POC, non-diatomC and faecal pellet

C) determined on the trap samples (results not shown).

3.4. Wind-induced mixing and phytoplankton aggre-

gation

The potential energy of the stratified layer [EP; J

m � 2; Eq. (2)] and the wind-generated energy avail-

able to homogenise that layer [EW; J m� 2; Eq. (3)]

were calculated for stations sampled before day 170

(end of the first sampling period) and after day 182

(beginning of the second sampling period). EP was

calculated using density profiles and wind data meas-

ured onboard the ship. No water-column data are

available between days 170 and 182, because the ship

was not INSIDE the polynya at that time. EW is the

energy generated by wind during a period of 3 days

following the sampling date, i.e. average wind veloc-

ity measured on board was used in Eq. (3) (Dt= 1200

s) and EW was integrated over 259200 s. Values of EP

and EW ranged from 39 to 5200 J m � 2 and from 6 to

436 J m� 2, respectively.

The abundance of diatoms in the field (AFLD; cells

m � 3) and the calculated critical abundance [ACR;

cells m � 3; Eq. (4)] were determined at several

stations sampled before day 170 and after day 182.

In the calculations of ACR, r= 25 mm (see Section

4.2.2) and z is the depth of maximum in vivo

fluorescence. Values of AFLD and ACR ranged from

2.2� 107 to 8.7� 109 cells m� 3 and from 6.3� 108

to 3.9� 1016 cells m� 3, respectively.

3.5. Vertical distribution of phytoplankton

The vertical distribution of phytoplankton in the

large-sized fraction was quantified by using the ratio

Fig. 6. Horizontal distributions of the five cases defined in Fig. 5. Hatched areas: to the north, the Ob Bank Ice Barrier, to the south, the Norske

Ice Barrier and, to the east, the pack ice. The location of the sediment trap is indicated by the large cross, which is the origin of the two axes. The

star identifies the only station where the settling velocity (SETCOL) was >10 m day� 1.

S. Pesant et al. / Journal of Marine Systems 31 (2002) 261–278270

of maximum chl a (BLmax) in the euphotic zone to the

average integrated chl a (BLavg) for the large-sized

phytoplankton, the latter being calculated as the

integrated value over the euphotic zone divided by

the depth of that zone. The remainder of the present

paragraph is devoted to the relationships between

BLmax/BLavg

and the two predictor variables EW/EP

and AFLD/ACR. Because terms used to calculate the

three ratios are estimates, we do not expect that a

ratio = 1 would necessarily correspond to an ecologi-

cally meaningful threshold. Instead, thresholds for

BLmax/BLavg

(3.0), AFLD/ACR (0.06) and EW/EP (0.75)

were set in order to obtain the most significant

relationships in two-way contingency tables between

BLmax/BLavg

and EW/EP (Fig. 5a; vP2 = 3.87, p-value

< 0.05), and between BLmax/BLavg

and AFLD/ACR (Fig.

5b; vP2 = 2.59, 0.25 > p-value > 0.20). Hence, in the

present study, BLmax/BLavg

> 3 corresponds to a deep

chlorophyll maximum (DCM), AFLD/ACR > 0.06 cor-

responds to potential aggregation and EW/EP > 0.75

corresponds to wind mixing. For stations with BLmax/

BLavg< 3.0, there was a significant positive relationship

between EW/EP and zm/zeu (ratio of mixed layer to

euphotic zone depths; Fig. 5d; vP2 = 20.5, p-value

< 0.001) so that EW/EP > 0.75 corresponded to the

excursion of cells below the euphotic zone. Arrows

Fig. 7. Frequency distributions of settling velocity, determined at 30 stations using SETCOLs, for phytoplankton in the small- and large-sized

fractions and for unfractionated (total) phytoplankton. (a) Negative settling velocities, i.e. buoyancy. (b) The settling velocity of unfractionated

phytoplankton was maximum (12.9 m day� 1) at one station located downstream from the sediment trap, at the ice-edge of the polynya (1 in

Fig. 6).

S. Pesant et al. / Journal of Marine Systems 31 (2002) 261–278 271

in Fig. 5c show how AFLD/ACR and EW/EP would be

influenced by an increasing wind velocity. Four com-

binations of the two predictors were investigated (Fig.

5c): potential aggregation alone (AFLD/ACR > 0.06,

EW/EP < 0.75; +), wind mixing alone (AFLD /ACR <

0.06, EW/EP > 0.75; w), potential aggregation and windmixing combined (AFLD/ACR > 0.06,EW/EP > 0.75;5),

and no aggregation or wind mixing (AFLD/ACR < 0.06,

EW/EP < 0.75; ). 6 in Figs. 5 and 6 correspond to

stations where AFLD/ACR was not determined and EW/

EP > 0.75, i.e. wind mixing occurred; however, there

is no information on the potential aggregation of

phytoplankton.

