Seasonal patterns in Arctic planktonic metabolism (Fram Strait – Svalbard region)

. The metabolism of the Arctic Ocean is marked by extremely pronounced seasonality and spatial heterogeneity associated with light conditions, ice cover, water masses and nutrient availability. Here we report the marine planktonic metabolic rates (net community production, gross primary production and community respiration) along three different seasons of the year, for a total of eight cruises along the western sector of the European Arctic (Fram Strait – Svalbard region) in the Arctic Ocean margin: one at the end of 2006 (fall/winter), two in 2007 (early spring and summer), two in 2008 (early spring and summer), one in 2009 (late spring–early summer), one in 2010 (spring) and one in 2011 (spring). The results show that the metabolism of the western sector of the European Arctic varies throughout the year, depending mostly on the stage of bloom and water temperature. Here we report metabolic rates for the different periods, including the spring bloom, summer and the dark period, increasing considerably the empirical basis of metabolic rates in the Arctic Ocean, and especially in the The results indicated that the observed GPP on % of the maximum and a maximum observed value of 57.8 % in one of the


Introduction
The climate of the Arctic marine environment is characterized by extreme seasonality in solar radiation, ice cover and atmospheric temperature and, to a lesser extent, water temperature Carmack and Wassmann, 2006). This variability should be reflected in significant variability in the pelagic metabolism of the Arctic Ocean during extreme transitions from complete darkness to continuous daylight, with negligible photosynthetic primary production during the extended dark period and respiration rates affected by the ensuing variability in the supply of organic matter and changes in water temperature from winter to summer. Hence, community respiration must prevail over primary production in the dark, while primary production can be quite high during the light period (Hodal and Kristiansen, 2008), when plankton communities receive photosynthetically-active radiation (PAR) 24 h per day (Sakshaug and Slagstad, 1991;Sakshaug et al., 1994). However, respiration rates are also expected to increase in the summer due to increased temperatures and increased supply of dissolved organic matter. Hence, both gross primary production and respiration rates are expected to show high seasonal variability in the Arctic Ocean. Additionally, increased advection of Atlantic waters into the Arctic generates high spatial variability and fronts (Dmitrenko et al., 2008;Ivanov et al., 2009), which may mask the seasonal signal of planktonic metabolism.
Although estimates of Arctic primary production are available (e.g. Rao and Platt, 1984;Sakshaug, 1997Sakshaug, , 2004Wassmann et al., 2006a;Pabi et al., 2008), reports of direct measurements of planktonic metabolism in the Arctic are sparse, much more so than those for Antarctic waters (e.g. Agusti et al., 2004;Agusti and Duarte, 2005;Dickson and Orchardo, 2001;Lefèvre et al., 2008;Robinson et al., 1999), and are limited to few publications, as one report of summer metabolism in the coastal waters of the Chukchi Sea sector (Cottrell et al., 2006), two reports from the Canadian Basin, reporting only respiration rates, just one of the components involved in the assessment of metabolic balance (Apollonio, 1980;Sherr and Sherr, 2003), four reports of summer primary production assessed using 14 C, two in the Chukchi Sea (Hameedi, 1978;Cota et al., 1996), one in the Baffin Bay (Harrison et al., 1982), one in the central Arctic (Olli et al., 2007), and one reporting summer metabolism (gross primary production, community respiration and net community production) in 2007 for the region studied here (Regaudie-de-Gioux and . This last study is included here to provide a more complete assessment of the metabolism in this area, as it was conducted in the same area using the same methods. There are a considerable number of studies reporting integrated values for planktonic metabolism (e.g. English, 1961;Sokolova and Solovyeva, 1971;Alexander, 1974; Subba Rao and Platt, 1984;Hodal and Kristiansen, 2008;Ardyna et al., 2011). However, as integration depths vary between studies, we have not included the data in our analyses. Whereas the previous observational data were insufficient, the set of estimates reported here provides the first empirical basis with which to establish patterns in the seasonal variability in planktonic metabolism in the European sector of the Arctic Ocean. Additionally it allows us to provide a first approximation at the annual balance between gross primary production and plankton respiration in these communities. Although the estimates are rough, the seasonal coverage at the regional scale provided here compares favourably with the state of knowledge available for any other ocean region in the world (Robinson and Williams, 2005).
The characterisation of the seasonal patterns of variability in plankton community metabolism in the Arctic Ocean is not only important to gain additional understanding on the functioning of these communities and their role in the regional carbon budget, but it is also essential to provide baseline data to detect changes in Arctic planktonic metabolism with climate change. The Arctic Ocean is warming at rates three times faster than the average rate of warming of the global ocean (ACIA, 2004;Trenberth et al., 2007) and is projected to continue to do so in the future (Houghton, 2005;Walsh, 2008). Indeed, impacts are already evident as the summer ice cover experienced a sudden decline resulting in a historical minimum in the summer of 2007, with a 43 % reduction in the minimum ice extent relative to the ice extent in 1979, a loss equivalent to more than twice the area of Alaska (Kerr, 2007), and a reduction of more than the 40 % of multiyear ice volume from 2005 to 2008 (Kwok et al., 2009). Recently, a new historical minimum has reached in September 2012, with a decrease of 760 000 km 2 below the previous record minimum extent in 2007 (http://nsidc.org/arcticseaicenews/). Reduced ice cover increases underwater irradiance to support primary production and may also, because of the enhanced supply of photosynthetic organic matter, leads to increased plankton community respiration in Arctic waters. Warming is also expected to directly affect metabolic rates, as temperature plays an important role in regulating metabolic processes (Iriberri et al., 1985;White et al., 1991), and metabolic rates are expected to increase exponentially with water temperature (Brown et al., 2004).
