Phosphorus dynamics in the Barents Sea

The Barents Sea is considered a warming hotspot in the Arctic; elevated sea surface temperatures have been accompanied with increased inflow of Atlantic water onto the shelf sea. Such hydrodynamic changes and a concomitant reduction of sea ice coverage enables a prolonged phytoplankton growing season, which will inevitably affect nutrient stoichiometry and the controls on primary production. During the summer of 2018, we investigated the role of phosphorus in mediating primary production in the Barents Sea. Dissolved inorganic phosphorus (DIP), its most bioavailable form, had an average net turnover time of 9.4 ± 4.8 d. The most southern Atlantic influenced station accounted for both the highest rates of primary production (655 mg C m2 d−1) and shortest net DIP turnover (2.8 ± 0.5 d). The fraction of assimilated DIP released as dissolved organic phosphorus (DOP) at this station was < 4% compared to an average of 21% at all other stations. We observed significant differences between phytoplankton communities in Arctic and Atlantic waters within the Barents Sea. Slower DIP turnover and greater release of DOP was associated with Phaeocystis pouchetii dominated communities in Arctic waters. Faster turnover rates and greater phosphorus retention occurred among the Atlantic phytoplankton communities dominated by Emiliania huxleyi. These findings provide baseline measurements of P utilization in the Barents Sea, and suggest increased Atlantic intrusion of this region could be accompanied by more rapid DIP turnover, possibly leading to future P limitation (rather than N limitation) on primary production.

Phosphorus (P) uptake by oceanic plankton has received little attention compared to nitrogen (N), mainly because N is considered limiting over biologically relevant time scales, and P over geological time scales (Hecky and Kilham 1988;Tyrrell 1999). Despite this, evidence is mounting to suggest P can be the proximal limiting macronutrient for primary production in various marine environments, particularly warm oligotrophic systems and coastal waters with enhanced N enrichment (Bjorkman and Karl 1994;Wu et al. 2000;Thingstad et al. 2005). While global primary production is expected to decrease as a result of warmer temperatures enhancing water column stratification (Falkowski et al. 1998;Behrenfeld et al. 2006), the Arctic is expected to see an increase in production due to reduced light limitation caused by declining sea ice extent and thickness (Carmack et al. 2006).
Extended areas of open water in Arctic systems could additionally lead to longer periods of phytoplankton growth and resulting nutrient limitation (Tremblay and Gagnon 2009). Thus, baseline knowledge of microbial P dynamics in Arctic waters is necessary to understand future productivity and associated changes in this region.
DIP is the most readily available form of P for phytoplankton; this pool is largely comprised of phosphate (PO 4 ). When PO 4 is depleted in the euphotic zone relative to other nutrients, primary production becomes P limited. Such conditions generally favor smaller phytoplankton species with a high surface area to volume ratio, as cellular P quotas and uptake rates scale with cell size (Edwards et al. 2012;Lomas et al. 2014). Hence, it has been widely reported that the smaller size fraction of the plankton community (< 3 μm) contribute to the majority of P uptake fluxes (Bjorkman and Karl 1994;Thingstad et al. 1998;Moutin et al. 2002;Tanaka et al. 2003). The availability of PO 4 can therefore impact the biological carbon pump, both in regulating the magnitude of primary production and influencing species composition, whereby low PO 4 conditions relative to other nutrients can favor smaller phytoplankton species with slower sinking velocities.
An alternative source of P for phytoplankton is in the form of dissolved organic phosphorus (DOP). The proportion of DOP in the total dissolved P pool varies regionally. For example, in shelf seas DOP accounts for 11-86% of the total dissolved P in the photic zone, this proportion is often greater in oligotrophic open ocean systems, ranging between $ 66% and 100% (Karl and Björkman 2015 and references therein). The chemical composition of DOP dictates its availability for biological assimilation. However, the chemical form and thus bioavailability of DOP is mostly unknown (Karl 2014). DOP originates from organic matter production with subsequent release via exudation, grazing, or cell lysis (Karl and Björkman 2015). Advective transport of DOP can provide downstream regions with an additional P source, and it has been postulated that this could be the case for Atlantic water entering the Barents Sea shelf (Torres-Valdes et al. 2009).
