The Arctic Ocean is warming (Huang et al., 2017), experiencing sea ice loss (Notz and Stroeve, 2016) and freshening (Coupel et al., 2015) as a direct response to climate change. It is an enclosed basin filled with waters from the Atlantic and Pacific oceans, which provide varying concentrations of nutrients (Torres-Valdes et al., 2013). In turn, these nutrient supply pathways influence the distribution and extent of primary production throughout the Arctic Ocean (Lewis et al., 2020). Approximately 50 % of the Arctic Ocean is made up of productive shelves that support large fisheries and diverse habitats (Dalpadado et al., 2014; Friedland and Todd, 2012). As the Arctic continues to warm and more sea ice is lost, phytoplankton growth will become less limited by light availability. Instead, nutrient availability, principally nitrate (Codispoti et al., 2013), may become the primary control on phytoplankton growth (Arrigo and van Dijken, 2015; Lewis et al., 2020). Further insight is required into how nitrate is supplied to Arctic shelves, the nutrient cycling processes that occur in situ and their sensitivity to climate change. A further understanding of these processes will help to inform on future changes to Arctic primary production and food web dynamics (de la Vega et al., 2020).

Atlantic Water (AW) is supplied to the Arctic via the Fram Strait and the Barents Sea Opening (BSO) and fills most of the deep basins of the Arctic. It supplies nutrients to the Eurasian shelves with nitrate and phosphate concentrations close to Redfield (15-16N:1P) and low concentrations of silicate, which can limit the extent of diatom growth (Hatun et al., 2017). AW is a mixture of nutrient-rich North Atlantic subpolar and nutrient-poor subtropical origin water advected into the Norwegian Sea. Over the last 2 decades, there has been a 7 % and 20 % decrease in nitrate and silicate concentrations, respectively, in the Barents Sea (Rey, 2012). This has been driven by shallower winter mixing in the subpolar gyre coupled with weakening and westward retraction of the gyre which has increased the proportion of subtropical origin water entering the Norwegian Sea (Rey, 2012; Hatun et al., 2017).

As warm and saline AW inflow water is transported across the Barents Sea, it is modified by atmospheric cooling and is mixed with cold, fresh Arctic origin water (ArW) and the Norwegian Coastal Current (NCC) (Fig. 1b). ArW found across the northern Barents Sea comprises fresh Arctic river runoff, sea ice melt and precipitation and contains the remnants of the winter mixed layer (Rudels et al., 1996). Less dense ArW isolates the sea surface and ice cover from warm AW below (Lind et al., 2016) and during the summer is capped by a well-mixed surface layer of fresh melt water (polar surface water) (Fig. 2b). Sea ice import from the Nansen Basin and Kara Sea is the most important source of freshwater in the northern Barents Sea (Lind et al., 2016; Ellingsen et al., 2009). The Barents Sea is a key mixing region of AW and ArWs (Porter et al., 2020). The transition between these water masses is marked by the polar front (PF) which can be identified from the sea surface temperature gradient (Barton et al., 2018; Oziel et al., 2016) (Fig. 1b).

Intense cooling of AW across the Barents Sea, reinforced by brine rejection due to ice formation, creates dense Barents Sea Water (BSW) that cascades into the deeper troughs of the central and eastern Barents Sea (Arthun et al., 2011; Oziel et al., 2016). BSW eventually leaves the shelf mainly through the St. Anna Trough (Smedsrud et al., 2013) where it is entrained into Arctic Intermediate Water and spreads further into the Arctic Basin (Schauer et al., 1997).

The Barents Sea is experiencing a rapid decline in winter and summer sea ice cover (Onarheim and Arthun, 2017; Arthun et al., 2012), full-depth warming driven by both increased ocean heat transport from the Atlantic and amplified atmospheric warming over the Arctic (Arthun et al., 2012; Onarheim et al., 2015; Serreze et al., 2009), and increases in salinity (Lind et al., 2018; Barton et al., 2018). The area occupied by AW is increasing, and the southern expression of the polar front is moving north (Oziel et al., 2016, 2020). In the northern Barents Sea, a reduction in sea ice import and therefore a loss of freshwater is weakening stratification and enhancing vertical mixing (Lind et al., 2018). The northern Barents Sea is therefore transitioning from a cold, salinity-stratified shelf into a warmer, less-stratified Atlantic-dominated climate regime (Lind et al., 2018), a process described as “Atlantification”. These changes may increase nutrient availability to phytoplankton over the growing season (Henley et al., 2020; Randelhoff et al., 2018), which is increasingly a control on Arctic net primary production (NPP) (Lewis et al., 2020).