Here we evaluate seasonal and spatial variability in planktonic gross primary production (GPP), net community production (NCP) and community respiration (CR) in the Fram Strait and Svalbard coastal waters of the European Sector of the Arctic Ocean. Here we address the questions of whether the Western European Arctic sector is net autotrophic at the annual scale and whether the excess production during the light period suffices to meet the respiratory requirements during the Arctic dark period. We do so on the basis of eight cruises conducted in four contrasting periods of the year, late fall-early winter 2006, spring 2007, 2008, 2010 and 2011, late spring-early summer 2009 and the summers of 2007 and 2008 (Fig. 1).

Research area
The Fram Strait, located between Greenland and Svalbard, connects the North Atlantic and the Arctic Ocean with an important heat and mass exchange, with large quantities of heat transported poleward by the extended North Atlantic Current; the West Spitsbergen Current (WSC), which influences the climate in the Arctic region as a whole (Fig. 1, Hop et al., 2006). Ice outflow from the Arctic occurs at the western part of the Fram Strait along the East Greenland Current (EGC, Schlichtholz and Houssais, 2002). The circulation is characterized by a generally southward EGC system on the western side along the Greenland slope and shelf, and a generally northward WSC system in the eastern side. The WSC and EGC exchange water though counter-clockwise recirculation (Schlichtholz and Houssais, 2002). The northward transport of warm Atlantic Water (AW) melts southward-drifting ice and maintains open waters north of Svalbard (Rudels et al., 2000). This area is hydrographically complex, including sharp gradients in plankton communities. During the cruise conducted in summer 2007 a pronounced intrusion of Atlantic waters was found north of Spitsbergen, with 71 % of the stations in this area containing AW. The Kongsfjorden-Krossfjorden fjord system is situated on the west coast of Spitsbergen (Svalbard), or at the eastern extreme of the Fram Strait (Fig. 1). This fjord system is mainly affected by the poleward transport of water in the WSC, and the mixing processes on the shelf result in transformed Atlantic water in the fjord (Hop et al., 2006). The West Spitsbergen Current plays a predominant role on the west coast of Svalbard, and directly influences open fjords. Advection of warm water masses during late autumn and winter, together with prevailing wind patterns and air temperatures, may prevent ice formation in the fjords (Hop et al., 2006;Cottier et al., 2007). During December 2006, at the time of one of our cruises, the Kongsfjorden was almost completely ice-free.
The Barents Sea is an advective shelf system where colder and less saline Arctic and modified Atlantic waters encounter and interact with warm and saltier Atlantic water, creating a mosaic pattern of water masses influencing biological production (Reigstad et al., 2002).

Methods
The cruises were conducted along the western European gateway of the Arctic Ocean, including the Fram Strait, the large Kongsfjorden-Krossfjorden fjord system in Svalbard, the western Barents Sea, the East Greenland Shelf, the Greenland Sea and North Spitsbergen waters (Fig. 1).
Samples were collected in eight different cruises across five different periods of the year: the dark period in the late fall-early winter, early spring, spring, late spring-early summer, and summer (Table 1). Cruises were conducted in December, in April, in April-May, in May, in May-June, in June, in July and in July-August, respectively. Seven stations were sampled in December 2006 on board R/V Jan Mayen  Table 1). The two early-spring cruises in 2007 and 2008 (4 and 3 stations respectively) were conducted in a pre-bloom situation, in heavily ice-covered waters on board the icebreaker KV Svalbard. Twenty-two stations were sampled in July 2007 on board R/V Hespérides. The remaining cruises were conducted on board the R/V Jan Mayen during summer 2008 (seven stations), June 2009 (8 stations), spring 2010 (seven stations) and spring 2011 (twelve stations, Fig. 1, Table 1).
Water samples were collected at different depths within the photic layer using a Rosette sampler system fitted with a CTD (Conductivity, temperature, depth recorder) for a total of 69 stations, during the cruise conducted in April 2007 a 30 L GO-FLO or Niskin bottle was used for 1 m samples. Samples were incubated for 48 h in December 2006 and in April 2007, when metabolic rates were particularly low, and for 24 h in the rest of cruises. Planktonic metabolism was evaluated from the changes in oxygen concentration in replicated (6 to 11 replicates, depending on season) narrow-mouth Winkler bottles. A set of bottles was fixed immediately to evaluate the initial oxygen content. During spring and summer cruises, bottles were incubated in the light and in the dark in water baths on deck, at in situ temperature, using neutral screens to reduce incident irradiance and to mimic the light environment in situ. In December 2006, the Winkler bottles were incubated for 48 h in the dark, in a controlled temperature room inside R/V Jan Mayen, as there were 24 h of darkness at the time of sampling. As incubation conditions were designed to mimic environment conditions the results are comparable with incubations performed in situ. In early spring cruises (April 2007 and2008), incubations were conducted in situ, deploying a buoy from the deck of the ship and mooring it to the ice edge. Winkler bottles were attached to methacrylate supports and suspended  (2) nd: no data at the same depth from which the samples had been sampled, thereby being exposed to the same light and temperature conditions. The work conditions were particularly challenging during the spring cruises, when low air temperatures (mean ± SE = −13.1 ± 0.3 • C) lead to frequent and rapid freezing and breakage of Winkler bottles during exposure and retrieval.
Community metabolism (gross primary production, community respiration and net community production) was evaluated at 3 or 4 different depths per station, depending on the cruise. During early-spring cruises the depths selected were 1, 5, 10 and 20 m. During the summer cruise in 2007, late spring-early summer cruise in 2009, and spring cruise in 2010 and 2011 the depths sampled were 1 m, the depth of the chlorophyll maximum layers (CML) and an intermediate depth between these two depths. In spring 2010, a fourth depth was sampled in three of the seven total stations, sampling two intermediate depths between the surface and CML. In summer 2008, the selected depths were 1, 10, 20 m and the CML; when CML was at or near 20 m, incubations were also conducted at 5 m. During late fall-early winter cruise only the surface (1 m) layer was assessed, as the temperature and irradiance (complete darkness) profile were uniform across the upper water column.