In order to utilize DOP which is not readily available for assimilation, microbes hydrolyse P from the organic moiety prior to uptake, and this is usually induced by the ectoenzyme alkaline phosphatase (AP). Thus, enhanced AP activity can be a proxy for DIP limitation (Rees et al. 2009;Lomas et al. 2010;Mahaffey et al. 2014;Labry et al. 2016). However, elevated AP activity has also been reported in DIP-replete conditions (Sebastian et al. 2004;Labry et al. 2016; Davis and Mahaffey 2017). Such ambiguity highlights the need for multiple, simultaneous indicators when investigating microbial P dynamics, including net and gross turnover time of phosphorus pools. Net turnover accounts for the fraction of DIP assimilated into the particulate fraction, while gross turnover includes the uptake of DIP and release of DOP. The difference between net and gross turnover gives an estimation of P retention. Low DOP release (high retention) and efficient recycling of P due to low ambient DIP concentrations has been demonstrated in Celtic Sea post spring phytoplankton bloom scenarios (Poulton et al. 2019).
The Barents Sea is a highly productive shelf sea experiencing seasonal ice cover. This region acts as a gateway between the Arctic and Atlantic Ocean, with an inflow of warmer saline Atlantic water from the south and colder fresher Arctic waters dominating the Northern region (Loeng et al. 1997). Over the past 30 yrs, the volume of Atlantic water entering the shelf sea has doubled (Oziel et al. 2016), while sea ice import has declined (Lind et al. 2018). This loss of freshwater and intrusion of saline Atlantic water has led to reduced sea ice formation (Lind et al. 2018). The Barents Sea accounts for approximately 40% of Arctic shelf sea primary production (PP) (Arrigo and van Dijken 2015). Ice algae account for $ 6% of annual PP in the region, with the majority being associated with pelagic phytoplankton (Wassmann et al. 2006). During spring, the seasonal ice zone retreats giving way to open water with a marginal ice zone on the boundary, this is accompanied by a northward progressing spring bloom (Sakshaug and Skjoldal 1989). The phytoplankton blooms succeeding the ice retreat are characterized by eutrophic conditions and new production, this is then followed by an oligotrophic summer period (Luchetta et al. 2000), during which nitrate is depleted to greater extent relative to DIP and silicate, and regenerated nutrients sustain phytoplankton growth (Erga et al. 2014).
Phytoplankton assemblages in the Barents Sea have been extensively studied, especially around the polar front, which is generally dominated by diatoms and the prymnesiophyte Phaeocystis pouchetii (Sakshaug et al. 1994;Wassmann et al. 1999;Degerlund and Eilertsen 2010). Observations in recent years, both in situ and satellite, have reported trends of extensive blooms of the coccolithophore Emiliania huxleyi (Smyth et al. 2004;Hegseth and Sundfjord 2008;Erga et al. 2014;Hovland et al. 2014;Oziel et al. 2017), a calcifying species typically associated with warmer temperate waters. E. huxleyi is known to occupy a niche of low nutrient conditions and increased stratification post spring bloom (Tyrrell and Merico 2004;Lessard et al. 2005). With rising temperatures and increased "Atlantification," E. huxleyi blooms are expected to increase in the Barents Sea in the future (Neukermans et al. 2018).
To our knowledge, there are no extensive studies of P uptake in the Arctic shelf marine environments. Therefore, the aim of this work was to build an understanding of P utilization in the Barents Sea, during the summer growing season where phytoplankton growth can be nutrient limited. We provide measurements of DIP uptake, DOP release and AP activity across a spatial gradient covering Atlantic and Arctic waters.

Study area and sampling
Experiments were conducted at six stations in the Barents Sea aboard the RRS James Clark Ross (JR17007), between the 11th and 29th July 2018, as part of the UK Changing Arctic Ocean Seafloor programme (ChAOS; Fig. 1). Depth profiles of temperature and salinity were determined using a SEA-BIRD CTD system mounted on a rosette sampler equipped with 24 Niskin bottles (20 L) to collect seawater at defined depths. At each station, six light depths were sampled corresponding to set photosynthetically active radiation (PAR) percentages (100%, 50%, 25%, 15%, 3%, 1%), calculated from the surface PAR (log). Samples were collected in 2 L polycarbonate bottles before immediate subsampling and processing as described below.