On the other side of the Arctic, the Pacific Ocean supplies high concentrations of nutrients onto the Chukchi and East Siberian shelves, fuelling productivity and nutrient uptake (Granger et al., 2011). Increases in volume transport through the Bering Strait in recent years (Woodgate, 2018) have increased Pacific nutrient supply to the Arctic Basin. These waters are relatively depleted of nitrate (in comparison to phosphate) and, combined with sedimentary denitrification on the shallow shelves (Fripiat et al., 2018; Granger et al., 2018), promote nitrogen limitation in the western Arctic Ocean (Mills et al., 2018).

Although many studies have found the western Arctic Ocean to be strongly N limited (Mills et al., 2018; Granger et al., 2018; Brown et al., 2015), we know less about the extent of N limitation and occurrence of sedimentary denitrification in the eastern Arctic Ocean. Nitrate isotope measurements can give integrated estimates of nitrogen cycling processes, yet there is currently no data on the North Atlantic inputs which provide nutrients to the Arctic Basin via the Barents Sea. The 15/14N and 18/16O in nitrate (δ15N-NO3 and δ18O-NO3, respectively) provide complementary information about nitrate uptake by phytoplankton and regeneration processes (Sigman et al., 2009b) and can be used to determine the relevant N cycling processes.

Nitrate consumption through algal uptake fractionates both N and O in a 1:1 ratio (Granger et al., 2004) with an isotope effect close to ∼ 5 ‰ (Sigman et al., 2009b). Nitrogen loss through denitrification in the water column leads to enrichment in N and O isotopes in the residual nitrate pool with a fractionation of 25 ‰–30 ‰ (Sigman et al., 2009a). In sediments, denitrification does not usually impart a signature on nitrate isotopes as the reaction goes to completion (Sigman et al., 2003; Lehmann et al., 2007). However, multiple studies from the western Arctic and Bering Sea show that benthic denitrification can impart a signature on the overlying water column (ε=0 ‰–5 ‰) when high δ15N-NH4 is released, a process termed coupled partial nitrification–denitrification (Brown et al., 2015).

Over most of the ocean, fixed nitrogen is efficiently recycled in surface waters, and the regeneration of nitrate in the water column retains a δ15N signature from the N source (Sigman et al., 2000). The δ15N signature imparted on nitrate can therefore be used to identify nutrient sources from a partial utilization of nutrients (Rafter et al., 2012), different water masses (Sigman et al., 2000; Tuerena et al., 2015), new N inputs (Knapp et al., 2008; Marconi et al., 2017), atmospheric inputs (Altieri et al., 2016) and rivers (Thibodeau et al., 2017).

In contrast to δ15N-NO3 in which N atoms are internally recycled during nitrification, oxygen atoms are sourced from ambient O2 and seawater, in general providing a nitrification signature of δ18O-H2O plus 1.1 ‰ (Buchwald et al., 2012; Sigman et al., 2009b). The contrasting sources of N and O atoms and thus their distinct isotopic signatures allow the relative importance of preformed and regenerated nitrate to be investigated (Rafter et al., 2013). The δ18O-NO3 value has been used to quantify the extent of regeneration on the Bering Sea shelf (Granger et al., 2011) and as evidence for the significance of nitrate regeneration in sustaining nutrient stocks on Arctic shelves (Fripiat et al., 2018; Granger et al., 2018). The tracer Δ(15–18) (δ15N-NO3 minus δ18O-NO3) captures the differences between these two isotopes, highlighting nitrate sources from different oceanic environments (Rafter et al., 2013).