The experimental standard errors (SE) of O 2 determinations among replicate bottles varied between 0.04 and 6.27 mmol O 2 m −3 , with a mean of 0.66 ± 0.03 mmol O 2 m −3 . These errors represent a mean of 0.19 % of the total value of the measurement, with the replicates of light bottles supporting a higher error than initial and dark bottle replicates. Although the lower range of these errors is close to the limit of analytical detection, reported to vary between 0.06 and 0.1 mmol O 2 m −3 (Robinson and Williams, 2005), the upper range of these errors is considerably higher.
Community respiration rates (CR) were calculated from the difference between the initial oxygen concentration and the oxygen concentration in the dark bottles after incubation. Net community production (NCP) was calculated from the difference between the oxygen concentration in the clear bottles after incubation and the initial oxygen concentration. Gross primary production (GPP) was calculated as the sum of NCP and CR rates. All the rates are reported in mmol O 2 m −3 d −1 and standard errors were calculated using error propagation. This method assumes equal respiration rates in the light and in the dark. This assumption may lead to underestimation of CR and GPP because respiration rates are likely to be higher during daylight than during night (Grande et al., 1989;Pace and Prairie, 2005;Pringault et al., 2007), but it does not affect NCP estimates (Cole et al., 2000).
Metabolic rates were integrated down to 20 m. The selection of an integration depth in the high Arctic is rather complicated. The two criteria most widely used in the literature, mixed layer and a light reference (e.g. 1 % PAR), are difficult to apply. Regarding the photic layer, no light penetrates to any depth during the dark winter period, ruling out the depth of a particular light penetration as integration criteria. The mixed layer is further complicated, as ice melting in spring and summer leads to very shallow pycnoclines and, correspondingly, a mixed layer of only 2-3 m depth, much shallower than the photic depth, and a water column that can be mixed to considerable depths (> 100 m) in the winter due to convective mixing. We chose to integrate down to 20 m across all cruises because this depth is close to both the chlorophyll a maximum layer (23.5 m) and to the mixed layer depth (17 m) located below the shallow thermocline in the summer. We assessed the sensitivity of our estimates to this choice of integration depth by also calculating metabolic rates integrated down to 30 m depth. This exercise showed integrated metabolic rates to be rather insensitive to the choice of either 20 or 30 m as integration depth (cf. Table S2).
Chlorophyll a was measured as detailed in Parsons et al. (1984) using a Turner Design AV-10 fluorometer, calibrated with pure chlorophyll a (Sigma 6041). Triplicate samples (100-500 mL) were filtered onto Whatman GF/F (glass fiber) filters.
Samples for dissolved organic carbon (DOC) were taken during the cruises conducted in summer 2007 and 2008 at the same depths sampled to estimate metabolic rates. Dissolved organic carbon (DOC) measurements were performed on 10 mL water samples sealed in precombusted glass ampoules (450 • C for 5 h) and kept acidified (pH 1-2) until analysis by high temperature catalytic oxidation on a Shimadzu TOC-5000A. Standards of 44-45 and 2 µmol C L −1 , provided by D. A. Hansell and Wenhao Chen (University of Miami), were used to assess the accuracy of the estimates.
Samples for total bacterial abundance (BA) were taken during the cruises conducted in summer 2007 and early-spring 2008, as well as in one station in the cruise conducted in the dark period of 2006. Total bacterial abundance (BA) samples were determined by flow cytometry by FACSCalibur (Fluorescence activated cell sorter) Flow Cytometer (Beckton Dickinson) as described in Ortega-Retuerta et al. (2008).
Samples for nutrient analysis (silicate, phosphate, nitratenitrite) were collected during early spring cruises ( Water masses were classified following descriptions from Rudels et al. (2000) (based on: Friedrich et al., 1995;Rudels et al., 1999). Polar Surface Waters (PSW) were defined as surface waters with a salinity lower than 34.4 and temperature below 0 • C. When PSW are warmed and the temperature increases beyond 0 • C these waters are called Warmed Polar Surface Waters (PSWw). Waters with a salinity higher than 34.4 and potential temperature above 2 • C are classified as Atlantic waters (AW) (Rudels et al., 2000). The mixed layer depth (MLD) was calculated from the vertical profile of density following the criteria outlined by Boyer Montegut et al. (2004). The mixed layer depth (MLD) was not always defined.
Quantile regression was used to describe the temperaturedependence of the volumetric and integrated metabolic rates. The relationship between metabolic rates and temperature was described by fitting the relationship between the 90, 50 (median) and 10 % quantiles of the distribution of metabolic rates and water temperature. Quantile regression estimates multiple rates of change (slopes), from the minimum to maximum response, providing a more thorough description of the relationships between variables, which are missed by other regression methods focused on prediction of the mean value (Cade and Noon, 2003;Koenker, 2005).
An estimate of the GPP threshold for metabolic balance was assessed using the relationship between the GPP to CR ratio (GPP/CR) and the GPP. As this relationship includes GPP in both its dependent and independent variables, the null hypothesis of this relationship is not that the slope equals zero, but that it equals one. A different approach to calculate the GPP threshold for metabolic balance free of this potential problem, was also used, based on inferring the GPP that equals respiration rates (i.e. NCP = 0) from the fitted relationship between Log CR and Log GPP. To calculate the GPP threshold for metabolic balance, the metabolic rates that were nonsignificant (i.e. < 2 × SE) were not included when calculating the above-mentioned relationships.