Inorganic nutrients and P*
Samples for nutrient analysis were collected and immediately stored at − 20 C, before returning to Plymouth Marine Laboratory for analysis. Dissolved inorganic nitrogen (DIN) is defined as the combined concentration of inorganic nitrate and nitrite, while DIP is assumed to be the concentration of soluble reactive phosphorus (Thomson-Bulldis and Karl 1998). Ambient DIN, DIP, and silicate concentrations were determined using a SEAL analytical AutoAnalyzer II, segmented flow colorimetric autoanalyzer, in accordance with the protocols outlined in Woodward and Rees (2001). To assess deviations of DIN : DIP ratios from the Redfield ratio, P* was calculated (Deutsch and Weber 2012): Dissolved organic phosphorus Ambient DOP concentrations were estimated as the difference between total dissolved phosphorus (TDP) and DIP. To measure TDP the persulfate oxidation method was used to liberate organically bound P, before measuring DIP (Menzel and Corwin 1965;Ridal and Moore 1990). After collection at sea samples were stored in polycarbonate bottles at − 20 C. Before analysis, samples were thawed and filtered (0.22 μm pore size, Millipore) into glass vials with PTFE lined caps for the oxidation step (Yoshimura 2013). Briefly, potassium persulfate solution was added to 20 mL acidified sample, 0.8% final concentration, and autoclaved for 2 h at 123 C. Samples were placed in a water bath at 90 C for 2 h to remove excess chlorine before measuring DIP. Triplicate measurements on select samples gave an approximate precision of AE 1%.

Dissolved inorganic phosphorus uptake and release of dissolved organic phosphorus
Daily rates of DIP uptake were determined using 33 P tracer experiments (Canellas et al. 2000). Water samples were aliquoted into 60 mL acid washed polycarbonate bottles (3 light, 1 dark). Each bottle was spiked with 5-12 kBq 33 P-orthophosphoric acid (Perkin Elmer, specific activity 5-7 kBq pmol −1 ) equating to an addition of < 3 pmol DIP. Bottles were then placed in on-deck incubators with light density filters corresponding to the PAR depths sampled. Incubations were carried out for 24 h, and temperature was maintained by a continuous supply of surface seawater. The contents of the bottles were filtered onto 47 mm diameter 0.2 μm polycarbonate filters presoaked and boiled in a lithium chloride phosphate buffer (1 mM PO 4 ); which prevents unincorporated 33 PO 4 adsorbing to filter surfaces. Filters were counted using a liquid scintillation counter (TriCarb 3100TR) using ProSafe FC+ liquid scintillation cocktail (Meridian Biotechnologies Ltd).
The release of DOP from cells was determined following Poulton et al. (2019). Before filtration at the end of the incubation, 10 mL aliquots were removed from incubation bottles at four light depths per station. The aliquot was filtered using a 25 mm diameter 0.2 μm polycarbonate filter (Millipore ™) and the filtrate collected in 20 mL glass vials. The filtrate was then , vortexed, and then centrifuged at 3500 rpm for 1 h. To measure the 33 P-DOP, 1 mL of the supernatant was placed in a 20 mL glass scintillation vial with 8 mL ProSafe HC+ liquid scintillation cocktail (Meridian Biotechnologies Ltd) prior to counting. The estimate of 33 P-DOP should be considered a minimum due to a possible loss by coprecipitation with NaOH.
To estimate the percentage of assimilated DIP released as DOP, the gross rate of DIP uptake was calculated (i.e., the rate of DIP uptake plus the rate of DOP production). The percentage release of DOP was therefore the fraction of gross DIP uptake measured as DOP release multiplied by 100. Gross and net turnover times were calculated accordingly.

AP activity
AP activity was measured fluorometrically with the substrate 4-MUF-phosphate (Hoppe 1983). The fluorogenic substrate was dissolved to a final concentration of 10 mM in 50% dimethyl sulfoxide. For each depth sampled, triplicate 3 mL aliquots were dispensed into 5 mL cuvettes and amended with a final concentration of 150 μM MUF-phosphate. Incubations were carried out in the dark at in situ temperatures. Fluorescence measurements were taken immediately after substrate addition and at 3-4 time points up to 24 h using a Turner biosystems TBS solid-state fluorometer using the "UV" channel (excitation 365 nm; emission 440-470 nm). Calibrations were carried out using six methylumbelliferone standards with a concentration range 2.5-100 nM L −1 .

Primary production
Rates of net primary production (PP) were estimated from the six light depths sampled according to Joint and Pomroy (1993). Triplicate samples were dispensed into 60 mL acid washed polycarbonate bottles, inoculated with 370 kBq (10 μCi) NaH 14 CO 3 and incubated on-deck at simulated PAR depths for 24 h, simultaneous with DIP uptake experiments. Incubations were terminated by sequential filtration through 20, 2, and 0.2 μm polycarbonate filters (Millipore™). Filters were fumed with HCl (37%) for 2 h and desiccated for 12 h prior to counting using ProSafe FC+ liquid scintillation cocktail (Meridian Biotechnologies Ltd).