When nitrate is not fully consumed in surface waters, δ15N-NO3 and δ18O-NO3 can indicate the extent of seasonal nitrate uptake by phytoplankton (DiFiore et al., 2009). The δ15N of surface water nitrate increases as nitrate is progressively utilized by phytoplankton through the preferential consumption of 14N (Sigman et al., 1999). Together with other N cycling processes, this can be described by Rayleigh fractionation systematics (Mariotti et al., 1981). Nitrate utilization by phytoplankton in an environment where there is no resupply of nutrients, i.e. a stratified upper ocean in summer, follows Rayleigh fractionation systematics for a closed system, with δ15N-NO3 falling on a fractionation trend for its isotopic effect (ε) (Granger et al., 2004). In combination with dissolved nutrients, the δ15N of particulate nitrogen (δ15N-PN) can track the extent of biological utilization, contrasting nutrient sources and the significance of new vs regenerated nutrients (Altabet and Francois, 1994).

In this study, we report the first stable isotope measurements of dissolved and particulate N in the Barents Sea and use them to understand the relative sources of nutrients fuelling contemporary Barents Sea productivity. We use stable isotope tracers to investigate how N cycling processes vary across the Barents Sea in contrast to other Arctic shelves and reflect upon the susceptibility of the ecosystem to climate change.

Samples were collected in the Barents Sea as part of the ARISE project (NERC Changing Arctic Ocean programme). Shipboard measurements were taken from the RRS James Clark Ross during July–August 2017 (JR16006). A 2200 km transect was completed, comprising 59 full-depth conductivity, temperature and depth (CTD) casts, starting from the northern tip of Norway and ending at the shelf edge north-east of Svalbard (Fig. 1a). The transect crossed the Barents Sea Opening (BSO) between Norway and the southern tip of Svalbard. Then, from Hopen Trench it continued north towards Kong Karls Land (KKL; along 30∘ E) and to the shelf edge and Nansen Basin.

Standard CTD measurements and water sampling were performed using a stainless steel rosette equipped with a full sensor array and 24 20 L OTE bottles. Conductivity, temperature and pressure were measured using a CTD system (Seabird 911+). Derived salinity was calibrated on-board with discrete samples using an Autosal 8400B salinometer (Guildline) (Dumont et al., 2019), and an SBE43 oxygen sensor was calibrated against oxygen samples analysed using the Winkler method. A Biospherical QCP Cosine PAR sensor measured downwelling photosynthetically available radiation (PAR; 400–700 nm). We define the base of the euphotic zone to be the depth where PAR decreased to 1 % of its surface value. The mean depth of the euphotic zone was 34.3 ± 11.9 m.

Dissolved inorganic nutrient concentrations were determined using a Bran and Luebbe QuAAtro 5-channel auto analyser (SEAL Analytical) and aACE operating platform (V 6.1) following standard colourimetric methods with a CRM precision of 0.3 %, 0.8 % and 1.9 % for nitrate+nitrite, phosphate and nitrite, respectively (Brand et al., 2020). Nitrate isotope samples were collected and filtered inline from the CTD using an AcroPak and were frozen at -20 ∘C until analysis. Of the 59 CTD casts in 2017, 23 were sampled for nitrate isotopes covering the full water column (Tuerena and Ganeshram, 2020). Particulate nitrogen (PN) and δ15N-PN samples were collected from 21 profiles in the upper 200 m at stations where nitrate isotope samples were also collected (Norman et al., 2020). Samples were gently vacuum filtered through combusted Whatman GF/F filters (450 ∘C, 4 h, 47 or 25 mm, nominal pore size 0.7 µm) until sufficient biomass was collected on the filter (8 to 12 L for the 47 mm diameter filters and 2 to 5 L for the 25 mm diameter filters). The filters were dried at 60 ∘C to remove all moisture and were stored folded and wrapped in combusted aluminium foil until their return to the home laboratory where they were placed in a -80 ∘C freezer until analysis.

To quantify the distinct nutrient concentrations, nutrient ratios and isotopic values of Atlantic Water (AW), Barents Sea Water (BSW) and Arctic Water (ArW), we define the water mass type of each sample using the water mass properties in Oziel et al. (2016). These are summarized in Table 1.