A first estimate of the annual metabolic rates in the western European Arctic sector was derived using the integrated metabolic rates presented here, classified into five distinct periods. Metabolic rates measured during fall/winter 2006 were R. Vaquer-Sunyer et al.: Seasonal patterns in Arctic planktonic metabolism used to estimate the period extending from the end of the 24 h daylight period to the end of the dark period (112 days). Stations visited during early-spring were used to estimate the period from the onset of the light period to the start of the 24 h daylight period (70 days). Stations visited during spring 2010, and some of the stations measured in 2011, were used as representative of a bloom stage (14 days). The latespring cruises and some stations measured in spring 2011 were used as data for a post-bloom stage during the 24 h daylight period (70 days). Finally, rates measured during summer cruises were used to estimate the summer period transition from 24 h daylight in the post-bloom stage to the onset of the polar night period, which includes the months of July, August and September (92 days). Metabolic rates were calculated for the duration of each of these periods (as the product of the mean rates and the period duration) and the rates derived from these periods were extrapolated to encompass a full year.
An estimate of the DOC needed to sustain community respiration during the dark period was derived using the mean volumetric community metabolism integrated during that period (112 days). Conversion from oxygen to carbon was made assuming a 1.25 molar stoichiometry between O 2 and C (Williams et al., 1979).

Hydrological data
The air temperature ranged from −25.  (Table 1). The average seawater temperature was lowest for the two early-spring cruises (mean ± SE = −1.78 ± 0.01 • C in 2007 and in 2008), which took place in the Arctic Ocean outflow, followed by the other 3 spring cruises, while temperatures exceeded 2.4 • C for all other cruises (Table 1 and Fig. 2). These significant (ANOVA (analysis of variance), F = 16.72, p < 0.0001) differences in water temperature between cruises can partly be attributed to seasonal differences but also to variability in the water masses sampled. Indeed, during early-spring cruises only Polar surface waters (PSW) were sampled, whereas during the other five cruises Atlantic water (AW) and warmed Polar surface water (PSWw) were also sampled. Differences in water temperature were also attributable to spatial differences, as there were significant differences in the temperature (F = 11.02, p < 0.001) among the various areas sampled (Barents Sea, North Spitsbergen, central Fram Strait, Svalbard Fjords, Greenland Sea, East Greenland Shelf and West Spitsbergen).  Table 1). The salinity differed significantly among cruises (ANOVA, F = 13.02, p < 0.0001). These differences reflect both the effects of ice melting and the distribution of Atlantic, saltier water, versus Arctic water at the stations sampled in the different cruises. Surface salinity differed significantly among sampled areas (ANOVA, F = 10.48, p < 0.0001), reflecting the presence of Polar surface waters transported southwards along the EGC and the ice melting on the Svalbard fjords during spring.
Chlorophyll a concentrations, at the stations and depths where metabolic rates were determined, were lowest during late fall-early winter 2006 (0.02 ± 0.02 µg chl a L −1 ), somewhat higher in early spring (0.03 ± 0.00 µg chl a L −1 in 2007 and 0.11 ± 0.02 µg chl a L −1 in 2008), higher in summer (2.43 ± 0.24 µg chl a L −1 in 2007 and 2.11 ± 0.34µg chl a L −1 in 2008), and highest in spring 2009 (2.55 ± 0.22 µg chl a L −1 , Table 1 and Fig. 2). Unfortunately, chlorophyll a analyses were not conducted for the cruises conducted in spring 2010 and 2011. Chlorophyll a content increased significantly with seawater salinity (R 2 = 0.20, p < 0.0001, N = 122) and seawater temperature (R 2 = 0.08, p < 0.002, N = 122) in the cruises and stations where data are available. Consequently, there were statistically significant differences in chlorophyll a concentration between water masses (F = 6.55, p < 0.003), with Atlantic water (mean ± SE = 2.90 ± 0.41 µg chl a L −1 ) having significantly higher chlorophyll a content than Polar surface waters (PSW, mean ± SE = 1.25 ± 0.31 µg chl a L −1 ), but comparable to warmed Polar surface water (PSWw, mean ± SE = 1.88 ± 0.21 µg chl a L −1 ). This partly reflects the bloom stage sampled in the different regions. Unfortunately we do not have data available for the spring cruise in 2010 where metabolic rates indicate that a spring bloom was sampled (see below). Mixed layer depth varied greatly between 5 m in summer 2007 and 67.7 m in the dark period of 2006, with a mean value of 17.0 ± 1.9 m for all stations and 25.8 ± 6.8 m for the cruise averages.

Volumetric metabolic rates
Net community production (NCP) ranged broadly from −21.7 ± 1.9 for strongly heterotrophic communities in summer 2007 to 81.6 ± 0.7 mmol O 2 m −3 d −1 for strongly autotrophic communities in spring 2011 (Tables 2 and S1, Supplement). NCP differed significantly between cruises, with higher NCP in spring 2010 and 2011 than for the other cruises (F = 15.32, p < 0.0001). The lowest, negative, NCP was measured in the dark period of late fall-early winter 2006 (average ± SE = −0.8 ± 0.3 mmol O 2 m −3 d −1 , Table 2, Fig. 3). In summer NCP tended to be negative, indicative of heterotrophic communities prevailing in this season. Most summer stations supported plankton communities in a post-bloom stage, when the CR of the planktonic community exceeded production. Consistently, the waters sampled tended to be undersaturated in oxygen in summer (mean ± SE = 89.3 ± 0.9 %). NCP values differed among water masses (F = 4.58, p < 0.02), with communities sampled in Atlantic water having statistically significant higher values (mean ± SE = 11.1 ± 1.7) than in warmed Polar surface waters (mean ± SE = 3.2 ± 2.0 mmol O 2 m −3 d −1 ), but comparable to those sampled in Polar surface waters (meanmean ± SE = 1.7, Fig. 4). NCP also differed significantly among regions (F = 9.32, p < 0.0001), with the East Fram Strait having higher NCP values (mean ± SE = 44.5 ± 7.5 mmol O 2 m −3 d −1 ) than the other sampled areas.