Phytoplankton community composition
Samples for phytoplankton enumeration were taken at four depths per station representing the surface, sub surface, chlorophyll maximum and 1% PAR depth. Duplicate samples were fixed, one with acid Lugol's solution (2% final concentration) and the other with neutral formalin (4% final concentration). Analysis was carried out using inverted settlement microscopy (Widdicombe et al. 2010), with identification to species level where possible using Lugol's fixed samples for noncalcifying species and formalin fixed samples for calcifying coccolithophores.

Satellite chlorophyll and sea ice extent
To contextualize our sampling period in terms of the summer phytoplankton growing season we used satellite chlorophyll data processed by the NERC Earth Observation Data Acquisition and Analysis Service (NEODAAS) from NASA OBPG L2 data using SeaDAS version 7.3. Chlorophyll a (Chl a) measurements were extracted from rolling 28-day 1.1 km resolution mean surface composites for each cruise station from March to October 2018, using the NASA Suomi-VIIRS Standard Chlorophyll-A Ocean Color Index algorithm. The OCI algorithm merges the color index product outlined in Hu et al. (2012)  Sea ice area fractions were extracted from the Operational Sea Surface Temperature and Sea Ice Analysis dataset (Donlon et al. 2012), provided by the "Group for High Resolution Sea Surface Temperature," the Met Office and the Copernicus Marine Environment Monitoring Service and all processed by NEODAAS. Sea ice area fractions were extracted for each cruise station from daily 5 km resolution composite data. Days that were ice free prior to sampling were calculated from the first day the station area had < 10% ice cover to the day of sampling.

Statistical analysis
One-way ANOVA was used to determine variance of environmental data and rate measurements between Atlantic and Arctic influenced stations, performed using Matlab R2018a. Multivariate patterns of phytoplankton community assemblages and their relationships to environmental variables were analyzed using the statistical package PRIMER (Clarke and Green 1988;Clarke 1993;Clarke and Ainsworth 1993). Phytoplankton abundances were square root transformed, from which a Bray-Curtis similarity matrix was generated and used to create a two-dimensional nonmetric multidimensional scaling (NMDS) ordination, to visualize multivariate patterns of phytoplankton community composition. ANOSIM (Analysis of similarity) was employed to test the significance of the NMDS. It was then analyzed for statistical correlation with several environmental variables using BEST; coupling Spearman rank correlation (BIO-ENV) with permutation analysis.

Environmental parameters
The cruise transect represented a clear gradient in the Barents Sea from Atlantic water in the south to Arctic water with overlying melt water at the northern three stations (Table 1), with the polar front situated between Sta. B14 and B15. Hence, we will refer to the Atlantic region as Sta. B3, B13, and B14 and the Arctic region being Sta. B15-B17. All stations were free of sea ice at the time of sampling; B14 was the only Atlantic influenced station to experience winter sea ice cover, but remained ice free for 82 d prior to sampling, compared to 43-46 d for the Arctic stations (Table 1).
For the Atlantic stations; DIN, DIP and silicate concentrations in the upper water column (0-50 m) followed similar latitudinal patterns with a northward gradient of decreasing surface nutrients (Fig. 2). The highest ambient concentrations were observed at Sta. B3, with DIN ranging from 4.70-7.31 μM, DIP: 0.31-0.44 μM, and silicate: 2.75-3.17 μM ( Fig Table 2). Compared to much lower concentrations at the northern most Atlantic Sta. B14 with DIN, DIP and silicate ranging from 0.02-0.04, 0.09-0.09, and 0.53-0.67 μM, respectively ( Fig. 2A-C; Table 2). At the Arctic stations sampled nutrient concentrations varied considerably. The lowest values measured were at Sta. B16, DIN ranged between 0.04-2.86 μM and DIP between 0.08-0.32 μM, compared to much higher concentrations at Sta. B17, 0.04-6.60 and 0.10-0.88 μM, respectively (Table 2; Fig. 2A,B). Depth profiles reveal a much deeper nutricline depth for DIN and DIP at B16 opposed to B15 and B17, being more pronounced for DIN compared to DIP ( Fig. 2A,  B). Furthermore, at Sta. B17, high ambient DIP (0.88 μM) measured at 39 m depth corresponds with high DIN (6.6 μM) and a hotspot in DOP concentration (0.50 μM) ( Fig. 2A,B,D). These high nutrient concentrations at 39 m coincided with a peak in primary productivity (Fig. 3F). With the exception of Sta. B3, silicate was depleted in the surface waters, ranging between 0.36 and 2.55 μM, including a notable hotspot in overlying melt water at B16 (Fig. 2C; Table 2). DOP generally remained higher in the surface waters and decreased with depth. Above the nutricline DOP accounted for 37-78% of the TDP pool, with no significant difference between the Atlantic and Arctic stations sampled (ANOVA, F = 1.75, p < 0.19).

