An Arctic Strait of Two Halves: The Changing Dynamics of Nutrient Uptake and Limitation Across the Fram Strait

The hydrography of the Arctic Seas is being altered by ongoing climate change, with knock‐on effects to nutrient dynamics and primary production. As the major pathway of exchange between the Arctic and the Atlantic, the Fram Strait hosts two distinct water masses in the upper water column, northward flowing warm and saline Atlantic Waters in the east, and southward flowing cold and fresh Polar Surface Water in the west. Here, we assess how physical processes control nutrient dynamics in the Fram Strait using nitrogen isotope data collected during 2016 and 2018. In Atlantic Waters, a weakly stratified water column and a shallow nitracline reduce nitrogen limitation. To the west, in Polar Surface Water, nitrogen limitation is greater because stronger stratification inhibits nutrient resupply from deeper water and lateral nitrate supply from central Arctic waters is low. A historical hindcast simulation of ocean biogeochemistry from 1970 to 2019 corroborates these findings and highlights a strong link between nitrate supply to Atlantic Waters and the depth of winter mixing, which shoaled during the simulation in response to a local reduction in sea‐ice formation. Overall, we find that while the eastern Fram Strait currently experiences seasonal nutrient replenishment and high primary production, the loss of winter sea ice and continued atmospheric warming has the potential to inhibit deep winter mixing and limit primary production in the future.

. Here, we use these tools to decipher how nutrient limitation is changing in the Fram Strait.
The Fram Strait is both an inflow and outflow gateway to the Arctic Ocean. In the east, warm and saline Atlantic Water (AW) that originates from the subpolar and subtropical North Atlantic gyres is transported northward within the West Spitsbergen Current (WSC) and enters the Arctic Ocean north of Svalbard. The WSC is a complex, eddy-shedding , multibranch system with both barotropic and baroclinic components (Beszczynska-Möller et al., 2012;Richter et al., 2018) that keeps the eastern Fram Strait ice free for most of the year. In the upper layers of the WSC, the AW has been observed to be warming since the mid-1990s at a rate of 0.06°C year −1 and also increasing in salinity (Beszczynska-Möller et al., 2012;Larsen et al., 2016;Polyakov et al., 2017;Tsubouchi et al., 2021;Walczowski et al., 2017). Warming of AW has been driven by a combination of both a local response to increasing air temperatures and reduced heat loss (Furevik, 2001;Karcher et al., 2005), and a greater proportion of subtropical water being transported into the Nordic Seas (Hatún et al., 2005). Carried by the boundary current in the Eurasian Basin, this warming signal is progressing into the Arctic Ocean and increasing heat fluxes to overlying water, which weakens the halocline, increases winter ventilation of the ocean interior, and accelerates sea-ice decline (Polyakov et al., 2017(Polyakov et al., , 2020Steele & Boyd, 1998). These changes drive an increase in nutrient supply, supporting the view that areas experiencing "Atlantification" may become more productive (Ardyna et al., 2014;Randelhoff et al., 2015Randelhoff et al., , 2018Randelhoff et al., , 2020. However, further warming, notably in regions that have already become ice free, could lead to increased stratification and decreased nutrient supply, ultimately controlling nutrient availability to Arctic phytoplankton with subsequent influence on productivity. On the western side of the Fram Strait, the East Greenland Current (EGC) system (Havik et al., 2017) transports cold, fresh Arctic origin water and sea ice southward into the subpolar North Atlantic Ocean. The Polar Surface Water (PSW) in the upper 200 m of the EGC consists of riverine input, sea-ice melt, and Pacific origin water (Dodd et al., 2012;Haine et al., 2015). The shelf break and offshore branches of the EGC also transport deeper Arctic Atlantic Water (AAW), an Atlantic derived water mass that has entered the Arctic Ocean and been modified during transit within the boundary current and passage across Arctic shelves (Wefing et al., 2019). Local westward recirculation of AW within Fram Strait (RAW; recirculated Atlantic Water), mediated by high eddy activity, joins the AAW and significantly increases the southward transport of the shelf break branch of the EGC (de Steur et al., 2014;Havik et al., 2017;Richter et al., 2018). Modeling studies reveal interannual to decadal scale variability in freshwater export from the Fram Strait. However, improvements to the simulation of salinity anomalies are needed in order to make robust estimates of longterm trends (Jahn et al., 2012;Karcher et al., 2005). Observations suggest that the freshwater flux within the EGC was relatively stable between 1998 and 2008, but transport over the wide East Greenland Shelf where the freshest waters are found was not accounted for . The salinity of polar water over the East Greenland Shelf may have been decreasing since the mid-2000s (Larsen et al., 2016). At the surface, the EGC transports relatively high silicate and phosphate, but low nitrate, a signature that is due to Pacific-water influence and amplified by benthic denitrification on Arctic shelves (Chang & Devol, 2009;Devol et al., 1997). The depleted nitrate concentrations in PSW are the main limiter to net primary production (NPP) in this region (Krisch et al., 2020). Higher nitrate concentrations are found in Atlantic origin water that intrudes across the shelf along the bottom . Any increases in freshwater input and/or surface warming that strengthen stratification across the western side of Fram Strait may suppress vertical mixing and limit the supply of bottom water nitrate to the surface.