Community respiration (CR) varied from a minimum value of 0.0 ± 0.4 mmol O 2 m −3 d −1 measured in spring 2007 to 40.9 ± 0.6 mmol O 2 m −3 d −1 measured in spring 2011. The respiration rates were similar among cruises, although the respiration rate in the spring 2011 cruise was significantly higher (mean ± SE = 7.2 ± 1.6 mmol O 2 m −3 d −1 ) than that measured during the summer of 2008 and that measured in spring 2010 (F = 3.76, p < 0.001; Fig. 3). CR did not show statistically significant differences between water masses (F = 0.16, p = 0.85) or between sampled areas (F = 1.86, p = 0.08). CR varied greatly, over 2 orders of magnitude, between stations from the same cruise in four of the eight cruises (Table 2). This high variability between stations sampled in the same cruise masks any existing seasonal variability in respiration rates. There were no significant relationships (p > 0.05) between metabolic rates and nutrient concentrations.
The ratio of GPP to CR (GPP/CR) describes the metabolic status of the community, which is net heterotrophic when GPP/CR < 1, net autotrophic when GPP/CR > 1 or in metabolic balance when GPP/CR = 1 (i.e. GPP = CR). GPP/CR varied between 0, for the late fall-early winter cruise in the dark, when no primary production occurred, to 33.64 ± 1.64 measured at 35 m depth in the  Barents Sea in spring 2009, the highest value reported here. There were significant differences in the GPP/R ratio between cruises (ANOVA, F = 3.19, p < 0.004), with the cruise in spring 2010 having the highest GPP/R ratio (mean ± SE = 49.53 ± 25.65), indicative of the overwhelming dominance of autotrophic production characteristic of the spring bloom stage (Fig. 3). GPP/CR did not show statistically significant differences between water masses (F = 1.33, p > 0.05) or between sampling areas (F = 1.73, p > 0.05). During the cruise conducted in summer 2008, CR increased linearly with GPP as described by the fitted regression equation: CR = 0.52 + 0.62 (± 0.13) GPP (R 2 = 0.54, p < 0.0001, N = 22), but no such relationship was found for the other cruises. For the entire data set there was a weak, albeit significant relationship between CR and GPP as described by the fitted regression equation: CR = 3.29 + 0.08 (± 0.03) GPP (R 2 = 0.04, p < 0.01, N = 165). There was also a weak, albeit significant relationship between CR and DOC and Bacterial Abundance (AB), described by the fitted regression equations: log CR = −10.37 (± 3.69) + 2.50 (± 0.82) log DOC (µM) (R 2 = 0.19, p < 0.005, N = 41) and log CR = −3.15 (± 2.13) + 0.31 (± 0.16) log BA (R 2 = 0.06, p < 0.05, N = 64). These results point at a higher depen- dence of community respiration rates on DOC content than on GPP rates or bacterial abundance. The GPP/CR ratio increased significantly with GPP ( Fig. 5)  . ( Because GPP is present both in the dependent and independent variables, we used a Monte Carlo approach to compare the observed slope and threshold against those expected by chance. This was done by randomizing the paired variables twenty times and calculating the slope and threshold for each random configuration of variables. The observed threshold and slope (3.13 ± 0.07 mmol O 2 m −3 d −1 and 0.80 ± 0.07) are significantly different from those expected by chance (2.04 ± 0.04 mmol O 2 m −3 d −1 and 0.94 ± 0.01, p < 0.05), confirming that this analysis describes a functionally meaningful, not spurious (Prairie and Bird, 1989), relationship between the variables analysed. Community respiration rates increased with increasing gross primary production as described by the fitted ordinary least squares regression equation: where the slope is significantly < 1 (p < 0.0001), indicating that community respiration is highest relative to GPP in communities with low GPP. Both volumetric and integrated NCP and GPP tended to decrease with increasing temperature. Examination of the relationship between production rates (both NCP and GPP) and temperature showed that the range of production rates become narrower with increasing temperature, with most production rates being low at higher temperatures (Fig. 6). Conversely, volumetric and integrated CR tended to increase with increasing temperatures, with the range of respiration rates becoming wider with increasing temperature (Fig. 6).
GPP increased significantly with increasing chlorophyll a (R 2 = 0.38, p < 0.0001, N = 98) for the stations and cruises where the data were available (Fig. 7).

Integrated metabolic rates
Depth-integrated metabolic rates, integrated down to 20 m, were calculated for each station (  (Table 3). Depth-integrated metabolic rates were also calculated for an integration depth of 30 m where data were available (Table S2). There were no significant differences between the metabolic rates integrated to 20 or 30 m depth (p > 0.05).
In the late fall-early winter cruise, in absence of light, all stations supported net heterotrophic communities. In spring, at the onset of the 24 h daylight period, communities are expected to be strongly autotrophic. Indeed, all stations had net autotrophic communities in early spring 2007, but the community at one of the three stations sampled in 2008 was net heterotrophic. The extreme low temperature and heavy ice cover encountered during early spring did not yield the appropriate conditions for bloom development. In May all stations were net autotrophic and the GPP/CR ratio was very high, with high production and low respiration rates, indicative of a bloom development. In the late spring-early summer cruise conducted in 2009 one of the eight stations sampled was found to be net heterotrophic. In the summer cruises a total of 40 % (N = 22) and 33 % (N = 7) of the stations were found to support net heterotrophic communities in 2007 and 2008 respectively.

Methods used
The Winkler method estimates planktonic metabolism in closed systems and it is subject to possible "bottle effects". The "bottle effect" refers to the concern that phenomena observed in confined assemblages derive from the consequences of the confinement of the community and could be different than under natural conditions (Pernthaler and Amann, 2005;Hammes et al., 2010). Some of the artefacts derived from bottle incubation are produced by substrates and bacteria adsorption and bacterial proliferation on the glass surface. Long incubation periods can also imply modifications in bacterial activity and diversity (Massana et al., 2001). However, several authors did not find any difference in microbial metabolism and/or growth (Fogg and Calvario-Martinez, 1989;Hammes et al., 2010;Garcia-Martin et al., 2011) when using different bottle sizes, which is one of the components determining the "bottle effect", when existing. Thus, although structural changes may occur, the metabolic rates measured through incubation bottles are considered to be meaningful (Gasol et al., 2008).