The horizontal distributions of the five cases de-

fined above are shown in Fig. 6. Aggregation alone

(crosses) corresponds to a wide range of BLmax/

BLavg, which encompasses both homogeneous vertical

distributions of chl a and occurrences of a DCM

(Fig. 5b). This situation was restricted to the northern

part of the polynya, downstream of the trap. Con-

ditions of wind mixing (5, w and 6) correspond to

relatively homogeneous vertical distributions of chl a

(Fig. 5a). This situation occurred near the trap and

upstream from it and also at the northern edge of the

polynya (Fig. 6). Coincidence of wind mixing and

aggregation (5) was observed at four stations near the

trap location during the first wind event only. Coin-

cidence of no aggregation and no wind mixing oc-

curred almost everywhere INSIDE the polynya, in-

cluding at the location of the trap after the first wind

event.

The calculation of EW [Eq. (3)] assumes that there

is no ice cover, which was not true in some parts of

the study area. The presence of a ice cover reduces the

fetch, which means that, under such conditions, EW

would have been overestimated. Thus, if the influence

of the ice cover had been considered, results shown in

Figs. 5 and 6 would not be different for EW/EP < 0.75,

but could be changed at stations where EW/EP > 0.75.

Because the latter stations were always in areas with

low ice cover (comparison of our Fig. 6 with Fig. 6 in

Schneider and Budeus, 1997), our results likely reflect

the situation in the NEW.

3.6. Phytoplankton settling velocity

The results of settling columns experiments are

summarised in Fig. 7. The settling velocities of

phytoplankton in the total and large-sized fraction

were generally between 0.1 and 10 m day � 1; how-

ever, in a few experiments, the cells were buoyant, i.e.

they had negative settling velocities ranging from

� 0.01 to � 1 m day � 1. Phytoplankton in the small-

sized fraction were buoyant in most experiments

(66%); however, they sometimes settled as rapidly

as those in the total and large-sized fraction. The only

settling velocity >10 m day � 1 (12.9 m day � 1) was

recorded for phytoplankton sampled at the ice-edge on

day 200 (Fig. 6; 1). Because delicate aggregates are

potentially disrupted during the setting of SETCOL

experiments, the settling velocity of phytoplankton

was probably underestimated, especially at stations

characterised by potential aggregation (Fig. 6). Be-

cause the magnitude of that error is not known, the

results of SETCOL experiments were not used in any

calculation; however, their range is considered in the

interpretatin of the results.

4. Discussion

4.1. Efficiency of the sediment trap

It may be argued that shallow-depth deployed

conical traps, like the one used here, may under-

estimate downward fluxes (Laws et al., 1989; Baker

et al., 1988). The downward fluxes reported here are

indeed one order of magnitude lower than those

determined in the NEW based on 234Thorium fluxes

(Cochran et al., 1995), nutrient depletion (Wallace et

al., 1995) and dissolved inorganic carbon (Yager et

al., 1995). One possible explanation, in addition to

trap configuration, is the use of HgCl2 in the collect-

ing cups of the trap used in the present study. HgCl2 is

a poison, not a preservative, so that it does not prevent

solubilisation of diatom particulate carbon in the cups.

The low values of the diatom C to particulate Si ratio

in the present study (Table 1) indicate that solubilisa-

tion could indeed have occurred. Yet, the issue of

undercollection by conical traps is of little importance

to the present paper because the absolute values of the

downward flux were never used in the time series

analyses. Only relative values of the downward flux

were used so that the time series does not reflect the

intensity of the flux, but only the presence or absence

of events (higher-than-median fluxes) in time. There

S. Pesant et al. / Journal of Marine Systems 31 (2002) 261–278272

is no indication in the literature that the collection

efficiency of conical sediment traps varies in time,

especially in a low advective system as the NEW.

4.2. Mechanisms of the wind-induced downward flux

of phytoplankton

The balance (or imbalance) between turbulent dif-

fusion, which maintains cells in surface waters, and

settling of phytoplankton, which favours downward

export, is sometimes described by calculating the

parameter D [Eq. (1)]. When D < 1, settling is the

dominant mechanism of vertical transport and, when

D>1, diffusion governs the vertical movements of

phytoplankton. In Fig. 8, the arrows depict two theo-

retical wind events consisting in an increase and a

subsequent decrease of wind velocity. In one case

(arrow a), the event is associated with an increase in

the settling velocity of phytoplankton and leads to a net

decrease in D. This corresponds to a shift from dif-

fusion-dominated to settling-dominated vertical move-

ments of phytoplankton, which is considered here to

cause a downward flux. In the other case (arrow b), the

phytoplankton settling velocity remains unchanged

during the wind event so that there is no net change

in D. Below, we discuss two mechanisms that could

increase phytoplankton settling velocity during a wind

event and, thus, lead to a reduction of D (arrow a), i.e.

the aggregation of phytoplancton and the breakup of

seasonal stratification, which favours the excursion of

phytoplankton below the euphotic zone.

4.2.1. Breakup of seasonal stratification

Because silica frustules are much denser than sea-

water, e.g. Smayda (1970),Villareal (1988), healthy

diatoms should settle in the absence of physiological

buoyancy. The only buoyancy mechanism demonstra-

ted to date in marine diatoms is the ionic pump

(Anderson and Sweeney, 1977) that influences the cell

density by transferring ions across the cell membrane.

