CO 2 increases 14 C primary production in an Arctic plankton community

Responses to ocean acidification in plankton communities were studied during a CO 2-enrichment experiment in the Arctic Ocean, accomplished from June to July 2010 in Kongsfjorden, Svalbard (78 562 N, 11536 E). Enclosed in 9 mesocosms (volume: 43.9–47.6 m 3), plankton was exposed to CO2 concentrations, ranging from glacial to projected mid-next-century levels. Fertilization with inorganic nutrients at day 13 of the experiment supported the accumulation of phytoplankton biomass, as indicated by two periods of high chla concentration. This study tested for CO 2 sensitivities in primary production (PP) of particulate organic carbon (PP POC) and of dissolved organic carbon (PP DOC). Therefore,14C-bottle incubations (24 h) of mesocosm samples were performed at 1 m depth receiving about 60 % of incoming radiation. PP for all mesocosms averaged 8.06 ± 3.64 μmol C L−1 d−1 and was slightly higher than in the outside fjord system. Comparison between mesocosms revealed significantly higher PP POC at elevated compared to low pCO2 after nutrient addition. PPDOC was significantly higher in CO 2-enriched mesocosms before as well as after nutrient addition, suggesting that CO 2 had a direct influence on DOC production. DOC concentrations inside the mesocosms increased before nutrient addition and more in high CO2 mesocosms. After addition of nutrients, however, further DOC accumulation was negligible and not significantly different between treatments, indicating rapid utilization of freshly produced DOC. Bacterial biomass production (BP) was coupled to PP in all treatments, indicating that 3.5± 1.9 % of PP or 21.6± 12.5 % of PPDOC provided on average sufficient carbon for synthesis of bacterial biomass. During the later course of the bloom, the response of14C-based PP rates to CO 2 enrichment differed from net community production (NCP) rates that were also determined during this mesocosm campaign. We conclude that the enhanced release of labile DOC during autotrophic production at high CO2 exceedingly stimulated activities of heterotrophic microorganisms. As a consequence, increased PP induced less NCP, as suggested earlier for carbon-limited microbial systems in the Arctic.


Introduction
The Arctic Ocean is predicted to be among the most affected marine ecosystems with respect to consequences of anthropogenic emissions of carbon dioxide (CO 2 ), such as ocean acidification and warming.Temperature increase in the Arctic is about twice as fast as the global rate, yielding an average of 1-2 • C yr −1 (Anisimov et al., 2007).Warming accelerates the melting of sea ice and Greenland's glaciers.Satellite data revealed that the loss of Arctic sea ice has tripled over the last 10 yr (Comiso et al., 2008).Freshening of seawater due to ice melt along with an enhanced uptake of CO 2 due to shrinking sea-ice coverage is predicted to amplify CO 2induced acidification of Arctic seawater (Steinacher et al., 2009), with so far unknown consequences on the pelagic ecosystem dynamics and productivity.