These east-west differences across Fram Strait produce gradients in the physical and biogeochemical processes which control NPP. In the now ice-free eastern Fram Strait, annual NPP is ∼80 g C m −2 (Hop et al., 2006) compared to 50-100 g C m −2 in partially ice-covered waters in the middle of the strait (Wassmann et al., 2010) and ∼15 g C m −2 (Codispoti et al., 2013) in the ice-covered PSW in the western Fram Strait. As the freshwater, heat and volume transports both into and out of the Arctic Ocean via the Fram Strait continue to be altered and the biogeochemical composition of these waters changes, there is a need to understand the impact on nutrient supply and the implications for NPP, both now and in the future. Nitrogen isotope measurements provide a tool to help address this challenge.
Nitrate isotope measurements estimate the relative contributions of NPP, remineralization, and denitrification to the nitrate present in the water column. The 15/14 N and 18/16 O in nitrate (δ 15 N-NO 3 and δ 18 O-NO 3 , respectively) provide complementary information about nitrate uptake and regeneration processes (Rafter et al., 2013;). Fractionation of nitrogen and oxygen atoms occurs at the same rate (ε) during uptake of nitrate by phytoplankton (∼5‰) (Tuerena et al., 2015;Waser et al., 1998) and during denitrification (25‰-30‰) (Sigman, DiFiore, Hain, Deutsch, Wang, et al., 2009). When nitrate is newly nitrified (i.e., produced from recently remineralized organic matter), the nitrogen atoms are unchanged and reflect the isotopic signature of organic matter, which varies in response to local conditions. In contrast, oxygen atoms are sourced from the oxidation of ambient O 2 and seawater, providing a predictable nitrification signature of δ 18 O-H 2 O plus 1.1‰ (Buchwald et al., 2012;. Isotopic signatures of nitrogen and oxygen in nitrate therefore diverge in waters with strong recycling, with a relative decrease in δ 18 O-NO 3 . The tracer Δ(15-18) (δ 15 N-NO 3 minus δ 18 O-NO 3 ) captures the differences between these two isotopes, highlighting nitrate sources from different oceanic environments (Rafter et al., 2013). A specific process in the Arctic affecting Δ(15-18) is benthic denitrification, which can release high δ 15 N-NH 4 into the water column, a process termed coupled partial nitrification-denitrification (Brown et al., 2015;Fripiat et al., 2018;Granger et al., 2018). This process leads to an increase in δ 15 N-NO 3 , a decrease in δ 18 O-NO 3 (via water column nitrification), and a concomitant increase in Δ(15-18).
The uptake of nitrate by phytoplankton can often be described by Rayleigh fractionation systematics (Mariotti et al., 1981). Nitrate utilization by phytoplankton in an environment where there is no resupply of nutrients, that is, a stratified upper ocean in summer, follows Rayleigh fractionation systematics for a closed system, with δ 15 N-NO 3 and δ 18 O-NO 3 increasing in a 1:1 relationship as nitrate is consumed. The slope of this line is dependent on the isotopic effect (ε) (Granger et al., 2004). As nitrogen is consumed to completion, the N isotopic signature in organic matter (δ 15 N-PN) will approach that of the inorganic nitrogen at the beginning of the growing season plus the isotopic effect, which for phytoplankton assimilation is 5‰. When the upper ocean is less stratified, δ 15 N-NO 3 and δ 18 O-NO 3 can represent open system dynamics where there is continual resupply of nutrients through the growing season. In this situation, the isotopic enrichment of δ 15 N-PN is weaker as new nitrogen with a lower isotopic signature is continually supplied. Applying closed system assumptions to an open system will result in an apparent isotope effect weaker than 5‰. Thus, in combination with nitrate isotopes, δ 15 N-PN can track the extent of biological utilization, contrasting nutrient sources, and the significance of new versus regenerated production.
This study characterizes the nitrogen cycling processes occurring in the upper waters of Fram Strait in spring and summer. We focus on the AW and PSW entering and exiting the Arctic, respectively. We utilize a combination of biogeochemical tracers, δ 15 N-NO 3 , δ 18 O-NO 3 , Δ(15-18), δ 15 N-PN, N* (N* = nitrate-phosphate•16 Gruber & Sarmiento, 1997) and Si* (Si* = silicate-nitrate Sarmiento et al., 2004), alongside hydrographic data (temperature, salinity, and mixed layer depth [MLD]) from two cruises in spring and summer to explore nitrogen availability to phytoplankton and the implications for primary production. Different nutrient supply mechanisms in the region are more broadly characterized using results from a historical hindcast simulation with a global ocean biogeochemical model from 1970 to 2019. We draw upon the simulations to discuss the implications of warming to Arctic nutrient availability and primary production in the future.