Additional sampling was conducted in collaboration with the Norsk Institutt for Vannforskning (NIVA, Oslo) during transits made by the general cargo vessel M/S Norbjørn between Tromsø, Norway, and Longyearbyen, Svalbard. The M/S Norbjørn is a “ship of opportunity” onto which NIVA has fitted a Ferrybox system that measures a variety of parameters including temperature and salinity at approximately 4 m depth. In addition, seawater can be collected directly from the system for further analysis. During each 4 d transit in March, June, August and November 2018, surface seawater samples were collected for the analysis of δ15N of particulate organic nitrogen (δ15N-PN), the δ15N and δ18O of nitrate (δ15N-NO3, δ18O-NO3), and inorganic nutrients (nitrate+nitrite, nitrite, silicate and phosphate) from 15 stations at pre-determined latitudes (Fig. 1a). Seawater was filtered through combusted GF/F filters for δ15N-PN analysis, and aliquots of the filtrate were placed into acid-cleaned high density polyethylene (HDPE) bottles and stored at -20 ∘C for the analysis of δ15N-NO3 and inorganic nutrients.

The isotopic composition of nitrate+nitrite (δ15N-NO3 and δ18O-NO3) was determined by the denitrifier method (Sigman et al., 2001; Casciotti et al., 2002) and following GEOTRACES protocols (Schlitzer et al., 2018). Samples were corrected using international reference standards N3 and USGS-34 (Weigand et al., 2016) and expressed in delta notation: δ15N-NO3 (‰ vs AIR) = (Rsam/Rstd-1)×1000; δ18O-NO3 (‰ vs VSMOW) = (Rsam/Rstd-1)×1000. Standards were run in triplicate with a reproducibility (σ) of δ15N ± 0.1 ‰ and δ18O ± 0.3 ‰. Internal standards were analysed in each run and corrected using N3 and U34, with an inter-run standard deviation of δ15N ± 0.1 ‰ and δ18O ± 0.3 ‰. Nitrite concentrations in our study region ranged from 0–0.66 µM, the highest concentration contributing 6 % of the nitrate+nitrite pool. Our isotopic measurements are compared to studies in which the nitrite in a sample has been removed using sulfamic acid (Granger and Sigman, 2009); to account for this, when nitrite was > 2.5 % of nitrate+nitrite, samples were re-run with sulfamic acid removal. For samples when nitrite was < 2.5 % of nitrate+nitrite, we correct our δ18O-NO3 data for nitrite interference following Kemeny et al. (2016). The δ15N-NO3+NO2 samples were also corrected assuming a δ15N-NO2 of -24 ‰ (Kemeny et al., 2016; Henley et al., 2017).

The δ15N-PN value was determined by elemental analysis with isotope ratio mass spectrometry (EA-IRMS) using a Costech Instruments elemental analyser coupled to a Thermo Scientific DELTA V Advantage mass spectrometer fitted with a ConFlo IV gas handling system. The instrumentation was operated using ISODAT 3.0 isotope ratio mass spectrometry software. Prior to analysis, the filters were wrapped in tin foil cones (OEA Laboratories) and pelletized. L-glutamic acid standards USGS 40 and USGS 41A were used as calibration standards during each analysis run. The δ15N values obtained for USGS 40 were -4.52 ± 0.08 ‰, n=28, and for USGS 41A 47.56 ± 0.18  ‰, n=21. A 10-point calibration using standard USGS 40 was measured to provide the linear regression equation (peak area vs expected N concentration) which was used to derive PN concentrations from the measured peak areas. Micrograms per litre (µg/L) concentrations were then calculated using the concentration obtained from the whole filter and volume of seawater filtered. The detection limit for PN was 10 µg.

The Barents Sea Opening (BSO) between Norway and Svalbard was dominated by saline (S> 34.8) Atlantic inflow, notably in Bear Island Trough (BIT; Figs. 1a, 2b). South of the Spitsbergen Bank, the water column was thermally stratified, with temperature exceeding 6 ∘C in the euphotic zone (Fig. 2a). South of 72∘ N, low-salinity (S< 34.7) water from the Norwegian Coast Current (NCC) occupied the near-surface layer (Fig. 2b). The water column was fresher, colder and well mixed over the shallow Spitsbergen Bank. This marks the westernmost extent of ArW and the polar front (Fig. 1b) and coincides with strong tidal currents and topographically steered flows (Oziel et al., 2016; Sundfjord et al., 2007; Vage et al., 2014).