Alternative methods to estimate planktonic metabolism, avoiding "bottle effects" include the assessment of the biological O 2 saturation, which refers to the differences between O 2 and Ar saturation (Quay et al., 1993), and the triple oxygen isotope composition ( 16 O, 17 O, and 18 O) of dissolved O 2 (Luz and Barkan, 2000). O 2 /Ar gas ratios measured in situ can be combined with the oxygen triple isotope composition to estimate rates of NCP (Bender, 2000;Hendricks et al., 2004;Reuer et al., 2007). The combination of these methods to estimate community metabolism remove the "bottle effect" and integrate metabolic rates over a period of weeks to months, but has a high associated error, from 30 to 40 % (Juranek and Quay, 2005;Robinson and Williams, 2005). Estimation of NCP in the upper water column can also be made from direct analysis of decreases in total dissolved inorganic carbon (DIC) after correcting for CO 2 exchange with the atmosphere (Ishii et al., 1998). Moreover, the use of incubation-free techniques in the Arctic is rendered difficult by the presence of pycnoclines in the summer (typically at 2 m), derived from ice melting, so that the assumption of mixing in the photic layer inherent to these techniques (Duarte et al., 2012) is violated.

Metabolic rates
There is a remarkable paucity of direct measurements of planktonic metabolic rates in the Arctic Ocean, with most available studies reporting only one of the components involved in the assessment of metabolic balance (Table 3) or deriving metabolic rates from models. The rates reported in this study are within the rates reported in the past, except for: the NCP we report for the winter, which is the only negative  Fig. 6. Relationship between volumetric and integrated metabolic rates and water temperature. Upper panels show volumetric metabolic rates: volumetric net community production (NCP), volumetric gross primary production (GPP), volumetric community respiration (CR). Lower panels show integrated metabolic rates: integrated NCP, integrated GPP and integrated CR. Solid lines represent the fitted regression for the median or the 50 % quartile. Dashed lines represent the fitted regression for the 90 and 10 % quantiles. Statistics for the regression lines are summarized in Table S3. rate reported so far (Table 3) since, in the past, NCP had not been assessed for Arctic communities during winter; and for the GPP values reported for the spring 2010, which are well above previous estimates reported for the Arctic Ocean. Planktonic metabolism in the Arctic Ocean margins exhibits, as expected, important annual variability, which is compounded with considerable spatial variability, partially masking the seasonal signal. The absence of sunlight and photosynthetic activity in winter renders Arctic planktonic communities heterotrophic, consuming the excess dissolved organic matter produced during the light period of the year and acting as CO 2 sources in winter. The productive photic period may generate slow-to-degrade dissolved organic matter (DOM), which could support bacterial production during winter, as it has been demonstrated in Antarctic waters (Azam et al., 1991(Azam et al., , 1994. We examined whether the DOC pool is sufficient to subsidize winter respiration when darkness prevents the inputs of a fresh photosynthetic period. We estimated, using the respiration rate measured in winter (Table 2), a preliminary respiratory carbon demand in the Fram Strait region of 75.26 ± 100.35 µmol L −1 during the dark period. This is below the average DOC pool in the area studied (89.01 ± 2.46 µmol C L −1 ; Kritzberg et al., 2010;Tovar-Sánchez et al., 2010 and this study), suggesting that the large DOC pool in Arctic waters would suffice to maintain significant respiration rates in the plankton community across Table 3. Average planktonic metabolic rates (mmol O 2 m −3 d −1 ) for different studies of planktonic community metabolism in the Arctic Ocean. Rates given as gross primary production (GPP), net community production (NCP) and respiration (R). Number of measurements included for each rate is given (N).

Authors
Region Date Season GPP NCP CR Cota et al. (1996) (7) a Data reported in carbon units converted to oxygen units assuming a 1.25 molar stoichiometry between O 2 and C (Williams et al., 1979).
the dark period, assuming all this DOC was labile. However, the resulting DOC concentration would be below that ever recorded in the ocean unless resupplied by convective mixing from deeper layers. Hence, respiration rates in the plankton community across the dark period may be partially supported by allochthonous DOC inputs. However, any assessment of the sources of organic carbon supporting community respiration in the winter is, at this stage, speculative. Spring, with the increase in PAR and the onset of melting of seasonal ice and surplus nutrients, is the most productive time of the year, when algal blooms occur (mainly in May) ( Table 2). The spring bloom in Arctic water can account for a 40 % of the total annual primary production (Lavoie et al., 2009). The highest NCP and GPP are both reached in spring (in a bloom stage), when water temperatures remain low and ice cover is reduced (Table 2), with an extremely high GPP/CR ratio, indicative of a spring bloom development, when production increases sharply and respiration rates remain low. In a previous study, (Cottrell et al., 2006) also reported higher metabolic rates in spring than in summer, but their production values were lower than the values reported here (Table 3). These differences can be attributed to differences in the stage of the bloom when the spring sampling was made. Whereas our spring samples were taken in a bloom situation (in May), the Cottrell et al. (2006) samples where probably taken during a post-bloom situation, as their GPP/CR ratios are lower than those measured here. In addition, our study was conducted mainly in the Fram Strait, whereas their study was conducted in the Chukchi Sea, at lower latitude than our study area, which may affect seasonal development.
NCP and GPP tended to decrease with increasing temperatures, concurrent with recent experimental work (Holding et al., 2013). At low temperatures, high GPP and NCP are reached during the spring bloom, and low GPP and NCP at stages previous to the development of the bloom. Thus, at low temperatures we found a high variability of NCP and GPP rates (Fig. 6), whereas at higher temperatures these metabolic rates tended to decrease and be confined within narrower ranges. This suggests that the NCP and GPP are related to the stage of the bloom at lower temperatures, while at higher temperatures temperature-dependence controls the relationship.