This mechanism requires energy, which is produced by

photosynthesis. Limitation of photosynthesis by light

or nutrients would disable the buoyancy control of

phytoplankton. Laboratory and field studies have

indeed shown that the settling velocity of diatoms is

influenced by nutrient and light conditions, e.g. Bien-

fang (1981), Harrison et al. (1986), Waite et al.

(1992a,b) and Waite et al. (1997). Under nutrient de-

plete conditions, phytoplankton would settle to the

nutricline and recover buoyancy at that depth unless

the nutricline is below the euphotic zone. The former

case would contribute to the formation of a DCM,

whereas, in the latter case, irradiance would be limiting

and phytoplankton would settle to depth regardless of

nutrient concentration. In the NEW, a continuous

supply of nutrient-rich water occurred upstream from

the sediment trap (Kattner and Budeus, 1997) so that

the latter case probably did not often occur in the pre-

sent study. Downstream from the trap, nutrients were

very low and, thus, could have influenced the buoy-

ancy control and the aggregation of phytoplankton;

however, that area of the polynya is of little importance

with respect to the aim of the present paper.

Irradiance limitation can occur during night; how-

ever, during our sampling season, the sun never set.

Irradiance can also become limiting during a wind

event, because of the deepening of the surface mixed

layer and accompanying reduction of the average

irradiance experienced by phytoplankton during their

vertical excursion (Estrada and Berdalet, 1997). More-

over, in seasonally ice-covered seas, the strong input

Fig. 8. Theoretical influence of wind velocity and particle settling

velocity on D (isolines). The arrows depict two theoretical wind

events with, in both cases, an increase and a subsequent decrease of

wind velocity. Arrow a: The event is associated with an increase in

the settling velocity of phytoplankton and leads to a net decrease in

D. Arrow b: The phytoplankton settling velocity remains unchanged

during the wind event and leads to no net change in D.

S. Pesant et al. / Journal of Marine Systems 31 (2002) 261–278 273

of melt water from the continent and sea–ice accel-

erates the restratification following wind events so

that phytoplankton entrained to depth during the wind

event can be trapped below a newly formed pycno-

cline. The trapped cells would experience a strong

reduction in irradiance and sink. The potential dis-

ruption of shallow stratification by wind in the NEW

was quantified by comparing the potential energy of

the stratified layer [EP; J m � 2; Eq. (2)] to the wind

generated energy that is available to homogenise that

layer [EW; J m� 2; Eq. (3)]. In the NEW, EW/EP > 0.75

corresponded to vertically homogeneous distributions

of chl a in the mixed layer (Fig. 5a) and to mixing

below the euphotic zone (Fig. 5d).

4.2.2. Aggregation of phytoplankton

Most studies assume that, according to Stokes’ law,

aggregation should increase the downward flux of

particulate matter. This is, however, not always true as

discussed below after considering the calculation of

potential aggregation. Jackson (1990) developed a

model that provides a simple way to estimate the

aggregation potential in a system, by calculating a

critical phytoplankton abundance (ACR) for which the

growth of a size fraction is balanced by transfer to a

larger size fraction due to coagulation. When phyto-

plankton concentrations are >ACR, the model predicts

a rapid coagulation of cells into larger aggregates. ACR

is sensitive to cell size and shear rate [see Eq. (3)], the

latter being a function of wind velocity and depth [see

Eqs. (4) and (5)] so that ACR decreases with increasing

wind velocity and increases with depth. Using a

diatom radius of 10 mm (reasonable for a single cell)

and the highest wind velocity observed during sam-

pling (9.8 m s� 1), ACR was >1010 cells m� 3 in the

NEW. Diatom abundances determined on water sam-

ples collected in the NEW (AFLD) were always < 1010

cells m � 3 (not shown) so that, according to Jackson’s

model, diatoms in the NEW were never abundant

enough to trigger rapid aggregation of phytoplankton.

Jackson’s model, however, assumes that all phyto-

plankton settle at the same rate, which is the case in

monospecific blooms. Because the model does not

take into account differential settling (i.e. collision

between particles settling at different rates), it is

expected that Eq. (3) would overestimate ACR in the

case of mixed diatom assemblages such as in the

NEW (Pesant et al., 1996; Booth and Smith, 1997;

Hellum von Quillfeldt, 1997). To correct for this, we

consulted the work of Riebesell (1992). During a

diatom bloom in the North Sea, he estimated ACR to

be 1.2� 109 cells m � 3, based on measurements of

diatom concentrations and total aggregate volumes. In

order to obtain the observed ACR values with Jack-

son’s model, Riebesell had to use a cell radius of 25

mm. We used that value to calculate ACR for diatoms in

the NEW (see Materials and Methods).