A. Engel et al.: CO 2 increases 14 C primary production
The Kongsfjorden is part of the Arctic archipelago Svalbard and situated on the west coast of Spitsbergen.It is a relatively well-studied system, compared to other areas in the Arctic, as several research stations are located in the village of Ny-Ålesund.A review by Hop et al. (2002) provides a compilation of current knowledge obtained for the Kongsfjorden ecosystem.For the phytoplankton community, a total of 148 taxa have been reported and showed the numerical dominance of diatoms and dinoflagellates (Eilertsen et al., 1989;Hasle and Heimdal, 1998;Keck et al., 1999;Wiktor, 1999).Primary production in Kongsfjorden was determined during several field studies, focusing mainly on the spring period (Piwosz et al., 2009;Rokkan and Seuthe, 2011;Hodal et al., 2011), when availability of nutrients and light after the polar night support a substantial fraction of annual productivity (Sakshaug, 2004).
Phytoplankton primary production is based on CO 2 as the main substrate, and since the CO 2 -binding enzyme RubisCO has a low affinity for its substrate (Badger et al., 1998), an increase in seawater pCO 2 was hypothesized to stimulate primary production (Riebesell et al., 2000;Schippers et al., 2004;Rost et al., 2008).The impact of increased pCO 2 on primary production has been investigated theoretically as well as experimentally.Some authors report small, if any, effects (Clark and Flynn, 2000;Tortell et al., 2002), whereas others document a clear increase in primary production with increasing pCO 2 (Hein and Sand-Jensen, 1997;Schippers et al., 2004;Riebesell et al., 2007).The effect of seawater carbonate chemistry on photosynthesis rates thereby strongly depends on the presence and characteristics of cellular CO 2concentrating mechanisms (CCMs; Rost et al., 2003, Giordano et al., 2005).Phytoplankton species that are able to enhance their CO 2 supply by CCMs (Raven, 1991) may exhibit no or minimal sensitivity to CO 2 enrichment (Raven and Johnson, 1991;Rost et al., 2003;Giordano et al., 2005).Others, such as the coccolithophore Emiliania huxleyi, respond to CO 2 enrichment with an increase in primary production (Rost and Riebesell, 2004).This suggests that the efficiency and regulation of CCMs differ among phytoplankton functional groups and species.Moreover, the capability of the phytoplankton cell to express a CCM relies on the availability of light and nutrients (Young and Beardall, 2005;Beardall et al., 2005), and may thus be restrained under sub-optimal conditions.Changes in CO 2 availability might therefore affect competition and succession of phytoplankton species (Burkhardt et al., 2001;Rost et al., 2003;Tortell et al., 2002).
Effects of elevated pCO 2 on phytoplankton are of major interest for understanding global biogeochemical cycles, since primary production mediates the transformation of CO 2 into organic carbon with variable stoichiometric relationships to other major elements, for example phosphorus (P) and nitrogen (N).If CO 2 assimilation is decoupled from other major elements, changes in the stoichiometric composition of organic material and altered biogeochemical path-ways through the microbial food web are potential consequences.A particular increase in C assimilation relative to the uptake of N and P and compared to Redfield stoichiometry of 106C : 16N : 1P is referred to as carbon overconsumption (Toggweiler, 1993).This imbalance in carbon and nutrient assimilation has been related to nutrient limitation of the cell (Wood and van Valen, 1990;Engel et al., 2002;Schartau et al., 2007) and also to enhanced CO 2 concentration (Engel, 2002;Riebesell et al., 2007;Kim et al., 2011;Borchard and Engel, 2012).Carbon overconsumption is often accompanied by a release of dissolved organic carbon (DOC) from the cell, either by passive (leakage) or active processes (exudation) (Fogg, 1966;Bjørnsen, 1988;Engel et al., 2004a, b;López-Sandoval et al., 2011).The extracellular release of DOC is a normal function of algal cells (Fogg, 1966) and represents with ∼ 3-40 % (percentage of extracellular release, PER) a significant fraction of primary production (Myklestad, 1977;Mague et al., 1980;Baines and Pace, 1991).Factors influencing primary production, such as light and temperature, were shown to also affect the production of DOC (Zlotnik and Dubinsky, 1989;Baines and Pace, 1991;Engel et al., 2011).
Release from phytoplankton cells is the major source of labile and semi-labile DOC in the ocean and drives the microbial loop (Azam et al., 1983), whereby DOC is either transferred to higher trophic levels or respired back to CO 2 (Ducklow et al., 1986).Microbial respiration represents an important loss for DOC globally (Williams, 2000;Hansell et al., 2009).Under a "malfunctioning" of the microbial loop, DOC accumulates (Thingstad et al., 1997) and may be subject to abiotic aggregation into gel particles, such as transparent exopolymer particles (TEP) (Alldredge et al., 1995).TEP formation thereby represents a repartitioning of dissolved organic carbon into particulate organic carbon (POC) without loss of mass (Engel et al., 2004b).An increase in TEP-C may raise C : N or C : P ratios in particulate organic matter, potentially providing an enhanced sinking flux of carbon to depth (Schneider et al., 2004).
In Arctic ecosystems, heterotrophic microbes are often limited by the amount of labile DOC (Kirchman et al., 2009) and co-limited by nutrients (Cuevas et al., 2011).An increased input of labile DOC (glucose) was rapidly consumed by bacteria and other osmotrophs during an earlier mesocosm study at Svalbard, resulting in enhanced competition for inorganic nutrients between phyto-and bacterioplankton, and in an overall reduction of autotrophic productivity of the system (Thingstad et al., 2008).A hypothesis that came out of the study of Thingstad et al. (2008) was that stimulation of the microbial loop in Arctic waters by increased DOC release under high pCO 2 may result in a counterintuitive carbon cycling (i.e., "more organic carbon gives less organic carbon") and not necessarily enhance carbon export to the deep sea.
In order to address potential consequences of the ongoing seawater pCO 2 increase in Arctic pelagic ecosystems, a mesocosm study was conducted in the framework of the European Project on Ocean Acidification (EPOCA).
Several methods were applied during this mesocosm study to investigate the sensitivity of plankton productivity to CO 2 perturbation, including in vitro O 2 measurements at 4 m depth outside the mesocosms (Tanaka et al., 2013), as well as changes in dissolved inorganic carbon (DIC) concentration (Silyakova et al., 2012) and uptake of 13 C-labelled DIC inside the mesocosms (de Kluijver et al., 2012).