Fieldwork
This work is based on samples collected on two cruises in the Fram Strait region. Cruise PS100 took place from June 18 to September 6, 2016, and JR17005 took place between May 8 and June 8, 2018 (Figure 1a). Cruise PS100 was conducted on the R/V Polarstern as part of the larger GEOTRACES program (Schlitzer et al., 2018). Cruise JR17005 took place on the RRS James Clark Ross as part of the UK Changing Arctic Oceans program. Although the two seasons sampled were not in the same year, they were both in years of relatively small April sea-ice area export through the Fram Strait (Mayot et al., 2020). Although we treat these cruises as representative of late spring and late summer, we acknowledge that this is an assumption and cannot account for interannual variability in the onset of seasonal changes. On both cruises, seawater samples were collected from Niskin bottles mounted on a rosette equipped with a SBE911plus CTD system recording conductivity, temperature, and pressure at 24 Hz (Hopkins et al., 2019;Kanzow et al., 2017).
Both cruises completed a transect across the Fram Strait at ∼79°N (F-Line, Figure 1a). On JR17005, a crossshelf transect running up the Norske Trough (NT Line) was also sampled. On PS100 sampling across the Norske Trough, further on-shelf (NT3 and NT4) was possible and a section along the front of the 79°N glacier (79NG) was completed. The extra sections are included in Figures S1 and S2 and are included in our statistical comparisons of AW and PSW (Table 1).
We quantified the stability of the upper water column to vertical displacements according to the Brunt-Väisälä frequency, N 2 : where g is the gravitational acceleration, ρ is the potential density, and z is the depth. assessed by calculating the difference in potential density between 10 and 200 m depth, Δρ (Figures 1b  and 1c). The thermal (Δρ T ) and haline (Δρ S ) contributions to this density difference, where Δρ ∼ Δρ T + Δρ S , were estimated as follows: where α and β are the thermal expansion and haline contraction coefficients, respectively, calculated at the average temperature, salinity, and pressure between 10 and 200 m. Subscripts 1 and 2 refer respectively to surface (10 m) and 200 m depth values of temperature (T) and salinity (S) and ρ is the average potential density between 10 and 200 m. The surface MLD was calculated following Peralta-Ferriz and Woodgate (2015), defined as the maximum depth at which the potential density was within 0.01 kg m −3 of the shallowest measurement. On both cruises, the shallowest measurement was always within 5 m of the surface. The MLD did not exceed 200 m on either cruise.
Dissolved inorganic nutrient concentrations from JR17005 were determined using a Bran and Luebbe QuAAtro 5-channel autoanalyser (SEAL Analytical) and AACE operating platform (V 6.1) following standard colorimetric methods. Nutrient concentrations from PS100 are publicly available in the Pangaea database (Graeve & Ludwichowski, 2017). Nitrate isotope samples were filtered inline from the Niskin bottles (0.7 μm pore size filters) and frozen at −20°C until analysis on land. Nitrate samples were measured for δ 15 N-NO 3 and δ 18 O-NO 3 using the denitrifier method and bacterial strain Pseudomonas aureofaciens (Casciotti et al., 2002;Sigman et al., 2001). Samples were corrected using international reference standards IAEA-N3 and USGS-34 and an internal standard of North Atlantic Deep Water was run with each batch to check for interrun comparability. Final values for both cruises were calculated using the correction scheme of Weigand et al. (2016) with reproducibility of 0.1‰ and 0.3‰ for δ 15 N-NO 3 and δ 18 O, respectively. Nitrite concentrations in our study region ranged from 0 to 0.32 μM and 91% of nitrate isotope samples had a <2% contribution of nitrite to the nitrate + nitrite pool (215/235 samples). Our isotopic measurements are compared to studies in which the nitrite in a sample has been removed using sulfamic acid (Granger & Sigman, 2009); to account for this, we correct our δ 18 O-NO 3 data for nitrite interference following Kemeny et al. (2016). The δ 15 N-NO 3 + NO 2 samples were also corrected assuming a δ 15 N-NO 2 of −24‰ (Kemeny et al., 2016;Tuerena et al., 2021). All samples with a high contribution (>2%) of nitrite were in the upper 50 m and were therefore not used for water mass averages. Note. AW, Atlantic Water; PSW, Polar Surface Water. Boundaries of potential temperature (θ) are in °C and potential density (σ θ ) are in kg m −3 . This definition also captures AW that has recirculated locally within the Fram Strait and has been entrained within the EGC system. No values are added to δ 15 N-NO 3 , δ 18 O-NO 3 , and Δ(15-18) during spring and summer, as their water mass characteristics are best described below the mixed layer and away from the effects of spring/ summer uptake; δ 15 N-PN values are not shown for below the mixed layer as these values would be influenced by regeneration. Samples collected from two cruises in summer 2016 were used to estimate δ 15 N-PN to increase the confidence of our statistics, those from July to August were collected from cruise PS100 and those from August to September were collected from cruise NPI-16. All other values in summer are from cruise PS100.  Richter et al. (2018) δ 15 N-PN from JR17005 was determined by EA-IRMS using a Costech Instruments Elemental Analyzer coupled to Thermo Scientific Delta V Advantage mass spectrometer fitted with Conflo IV gas handling system. The instrumentation was operated using ISODAT 3.0 isotope ratio MS software. Prior to analysis, the filters were wrapped in tin foil cones (OEA Laboratories) and pelletized. δ 15 N-PN from PS100 were analyzed with elemental analyzer (Thermo Flash 2000) coupled to an isotope ratio mass spectrometer (Thermo, Delta V Plus). Additional δ 15 N-PN samples were collected from the Norwegian Polar Institute cruise of Fram Strait in August-September 2016 and analyzed using a Carlo Erba NA 2500 elemental analyzer inline with a VG PRISM III isotope ratio mass spectrometer. The inclusion of these additional summer samples provides 125% more summer measurements and increases confidence in statistics (see Table 1).