The settling velocity of phytoplankton depends on

the sizes and shapes of cells or aggregates and on the

difference of density between water and cells. The

first wind-induced mechanism (breakup of the sea-

sonal stratification), which is quantified here using

EW/EP, acts mainly on the density of phytoplankton,

whereas the second (cell aggregation), which is quan-

tified here using AFLD/ACR, acts mainly on the sizes

and shapes of the settling particles. There is increasing

evidence in laboratory and field studies that there is no

obligate relationship between particle size and settling

velocity for metabolically active diatoms (e.g. Riebe-

sell, 1992; Waite et al., 1992b; Diercks and Asper,

1997; Waite et al., 1997). Hence, in spite of their large

sizes, aggregates composed of metabolically active

cells can be buoyant so that the effect of aggregation

on the settling velocity of phytoplankton should be

effective only when there is no buoyancy control, e.g.

when EW/EP > 0.75. We conclude that aggregation

increases the range of possible settling velocities for

phytoplankton; however, potential aggregation alone

was not a sufficient condition for a downward flux of

phytoplankton in the NEW.

We propose that, in our study, the disruption of the

seasonal stratification by winds triggered the events of

downward flux, whereas aggregation determined the

settling velocity of phytoplankton once exported

below the euphotic zone. In the NEW, EW/EP > 0.75

combined with AFLD/ACR > 0.06 occurred at four sta-

tions (Fig. 5a), which were located near the trap (Fig.

6; 5), where the two mechanisms probably contrib-

uted to a rapid downward flux of intact diatoms. This

can explain, however, only the first partially sampled

flux event because these four stations were all

sampled during the first wind event. Other studies

conducted in the Baltic and Mediterranean Seas (Sjo-

berg and Wilmot, 1977; Miquel et al., 1994) have

come to the conclusion that wind events initiate the

downward flux of phytoplankton. Legendre and Ras-

S. Pesant et al. / Journal of Marine Systems 31 (2002) 261–278274

soulzadegan (1996) proposed that the proportion of

primary production that is exported downwards

should be proportional to the frequency of destabili-

sation–stabilisation of the surface mixed layer. It

follows that the frequency of wind events should

determine the magnitude of the summer downward

flux. In the NEW, time series of winds during sum-

mers 1990 to 1993 (Schneider and Budeus, 1997)

show a consistent pattern of fluctuating wind condi-

tions, with a frequency of ca. one event every 15 to 25

days. Unfortunately, sediment traps were only de-

ployed in 1993 so that interannual comparison of

wind and downward flux time series was not possible.

4.3. Contributions of phytoplankton production and

faecal pellets to the downward flux

The contribution of the diatom C flux to the POC

flux indicates that the major events of downward flux

recorded in the sediment trap mostly consisted of intact

diatoms from the euphotic zone, whereas between

events, the C flux mostly consisted of degraded mate-

rial (Table 1). Lower values of the diatom C: PSiO2

flux ratio (Table 1) between the events of downward

flux also indicate the dominance of empty diatom frus-

tules over intact diatoms, at these times. The contri-

bution of faecal pellets to the downward C flux was

almost negligible during the flux events (Table 1),

which may have been caused by the disaggregation of

faecal pellets, whose content thus contributed to the

flux of diatoms. This possibility could not be assessed

in the present study. Contrary to intact diatoms, debris

and faecal pellets are inert particles which should settle

according to Stokes’ law. This is consistent with the

observed constancy of the ‘non-diatom’ and faecal

pellet C fluxes (Table 1), which suggests that the small

downward flux of debris and pellets from the euphotic

zone was continuous and independent of wind events.

Near the trap location, the faecal pellet C flux repre-

sented ca. 0.2% to 0.04% of the potential C egestion of

copepods in the top 50 m (i.e. 27–170 mg C m � 2

day � 1; Daly, 1997), which suggests that faecal pellet

C was mostly recycled within surface waters. Several

mechanisms of pellet degradation have been proposed

and demonstrated in the literature, including ingestion

and disruption by copepods and membrane lysis (Paf-

fenhofer and Knowles, 1979; Lampitt et al., 1990; Noji

et al., 1991).

A recurrent problem in estimating the proportion of

phytoplankton production collected in sediment traps

is to account for the spatial and temporal decoupling

between production, in the surface waters, and down-

ward flux, at depth. In most studies, a close coupling is

assumed, i.e. the sedimentation flux is compared to

phytoplankton production sampled simultaneously at

the trap location. In some studies, however, phyto-

plankton characteristics at the trap location were found

to differ from those in the area where trapped particles

originated, e.g. Jaeger et al. (1996) and Reigstad and

Wassmann (1996). In the present study, lag-correlation

analyses indicate that wind led the diatom C flux by ca.

16 days. According to the average current velocity

along the NEW gyre (ca. 10 cm s� 1; Johnson and

Niebauer, 1995), a lag of 16 days corresponded to

lateral advection of ca. 130 km, upstream from the se-

diment trap. Assuming that phytoplankton settled over

90 m, i.e. from the bottom of the surface mixed layer

(average 40 m) to the depth of the trap (130 m), the 16-

day lag corresponded to settling velocities of ca. 6 m

day � 1. Hence, a first scenario explaining the events of

diatom C flux measured in the trap is that phytoplank-

ton originated from the southern limit of the polynya

(ca. 130 km upstream from the trap) and settled slowly

(ca. 6 m day � 1). A second scenario, which assumes a

tight coupling, would be that phytoplankton originated

near the location of the trap (e.g. 10 km) and settled

rapidly (e.g. 90 m day � 1). The second scenario is only

hypothetical because it is not based on significant data.