The GPP observed during the summer cruise in 2007 (the only cruise where all necessary data were available) was compared with the upper limit imposed by the underwater PAR, the light absorbed, calculated from chlorophyll a using the specific absorption coefficient for Arctic communities by Matsuoka et al. (2009), and the quantum yield (from Kirk, 1983). The results indicated that the observed GPP represents, on average, 4.6 ± 1.3 % of the maximum possible rates, and a maximum observed value of 57.8 % in one of the stations. GPP for the spring bloom is expected to approach more closely the biophysical maximum imposed by light and the quantum yield. Unfortunately, we lack the data needed to make comparable calculations.
The GPP/CR ratio increased with increasing GPP, as observed elsewhere in the ocean (see Duarte and Agusti, 1998;Duarte and Regaudie-de-Gioux, 2009), implying that unproductive Arctic communities tend to have a low GPP/CR, thus tending to be heterotrophic. The fitted regression equation implies that the average GPP required to balance Arctic planktonic metabolism is 3.13 mmol O 2 m −3 d −1 , when using ordinary least squares (OLS) regression and of 3.94 mmol O 2 m −3 d −1 when using model II regression. Fitting the relationship between Log CR and Log GPP yields a similar result (3.01 mmol O 2 m −3 d −1 ) to that obtained using ordinary least squares regression. These rates are higher than average rates for oceanic communities (1.07 mmol O 2 m −3 d −1 ) and proposed estimates for threshold GPP in cold environments derived from a theoretical model (< 1 mmol O 2 m −3 d −1 ; López-Urrutia et al., 2006), but lower than a previously reported value for the Arctic Ocean based on a more limited data set collected in summer (5.45 mmol O 2 m −3 d −1 ; Duarte and Regaudie-de-Gioux, 2009).
Pelagic respiration in the Arctic may be subsidised by riverine inputs of organic carbon, as the Arctic receives the discharge of some of the world's largest rivers, delivering 30 × 10 6 t C yr −1 of organic carbon to the Arctic Ocean (Rachold et al., 2004), as a consequence the Arctic Ocean supports the highest concentration of terrestrial DOM in any ocean . Use of terrestrial DOM by marine bacterial communities will largely depend on its chemical composition and lability (Sondergaard et al., 2003). The prevailing paradigm has been that terrestrial DOC discharged by Arctic rivers is highly refractory. However, recent evidence suggests that between the 20-40 % of terrestrial DOC is labile (Cooper et al., 2005;Holmes et al., 2008;van Dongen et al., 2008). Hansell et al. (2004) determined that terrestrial DOC in the Beaufort Gyre was mineralized with a half-life of 7 yr, allowing only 21 to 32 % to be exported to the North Atlantic Ocean. Holmes et al. (2008) demonstrated that DOC concentrations and lability changed drastically between seasons, with the spring freshet having major DOC flux and lability. Letscher et al. (2011) also reported a rapid removal of terrigenous DOC over the Eurasian shelves of the Arctic Ocean, suggesting that terrestrial DOC is composed of multiple compartments of different reactivity and reinforcing the idea of a dynamic terrestrial DOC pool of biolabile components that can support the microbial loop. Glaciers can also be a considerable source of labile organic matter to the marine environment in the Gulf of Alaska, with 66 % of the total DOC being bioavailable (Hood et al., 2009). This study reported bioavailable DOC to range between the 23 and 66 % in different watersheds of the Gulf of Alaska. Furthermore, there are also considerable inputs of allochthonous organic matter with the AW flowing to the north (Wassmann, 2001).
Although, diatoms are expected to represent an important component of the phytoplankton community in the marginal ice zone and in waters influenced by ice melting (von Quillfeldt, 1997(von Quillfeldt, , 2000Falk-Petersen et al., 1998), during our summer cruise in 2007, the prymnesiophyte Phaeocystis pouchetti, in its colonial form, dominated the phytoplankton community and diatoms represented only 7.3 % of the phytoplankton biovolume . In the only station where diatom abundance exceeded that of P. pouchetti the lowest NCP and the highest CR rates were measured (in this station the water temperature was the warmest measured in the cruise). Diatoms were found to be scarce in colder and low salinity waters, indicating that this group was more affected by ice melting . During the spring cruise in 2008, the phototrophic protist biomass dominated over that of heterotrophic protists in the stations with autotrophic metabolism, suggesting that protists strongly contributed to the metabolism of the communities (Seuthe et al., 2011). In contrast, bacterial respiration appeared to be small during this cruise, as indicated by very low rates of bacterial production (Seuthe et al., 2011). During the pre-bloom stage, in heavily ice-covered waters, protists are believed to greatly contribute to community metabolism (Seuthe et al., 2011).
An approximation to the annual metabolic rates in the western European Arctic sector can be attempted with the integrated metabolic rates presented here. However, this exercise must be considered a tentative one, due to the sparse sample density over time, particularly during wintertime and transition periods between the polar night and midnight sun. The mean annual GPP was calculated to be 32 mol O 2 m −2 yr −1 (305 g C m −2 yr −1 ) and the mean annual CR was estimated at 20 mol O 2 m −2 yr −1 (197 g C m −2 yr −1 ), lower than the GPP estimate. Accordingly, these calculations indicate that the mean annual NCP (NCP = GPP − CR) across the study area is expected to be positive at 11 mol O 2 m −2 yr −1 (108 g C m −2 yr −1 ), implying that the planktonic community in the European sector of the Arctic is likely to be net autotrophic on an annual scale, thereby acting as a significant atmospheric carbon sink. The spring bloom, with a duration of 14 days contributed to 26 % of the total annual gross primary production. The GPP estimate reported here is 69 % higher than previous estimates of annual production for this area (average of 93 ± 18 g C m −2 yr −1 , Wassmann et al., 2006b). The annual NCP value derived here is slightly lower than NPP values derived from satellite data for the Bering Sea (124 g C m −2 yr −1 ), and bellow the global mean NPP of 140 g C m −2 yr −1 (Brown et al., 2011).