Within regions on the eastern (3°-8°E) and western (6°-10°W) sides of the Fram Strait that were sampled on both cruises, we calculate the depth of the top and bottom of the nitracline following Randelhoff et al. (2016). First, the nitrate samples were interpolated onto a 1-m-resolution vertical grid. The top and bottom of the nitracline were then defined as the depth where the nitrate concentration was 20% and 80%, respectively, of the difference between surface and deep reference values. The surface reference value was taken as the average concentration over the top 15 m of the water column. The deep reference value was taken as the average concentration between 90 and 160 m on the eastern side of the Fram Strait and between 140 and 200 m depth on the western side of Fram Strait, a depth range occupied by Atlantic origin water.
With the exception of the eastern side of the Fram Strait in May-June, these deep references were below the base of the mixed layer and the depth of maximum N 2 . In May-June, although the mixed layer over the deep basin on the eastern side of the Fram Strait was often greater than 160 m, nitrate within the 90-160 m reference interval was homogeneously distributed and the average concentrations were above 10 μM.
To quantify the different nutrient concentrations, nutrient ratios, and isotopic values of AW and PSW, we defined the water mass type of each sample using the water mass properties defined in Richter et al. (2018) and summarized in Table 1. We explicitly exclude samples collected within AAW whose signatures may have been modified within the Arctic basin. Note also that the definition of AW used here (see Table 1) includes Atlantic origin water that has recirculated locally within the Fram Strait (RAW) and has been entrained within the EGC system.

Hindcast Simulation
We used the Pelagic Interactions Scheme for Carbon and Ecosystem Studies version 2 (PISCES-v2) biogeochemical model (Aumont et al., 2015), attached to the Nucleus for European Modeling of the Ocean version 4.0 (NEMO-v4) general ocean circulation model and the Sea Ice modeling Integrated Initiative (SI 3 ) sea-ice model. The biogeochemical model carries two phytoplankton types (nanophytoplankton and diatoms), two zooplankton types (microzooplankton and mesozooplankton), large and small particulate organic matter, dissolved organic matter, oxygen, carbon chemistry tracers, and five nutrients (nitrate, ammonium, silicate, phosphate, and iron). The nitrogen cycle accounted for nitrogen fixation, atmospheric deposition, river inputs, nitrification, sedimentary and water column denitrification, and burial. Nitrogen isotopes were integrated within PISCESv2 to enable direct comparison with field data (Text S1). Latitudinal resolution on the tripolar grid was nominally 2° but increased to 0.5° at the equator and 1° poleward of 50° (in Fram Strait the resolution was ∼5° longitude by 0.9° latitude [∼100 km 2 ] which is not eddy-resolving), while vertical resolution varied between 10 and 500 m thickness over 31 levels. We forced the coupled NEMO-SI 3 -PISCES model with the Japanese atmospheric reanalysis (JRA-55) over the years 1958-2019 (Tsujino et al., 2018) following the Ocean Modeling Intercomparison Project protocols (Orr et al., 2017). Six repeat cycles of this 62-year forcing (372 years) were made beginning on the January 1, 1648 CE, ending on December 31, 2019, with the ocean biogeochemical initial state taken from a 5,000 years spin-up simulation under preindustrial conditions. Only output in the final cycle was used in analysis, and because the final cycle was initialized with the end of the fifth cycle, we only included the period 1970-2019 in our analysis to avoid unrealistic trends due to adjustment at the beginning of the final cycle.
While the resolution of the model did not resolve the observed eddy kinetic energy field, the baroclinic component of the boundary currents or fine-scale AW recirculation pathways (Beszczynska-Möller et al., 2012;Richter et al., 2018;von Appen et al., 2016), it did simulate realistic trends in temperature and salinity in the west and east of the northern Greenland Sea. However, due to a weak WSC, the east Fram Strait at 79°N was dominated by PSW rather than AW. To overcome this local inconsistency, we assessed hydrographic and biogeochemical properties within the AW and PSW water masses over the Greenland Sea between 22°W-15°E and 70°N-80°N. AW was defined as warmer than 2°C and saltier than 34.5 psu, while PSW was defined as cooler than 0°C and fresher than 33.5 psu. These property bounds are slightly different to those in the field data (Table 1) but were chosen to best represent these water masses in the model ( Figure S3). Temperature and salinity trends within each water mass were assessed in the mixed layer, defined by the 0.01 kg m −3 density change. We assume that the depth of the mixed layer in winter is a reasonable proxy for the maximum depth of winter mixing. Simulated nitrate concentrations and its isotopes were assessed in the upper 50 m because summer euphotic and MLDs were consistently shallower than 50 m and 98% of all NPP occurred within the upper 50 m. This water mass approach, rather than a transect approach, enabled a more direct comparison between simulated and observed AW and PSW.