The following comparison of the two scenarios pro-

vides evidence that the usual assumption of a tight

coupling between the surface and the trap can mislead

scientists when assessing the proportion of phyto-

plankton production collected in sediments traps.

According to the first scenario, the origin of algae

collected in the trap was 130 km upstream from the

trap, where the wind-induced vertical excursion of

phytoplankton below the euphotic zone could have

triggered the downward flux of phytoplankton; how-

ever, the settling velocity of cells was probably low

because the formation of fast settling aggregates did

not take place (Fig. 6). The range of phytoplankton

settling velocities in our SETCOLs (Fig. 7) and

reported in the literature for actively growing cells

are consistent with the settling velocities calculated

from the results of lag-correlation analyses (ca. 6 m

day � 1). Hence, the results derived from three inde-

S. Pesant et al. / Journal of Marine Systems 31 (2002) 261–278 275

pendent methods, i.e. lag-correlation, horizontal dis-

tributions of EW/EP, AFLD/ACR, and SETCOL deter-

minations, support the first scenario.

According to the second hypothetical scenario, the

algae collected in the trap would have come from

neighbouring waters, where aggregation potentially

occurred, and the seasonal stratification could have

been disrupted by winds (Fig. 6; 5). This last condi-

tion, however, was met during the first wind event

only (Fig. 3). The size of aggregates in the NEW was

potentially as large as those observed by Riebesell in

the North Sea (estimated spherical diameter up to 3.5

mm or 1010 mm3) because the conditions of wind,

mixed layer depth and maximum cell abundance were

similar. According to Fortier et al. (1994), aggregates

of that size would settle with velocities of 10 to 100 m

day � 1, which is consistent with the settling velocity

predicted for the second scenario (90 m day � 1). The

second scenario disagrees, however, with settling

velocities determined in the present study but, as

explained in Materials and Methods, our experiments

probably underestimated the settling velocity of phy-

toplankton. Because the combination of wind-induced

aggregation and disruption of stratification at the trap

location occurred only during the first wind event (not

shown), the second scenario could not have caused

the three downward flux events in the present study,

but could have contributed to the flux around day 155

(i.e. synchronised with the first wind event), which

was only partly sampled. The maximum flux of faecal

pellets (17% POC flux), which occurred at that time,

could have also been caused by aggregation in the

euphotic zone.

We have shown (first scenario) that particles col-

lected in the trap would have originated from an area

where there was a nonbloom regime (Pesant et al.,

1996, their Regime 5), characterised by low produc-

tion rates of large-sized phytoplankton (PL; median

value = 49 mg C m� 2 day � 1). Using this value, the

three diatom C flux events corrected using a factor 10

(see beginning of Discussion) corresponded to 60%,

48%, and 21% of PL, respectively, whereas the down-

ward flux between events was < 10% of PL. In con-

trast, the location of the sediment trap was charac-

terised by relatively high PL, (median value = 451 mg

C m � 2 day � 1; Pesant et al. (1996), their Regime 1) so

that if we had assumed a tight coupling between the

surface and the trap (second scenario), the downward

diatom C fluxes during wind events would have repre-

sented only 2% to 7% of PL. The difference between

the two scenarios has great implications for the cycling

of phytoplankton C since the proportion of primary

production exported below the euphotic zone differs

by one order of magnitude between the two scenarios.

5. Conclusions

The recent literature suggests that the residence

time and the excursion of phytoplankton in the sea-

sonal mixed layer has a significant effect on the

pathways of carbon cycling (e.g. Mann and Lazier,

1991; Estrada and Berdalet, 1997; Margalef, 1997). If

phytoplankton accumulate in the mixed layer, a large

fraction of particulate primary production can be

consumed by zooplankton and transferred to the

dissolved carbon pool, which favours recycling within

the mixed layer. In contrast, when phytoplankton

settle below the mixed layer, the likelihood of deep

export increases. Our results show that 21% to 60% of

the diatom C produced in the euphotic zone was ex-

ported downwards during wind events, whereas the

diatoms were maintained in the mixed layer during

periods of low winds. Thus, in the NEW, wind events

modified the pathways of phytoplankton C cycling,

i.e. recycling within vs. export below the mixed layer.

Our results also stress the importance of investigating

the spatial coupling between surface and trap data

when assessing the pathways of carbon cycling.

Acknowledgements

We thank the master and crew of the R.V.

‘Polarstern’ for their efficient support, G. Bergeron,

C. Fraiken, O. Haupt, M. Krumbholz, S. Lessard, F.

McGuiness for work in the field, and C. Belzile and B.