Primary production, phytoplankton, and chlorophyll
Satellite surface chlorophyll estimates suggest that our sampling periods represented post bloom scenarios at all stations (Fig. 4). The spring bloom occurred in Atlantic waters during May, being more pronounced at Sta. B13 and B14 compared to B3. Bloom timing was later at the Arctic stations, occurring in June, and B17 showed a second chlorophyll peak during September. Late summer/early autumn blooms also occurred at the Atlantic stations albeit at varying timings (Fig. 4). However, satellite chlorophyll estimates typically capture the first optical depth only, and are therefore not able to account for sub surface chlorophyll peaks. The experimentally derived depth profiles show distinct subsurface peaks in PP for the three Arctic stations (Fig. 3C,D,F) compared to the Atlantic stations where PP was high at the surface and decreased with depth ( Fig. 3A-C). Thus when using satellite estimates to contextualize in situ data, in terms of the timing of phytoplankton blooms, there will be greater uncertainty for the Arctic stations where more discrete nonsurface blooms will be missed. However, overall the temporal variations in surface chlorophyll demonstrate that our ship derived measurements represent summer phytoplankton growth occurring post spring bloom and before any late summer/autumn blooms.
Depth integrated primary production (Table 1; Fig. 3) estimates show the southern station (B3) accounted for the highest PP in Atlantic waters (655 mg C m 2 d −1 ) and the northern most station (B17) having the highest rates in Arctic waters (605 mg C m 2 d −1 ). Depth profiles reveal a defined subsurface peak in PP at the 15% PAR depth for the three Arctic stations (Fig. 3D-F), which occurs below the overlying melt water. The dominant phytoplankton species at these stations was P. pouchetii (Table 1) belonging to the nano (2-20 μm) size class, which accounted for 55-77% of PP at the Arctic stations. At Sta. B15 and B16 the subsurface peak in PP was shallower ($ 20 m) and corresponded to lower nutrient availability compared to a deeper peak in PP ($ 40 m) observed at Sta. B17, with much higher ambient DIN, DIP, and DOP concentrations (Figs. 2A-C, 3D-F). Furthermore, the magnitude of subsurface PP at B17 (27.8 AE 10.9 mg C m −3 d −1 ) was considerably higher than that measured at B15 and B16, 4.33 AE 1.03 and 1.54 AE 0.50 mg C m −3 d −1 , respectively. The stations in Atlantic waters, with no overlying melt water, were typically characterized by the highest PP in the surface, decreasing with depth. The nanosize class accounted for 76% of PP at B3, and was comprised mainly of phytoflagellates and E. huxleyi (Table 1). To a lesser extent B13 also had a large contribution of the nanosize class to total PP (57%), which was, similar to Sta. B3, dominated by E. huxleyi. Sta. B13 also had a large contribution of the micro size class (> 20 μm), accounting for 35% of PP and being comprised mainly of dinoflagellates e.g., Prorocentrum f.minimum (Table 1). Similarly, micro sized phytoplankton accounted for 33% of PP at B14, again being comprised mainly of dinoflagellates e.g., Peridiniales and Gymnodiniales. Overall the phytoplankton species composition showed a trend from autotrophic nanosized plankton dominating the high rates of PP at B3, with increasingly more micro sized (> 20 μm) dinoflagellates capable of mixotrophy toward B14 where nutrients were increasingly depleted. E. huxleyi was confined to the warmer waters (≥ 6 C) at Sta. B3 and B13. Whereas, P. pouchetii was present in Arctic waters only, being the dominant species at Sta. B15-B17.
Patterns of similarity between phytoplankton species composition and abundance can be visualized with NMDS ordination (Fig. 5). Assemblages within Arctic and melt water show a defined cluster separate from the Atlantic communities, which form individual clusters according to stations. The ANOSIM model was employed to test the degree of significance of the NMDS giving a dissimilarity of 61.1% (p < 0.008) between the Arctic and Atlantic communities. Results from BEST analysis show the environmental variables that significantly correlated (p < 0.02) with the multivariate pattern of community composition across all stations (Table 3). Here silicate and temperature were in all the combinations of environmental variables correlating with patterns of species composition. We repeated the same analysis separately for the Atlantic and Arctic stations. For the three Arctic stations there was no significant correlation (p < 0.69), this was expected due to small differences in the species composition between Arctic stations (Fig. 5). Whereas the Atlantic stations had a significant correlation (p < 0.001) with the environmental variables (Table 4), and again the temperature was part of every combination and showed strong latitudinal variation at these stations. Also, DIP concentration and its relative availability was part of the best correlated variables (Table 4), suggesting a more important role of DIP in forcing patterns of phytoplankton distribution in Atlantic waters alone.
