Dissolved Fe in the Deep and Upper Arctic Ocean With a Focus on Fe Limitation in the Nansen Basin

Global warming resulting from the release of anthropogenic carbon dioxide is rapidly changing the Arctic Ocean. Over the last decade sea ice declined in extent and thickness. As a result, improved light availability has increased Arctic net primary production, including in under-ice phytoplankton blooms. During the GEOTRACES cruise PS94 in the summer of 2015 we measured dissolved iron (DFe), nitrate and phosphate throughout the central part of the Eurasian Arctic. In the deeper waters concentrations of DFe were higher, which we relate to resuspension on the continental slope in the Nansen Basin and hydrothermal activity at the Gakkel Ridge. The main source of DFe in the surface was the Trans Polar Drift (TPD), resulting in concentrations up to 4.42 nM. Nevertheless, using nutrient ratios we show that a large under-ice bloom in the Nansen basin was limited by Fe. Fe limitation potentially prevented up to 54% of the available nitrate and nitrite from being used for primary production. In the Barents Sea, Fe is expected to be the first nutrient to be depleted as well. Changes in the Arctic biogeochemical cycle of Fe due to retreating ice may therefore have large consequences for primary production, the Arctic ecosystem and the subsequent drawdown of carbon dioxide.


Introduction
The Arctic Ocean is the most rapidly changing region of our planet due to recent global warming (IPCC, 2013); yet the central Arctic belongs to the least studied parts of the Earth. Over the last decade, Arctic sea ice has been observed to decline in extent (Stroeve et al., 2012;IPCC, 2014;Serreze et al., 2016) and thickness (Haas et al., 2008;Serreze and Stroeve, 2015) and changed from multi-year sea ice into more first-year sea ice (Maslanik et al., 2011). The Increased light penetration and nutrient availability during spring from earlier ice breakup enhances primary production in the Arctic Ocean and its adjacent shelf seas (Bhatt et al., 2014). The assumption has been for a long time that primary productivity is negligible in waters beneath ice because of insufficient light (Arrigo and Van Dijken, 2011). However, large under-ice blooms have been observed in the Barents Sea, Beaufort Sea, Canadian Arctic Archipelago, Chukchi Sea and also in the Nansen basin (Strass and Nöthig, 1996;Fortier et al., 2002;Mundy et al., 2009;Arrigo and Van Dijken, 2011;Ulfsbo et al., 2014) suggesting that under-ice blooms are widespread. Most studies expect nitrate to be the next limiting factor to determine primary productivity (Nishino et al., 2011;Vancoppenolle et al., 2013;Fernández-Méndez et al., 2015;Tremblay et al., 2015). However, in the Eurasian basin surface nitrate concentrations are significant and even persist in summer (Codispoti et al., 2013). Light limitation and/or grazing pressure have been suggested to prevent the full use of surface nitrate here (Olli et al., 2007;Bluhm et al., 2015).
Another consequence of global warming is the increase in river discharge (Peterson et al., 2002(Peterson et al., , 2006 which, combined with net loss of the Greenland ice-cap and melting of sea ice, can contribute to freshening of surface waters and increased stratification. However, the accumulation and distribution of this fresh water is strongly influenced by the Arctic circulation (Rabe et al., 2014).
River discharge in the shelf-surrounded Arctic Ocean is a source of nutrients.
These nutrients, including dissolved Fe (DFe) complexed by humic organic ligands (Laglera and van den Berg, 2009;Slagter et al., 2017), are transported in the upper 50m through the Arctic Ocean by the transpolar drift (TPD) from the Eurasian rivers over the central Arctic to Fram Strait (Gordienko and Laktionov, 1969;Gregor et al., 1998). The TPD track varies annually depending on the Arctic Oscillation index (Macdonald et al., 2005).
These climate-induced changes will change the biogeochemical cycling and therefore the distribution of many trace elements and isotopes. Of these Fe has proven to be the most important trace element as its low concentrations limit primary production in 30-40% of the global surface ocean and therefore regulates ocean processes such as marine ecosystem dynamics and carbon cycling (de Baar et al., 2005;Boyd et al., 2007). We do not directly expect Fe limitation in the Arctic Ocean because of the above-mentioned input and transport of DFe, confirmed by Klunder et al. (2012a). However, the results of the present investigation proved otherwise. Arctic Ocean waters form an important part of the global thermohaline circulation (Aagaard et al., 1985;Rudels, 2015;Carmack et al., 2016). Changes in the Arctic biogeochemical cycle of Fe will not only affect the Arctic ecosystem but will affect also the chemical composition of for example the North Atlantic Deep Water. Gerringa et al.
(2015) assumed that the Arctic is a source of Fe-binding dissolved organic ligands, since these decreased with distance from Fram Strait.
The deep water composition of DFe depends on the input of Atlantic water (200-900 m) with more elevated DFe in the Nansen Basin, and on slope processes with downwards convection of Fe released from resuspended sediments (Klunder et al., 2012b). A major source in the Nansen Basin is the hydrothermal activity at the Gakkel Ridge (Edmonds et al., 2003;Baker et al., 2004). These three processes are not expected to be rapidly affected by climate change.
Scavenging, however, might be prone to changes due to a higher biological activity in the euphotic zone, influencing the flux of sinking particles. According to Klunder et al. (2012b)