Klein for their assistance in laboratory or data ana-

lyses. The Danish Meteorological Institute, the Alfred

Wegener Institute for Polar and Marine Research and

the Ottawa microwave group for providing data. The

authors also thank U. Bathmann and U. Riebesell for

comments at an early stage of this work, and three

anonymous reviewers for helpful comments on the

manuscript. This research was funded by a Collabora-

tive Special Project grant from the Natural Sciences

S. Pesant et al. / Journal of Marine Systems 31 (2002) 261–278276

and Engineering Research Council of Canada

(NSERC) and by grants to L.L. and M.G. from

NSERC, to GIROQ (Groupe interuniversitaire de re-

cherches oceanographiques du Quebec) from NSERC

and the Fonds (FCAR) of Quebec, and to S.P. from the

Deutscher Akademischer Austauchdienst (DAAD)

and the Sonderforschungsbereich (SFB) 313. This is

a contribution to the programme of GIROQ and of the

Alfred Wegener Institute for Polar and Marine

Research.

References

Alldredge, A.L., Jackson, G.A., 1995. Aggregation in marine sys-

tems. Deep-Sea Res. 42, 1–7.

Alldredge, A.L., Gotshalk, C.C., MacIntyre, S., 1987. Evidence for

sustained residence of macrocrustacean fecal pellets in surface

waters off Southern California. Deep-Sea Res. 34, 1641–1652.

Anderson, D.M., Sweeney, B., 1977. Diel changes in the sedimen-

tation characteristics of Dityllum brightwellii, a marine centric

diatom: changes in cellular lipids and effects of respiratory

inhibitors and ion-transport modifiers. Limnol. Oceanogr. 22,

539–552.

Baker, E.T., Milburn, H.B., Tennant, D.A., 1988. Field assessment

of sediment trap efficiency under varying flow conditions. J.

Mar. Res. 46, 573–592.

Bauerfeind, E., Garrity, C., Krumbholz, M., Ramseier, R.O., Voß,

M., 1997. Seasonal variability of sediment trap collections in the

Northeast Water Polynya: Part 2. Biochemical and microscopic

composition of sedimenting matter. J. Mar. Syst. 10, 371–389.

Bienfang, P.K., 1981. SETCOL: a technologically simple and reli-

able method for measuring the phytoplankton sinking rates.

Can. J. Fish. Aquat. Sci. 38, 1289–1294.

Bienfang, P.K., Syper, J., Laws, E., 1983. Sinking rate and pigment

responses to light-limitation of a marine diatom: implications to

dynamics of chlorophyll maximum layers. Oceanol. Acta 6,

55–62.

Bodungen, B.V., Wunsch, M., Furderer, H., 1991. Sampling and

analysis of suspended particles in the North Atlantic. In: Hurt,

D.C., Spencer, D.W. (Eds.), Marine Particles: Analyses and

Characterization. Am. Geophys. Union, Geophys. Monogr.

Ser., vol. 63, pp. 47–53.

Booth, B.C., Smith Jr., W.O., 1997. Autotrophic flagellates and

diatoms in the Northeast Water Polynya, Greenland: summer

1993. J. Mar. Syst. 10, 241–261.

Budeus, G., Schneider, W., 1996. The 1993 Northeast Water Ex-

pedition data report on CTD measurements of RV ‘Polarstern’

cruises ARKTIS IX/2 and 3. Ber. Polarsforch. 199, 1–340.

Budeus, G., Schneider, W., Brunßen, J., Damm, M., Plugge, R.,

1994. Ship-borne physical oceanography. Ber. Polarforsch.

142, 16–105.

Cochran, J.K., Barnes, C., Achman, D., Hirchberg, D.J., 1995.

Thorium-234/Uranium-238 disequilibrium as an indicator of

scavenging rates and particulate organic carbon fluxes in the

Northeast Water Polynya, Greenland. J. Geophys. Res. 100,

4399–4410.

Daly, K.L., 1997. Flux of particulate matter through copepods in the

Northeast Water Polynya. J. Mar. Syst. 10, 319–342.

Denman, K.L., Miyake, M., 1973. Upper layer modification at

ocean station Papa: observations and simulation. J. Phys. Ocean-

ogr. 18, 185–196.

Diercks, A.-R., Asper, V.L., 1997. In situ settling speeds of marine

snow aggregates below the mixed layer: Black Sea and Gulf of

Mexico. Deep-Sea Res. 44, 385–398.

Edler, L., 1979. Recommendations on the methods for marine bio-

logical studies in the Baltic Sea. Phytoplankton and chlorophyll.

Balt. Mar. Biol. Publ. 5, 1–38.

Eppley, R.W., 1972. Temperature and phytoplankton growth in the

sea. Fish. Bull. 70, 1063–1085.

Estrada, M., Berdalet, E., 1997. Phytoplankton in a turbulent world.

Sci. Mar. 61, 125–140.

Fortier, L., Le Fevre, J., Legendre, L., 1994. Export of biogenic

carbon to fish and to the deep ocean: the role of large planktonic

microphages. J. Plankton Res. 16, 809–839.

Franks, P.J.S., Marra, J., 1994. A simple new formulation for phy-

toplankton photoresponse and an application in a wind-driven

mixed-layer model. Mar. Ecol.: Prog. Ser. 111, 143–153.