Water Mass Characterization, Mixed Layer Depths, and Stratification
The strength of stratification across the Fram Strait within the upper 200 m of the water column increased from east to west in both seasons (Figures 1b and 1c). On the eastern side of the strait, relatively weak upper ocean stratification in Atlantic origin water was maintained by the thermal contribution to vertical density differences. This thermal control leads to deep convection during the cold winter months and thermal stratification in the summer. To the west of the AW-PSW front, the fresher Arctic sourced surface waters resulted in a much more strongly stratified water column. Note that during May-June, the vertical temperature gradient on the western side of the Strait was unstable (i.e., it made a negative contribution to stability). Spring to summer changes in MLD, the depth of maximum N 2 , and the depth of the top and bottom of the nitracline on the eastern and western sides of Fram Strait are summarized in Figure 4. Early in the season between May and June, the MLD on the eastern side of the Strait ranged between 120 and 180 m, with some shallower and deeper instances (Figures 2a and 4e). Although surface warming had started to establish a weak peak in N 2 at around 30 m (Figure 4e), the surface temperature was <4°C (Figure 2a), and the resulting vertical density differences associated with this near surface warming were not sufficient to be identified as the MLD according to the threshold used here. The nitracline was found between 40 m (top) and 70 m (bottom), just below the maximum in N 2 , but above the MLD (Figure 4e). The surface chlorophyll-a fluorescence signal in spring was well mixed down to the top of the nitracline (maximum in N 2 ) and then gradually decreased toward the base of the nitracline (Figure 4d). In the summer, surface water temperatures increased above 6°C (Figure 3a) and stratification strengthened (Figures 4a and 4b), leading to a much more clearly defined mixed layer at ∼20 m, coincident with the maximum in N 2 (Figure 4e). Both the top and bottom of the nitracline also shoaled by ∼10 m but remained below the MLD. A subsurface chlorophyll-a fluorescence maximum was observed at the base of the surface mixed layer (Figure 4d).
In contrast to the large seasonal change in MLD observed on the eastern side of Fram Strait, the spring to summer change in MLD on the western side of Fram Strait (in PSW) was smaller. In spring, a 20-40 m well-mixed surface layer overlaid a broad (60-70-m-thick) pycnocline, which typically extended below the top of the nitracline (Figure 4j). During the summer, warming of a thin PSW layer (potential temperature >0°C) and the addition of sea-ice melt established a MLD and maximum value of N 2 between 10 and 20 m (Figure 4j). This shallow surface layer was decoupled from the nitracline below (30-110 m). There was little change in the depth of the top and bottom of the nitracline between spring and summer. The surface chlorophyll-a fluorescence signal in spring was well mixed throughout the surface mixed layer and gradually decreased within the broad pycnocline. During the summer, two subsurface peaks in chlorophyll-a fluorescence were observed: the shallowest just below the surface mixed layer and N 2 maximum and a second, deeper and weaker peak within the nitracline.

Nutrient Concentrations and Isotopic Ratios
Surface water nitrate was low from nitrate uptake by phytoplankton (Figures 2 and 3c). To the east of the Fram Strait, nitrate below the mixed layer in Atlantic origin water was 11.6 ± 0.8 μM (Figures 2 and 3c and Table 1). In the western Fram Strait, nitrate in PSW found beneath the mixed layer was lower, 7.2 ± 2.3 μM, demonstrating a transfer of low nitrate waters from within the central Arctic basin . Nitrate within the deeper Atlantic origin water on the western side of the Fram Strait was higher (∼11 μM). Silicate concentrations were depleted to below 4 μM through the upper 200 m of the AW, a contrast to the PSW where concentrations often exceeded 6 μM in the upper 100 m. During summer, nitrate in the mixed layer was increasingly depleted in both water masses, in the AW nitrate decreased from 5.1 ± 4.0 to 0.9 ± 1.3 μM and in the PSW nitrate decreased from 1.6 ± 1.3 to 0.5 ± 0.4 μM.
In the east, the nitrate transported in AW (below 100 m depth) had a δ 15 N-NO 3 value of 5.1‰ ± 0.2‰, which is comparable to values entering the Barents Sea and those from the North Atlantic Ocean (Peng et al., 2018;Van Oostende et al., 2017). This suggests that δ 15 N-NO 3 has not been modified in transit from the subpolar North Atlantic through uptake and regeneration processes. δ 18 O-NO 3 was elevated to 2.4‰ ± 0.4‰ in AW. This is comparable to preformed values across the Barents Sea Opening  and results from partial nitrate assimilation followed by nitrification and mixing occurring in transit from the subpolar North Atlantic, causing δ 18 O-NO 3 to drift upward as the δ 18 O-NO 3 consumed is lower than the δ 18 O-NO 3 produced (Peng et al., 2018).