Sampling
A total of 28 stations were sampled for DFe during the GEOTRACES TransARC II cruise (PS94) on the German icebreaker RV Polarstern between 17 August and 14 October 2015 (Figure 1). Two CTD systems were used, a standard rosette sampling system equipped with a fluorometer for Chl a fluorescence and an "ultraclean CTD" sampling system; both were equipped with a SEABIRD 911 CTD. The temperature and salinity data from the standard sampling system were used for their higher spatial resolution (Rabe et al., 2016). The standard system also employed an uncalibrated fluorometer for CDOM measurement in arbitrary units (a.u.; BackScat, Dr. Haardt). Samples for DFe and nutrients were taken using 24 ultra-trace-metal clean polypropylene samplers of 24 L each mounted on an all titanium frame with a SEABIRD 911 CTD system and deployed on a 11 mm Dyneema cable without internal signal transduction cables. We used the SBE 17plus V2 Searam in a titanium housing to provide power, save the CTD data and close the sampling bottles at pre-programmed depths. After deployment, the complete "ultraclean CTD" was immediately placed in an ISO Class 6 clean room container, where samples for dissolved metals were filtered directly from the polypropylene samplers over < 0.2µm Sartobran 300 cartridges (Sartorius) under pressure of filtered N2 (0.7 bar) applied via the topconnector of the polypropylene sampler (de Baar et al., 2008;Rijkenberg et al., 2015). Filters were rinsed with approximately 700 mL seawater before use.

Calibration and detection limits
A five-point calibration in low iron containing seawater and blank determination were made daily. The blank was determined as the intercept of the signals of increasing preconcentration times (5, 10, 15 s) of the seawater used for the calibration. The analytical blank was on average 0.007 ± 0.004 nM DFe (N = 32) and the average detection limit (defined as 3σ of the blank pre-concentrated for 5 s) was 0.012 ± 0.007 nM DFe (N = 32). The Fe added by the Seastar acid (maximum ∼0.4 pM) was ignored. Table 2 Table 1 Detection limits, precision and accuracy in μM of the nutrients. The precision in one run is calculated to check whether small differences in one run are significant. The accuracy of the measurements of Si in the reference material seems high, but the concentration is high.

PO4
where (Fe:GLnut) is the average biological uptake ratio of Fe over the growthlimiting macronutrient (GLnut).Which of the macronutrients is expected to be growth limiting can be determined from N/P plots. Surface waters in many samples outside the TPD are depleted in N, or N will be the first macronutrient to deplete when extrapolating the N/P slope. Therefore, the ratio Fe:GLnut will be the uptake ratio of Fe:N (RFe:N).
As  , 1997). As N is the limiting macronutrient we have converted the reported RFe:P to RFe:N using 14.14, being the slope of the N/P plot at depths below 100 m in the Nansen Basin. A cut-off of 100 m was chosen since this is the maximum depth for primary production due to light limitation.

Calculation Percentage Unused Nitrogen due to Fe-Limitation
To calculate the percentage unused nitrogen due to Fe-limitation we integrated the sum of nitrate and nitrite concentrations over the upper 100m along our transect in the Nansen basin (PS94, Station 40, 50, 54, and 58) and subtracted the nitrogen used when all DFe would be fully used using the RFe:N of scenario 1 and 2 (Table 3)

Hydrography
The hydrography of the Transects I and II is described in detail by Slagter et al.      respectively. These data compare well although our data were lower in general.
Apparently Klunder et al. (2012b) sampled in more hydrothermally influenced areas in the Nansen Basin, than we did in the Amundsen Basin, as shown by higher average concentrations and standard deviations.