Here, we report on sensitivities in primary production to increasing pCO 2 , for both the production of POC and of DOC based on the classical Steemann Nielsen in vitro 14 Ctracer approach (Steemann Nielsen, 1952).Bottle incubations outside of the mesocosms were performed at 1 m depth, equivalent to approximately 60 % of incoming light, and over a period of 24 h.
The 14 C-tracer approach has the advantage of being highly sensitive, and thus ideally suited for fieldwork, when there is low photosynthetic activity.One drawback of this method, however, is that C-uptake rates cannot be attributed precisely to either net or gross primary production (Peterson, 1980;Dring and Jewson, 1982).Short-term incubations are expected to provide gross rates of C-fixation, whereas longer incubations tend to measure net production, depending, however, on the metabolic activity of the microbial community included.
Primary production was compared to changes in the concentration of DOC and to the production of bacterial biomass in order to infer the fate of freshly produced organic compounds at different pCO 2 in this Arctic ecosystem.
Sampling of seawater from the mesocosms was conducted with a depth-integrated water sampler (Hydro-Bios).The sampler is equipped with a motor and continuously collects water (5 L volume) while being lowered from surface to 12 m depth.Samples were collected in the morning (09:00 a.m.-11:00 a.m.local time).
Triplicate light and one dark incubation were performed for each of the nine mesocosms and for the fjord on days −1, 2, 5, 7, 10, 12, 14, 16, 18, 20, 22, 24, 26, and 28 of the experiment.Dark incubation was conducted with black taped bottles.All samples were incubated for 24 h.The incubation length was chosen for two reasons.First, we expected an overall low productivity of the Arctic phytoplankton www.biogeosciences.net/10/1291/2013/Biogeosciences, 10, 1291-1308, 2013 community at low temperatures, low biomass density, and low nutrient concentrations at the start of the experiment.Under these conditions, short-term incubations of only a few hours may underestimate primary production, because carbon assimilation by algal cells may be too low to discriminate against 14 C adsorption as determined in blank dark incubation.Moreover, release of freshly assimilated carbon into the pool of dissolved organic matter has a time scale of several hours, because of equilibration of the tracer and because metabolic processes of organic carbon exudation follow those of carbon fixation inside the cell.Another reason was to cover the daily photoperiod for the cells.Since the experiment was conducted at high latitude (78 • 56 N) and around the time of summer solstice, light availability was high (> 100 µmol photons m −2 s −1 ) even during the middle of the night (Schulz et al., 2013), and supported autotrophic production over a 24 h period.Other studies in the Svalbard area therefore also used 24 h incubations for measurements of primary production when working with the 14 C-tracer (Iversen and Seuthe, 2011;Hodal et al., 2011).
Incubations were performed close to the marine lab at 1 m depth, receiving about 60 % of the incoming photosynthetically active radiation (PAR) during most time of the study.For comparison, in vitro O 2 -measurements were performed at 4 m depth, equivalent to 20 % PAR, whereas productivity estimates directly in the mesocosms obtained from DIC changes (Silyakova et al., 2012) or 13 C incorporation (de Kluijver et al., 2012) yielded integrated values over a 12 m water column that received 100-17 % of incoming light, with a median value of 23 % (see Fig. 2).Hence, 14 C primary production rates were obtained at a relatively high light level.This level was chosen to ensure that (i) cells would not become light-limited in the course of the study, and (ii) cell would receive enough light for determinable exudation of DOC, since exudation in marine phytoplankton has been reported to increase with light availability (Zlotnik and Dubinsky, 1989).
Incubations were stopped by filtration of a 50 or 100 mL sub-sample onto 0.4 µm polycarbonate filters (Nuclepore).Primary production of POC (PP POC ) was determined from material collected on the filter, while the filtrate was used to determine primary production of DOC (PP DOC ).After removing the vials collecting the filtrate of the associated filter, all filters were rinsed with 10 mL sterile filtered (< 0.2 µm) seawater, and then acidified with 250 µL 2N HCl in order to remove inorganic carbon (Descy et al., 2002).Filters were transferred into 5 mL scintillation vials, 200 µL of 2N NaOH, and 4 mL scintillation cocktail (Ultima Gold AB) were added.For determination of PP DOC , 4 mL of filtrate were transferred to 20 mL scintillation vials, acidified (100 µL 1N HCl) and left open in the fume hood to remove inorganic carbon.Then, 800 µL of 2N NaOH and 15 mL scintillation cocktail were added.All samples were counted the following day in a liquid scintillation analyzer (Packard Tri-Carb, model 1900 A).Primary production of organic carbon was calculated from scintillation data according to Gargas (1975): where a 1 and a 2 are the activities (DPM) (disintegrations per minute) of the added solution and of the sample corrected for dark samples, respectively, and DI 12 C is the concentration (µmol L −1 ) of dissolved inorganic carbon (DIC) in the sample.The value 1.05 is a correction factor for the discrimination between 12 C and 14 C, as the uptake of the 14 C isotope is 5 % slower than the uptake of 12 C, k 1 is a correction factor for subsampling (bottle volume/filtered volume) and k 2 is the incubation time (d −1 ).Total primary production (PP; µmol C L −1 d −1 ) was derived from the sum of the production of PP POC and PP DOC according to PP = PP POC + PP DOC . (2) The percentage of extracellular release (PER) was calculated as Based on 14 C primary production, the cumulative production of POC and DOC was calculated from the sum of daily production.Values for days between measurements were calculated by linear interpolation of adjacent data points.
Primary production estimates obtained with the 14 Cmethod at 1 m depth exceeded O 2 gross community production (GCP) determined at 4 m depth by a factor of ∼ 2 (Tanaka et al., 2013), and O 2 -and DIC-based net community production (NCP) by a factor of 3-4 (Silyakova et al., 2012).These discrepancies were mainly due to the different amount of light that cells received during the various measurements, and are comparable to differences observed for polar phytoplankton along comparable depth and light gradients (Yun et al., 2012).