In the west, PSW exiting the Arctic via the Fram Strait contains waters which have been transported through the Barents, Kara, and Laptev Seas, where they are susceptible to modification processes such as nitrate regeneration and denitrification occurring on these shelves (Wefing et al., 2019). The outflowing PSW therefore has an imprint of the eastern and western Arctic shelf signals and that of the Transpolar Drift; their proportions dependent on the strength of the Beaufort Gyre, the pathway of the Transpolar Drift, and the location of the Pacific-Arctic front located near the Lomonosov Ridge (Proshutinsky & Johnson, 1997;Steele et al., 2004). The PSW has relatively high concentrations of phosphate and silicate but relatively low concentrations of nitrate (Torres-Valdes et al., 2013). Below 75 m depth (below the surface mixed layer), we measured δ 15 N-NO 3 and δ 18 O-NO 3 of 5.6‰ ± 0.4‰ and 1.5‰ ± 0.5‰, respectively (Table 1), which are similar to values measured by Fripiat et al. (2018) in the northern Fram Strait (∼82.5°N). High δ 15 N-NO 3 and low δ 18 O-NO 3 represent an Arctic signal from Pacific origin waters alongside the influence of benthic denitrification and nitrate regeneration from the Arctic basin (Granger et al., 2018). The lower δ 15 N-NO 3 and higher δ 18 O-NO 3 compared to the Pacific source and East Siberian Sea (Figure 5b) represent a more widespread nitrate source from both Pacific and Atlantic influenced shelves (Fripiat et al., 2018).
Nitrate isotope values were enriched in all surface waters across the Fram Strait, revealing the importance of nitrate uptake by phytoplankton elevating δ 15 N-NO 3 and δ 18 O-NO 3 above deep water values (Sigman et al., 1999) (Figures 2, 3g and 3h). During summer, nitrate concentrations in surface waters were often below our measurement limit for nitrate isotopes, preventing the expected higher values being measured.
In general, the more elevated nitrate isotope signatures were associated with higher δ 15 N-PN, as the heavier isotope signature was progressively taken up by phytoplankton as the season progressed ( Figure 5).
By considering the tracer Δ(15-18), the effects of assimilation are removed allowing the source signals to be viewed more clearly (Figure 5b and Table 1). The Δ(15-18) tracer shows a gradient from lower Δ(15-18) in the east to higher Δ(15-18) in the west (AW 2.7‰ ± 0.5‰ and PSW 4.2‰ ± 0.7‰). The low Δ(15-18) in AW has been measured in the subpolar North Atlantic and Barents Sea (Peng et al., 2018;Tuerena et al., 2021). We identified a significant negative correlation (r 2 = 0.515, df = 21, p ≤ 0.0002) between Δ(15-18) and N* (Figure 5b). The linear relationship suggests that the signal of low N* and high Δ(15-18) in the PSW is a farfield signal which has been exported from the Arctic basin and carried south within the EGC. The Δ(15-18) measured in the PSW is lower than measurements from the East Siberian Sea ( Figure 5, Fripiat et al., 2018), likely from mixing with Eurasian influenced shelves suggesting a diluted Pacific source signal exiting the Fram Strait.

Present Day Fram Strait Dynamics
We have identified strong west-east differences in our data originating from the far-field transfer of hydrographic and biogeochemical properties, from the North Atlantic (eastern Fram Strait) and central Arctic basin (western Fram Strait). Our isotopic data can be further used to determine the effects of these properties on N uptake and limitation.
In AW, weak stratification, a deep mixed layer, and a shallow nitracline during spring (Figure 4e) are likely to have facilitated wind-driven vertical mixing of nitrate into the euphotic zone and the even distribution of chlorophyll-a between the surface and top of the nitracline (Figures 4c-4e). The AW was most represented by an open system for samples in the upper 200 m, suggesting ample replenishment of nitrate used by phytoplankton (Figure 5c). In spring, δ 15 N-PN was 3‰ ± 1.2‰, the low δ 15 N-PN (lower than the nitrate source of ∼5‰) reflects incomplete nitrate utilization and resupply of nutrients within the deep mixed layer. In summer, increased stratification and shoaling of the MLD to ∼20 m restricted the vertical supply of nutrients. However, average δ 15 N-PN remained low in summer (3.1‰ ± 2.5‰ during July-August and 3.6‰ ± 0.5‰ in August-September, Figure 5d and Table 1), suggesting that even in late summer, NPP was not strongly limited by N. This finding corroborates recent work that AWs in the Fram Strait are colimited by Fe and Si (Krisch et al., 2020). Overall, in the eastern Fram Strait, there are more nutrients available to phytoplankton, likely supporting higher NPP compared to the western Fram Strait.