Dense Barents Sea Water (BSW) was observed near the seabed in Hopen Trench (Fig. 2a; HT). Above it lay cooled Atlantic origin water and a thermally stratified surface layer. North of the narrow sill joining the Spitsbergen and Great banks (Fig. 2a; SB-GB), the approximate location of the polar front, colder (< 0 ∘C) and fresher (S< 37.7) ArW occupied depths below 50 m (Fig. 2b). This was capped with an even fresher (S< 34) layer of sub-zero temperature melt water. This polar surface layer extended southwards from the Nansen Basin, becoming progressively thinner. Below 100 m depth, over the shelf break and continental slope of the Nansen Basin, high-salinity (cooled) Atlantic origin water was observed within the boundary current that entered the Arctic via the Fram Strait (far right of Fig. 2a and b).

In the AW, nitrate concentrations were relatively homogenous below the mixed layer (11.8 ± 1.8 µM) but low or below the limits of detection in the euphotic layer (Fig. 2c). NH4+ concentrations were highest close to the seafloor over the Spitsbergen Bank (Figs. 2d, 4c). Both δ15N-NO3 and δ18O-NO3 were relatively homogenous in the deeper AW (δ15N-NO3=5.1 ± 0.1 ‰, δ18O-NO3=2.8 ± 0.3 ‰; Fig. 3a and b; Table 1), and N* was close to Redfield (N*-0.2 ± 0.8 µM). As nitrate concentrations decreased into the euphotic zone, both δ15N-NO3 and δ18O-NO3 increased as a result of nitrate utilization by phytoplankton (Fig. 4e and f).

The cooler ArW in the north of the Barents Sea had slightly lower (although not significantly different) nitrate concentrations of 10 ± 1.1 µM (Table 1). NH4+ concentrations were high close to the seafloor in Hopen Trench but decreased with increasing latitude. There was no significant difference in δ15N-NO3 or N* between AW and ArW (ArW δ15N-NO3=5.1 ± 0.1 ‰, N*=-1.0 ± 0.7 µM, Table 1), suggesting these nutrients also originated from the Atlantic. In contrast, δ18O-NO3 was 1.2 ‰ lower in ArW compared to AW (Figs. 3b, 4f).

In the BSW, δ15N-NO3, nitrate and N* were comparable to AW and ArW (δ15N-NO3=5.1 ± 0.4 ‰, nitrate =10.4 ± 1.2 µM, N*=-1.1 ± 1.1 µM). BSW δ18O-NO3 was 2.1 ± 0.5 ‰, lower than AW but higher than ArW, reflecting a mix between these two water masses.

We show that nitrogen availability in the Barents Sea is supported through AW supply and the efficient replenishment of nutrients through seasonal cycling processes. By the end of the growing season, all nitrate is consumed in surface waters, and the δ15N of PN reflects the AW source. The N inventory is also dependent on the NH4+ release from sediments and nitrification. In contrast to other Arctic shelf regions, we find no evidence for benthic denitrification interacting with the water column (and no loss of N relative to P). Our results indicate that although nutrients are regenerated in the western Barents Sea, δ15N-NO3 does not increase, suggesting that δ15N-NO3 supplied to Arctic Intermediate Water may be comparable to the AW source values. Our findings suggest δ15N-NO3 is unmodified in transit through the western Barents Sea. Additional samples collected in the eastern Barents Sea and at the primary export gateway (St. Anna Trough) are needed to confirm this.

Previous work has suggested that increasing NPP on Arctic shelves would increase organic matter supply to sediments and thus increase sedimentary denitrification rates (Arrigo and van Dijken, 2015). As N is the primary limiting nutrient to Arctic phytoplankton (Mills et al., 2018), this would have downstream consequences for NPP in the central Arctic Basin. Given the Barents Shelf is not currently a locale that hosts significant sedimentary denitrification and NPP here is limited by N, the future changes are likely to be different from those envisioned for other Arctic shelves. We suggest that N supply through the Barents Sea to the Arctic is likely to be determined by variability in AW inflow. Future changes in this inflow could impact the nutrient inventory transported through the Arctic Intermediate Water, impacting productivity in the central Arctic Basins where AWs are transported.