Light and temperature during the 14 C incubations
PAR, for practical reasons defined as radiation in the wavelength range 400-700 nm, and temperature were determined at the incubation site in the afternoon (between 03:00 and 04:00 p.m.) from day 7 onwards by the use of a CTDmounted LICOR spherical quantum sensor .
Seawater temperature increased during the mesocosm study from 2.0 • C at the beginning of June to 5.2 • C at the end of the study.No temperature differences were observed among the nine mesocosms, and between the mesocosms and the fjord.At the site and depth of 14 C incubations, temperature was on average 1-1.5 • C higher than at the location of mesocosm deployment.
PAR ranged between 130 and 800 µmol photons m −2 s −1 at the incubation site (1 m), representing cloudy and clear sky, respectively, and corresponded to approximately 60 % of surface light for most time (range: 45-85 %) (Fig. 2).PAR at the incubation site was not significantly different from the 1 m depth horizon in the fjord at the mesocosm site (p = 0.09).

Chlorophyll a
Concentration of chlorophyll a (chl a) in the mesocosms and in the fjord was determined from 500 mL seawater filtered onto glass fiber filters (Whatman GF/F 25 mm, pre-combusted 450 • C for 5 h) by low vacuum filtration (< 200 mbar) and stored frozen at −20 • C. Chl a was determined fluorometrically according to Welschmeyer (1994) using a Turner fluorometer 10-AU (Turner BioSystems, CA, USA).

Dissolved organic carbon (DOC)
Samples for dissolved organic carbon (DOC) were collected in combusted glass ampoules after filtration through combusted GF/F filters.Samples of 20 mL were acidified with 100 µL of 85 % phosphoric acid and stored at 4 • C in the dark until analysis.DOC samples were analysed using the hightemperature combustion method (TOC-VCSH, Shimadzu) (Qian and Mopper, 1996).A multi-point calibration curve was constituted for each day of measurement using potassium hydrogen phthalate standard, which was prepared in Milli-Q water.Additionally, two reference seawater stan-dards (Hansell laboratory, RSMAS, University of Miami) were used to determine the instrument blank.Each sample was measured in quadruplets.
Considerable day-to-day variations of DOC concentrations of up to 30 % were observed on some days in all mesocosms and in the fjord samples.These variations may partly be attributed to contamination of samples during sample collection and transport as well as during instrument deployment inside the mesocosms.We assume that this methodological error occurred randomly and was not discriminating between CO 2 treatments.Thus, although the absolute concentration of DOC may have been defective on individual days, averaged and time-averaged differences in DOC concentration between treatments should be reliable.
In order to identify treatment related differences, we calculated mean deviations of DOC concentration (MD-DOC) in the mesocosms.We did not include fjord samples in this analysis, because temporal variations of DOC concentration in the fjord may have been due to processes other than biological activity, such as glacial melting and terrestrial meltwater run-off.The latter was at times indicated in the area of mesocosm deployment by a brownish color of surface water.

Bacterial secondary production
Bacterial production (BP) was estimated from the uptake of 14 C leucine during < 24 h incubations in 2 mL vials at 2 • C in the dark.Duplicate incubations revealed an analytical error ≤ 10 %.Rates of 14 C leucine incorporation were converted into BP applying a conversion factor of 1.5 kg C mol −1 leucine (Ducklow et al., 1999).For more information see Piontek et al. (2013).

Data treatment
Differences in data as revealed by statistical tests (t-test, ANOVA, Kolmogorov-Smirnov test) were accepted as significant for p < 0.05.Average values for total concentrations are given by their arithmetic mean, averages for ratios by their geometric mean.
For identifying differences between the pCO 2 treatments, absolute deviations (AD) were calculated for each mesocosm for PP POC , PP DOC and DOC.Therefore, the arithmetic mean of all mesocosm observations per time-point was subtracted from each mesocosm observation at that time-point.The mean deviation (MD) represents the arithmetic mean of AD for a specific time interval, and is expressed in a relative value ( %).Three time intervals were considered: total period of CO 2 treatment (day 5-day 28), before nutrient addition (day 5-day 12), and after nutrient addition (day 14-day 28).MD values thus illustrate how one mesocosm deviates from the mean development in all mesocosms, i.e., the anomaly of a mesocosm.
Table 1.Time averaged (day 5-day 28) rates (µmol C L −1 d −1 ) of total primary production (PP), primary POC production (PP POC ), and primary DOC production (PP DOC ), based on 14 C bottle incubations, as well as ratios of PP normalized to chlorophyll a concentration (µmol C µg −1 chl a d −1 ).Averages (Avg.) and standard deviations (SD) were calculated from n = 12 observations for each mesocosm and for the fjord, respectively.Calculations, statistical tests and illustration of the data were performed with the software packages Microsoft Office Excel 2010 and SigmaPlot 12.0 (SYSTAT).

Bloom development
Changes in chl a concentration (range: ∼ 1-3 µg L −1 ) during the study indicated the development of one smaller phytoplankton bloom before day 13, i.e., before addition of nutrients to the mesocosms, as well as two bloom phases thereafter (Fig. 3).Thereby, the bloom directly following nutrient addition (day 14-day 22) developed faster and more pronounced in the high pCO 2 mesocosms, while the second bloom phase after day 23 was characterized by higher chl a concentrations in the lower pCO 2 mesocosms.For more in-