In contrast, strong salinity stratification in PSW prevented MLDs from reaching the nitracline in both spring and summer (Figure 4j). Decoupling of surface waters from the nitracline reduced the potential vertical supply of new nitrate to phytoplankton (Figure 4). This was reflected in our isotopic measurements, which showed closed system dynamics and a low isotope effect of ∼2‰ (Figures 5d and 5e). The lower isotope effect is a common feature of N-limited algae  and is present in the mixed layer of salinity stratified Arctic waters . δ 15 N-PN was high throughout spring and summer and matched the nutrient source by late summer (Table 1), highlighting the stratified environment with stronger N limitation to local phytoplankton.

Simulated Changes in Atlantic and Polar Surface Water
Using a biogeochemical model hindcast simulation from 1970 to 2019 (see Section 2), we assessed how intra-and inter-annual changes in stratification and winter MLDs (taken as a proxy for the depth of maximum winter mixing) controlled nitrate cycling and its isotopes in PSW and AW. It should be noted that the coarseness of our model substantially degraded the local circulation, causing a weak WSC that was overcome by PSW in the eastern Fram Strait. The importance of eddy-driven transport across the Fram Strait is clear (Fieg et al., 2010;Hattermann et al., 2016;Wekerle et al., 2017), but coarse resolution was a consequence of resolving the complex cycling of nitrogen and its isotopes, requiring dozens of tracers. To alleviate the physical biases, we widened our region of focus to include the Greenland Sea where PSW and AW dominated the west and east, respectively.
The hydrography of PSW and AW over the Greenland Sea was well reproduced by the simulation and showed strong qualitative agreement with field data taken in 2016 and 2018 ( Figures S3 and S4), as well as strong multiannual trends in the region that replicated observations. Surface water over the Greenland Shelf, west Greenland Sea, and Iceland Sea freshened at rates of roughly −0.05 psu per decade (Figure 6a), consistent with observed changes in salinity of the EGC since the mid-1980s (Larsen et al., 2016). On the eastern side of the Greenland Sea and in the Loften and Norwegian Basins, there was an increase in salinity that exceeded 0.02 psu per decade (Figure 7a), also consistent with observational evidence since the 1990s (Beszczynska-Möller et al., 2012;Larsen et al., 2016;Lauvset et al., 2018). The salinity trends were strongest over the upper 200 m and were accompanied by warming trends across the region (Figure 6b). Warming was, however, strongest in the east, with trends in temperature as high as 0.5°C per decade, similar to observations (Beszczynska-Möller et al., 2012;Lauvset et al., 2018;Skagseth & Mork, 2012;Tsubouchi et al., 2021;Walczowski et al., 2017). Amplified warming in the east relative to the west at similar latitudes strongly suggested that lateral heat transport made an important contribution to local warming. The strongest west-east gradients in temperature and salinity were established between 2010 and 2019 in our model. Freshening strengthened stratification of PSW in the west, whereas in the east the compensating effects of warming and increasing salinity caused small changes to already weak vertical density gradients in AW (Figures 6c and S3).

10.1029/2021GB006961
14 of 20 The contrasting PSW and AW hydrography impacted nitrate supply to surface waters. Intense stratification in PSW inhibited vertical supply ( Figure 7a). Meanwhile, lateral supply decreased at the beginning of the 1990s and culminated in a loss of roughly 2 μM of nitrate from PSW by 2019 (Figures 7c and 7g). Nitrate limitation in PSW, as observed in our field data, may therefore be a recent consequence of far-field nitrate loss originating in the Arctic, likely driven by upstream increases in primary production (Lewis et al., 2020) and denitrification (Granger et al., 2018) that strip Arctic surface waters of nitrate. Deep winter mixing in AW in the east provided a strong contrast to stratified PSW in the west. Winter mixing consistently intersected the nitracline and replenished nitrate to support the spring bloom ( Figure 7b). Phytoplankton within AW experienced nonlimiting nitrate concentrations as a result, and this was also evident in the greater difference between simulated δ 15 N-NO 3 and δ 15 N-PN in the east compared to the west (Figures 7e and 7f). However, despite nonlimiting nitrate concentrations, a clear decline in AW nitrate of ∼0.5 μM was apparent since the 2000s and paralleled declines in the depth of winter mixing. The relationship between winter-spring (January-May) MLDs and surface nitrate was strong (Spearman's rank correlation; r 2 = 0.72), with deep mixing (>400 m) in the 1970s and early 1990s providing as much as 0.8 μM of nitrate more than during shallower winter mixing (250-300 m) in the 1980s and post-2000.
Changes in the depth of winter mixing in AW in our model were primarily driven by a reduction in the influence of sea ice in the region. As sea-ice concentrations declined in the North Greenland Sea and eastern Fram Strait, salt fluxes associated with freezing (i.e., brine rejection) decreased. The contribution of salinity to winter density gradients changed over time during our model experiment, transitioning from a net contribution to mixing (negative downward gradient) to a net contribution to stratification (positive downward gradient) (Figure 7h). By the final years of the simulation, the winter halocline reversed from a TUERENA ET AL.