Dissolved inorganic phosphorus uptake, dissolved organic phosphorus release, and alkaline phosphatase activity DIP net uptake fluxes (determined using 33 P) were higher in the Atlantic section (8.50-158 nM d −1 ) compared to the Arctic (5.40-59.6 nM d −1 ; Table 2). An ANOVA test (F = 5.36, p = 0.031) revealed these differences between the Atlantic and Arctic water masses were statistically significant. The highest DIP uptake fluxes were at Sta. B3, which ranged between 93.2 and 157 nM d −1 . High rates of DIP uptake and PP at this station, were reflected in the low P* values (0.01 AE 0.04), suggesting ambient DIP was close to Redfield ratio. In contrast, the high P. pouchetii driven PP at B17 (Fig. 3F) corresponded to a high P* value of 0.47 at the PP peak (Fig. 6A), due to excess DIP compared to DIN.
Significant differences between the Atlantic and Arctic waters were also found for extracellular release of DOP ( Fig. 6C; ANOVA, F = 5.36, p = 0.03). The fraction of assimilated DIP released as DOP was much greater in Arctic waters than Atlantic ( Fig. 6B; Table 2). The lowest values were at Sta. B3, 3.2-4.1%, compared to 4.7-26% at the other Atlantic stations (B13 and B14) and 17-42% for the Arctic stations ( Table 2).
The gross turnover time of DIP showed large variability, ranging from 1.9 to 13 d in surface waters above the nutricline depth (Figs. 6C, 7A), where high DIP uptake fluxes (8.5-158) in the Atlantic waters corresponded to faster gross turnover times (1.9-8.8) ( Table 2). The most rapid average gross turnover time was at B3, 2.7 AE 0.5 d (Fig. 7A), despite having the greatest ambient DIP concentration of 0.36 AE 0.05 μM (Table 2; Fig. 2B). The ANOVA test revealed both net turnover (F = 5.17, p = 0.03) and gross turnover (F = 11.29, p = 0.003) were significantly different between Arctic and Atlantic waters. Differences between gross and net DIP turnover times can be attributed to the release of DOP (Fig. 7B). The relationship between the release of DOP and the gross turnover time (Fig. 8) shows a linear correlation coefficient of 0.77 (R 2 = 0.593). Although greater DOP release from cells results in

Downes et al.
Phosphorus in the Barents Sea faster gross turnover times compared to net turnover times (Fig. 7), DOP release and gross turnover time show a positive correlation, i.e., a high percentage of DOP release corresponds to slower turnover of the DIP pool. AP activity ranged from 0.62 to 4.21 nM h −1 (Table 2). There was no significant difference between the Atlantic and Arctic stations (ANOVA, F = 0.98, p = 0.331). The highest rates were in the surface at Atlantic Sta. B3 (4.21 AE 0.08 nM hr −1 ; Table 2) corresponding with the high PP (69.7 mg C m 3 d −1 ; Fig. 3A) and high DIP uptake (96.92 nM d −1 ), low percentage DOP release (4.0%) and a low P* (0.01) (Fig. 6). The lowest AP activity at Arctic Sta. B16 (0.86 AE 0.18 nM hr −1 ; Table 2) was matched with very low rates of PP (41.4 mg C m 2 d −1 ) along with low ambient DOP concentrations (0.16 AE 0.04 μM).

Discussion
There are distinct differences in nutrient stoichiometry between Arctic and Atlantic waters in the Barents Sea. Overlying melt water in the Arctic section results in enhanced vertical stratification and causes a subsurface peak in PP, with an apparent excess of DIP, denoted by relatively high P* values. At the Arctic stations silicate remained below the concentration limit (2 μM) required to support diatom dominated primary production (Egge and Aksnes 1992). Silicate limitation on PP has been reported for subarctic provinces, terminating diatom blooms (Krause et al. 2019). Our BEST analysis indicates that silicate and temperature are the dominant factors driving patterns in species distribution across both Atlantic and Arctic waters combined. It is likely that the early summer blooms indicated from satellite imagery results in silicate depletion in the surface layer, limiting diatom driven PP, resulting in phytoplankton species succession whereby P. pouchetii are the abundant species during the low nutrient summer period.