Primary production of organic carbon
Primary production (PP) during the time of the experiment (day 5-day 28) averaged 8.1 ± 3.6 µmol C L −1 d −1 in all mesocosm samples and was slightly higher than in the fjord samples with 6.5 ± 2.5 µmol C L −1 d −1 (Table 1).PP varied significantly between mesocosm samples (ANOVA; p < 0.001), with highest rates observed in the high CO 2 mesocosm (M9: 1136 µatm) and lowest rates in the low CO 2 mesocosm (M7: 180 µatm).
We observed that PP POC on the first day of incubation (day −1), i.e., after first salt addition but before pCO 2 perturbation, was not equal among samples that were collected from the mesocosms.While mesocosms 1-3 had a similarly high primary production of POC (PP POC ) in range of 4.1-6.1 µmol C L −1 d −1 , comparable to PP POC observed in the fjord, mesocosm 4-9 clearly showed lower productivity.This difference in the initial conditions between mesocosms disappeared during the following days and was already absent at day 2 (Table 2).
PP POC as well as PP DOC increased with increasing phytoplankton biomass after nutrient addition on day 13 (Tables 2, 3).Response of PP POC to nutrient addition was clearly faster in the high pCO 2 mesocosms; i.e., between day 12 and day 16, PP POC increased by 74 % in the high CO 2 mesocosms, by 48 % in the medium, and by only 21 % in the low CO 2 mesocosms.
For the total period of the experiment (day 5-day 28), a cumulative PP POC between 84 and 174 µmol C L −1 was obtained for the three lowest, between 94 and 203 µmol C L −1 for the three medium, and between 196 and 222 µmol C L −1 for the three highest pCO 2 mesocosms.For comparison, cumulative PP POC in the fjord was 138 µmol C L −1 and therewith in the range of data observed in mesocosms with a similarly low pCO 2 .Cumulative PP POC of the au- totrophic community clearly increased with CO 2 concentration (p < 0.01), while the variability between mesocosms decreased.Hence, highest variability of cumulative POC production was observed at the lower end of the pCO 2 range (Fig. 5a).The difference in cumulative PP POC between low and high CO 2 treatments covered a relatively broad range, i.e., 29 µmol C L −1 comparing M3 (178 µatm) and M5 (892 µatm), or 138 µmol C L −1 comparing M7 (180 µatm) and M9 (1136 µatm).Thus, the applied pCO 2 induced an increase in PP POC by 10-60 %.The cumulative production (day 5-day 28) of DOC was estimated in a similar way and ranged between 19 and 36 µmol C L −1 in the low, 20-34 µmol C L −1 in the medium, and 32-40 µmol C L −1 in the high pCO 2 mesocosms (Fig. 5b).Cumulative PP DOC in the fjord, was estimated to 19 µmol C L −1 , and thus at the lower end of values observed in the mesocosms.Similar to cumulative PP POC , cumulative PP DOC increased significantly with CO 2 concentration (p < 0.05).Maximum difference in cumulative PP DOC was observed between M7 (180 µatm) and M9 (1136 µatm) with 21 µmol C L −1 , equivalent to an increase by about 50 %.However, variability of cumulative PP DOC was high at the lower pCO 2 range also.The low pCO 2 mesocosm M3 (178 µatm) even yielded about 11 % higher cumulative PP DOC than the high CO 2 treatments M5 (892 µatm) and M6 (701 µatm).
PP DOC generally increased after nutrient addition, following the course of PP (Table 1).The percentage of extracellular organic carbon release (PER), however, decreased immediately after nutrient addition in all mesocosms (Fig. 6).Un-til day 12 PER ranged between 21 and 23 %.After nutrient addition, PER was 18 ± 6 % in the three low pCO 2 mesocosms and decreased with increasing pCO 2 to 14 ± 5 % in the three high CO 2 mesocosms.Thus, nutrient addition suppressed exudation at higher pCO 2 more than at low pCO 2 (t-test, p < 0.05), suggesting in turn that a higher proportion of PP was used for POC production at high pCO 2 .However, due to absolute higher PP, the total amount of DOC released by autotrophs was still higher at high CO 2 despite of lower PER.
In the fjord, PER was 14 ± 8 % until day 12, and also decreased -not impacted by nutrient addition -to 11 % by day 14.This suggests that nutrient addition was not the sole factor responsible for the PER decrease after day 12.Another factor that has often been reported to increase exudation of organic carbon is light (Zlotnik and Dubinsky, 1989).During this study, we also observed a moderate increase in PER with light intensity (Fig. 7, p < 0.05).Following this argument, light was likely not responsible for the reduction of PER observed on day 14, because PAR at that day was 325 ± 164 µmol photons m −2 s −1 and rather above than below the PAR of previous days.Temperature has also been suggested to affect exudation, yielding higher PER at higher temperatures (Zlotnik and Dubinsky, 1989;Moran et al., 2006;Engel et al., 2011).However, since temperature increased in the course of the mesocosms study, this also would favor rather than suppress PER.We do not know if the decreases in PER in fjord and in mesocosms samples around day 14 were related, or just coincided.Therefore, we cannot exclude a potential co-effect on PER besides nutrient availability.
Mean deviations (MD) of PP POC were positive for the three highest CO 2 mesocosms during all periods (Fig. 8a-c).This was most pronounced for the period after nutrient addition, when MD of PP POC in the high pCO 2 mesocosm (974 µatm) was 44 % higher than average.For the total period, a significant positive relationship was observed between MD-PP POC and average pCO 2 (p < 0.05) (Fig. 8a).This relationship was not seen during the time before nutrient addition, but clearly observed thereafter (p < 0.01) (Fig. 8b, c).
Again, relatively large differences were determined among the low CO 2 mesocosms, where MD-PP POC ranged from −49 % to +6 %.
For PP DOC , the relationships of MD to average pCO 2 during the respective periods were significant before as well as after nutrient addition (Fig. 9a-c).Largest negative values for MD-PP DOC were observed for the period after nutrient addition for the fjord (−57 %) and for the low pCO 2 mesocosm M7 (−41 %).Largest positive values for MD-PP DOC again were determined in samples of the high pCO 2 mesocosm M9 (+40 %).