10.1029/2021GB006961
15 of 20 negative to positive gradient for the first time, indicating that winter brine rejection from sea-ice formation was no longer influential. This link between sea-ice formation and deep winter mixing has been previously demonstrated by observations, with buoyancy losses tending to be greatest at the sea-ice edge where brine rejection combines with strong turbulent heat losses (Vage et al., 2018), and is apparent in other models (Brodeau & Koenigk, 2016).
It must be acknowledged, however, that the transition from sea-ice presence to absence occurred much earlier in reality than in our simulation. Like other coarse resolution models (e.g., Brodeau & Koenigk, 2016), we observed a cold bias and too much sea ice in the west and north of the Greenland Sea due to biases in the mean state circulation. In reality, the eastern Fram Strait has been ice free since the early 1980s (Bonnet et al., 2010), likely making heat fluxes the current dominant control on winter mixing rather than sea-ice-associated buoyancy losses. As rapid warming has weakened turbulent heat loss from surface waters in recent decades (Moore et al., 2015), the eastern Fram Strait may have initially experienced weaker winter mixing in the 1980s associated with the transition to sea-ice-free conditions, followed now by further stratification associated with atmospheric warming. Our results therefore suggest that both mechanisms of increasing stratification (less brine rejection and less sea-air heat loss) limit seasonal nitrate replenishment in the eastern Fram Strait. However, multidecadal decreases in mixing in the region are highly uncertain (Brodeau & Koenigk, 2016). Considering the coarseness of our model prevents an accurate representation of important circulation features (Fieg et al., 2010;Hattermann et al., 2016;Wekerle et al., 2017), and predictions of reduced mixing should be viewed with some skepticism. Whatever the direction of change, our field measurements and modeling show that continued monitoring using nitrogen isotopes will reveal if nitrogen is ample or limiting to primary producers.

Future Prospects for Fram Strait Nutrient Dynamics
In the near future, N limitation may intensify in the western Fram Strait as warming and sea-ice loss stimulate phytoplankton growth in the central Arctic basin and benthic denitrification on Arctic shelves (Arrigo & van Dijken, 2015;Lewis et al., 2020). In the eastern Fram Strait, nutrient supply to surface waters will likely remain stable if winter mixing continues to intersect the nitracline. Observed deep mixing in spring 2018 was sufficient for seasonal nitrate replenishment, indicating that the east-west gradient in nutrient concentrations may strengthen in coming years. Further declines in nitrate supply to the western Fram Strait will reduce NPP and further alter phytoplankton δ 15 N with implications for Arctic food web studies . It is important to note that we draw contrasts between distinct water masses on either side of the Fram Strait and acknowledge that the transition between these water masses is unlikely to be linear. The active eddy field and recirculations across the central Fram Strait introduce a host of nonlinear dynamics that mix AW and PSW in highly variable ways.
Over longer time scales of warming, however, winter mixed layers in the east may continue to shoal to depths above the nitracline (Brodeau & Koenigk, 2016;Moore et al., 2015). At present, AWs contribute the majority of the nitrate that supports current NPP in the Fram Strait, implying that any declines in AW nitrate supply could dramatically change NPP in this region. Nitrate in the east is already a primary limiting nutrient to phytoplankton (alongside iron and silicate) (Krisch et al., 2020). Nitrate decline as a consequence of shoaling mixed layers would therefore decrease NPP in the east and potentially increase community recycling of nutrients.
An important consequence of reduced winter mixing in the eastern Fram Strait would be to reduce summertime silicate at a faster rate than nitrate due to the deeper silicicline, potentially increasing silicate limitation in the eastern Fram Strait (Hatún et al., 2017). Although silicate is not a limiting nutrient to all phytoplankton, these changes would alter local phytoplankton assemblages. Diatom limitation and blooms of non-silicifying species, such as Emiliana huxleyi, are already observed in the Barents Sea and Fram Strait (Neukermans et al., 2018;Oziel et al., 2020). Diatoms account for much of polar NPP and carbon drawdown (Krause et al., 2019;Vaquer-Sunyer et al., 2013), and thus decreases in silicate availability will affect the strength of the biological pump in this region and also impact food web structure (Vernet et al., 2017).
Our results show that climate change in the short term is unlikely to greatly affect NPP across the Fram Strait in the coming years, with nitrate limitation already resulting in low NPP in the west and deep winter TUERENA ET AL.

10.1029/2021GB006961
mixing replenishing nitrate in the east. The decline in nutrient supply to the western Fram Strait is only a small proportion of the N that currently supports NPP across Fram Strait and so these projected declines will not greatly reduce total NPP. However, looking further into the future, if winter MLDs shoal above the nitracline due to rapid atmospheric warming, the ongoing increase in Arctic NPP from alleviated light limitation and terrestrial inputs (Lewis et al., 2020;Terhaar et al., 2021) may be curtailed.

Conflict of Interest
The authors declare no conflicts of interest relevant to this study.