Nutrient concentrations in Atlantic waters varied between stations, with the highest ambient concentrations at the southern Sta. B3. This station is located toward the Barents Sea opening receiving Atlantic waters inflowing onto the shelf. As surface nutrient concentrations decrease moving from B3 to B14 so does the magnitude of PP and the species composition, shifting from autotrophic nanoplankton, e.g., phytoflagellates and E. huxleyi, to more mixotrophic communities dominated by dinoflagellates. Although DIP concentrations at B3 are relatively high (0.36 AE 0.05 μM), concentrations were close to, or below the Redfield ratio. In addition, the BEST analysis on Atlantic stations only, reveal a more significant role of DIP, its relative availability (P*), as well as the temperature in shaping the community structure.
The magnitude of PP varied greatly between the Arctic stations, yet the phytoplankton communities showed significant similarities, all being dominated by P. pouchetii. A subsurface peak in PP occurred at the 15% PAR depth for the three stations. At B17 this depth was situated at $ 40 m, coinciding with much greater ambient nutrient availability with concomitant high PP; a factor of 5 and 15 greater than B15 and B16, respectively. At B15 and B16 the subsurface PP peak occurred at $ 20 m, here nutrients, especially DIN were depleted, perhaps indicating post bloom more grazed conditions. DIN limitation is further apparent in the high P* values for these stations, showing a deficit of DIN relative to DIP. Nitrogen limited PP and subsurface chlorophyll maximums have  previously been observed in Arctic waters surrounding the polar front region of the Barents Sea post spring bloom (Erga et al. 2014). When DIN is depleted, PP is predominately regenerated production, whereby phytoplankton growth is supported by the production of regenerated forms of N, such as ammonium and urea, by zooplankton grazing and degradation of organic matter by heterotrophic bacteria. Ammonium and urea uptake have shown to account for 75-93% of total N uptake during the Barents Sea summer (Kristiansen et al. 1994). Unfortunately no ammonium and urea data was available for this study. However, it is likely that the rate of N regeneration is a key factor explaining the differences in PP between Arctic stations e.g., B15 and B16 where DIN was depleted at the subsurface PP peak for both sites but PP was 2.8 times greater at B15 compared to B16. At Sta. B17, the more transparent waters (the 1% PAR depth was $ 70 m compared to $ 50 m for B15 and B16) allow for phytoplankton to occur deeper in the nutricline, with greater ambient DIN concentrations, supporting much higher rates of PP. Differences in DIP availability and phytoplankton P requirements, from Atlantic to Arctic waters are apparent in the turnover time of the DIP pool. Rather than comparing pool sizes and uptake rates alone, turnover time considers both, providing additional insight into the biogeochemical dynamics. Rapid biological uptake results in short turnover times (hours or days), while longer turnover times (weeks or months) implies lower cellular requirements or less bioavailability (Benitez-Nelson 2000). In general, the DIP turnover times we observed are consistent with that of open ocean systems ranging between days and weeks (Eppley et al. 1973;Harrison et al. 1977;Perry and Eppley 1981;Harrison 1983;Canellas et al. 2000;Sohm and Capone 2010). However, the  most rapid net turnover times observed in the southern Barents Sea at Sta. B3 (2.8 AE 0.5 d) could suggest that the phytoplankton communities are close to, or moving toward P deficiency.
Along with the shortest turnover time, we observed high P retention at Sta. B3 (expressed as a low fraction of assimilated DIP released as DOP). Here the average release of DOP is remarkably low (3.6 AE 0.4%), compared to all other stations (21 AE 9.7%) despite relatively high ambient DIP at Sta. B3 (0.36 AE 0.05 μM). This low extracellular release of DOP is comparable to summer time measurements (2-11%) in the Celtic shelf sea, whereby the planktonic community exhibited highly efficient P uptake and retention under low DIP conditions with ambient DIP concentrations ranging from 0.05 to 0.09 μM (Poulton et al. 2019). Furthermore, E. huxleyi was among the dominant phytoplankton species at B3, and it is known to have a large internal DIP pool, allowing them to proliferate in low DIP conditions (Riegman et al. 2000).
The portion of DIP assimilated and then released as DOP was significantly greater at the Arctic stations where the phytoplankton communities were comprised mainly of P. pouchetii, compared to the Atlantic stations with no P. pouchetii present. Previous field studies have observed high concentrations of dissolved organic carbon, dissolved carbohydrates and dissolved organic nitrogen during Phaeocystis blooms (Eberlein et al. 1985;Billen and Fontigny 1987). The life cycle of P. pouchetii has two main phases; solitary cells and colonial cells which are embedded in a matrix formed of mucoploysaccharides and heteropolysaccharies (Alderkamp et al. 2007). These carbon rich compounds can account for up to 90% of the total Phaeocystis biomass (Rousseau et al. 1990) resulting in high C/P and C/N ratios of organic matter produced by Phaeocystis. Hence, the higher rates of DOP release we observed in association with Phaeocystis communities are unlikely to be exclusively caused by these cells excreting colony mucus.