DOC concentration
Average DOC concentration in the mesocosm at day −1 was 76 ± 3 µmol C L −1 , and slightly higher than observed in the fjord at that day (71 µmol C L −1 ).DOC concentrations were thus lower than the annual range of 100-244 µmol C L −1 determined for the Kongsfjorden by Iversen and Seuthe (2011) Fig. 6.Exudation of DOC calculated as percentage of extracellular release (PER) and averaged for grouped treatments (low, medium, high pCO 2 ) for the time before nutrient addition (hatched bars, day 5-day 12, n = 12), and after nutrient addition (solid bars, day 14-day 28; n = 21).For color information see Fig. 3. PER was not significantly different before nutrient addition, but decreased thereafter with increasing pCO 2 (t-test, p < 0.05). .PER during the total period of observation (day 5-day 28) increased with average PAR received during the 24 h bottle incuba-For color information see Fig. 3. Cuevas et al. (2011), i.e., 61-84 µmol L −1 , and by Myklestad and Boersheim (2007), i.e., 87 ± 16 µmol C L −1 .DOC concentration increased significantly between day 4 and day 13 in all mesocosms, yielding a rate of 1.6 ± 5.4 µmol C L −1 d −1 (p < 0.01) (Fig. 10), equivalent to 15 ± 5.4 µmol C L −1 for this period.DOC accumulation before nutrient addition was thus comparable to cumulative PP DOC (range day 12: 8-13 µmol C L −1 ).For the period after nutrient addition, no further accumulation of DOC was observed, and values averaged 91 ± 7 µmol C L −1 .The ab-sence of DOC accumulation during the bloom periods was in contrast to the potential production of DOC by PP DOC , which was estimated for that period to amount to 11-27 µmol C L −1 .
For the mesocosms, a positive correlation between MD-DOC and average pCO 2 was observed only for the period before nutrient addition (p < 0.05) (Fig. 11), and in accordance with increasing PP DOC at higher pCO 2 observed during this period (Fig. 9b).After nutrient addition, no significant difference in MD-DOC between mesocosms was observed, despite CO 2 -related differences in carbon exudation.

Primary vs. bacterial production
Primary produced organic compounds directly fuel the heterotrophic food web, amongst which bacteria are the main consumers.Bacterial production (BP) during this study ranged between 0.04 and 0.54 µmol C L −1 d −1 in the mesocosms, and between 0.10 and 0.84 µmol C L −1 d −1 in the fjord samples.Detailed information is given in Piontek et al. (2013).BP was directly related to PP considering the entire duration of the experiment (day 5-day 28) and all mesocosms (n = 108, r 2 = 0.28, p < 0.001) (Fig. 12).Assuming that bacteria preferentially consume dissolved organic compounds, we calculated the ratio of BP : PP DOC Here, values ranged between 20 % and 50 % in the mesocosms (Fig. 13a), and were lower than in the fjord water outside the mesocosms.
Related to the total amount of organic carbon produced, the fraction of BP was much smaller.Averaged over all mesocosms, BP : PP was 3.5 ± 1.9 %, and lower than in the fjord at the same time (6.5 ± 4.0 %).BP : PP DOC as well as BP : PP did not differ significantly between mesocosms, nor over time (ANOVA; p > 0.1), and no significant influence of nutrient addition at day 13 was determined either (t-test; p > 0.1).However, lowest ratios were observed at highest pCO 2 (Fig. 13b).

Temporal variability of primary production during the experiment
The experiment started at a time when the natural autotrophic community in the Kongsfjorden experienced low nutrient concentrations (Schulz et al., 2013).Until the day of nutrient addition (day 13), PP POC and PP DOC in the mesocosms were low and similar to rates determined in the fjord.During this time, CO 2 -related differences were identified for PP DOC but not for PP POC .Higher exudation of DO 14 C at higher pCO 2 was in good accordance with higher accumulation of DOC.Addition of nutrients to the mesocosms on day 13 initiated phytoplankton bloom developments with a faster and more pronounced immediate response of the autotrophic community at high pCO 2 (Fig. 3; see also Schulz et al., 2013).In   accordance, higher values for PP POC and PP DOC were determined for the high pCO 2 mesocosms, also.
A positive response of primary production (PP POC ) to increasing seawater pCO 2 has been observed during earlier mesocosm experiments (Egge et al., 2009), as well as during laboratory studies for a variety of phytoplankton species and at different light and temperature conditions (Hein and Sand-Jensen, 1997;Schippers et al., 2004;Rost et al., 2008).A stimulation of photosynthesis by increasing pCO 2 is attributed to the Michaelis-Menten type relationship between photosynthesis rate and CO 2 concentration, showing high sensitivity of photosynthesis to changes in CO 2 at lower CO 2 concentration and little changes at high and saturating pCO 2 .During this study, larger differences of primary production rates were observed among the low pCO 2 mesocosms and may be explained by differences in the CO 2 affinity (K m value) between phytoplankton species (Reinfelder, 2011).Hence, the natural variability in species composition and physiology of the phytoplankton community likely translated into larger differences of primary production rates among the low pCO 2 mesocosms.
Overall, the temporal development of primary production of phytoplankton sampled from the mesocosms and the stimulation of PP POC and PP DOC by increasing pCO 2 met well with our expectations and earlier findings.