Gawarkiewicz, G., Chapman, D.C., 1995. A numerical study of

dense water formation and transport on a shallow, sloping con-

tinental shelf. J. Geophys. Res. 100, 4489–4509.

Gonzalez, H.E., Gonzalez, S.R., Brummer, G.-J.A., 1994. Short-

term sedimentation pattern of zooplankton, faeces and micro-

plankton at a permanent station in the Bornafjorden (Norway)

during April–May 1992. Mar. Ecol.: Prog. Ser. 105, 31–45.

Harrison, P.J., Turpin, D.H., Bienfang, P.K., Davis, C.O., 1986.

Sinking as a factor affecting phytoplankton species succession:

the use of selective loss semi-continuous cultures. J. Exp. Mar.

Biol. Ecol. 99, 19–30.

Hellum von Quillfeldt, C., 1997. Distribution of diatoms in the

Northeast Water Polynya, Greenland. J. Mar. Syst. 10, 211–240.

Hitchcock, G.L., Langdon, C., Smayda, T.J., 1987. Short-term

changes in the biology of a Gulf Stream warm-core ring: zoo-

plankton biomass and grazing. Limnol. Oceanogr. 32, 919–928.

Jackson, G.A., 1990. A model of the formation of marine flocs by

physical coagulation processes. Deep-Sea Res. 37, 1197–1211.

Jaeger, J.M., Nittrouer, C.A., DeMaster, D.J., Kelchner, C., Dunbar,

R.B., 1996. Lateral transport of settling particles in the Ross Sea

and implications for the fate of biogenic material. J. Geophys.

Res. 101, 18479–18488.

Johnson, M.A., Niebauer, H.J., 1995. The 1992 summer circulation

in the Northeast Water Polynya from acoustic doppler current

profiler measurements. J. Geophys. Res. 100, 4301–4307.

Kattner, G., Budeus, G., 1997. Nutrient status of the Northeast

Water Polynya. J. Mar. Syst. 10, 185–197.

Kiørboe, T., 1993. Turbulence, phytoplankton cell size, and the

structure of pelagic food webs. Adv. Mar. Biol. 29, 1–61.

Kiørboe, T., Hansen, J.L.S., 1993. Phytoplankton aggregate forma-

tion: observations of patterns and mechanisms of cell sticking

and the significance of exopolymeric material. J. Plankton Res.

15, 993–1018.

S. Pesant et al. / Journal of Marine Systems 31 (2002) 261–278 277

Kiørboe, T., Nielsen, T.G., 1990. Effects of wind stress on vertical

water column structure, phytoplankton growth, and productivity

of planktonic copepods. In: Barnes, M., Gibson, R.N. (Eds.),

Trophic Interactions in the Marine Environment. Proc. 24th Eur.

Mar. Biol. Symp., pp. 28–40.

Koenig-Langlo, G., Marx, B., 1997. The meteorological informa-

tion system at the Alfred Wegener Institute. In: Lautenschlager,

M., Reinke, M. (Eds.), Climate and Environmental Database

Systems. Kluwer Academic Publishing, Norwell, pp. 117–125.

Lampitt, R., Noji, T.T., Bodungen, B.V., 1990. What happens to

zooplankton faecal pellets? Implication for material flux. Mar.

Biol. 104, 15–23.

Laws, E.A., DiTullio, G.R., Betzer, P.R., Karl, D.M., Carder, K.L.,

1989. Autotrophic production and elemental fluxes at 26 degree

N, 155 degree W in the North Pacific subtropical gyre. Deep-

Sea Res. 36, 103–120.

Legendre, L., Legendre, P., 1998. Numerical Ecology. Elsevier,

Amsterdam, 853 pp.

Legendre, L., Rassoulzadegan, F., 1996. Food-web mediated export

of biogenic carbon in oceans: hydrodynamic control. Mar. Ecol.:

Prog. Ser. 145, 179–193.

Lund, J.W.G., Kipling, C., Le Cren, E.D., 1958. The inverted micro-

scope method of estimating algal numbers and the statistical

basis of estimation by counting. Hydrobiologia 11, 143–170.

Mann, K.H., Lazier, J.R.N., 1991. Dynamics of Marine Ecosystems.

Blackwell Scientific Publications, Boston.

Margalef, R., 1997. Turbulence and marine life. Sci. Mar. 61, 109–

123.

Miquel, J.C., Fowler, S.W., La Rosa, J., Buat-Menard, P., 1994. Dy-

namics of the downward flux of particles and carbon in the open

northwestern Mediterranean Sea. Deep-Sea Res. 41, 243–261.

Noji, T.T., Estep, K.W., MacIntyre, F., Norrbin, F., 1991. Image

analysis of faecal material grazed upon by three species of co-

pepods: evidence for coprophagy and coprochaly. J. Mar. Biol.

Assoc. U. K. 71, 465–480.

Paffenhofer, G.A., Knowles, S.C., 1979. Ecological implications of

fecal pellet size, production and consumption by copepods. J.

Mar. Res. 37, 35–49.