Alternatively, DOP release can result from herbivorous grazing and viral lysis where the labile cell contents are released as dissolved organic matter comprised of P rich compounds including phospholipids, nucleic acids and metabolic intermediates such as adenosine triphosphate. Unfortunately, there were no estimates of grazing rates during the current study. Despite a considerable amount of research there is still much uncertainty on the extent to which grazing pressure causes Phaeocystis mortality (Verity et al. 2007). The occurrence of P. pouchetii has previously been reported alongside low or insignificant herbivory rates (Calbet et al. 2011;Gifford et al. 1995;Saiz et al. 2013;Stoecker et al. 2015). Grazing dilution experiments conducted in Disko Bay, West Greenland, revealed a negative correlation between Phaeocystis spp. abundance and bulk grazing rate (Menden-Deuer et al. 2018). The authors hypothesized that this was likely a result of secondary metabolites, specifically polyunsaturated aldehydes, released by Phaeocystis which act as cytotoxic grazing deterrents Menden-Deuer et al. 2018). In addition, the size of the colonies and the colony mucus has been hypothesized to provide mechanical protection and disrupt feeding by zooplankton (Kamermans 1994;Weisse et al. 1994). Despite this, large copepods including Calanus spp. are able to feed on P. pouchetii colonies ). In the Barents Sea, Calanus spp. account for $ 80% of the mesozooplankton biomass (Aarflot et al. 2018). Hence, the higher rates of DOP release we observed among Phaeocystis communities, could to be the consequence of grazing pressure from larger copepod species such as Calanus spp. which, through sloppy feeding and leakage from fecal pellets release phytoplankton derived organic phosphorus to the dissolved pool.
Viral lysis is also a major cause of mortality during Phaeocystis blooms (Brussaard et al. 1995(Brussaard et al. , 1996(Brussaard et al. , 2005. Culture experiments have shown that within 3 d, the entire P. pouchetii carbon biomass gets released as DOC due to viral lysis, compared to $ 20% in control cultures with no virus present (Bratbak et al. 1998a). The host-virus system of P. pouchetii has a lytic cycle of 12-18 h (Jacobsen et al. 1996) and during most of the lytic cycle, primary production can continue (Bratbak et al. 1998b). Indicating that the 24 h incubations used in the present study could have undergone at least one, if not two lytic cycles, while infected cells were still metabolically active; assimilating DIP before cell lysis and the release of DOP. The correlation we observed between gross turnover time and the release of DOP could therefore be attributed to efficient P uptake and retention by the phytoplankton communities at the Atlantic stations (especially B3) and lower P requirements combined with a higher release of DOP via a combination of exudation, grazing and viral lysis at the Arctic stations (where P. pouchetii dominated phytoplankton communities).
AP activity remained relatively low, and within ranges previously reported for Atlantic and Arctic waters in the Fram Strait; 0-47.8 nM h −1 (Sala et al. 2010).To our knowledge there have been no previously published measurements of APA activity for the Barents Sea. The highest average value (3.38 AE 0.80 nM h −1 ) recorded at the surface of Sta. B3 coincided with the high primary production, fast DIP turnover and low release of DOP. Combined, this could suggest DOP is being utilized as an alternative P source by members of the phytoplankton community. Here, E. huxleyi was present in relatively high abundance and is known to synthesize AP, under DIP limited conditions (Riegman et al. 2000). However, we found no significant difference in AP activity between Atlantic and Arctic stations. In addition, we are unable to attribute AP activity to either the phytoplankton or bacterioplankton communities based on the bulk activity method used. Therefore, it is likely that the DIP concentrations we measured were above the threshold required to stimulate high AP activity in response to P stress.
The prevalence of E. huxleyi in the southern Barents Sea is now a common phenomenon during the summer months (Oziel et al. 2017). We observed the occurrence of E. huxleyi to be restricted to temperatures above $ 6 C. This is consistent with satellite observations, where a Gaussian distribution of coccolithophore abundance centered around 8 C has been reported for this region (Signorini and McClain 2009). Our results indicate that the E. huxleyi dominated phytoplankton communities occurring in the warmer Atlantic waters likely have greater P requirements than that of Arctic phytoplankton communities. This results in a more rapid turnover and greater retention of assimilated DIP compared to Arctic phytoplankton assemblages. If the "Atlantification" of the Barents Sea shelf continues to intensify, the region may experience much faster and tighter cycling of phosphorus associated with the Atlantic phytoplankton communities, and ultimately become more P limited.