Surface and Shelf
Sources of DFe to the surface Arctic Ocean are sea ice melt, atmospheric inputs, lateral input from land with rivers as the main contributor, Atlantic inflow and mixing with deep water.
Sea ice melt is found to be important in the Gulf of Alaska and the Bering Sea (Aguilar-Islas et al., 2008, 2016 which feeds the coastal blooms alleviating Fe limitation; but we did not find distinct proof of the importance of sea ice melt as an Fe source in the Arctic during our cruise. However, Lam et al. (2006) concluded that melting sea ice provided substantial DFe to the water column.
The additional DFe input from melting sea ice is thought to become important for biological uptake for under ice and ice edge blooms. Most of our stations in Transects 1 and 2 had sea ice cover (Stations 32-134 were ice covered, Figure   1), and DFe concentrations were low outside the TPD ( Figures 4A-C, 5).
Apparently microbial utilization hid any contribution from melting sea ice at these locations.
Although the potential influence of dust input is acknowledged not much is known from dust input in the Arctic Ocean (Bullard, 2017). In Transect 3 we did see increased DFe in the surface at Station 147 ( Figures 4D, 7). Although this might be due to dust input, as DFe concentrations were high throughout the depth profile, lateral transport from Svalbard, Bear Island (North of Station 147) or sediment resuspension may be more likely explanations. The CTD transmission data (not shown) was slightly lower at stations in the Barents Sea According to literature (Lam et al., 2006;Brown et al., 2012) lateral transport from the coast, shelves and land-fast-glaciers, with or without further transport by mesoscale eddies, is feeding the blooms over nearby Canadian Basin. In the present study the major driver of such lateral transport is the TPD (Gordienko and Laktionov, 1969;Gregor et al., 1998;Macdonald et al., 2005). The TPD is the major surface current over the Arctic Ocean transporting sea ice and river water from the Arctic shelf seas toward Fram Strait. Its influence on the distribution of Fe is distinct (Klunder et al., 2012a;Slagter et al., 2017).
According to Klunder et al. (2012a) ice melt resulted in a relatively small increase in DFe relative to the effect of the TPD. Outside the TPD, in the Nansen Basin (Stations 32,40,50,54,58,and 64), surface concentrations of DFe were very low and coinciding with under ice bloom (Figures 5, 8).
To investigate if Fe could become a limiting factor for this under ice bloom preventing the full use of the macronutrients, as described in the method section (Fe * <0), we calculated Fe * for both the Klunder data as well as our own (Parekh et al., 2005;Blain et al., 2008;Lannuzel et al., 2011). Nitrogen is generally considered to be the limiting macro nutrient in the Arctic Ocean (Nishino et al., 2011;Vancoppenolle et al., 2013;Fernández-Méndez et al., 2015;Tremblay et al., 2015) and was the limiting macronutrient during PS94 (Figures 3A-E , using that N:P ratio the area of Fe * <0 occurrence would be larger. Along Transect 3 DFe is also low (i.e., 0.08 nM at Station 169 at 10-25 m depth) and the macronutrient concentrations between Stations 153 and 169 are not depleted. The N:P slope for depths >100 m is inconclusive here. Extrapolating the N:P slope for all depths ( Figure 3E), nitrogen would be the first to be depleted, but phosphorous would also be low. Like Transects 1 and 2 we used nitrogen as macronutrient to calculate Fe * , and indeed Fe * <0 also occurs here, though only in scenario 1 (Figure 9). In a warming Arctic Ocean, ice cover decreases releasing primary producers from light limitation. We here show that Fe may become the next limiting factor.
In the Nansen Basin, depending on the RFe:N scenario and the estimate of the background concentration of nitrate + nitrite, between 5 to 54% of available nitrate + nitrite would remain unused for the production of biomass due to Fe limitation ( Table 3). increasing light availability will increase DFe uptake by increasing primary production. In the present study we found Fe to be limiting in the Nansen Basin.
Depending on the Fe requirements of the Arctic microbial community Fe has the potential to also limit primary production in the Barents Sea. So depending on how climate change will affect the role of the TPD in the provision of DFe to the surface of the Arctic Ocean Fe limitation may either be alleviated or exacerbated in a future Arctic Ocean.