An increased sampling frequency will be required to improve these estimates; an effort that will require increased international collaboration. While there is ample room for improvement, the annual estimate derived here for the studied region is based on a sampling effort unparalleled for any other polar region (Robinson and Williams, 2005), where plankton metabolism remains grossly undersampled.
The estimate provided here does not include production by ice algae, generally reported to contribute 5-10 % of overall primary production in shelf areas (Horner and Schrader, 1982;Gosselin et al., 1997;Lavoie et al., 2009) or microbial respiration in sea ice, which has been shown to be an important organic C sink in sea ice (Nguyen and Maranger, 2011). Ice algae production has been reported to average 36 mg C m −2 d −1 in the Beaufort Sea with a peak of 62 mg C m −2 d −1 in May (Horner and Schrader, 1982), at 28 mg C m −2 d −1 in the Chuckchi Sea (Gosselin et al., 1997) and at 14.5 mg C m −2 d −1 in the northern Barents Sea (Hegseth, 1998). The estimate provided here does not include zooplankton respiration rates, estimated to have requirements in the upper 200 m in summer of 2007 averaging 23.2 % of the 14 C primary production (Alcaraz et al., 2010).
Previous studies reported an increase of Arctic primary production in recent years. Arrigo et al. (2008) estimated that the net annual CO 2 -fixation by Arctic plankton has increased by 26 % (6.5 % per year) between  reported a 30 % increase in Arctic annual primary production between 1998 and 2006. This trend is expected to continue. However, close inspection of the data presented by Arrigo et al. (2008) shows that the primary production in the Atlantic sector of the Arctic Ocean did not increase in the summer of 2007. As the Arctic Ocean is very heterogeneous and exhibits a wide range of regional responses, responses to global warming will probably also vary across regions. Ellingsen et al. (2008) predict an increase of primary production in the Barents Sea of 8 % over the period 1995-2059. Both studies, Ellingsen et al. (2008) and Arrigo et al. (2008), support their statements on the predictions of ice melting and reduced ice surface, leading to an extended productive season.
Yet, respiration rates are also expected to increase with increasing temperature, more so than primary production Lopez-Urrutia et al., 2006). In the studied area, community respiration rates are predicted to increase by 62 % with a 6 • C warming (Vaquer-Sunyer et al., 2010), doubling the 30 % increment expected for primary production (Wassmann et al., 2008). Bacterial respiration is also predicted to increase faster than bacterial production in this area (Kritzberg et al., 2010). Thus, the net community production may not increase or may even decrease in the future. Warming can result in substantially weakening the role of Arctic communities as significant CO 2 sinks, and may even be reverted to becoming a CO 2 source to the atmosphere (Vaquer-Sunyer and Duarte, 2010) because warming is predicted to increase the carbon flow through bacteria and most of the carbon consumed would be released as CO 2 (Kritzberg et al., 2010). Indeed, a recent experimental assessment suggests the existence of a 5 • C threshold for Arctic waters, beyond which the metabolism (NCP) of plankton communities shifts from autotrophic to heterotrophic (Holding et al., 2013). This study also finds a similar threshold response where community respiration doubles at 5 • C, concurrent with previous work (Vaquer-Suyner et al., 2010). Warming will probably impact different aspects of the structure and functioning of marine ecosystems (Edwards and Richardson, 2004;Hinder et al., 2012;Montoya and Raffaelli, 2010), such as the heterotrophic to autotrophic biomass ratio. An experimental study reported that the heterotrophic to autotrophic biomass ratio increased 5 times and the carbon fixation to respiration ratio decreased six times when temperature was raised from 5 to 10 • C (Müren et al., 2005). Rising temperatures also affect ice melting, thereby affecting the production of ice algae, and may lead to higher DOC inputs to the Arctic Ocean (Cooper et al., 2005) through river discharge (Peterson et al., 2002), permafrost thawing ) and glacier melting, which will potentially support higher pelagic respiration rates.
Ice melting can also produce a decrease in primary production (Regaudie-De-Gioux and Duarte, 2010), as ice melting increases stratification, which could possibly limit nutrient supply to the photic layer (Wassmann et al., 2008). This is consistent with the positive relationship between chlorophyll a and salinity, and the negative relationship between production rates and temperature reported here. These results are in contrast with earlier findings for the Southern Ocean that suggest that freshwater discharge with ice melting should increase primary production due to increased stratification (Montes-Hugo et al., 2009. Global warming results in an "atlantification" of large regions in the Atlantic sector of the Arctic Ocean (Wassmann et al., 2004). Implications of "atlantifications" will be multiple, affecting vertical mixing and introducing Atlantic species that competitively displace Arctic species poleward, among others. However, the effects of "atlantification" of the Arctic metabolic rates are unknown. As atlantification is expected to reduce stratification, it will result in significant changes in phytoplankton composition, bloom size and development, and vertical flux possibly leading to a regime shift in the Arctic marine ecosystem (Wassmann et al., 2004).
The results presented here provide a first assessment of seasonal and spatial variability in planktonic metabolism in the Western European sector of the Arctic, allowing the evaluation of patterns in metabolic rates and a first, albeit rough, approximation of the annual metabolic balance of Arctic plankton communities. The estimates derived here can be improved further through efforts to resolve spatial variability in Arctic metabolic rates and increasing the research effort during fall and winter, when harsh weather conditions render oceanographic research in the high Arctic cumbersome. Particular efforts are required to capture the metabolic rates during the onset and subsequent development of the highly seasonal spring bloom period, which may last for only two weeks in marginal ice zones (Wassmann et al., 2006a, b). The results provided here provide a valuable baseline to assess future changes in plankton metabolism with warming and ice loss in the Arctic, which can affect the role of the Arctic Ocean in a warmer Earth system.