Primary production vs. net community production
While 14 C primary production increased with pCO 2 during all phases of the experiment, net community production (NCP) determined by in vitro O 2 measurements as well as by cumulative changes of DIC concentration inside the   Relationship between bacterial biomass production (BP) and primary production (PP) in the mesocosm samples was highly significant (p < 0.001, n = 108) for the total period of the experiment (day 4-day 28).
mesocosms was highest at low CO 2 concentration during the later phase of the experiment, i.e., after day 21 (Tanaka et al., 2013;Silyakova et al., 2012).A different or even anti-correlated response of PP and NCP to increasing pCO 2 would have important implications for carbon and oxygen cycling in the surface ocean.We therefore will try to find some explanations for the apparent discrepancies.First, the observed differences between PP and NCP may be due to methodological constraints.It has to be emphasized that the 14 C technique gives an estimate for the assimilation of carbon into POC and DOC that is lower than gross but higher than net production.Even under high heterotrophic activities, 14 C primary production rates cannot become negative, as respiration of abundant organic matter by heterotrophic organisms cannot be accounted for.Respiration, however, is included in NCP measurement based on O 2 or DIC, and negative NCP was determined on some days during this study (Tanaka et al., 2013), suggesting that "older" and previously abundant organic matter was respired by the plankton community.Thus, in vitro 14 C-PP measurements are biased towards autotrophic production, while NCP measurements rather estimate the net productivity of the autoand heterotrophic community.This general difference was even amplified in this study, because our 14 C incubations were performed at high light (1 m) and excluded larger zooplankton (> 200 µm), while in vitro O 2 and DIC measurements also included lower light levels (4 m and whole mesocosm) without pre-filtering.
Discrepancies between 14 C-PP and NCP were primarily observed during the second bloom phase after nutrient addition.Until day 21, highest cumulative NCP as estimated from DIC was determined for the high pCO 2 mesocosms (Silyakova et al., 2012; this study), in accordance with higher 14 C-PP.Thus, a potential cause for the discrepancy between 14 C-PP and NCP estimates likely involved the response of the heterotrophic community and evolved during the experiment.
We suggest that heterotrophic microbes were primarily responsible for differences in the response of PP and NCP to CO 2 .During this study, heterotrophic activity was closely coupled to PP, as derived from bacterial production and from hydrolytic enzyme activities (Fig. 11; see also Piontek et al., 2013).Prior to the experiment, bacterial growth was limited by the availability of labile organic carbon and co-limited by nitrogen (Piontek et al., 2013).It can therefore be assumed that bacteria directly responded to the release of labile organic carbon by phytoplankton.Nutrient addition at day 13 then not only provided substrate for autotrophic cells but likely fueled the growing community of heterotrophic bacteria also.
After nutrient addition, values of PER decreased in all mesocosms.Nutrient limitation has been shown earlier to increase PER in marine phytoplankton (Myklestad, 1977;Obernosterer and Herndl, 1995;Lopez-Sandoval et al., 2011).A reduction of PER in response to the elimination of nutrient limitation as observed during this study supports the idea of exudation being a discharge mechanisms for excess photosynthates (Wood and van Valen, 1990;Schartau et al., 2007).Nevertheless, as PP was higher at high CO 2 , the absolute rates of PP DOC were still higher in the high CO 2 mesocosm samples.The observation that PP DOC increased with pCO 2 after nutrient addition, but DOC concentration did not (Figs.5b and 10), suggests that the growing community of heterotrophic bacteria consumed DOC to a larger extent at high pCO 2 .
Higher PP POC in the high pCO 2 mesocosms translated into higher phytoplankton biomass directly after addition and before re-depletion of nutrients.This is in accordance with our expectation, as the utilization of photosynthetic products for biomass synthesis by heterotrophic as well as by autotrophic cells depends on the availability of nitrogen and phosphorus.Data on cell abundance as determined by flow cytometry suggest that particularly fast-growing picoautotrophic cells benefitted from nutrient addition at high pCO 2 (Brussaard et al., 2012).
However, with regard to the entire study period, the maximum yield of phytoplankton biomass in the low pCO 2 mesocosms exceeded the maximum biomass yield in the high CO 2 treatments, despite higher PP POC in the latter.This apparent difference between autotrophic POC production and accumulation of phytoplankton biomass at high CO 2 may be explained by either one, or a combination of the following processes: (i) enhanced settling loss of phytoplankton biomass from the water column, (ii) enhanced solubilization and remineralization of phytoplankton cells, or (iii) increased nutrient competition between auto-and heterotrophic microorganisms.(i) It has been suggested that under nutrient limiting conditions phytoplankton produce more exopolymer carbohydrates at high pCO 2 (Engel, 2002;Borchard and Engel, 2012).Since exopolymer carbohydrates are important agents in coagulation processes and enhance aggregate formation, a higher export of organic matter may be inferred.However, higher export fluxes and therewith a higher loss of organic matter from the water column in the high pCO 2 bags were not directly observed during this study (Czerny et al., 2012).(ii) Recent studies suggest that bacterial processes such as organic matter solubilization and hydrolysis by extracellular enzymes are enhanced by ocean acidification (Grossart et al., 2006;Piontek et al., 2010Piontek et al., , 2013;;Yamada and Suzumura, 2010;Endres et al., 2013).Higher activities of hydrolytic enzymes were observed at reduced pH also during side-experiments of this study (Piontek et al., 2013), and may have resulted in faster degradation of organic matter, including autotrophic biomass.(iii) It is well known that the release of organic substrates from phytoplankton fuels the microbial food web (Azam and Hodson, 1977;Azam et al., 1983).The higher production and release of DOC at high pCO 2 likely enhanced the utilization of organic carbon, oxygen and nutrients by marine bacteria also during this study.The higher demand for nitrogen and phosphorus in marine bacteria potentially exacerbated competition between phytoand bacterioplankton for inorganic nutrients and curtailed autotrophic growth.During this study, nutrient consumption directly after nutrient addition was faster in the high pCO 2 mesocosms.Although more autotrophic biomass, as indicated from chl a, was observed at higher pCO 2 initially, a much stronger phytoplankton bloom developed later during the experiment at low pCO 2 (Schulz et al., 2012).As a consequence the peak ratios of chl a to particulate organic nitrogen ([chl a] : [PON]) achieved during this study were smaller at high pCO 2 than at medium and low pCO 2 (data Schulz et al., 2012).This supports the idea that a higher amount of nutrients were partitioned into the heterotrophic food web under high pCO 2 .