Parsons, T.R., Maita, Y., Lalli, C.M., 1984. A Manual of Chemical

and Biological Methods for Seawater Analysis. Pergamon, Tor-

onto, Canada.

Pesant, S., Legendre, L., Gosselin, M., Smith, R.E.H., Kattner, G.,

Ramseier, R., 1996. Size-differential regimes of phytoplankton

production in the Northeast Water Polynya (77�–81�N). Mar.

Ecol.: Prog. Ser. 142, 75–86.

Pesant, S., Legendre, L., Gosselin, M., Ashjian, C., Booth, B., Daly,

K., Fortier, L., Hirche, H.-J., Michaud, J., Smith, R.E.H., Smith,

S., Smith Jr., W.O., 1998. Pathways of carbon cycling in the

euphotic zone: the fate of large-sized phytoplankton in the

Northeast Water Polynya. J. Plankton Res. 20, 1267–1291.

Ramseier, R.O., Bauerfeind, E., Garity, C., Walsh, I.D., 1997. Sea-

sonal variability of sediment trap collections in the Northeast

Water Polynya: Part 1. Sea– Ice parameters and particle flux. J.

Mar. Syst. 100, 359–369.

Reigstad, M., Wassmann, P., 1996. Importance of advection for

pelagic–benthic coupling in north Norwegian fjords. Sarsia

80, 245–257.

Riebesell, U., 1992. The formation of large marine snow and its

sustained residence in surface waters. Limnol. Oceanogr. 37,

63–76.

Ruiz, J., 1996. The role of turbulence in the sedimentation loss of

pelagic aggregates from the mixed layer. J. Mar. Res. 54, 385–

406.

Schneider, W., Budeus, G., 1997. Summary of the Northeast Water

Polynya formation and development (Greenland Sea). J. Mar.

Syst. 100, 107–122.

Simpson, J.H., Allen, C.M., Morris, N.C.G., 1978. Fronts on the

continental shelf. J. Geophys. Res. 83, 4607–4614.

Sjoberg, S., Wilmot, W., 1977. System Analysis of a Spring Phy-

toplankton Bloom in the Baltic. Stockholm Univ. Press, Stock-

holm.

Smayda, T.J., 1970. The suspension and sinking of phytoplankton

in the sea. Mar. Biol. Annu. Rev. 8, 353–414.

Smith Jr., W.O., 1995. Primary productivity and new production in

the Northeast Water (Greenland) Polynya during summer 1992.

J. Geophys. Res. 100, 4357–4370.

Smith Jr., W.O., Walsh, I.D., Booth, B.C., Deming, J.W., 1995.

Particulate matter and phytoplankton and bacterial biomass dis-

tributions in the Northeast Water Polynya during summer 1992.

J. Geophys. Res. 100, 4341–4356.

Steele, J.H., Yentsch, C.S., 1960. The vertical distribution of chlor-

ophyll. J. Mar. Biol. Assoc. U. K. 39, 217–226.

Strathmann, R.R., 1967. Estimating the organic carbon content of

phytoplankton from cell volume or plasma volume. Limnol.

Oceanogr. 12, 411–418.

van Aken, H.M., 1984. A one-dimensional mixed-layer model for

stratified shelf Seas with tide- and wind-induced mixing. Dtsch.

Hydrogr. Z. 37, 1–27.

Villareal, T., 1988. Positive buoyancy in the oceanic diatom Rhizo-

solenia debyana H. Peragallo. Deep-Sea Res. 35, 1037–1045.

Waite, A.N., Bienfang, P.K., Harrison, P.J., 1992a. Spring bloom

sedimentation in a subarctic ecosystem: I. Nutrient sensitivity.

Mar. Biol. 114, 119–129.

Waite, A.N., Thompson, P.A., Harrison, P.J., 1992b. Does energy

control the sinking rates of marine diatoms? Limnol. Oceanogr.

37, 468–477.

Waite, A.N., Fisher, A., Thompson, P.A., Harrison, P.J., 1997. Sink-

ing rate versus cell volume relationships illuminate sinking rate

control mechanisms in marine diatoms. Mar. Ecol.: Prog. Ser.

157, 97–108.

Wallace, D.W.R., Minnett, P.J., Hopkins, T.S., 1995. Nutrient, oxy-

gen, and inferred new production in the Northeast Water Poly-

nya, 1992. J. Geophys. Res. 100, 4323–4340.

Webster, I.T., Hutchinson, P.A., 1994. Effect of wind on the distri-

bution of phytoplankton cells in lakes revisited. Limnol. Ocean-

ogr. 39, 365–373.

Yager, P.L., Wallace, D.W.R., Johnson, K.M., Smith Jr., W.O., Min-

nett, P.J., Demings, J.W., 1995. The Northeast Water Polynya as

an atmospheric CO2 sink: A seasonal rectification hypothesis. J.

Geophys. Res. 100, 4389–4398.

Yamazaki, H., Kamykowski, D., 1991. The vertical trajectories of

motile phytoplankton in a wind-mixed water column. Deep-Sea

Res. 38, 219–241.

S. Pesant et al. / Journal of Marine Systems 31 (2002) 261–278278