Conclusion
Increasing CO 2 concentration can enhance the primary production of organic carbon by Arctic phytoplankton.Due to higher primary production, the amount of DOC released by phytoplankton at high pCO 2 may also increase.However, as activities of Arctic bacterioplankton seem closely coupled to primary production, bacteria will efficiently counteract a surplus of labile organic carbon during bloom and postbloom situations.The stimulation of bacterial activities, further supported by acceleration of extracellular enzyme activities, would exacerbate the competition between phyto-and bacterioplankton for inorganic nutrients.
As a consequence, net community production may decrease at high pCO 2 despite higher primary production of organic carbon.Such a counterintuitive cycling of carbon (i.e., higher autotrophic carbon fixation leads to less net production of the whole community) has been hypothesized for Arctic systems previously (Thingstad et al., 2008).
The Arctic Ocean at present is a net sink for atmospheric CO 2 on an annual scale (Arrigo et al., 2010), and an increase in primary production and biological CO 2 draw-down associated with the ongoing sea-ice loss have been predicted (Arrigo et al., 2008).This study reveals that primary production may increase in the wake of ocean acidification.However, the heterotrophic microbial community has a strong potential to diminish air-sea carbon fluxes and needs to be considered when estimating the response of the Arctic Ocean to future environmental changes.

Fig. 2 .
Fig. 2. Fraction of surface light received at different depths in the mesocosms in the course of the study as exemplified for M1.For comparison, bottle incubations were performed at 1 m depth ( 14 C incubations) and at 4 m depth (O 2 , Tanaka et al., 2013), while changes in DIC concentration were calculated from depth-integrated water sampling (0-12 m; de Kluijver et al., 2012; Silyakova et al., 2012).

Fig. 3 .
Fig. 3. Biomass changes of the phytoplankton community in the nine mesocosms as indicated by chlorophyll a (chl a) concentration.

Fig. 5 .
Fig. 5. Cumulative primary production of POC (a) and of DOC (b) as determined from 14 C-bottle incubations for the different mesocosms and for the fjord.Values for pCO 2 are the arithmetic mean of data over the full period of observation (day 5-day 27).
Fig. 7. PER during the total period of observation (day 5-day 28) increased with average PAR received during the 24 h bottle incuba-For color information see Fig. 3.

Fig. 8 .
Fig.8.Mean deviations of PP POC (MD-PP POC , %) for the nine mesocosms and for the fjord (left bar) calculated for (a) the total period of observation (day 5-day 28; n = 12), (b) the period before nutrient addition (day 5-day 12; n = 4), and (c) the period after nutrient addition (day 14-day 28; n = 8).Significance of relation between MD-PP POC and average pCO 2 at the time of observation was calculated by linear regression.

Fig. 9 .
Fig. 9.Mean deviations of PP DOC (MD-PP DOC , %) for the nine mesocosms and for the fjord (left bar) calculated for (a) the total period of observation (day 5-day 28; n = 12), (b) the period before nutrient addition (day 5-day 12; n = 4), and (c) the period after nutrient addition (day 13-day 27; n = 8).Significance of relationship between MD-PP DOC and average pCO 2 at the time of observation was calculated by linear regression.

Fig. 11 .
Fig. 11.Mean deviations of DOC concentrations (MD DOC , %) for the nine mesocosms, for (a) the total period of observation (day 4-day 27; n = 24), (b) the period before nutrient addition (day 4-day 12; n = 10), and (c) the period after nutrient addition (day 13-day 27; n = 14).Significance of relation between MD DOC and average pCO 2 at the time of observation was calculated by linear regression.Fjord samples were not included. Figure12

Fig. 13 .
Fig. 13.Box and whisker plots of the ratio of bacterial biomass production (BP) to (a) primary production of DOC (PP DOC ) and to (b) total primary production (PP = PP POC + PP DOC ) as derived from dark and light bottle incubations.Average values (day 5-day 28) for pCO 2 are shown.

Table 3 .
Production (µmol C L −1 d −1 ) of dissolved organic carbon (DOC), based on 14 C bottle incubations.Averages (Avg.) and standard deviations (SD) were calculated from on triplicate measurements of 24 h incubations.