Dissolved Trace Metals in the Ross Sea

The dissolved (D) trace metals zinc (Zn), cadmium (Cd), cobalt (Co), copper (Cu), iron (Fe), manganese (Mn), nickel (Ni), titanium (Ti), lanthanum (La), yttrium (Y), and lead (Pb) were analyzed via ICPMS in samples from the Ross Sea obtained during a cruise between 20 December 2013 and 5 January 2014. The concentrations of DZn, DCd, DCo, DCu, DFe, DMn, DNi, and DTi were significantly lower in the Antarctic surface Water (AASW) compared to the other deeper water masses, indicating biological uptake and possibly scavenging. In the AASW, DLa and DY were higher than in Winter Water (WW). This can be explained by a spring source from ice melt followed by loss during summer and autumn, probably due to passive adsorption. Dissolved Pb was low (16 pM) and no distinction between water masses was possible. Akin to the macro-nutrients nitrate and silicate, the modified Circumpolar Deep Water (mCDW) shows elevated DCd compared to the shelf water masses. Sea ice melt and ice sheet melt released DZn, DFe, DMn, DNi, DY, DLa, and probably DPb into the Ross Sea. However, only DFe, DMn, DY and DLa are transported into the Antarctic Circumpolar Current with the outflowing High Salinity Shelf Water (HSSW). The bottom nepheloid layer (BNL) released DFe, as well as DMn and DCu, into the HSSW whereas lateral transport from land formed a source of DMn and DFe. One station in the Ross Sea Polynya was resampled after two weeks, during which time the thickness of the BNL increased, with accompanying increases in DFe and DMn near the seafloor. In the surface layer nutrients (including micro-nutrients) were depleted further. The uptake slopes/stoichiometric ratios of DZn, DCd and DCo versus phosphate indicated that the distribution of these metals is related to uptake as well as the composition of the phytoplankton community. Estimated stoichiometric ratios of Zn and Cd relative to P were higher at a station dominated by Phaeocystis antarctica than at diatom-dominated stations, implying a higher utilization of these metals by P. antarctica.


INTRODUCTION
The growth of Southern Ocean phytoplankton is primarily limited by low dissolved Fe (DFe) concentrations and availability of light, making the Southern Ocean the largest of the world's high nutrient low chlorophyll areas (de Baar et al., 1990;Martin et al., 1990;Boyd et al., 2012). Polynyas, open waters amidst an ice-covered sea, are essential components of the Antarctic ecosystem since these are hotspots for primary productivity . Since phytoplankton comprises the base of the marine food web, the higher trophic levels depend on primary productivity within coastal polynyas. For example, in western Antarctica, 91% of all Adélie penguins are associated with productive coastal polynyas (Arrigo and Van Dijken, 2003). The Ross Sea Polynya is the most productive of the 46 Antarctic coastal polynyas that have been identified (Arrigo and Van Dijken, 2003;Arrigo et al., 2008Arrigo et al., , 2015 and is also a site of deep-water formation, resulting in significant CO 2 drawdown and sequestration of carbon (Dunbar et al., 1998;Arrigo et al., 2008). In the Ross Sea Polynya, DFe is often the primary nutrient that is limiting to phytoplankton growth (Sedwick et al., 2000Marsay et al., 2014;McGillicuddy et al., 2015;Hatta et al., 2017), even in early spring due to "early season depletion" (Sedwick et al., 2011). This early depletion was confirmed by additional shipboard DFe measurements made in the spring of 2013, with DFe in the upper mixed layer as low as 0.02-0.08 nM (Gerringa et al., 2015) when the polynya was still forming. Thus, phytoplankton depended on Fe from either internal cycling through remineralization of organic matter or from external sources. It has been established that apart from local sea ice melt at the margins of the polynya and lateral transport from land, upward fluxes from the sediment are the most important DFe source to surface waters of the Ross Sea (Marsay et al., 2014;Gerringa et al., 2015;Hatta et al., 2017).
Because Fe is the main limiting nutrient in the Ross Sea, this element has been studied extensively. However, other trace elements may play an important role in regulating rates of primary production as well as determining community composition (Bruland et al., 1991;Sunda and Huntsman, 2000;Bertrand et al., 2007;Hoppema et al., 2007;Twining and Baines, 2013;Saito et al., 2017;Alderkamp et al., 2019). For example, Wu et al. (2019) demonstrated that Mn was co-limiting with Fe in cultures of Phaeocystis antarctica from McMurdo Sound, a key member of the phytoplankton community in the Southern Ocean, indicating late-season Mn stress to be important for primary production in the Ross Sea, warranting further investigation.
The stoichiometric relationships between dissolved (D) metals [e.g., zinc (Zn), cadmium (Cd), cobalt (Co), copper (Cu) and nickel (Ni)] and macro-nutrients {e.g., nitrate (NO 3 − ), silicate (Si(OH) 4 ), and phosphate (PO 4 3− )} indicate that their concentrations are regulated by phytoplankton uptake (de Baar, 1994;Löscher et al., 1998;Cullen and Sherrell, 2005;Cullen, 2006;Lai et al., 2008;Bruland et al., 2014;Dulaquais et al., 2014a;Wyatt et al., 2014;Saito et al., 2017;Till et al., 2017;Middag et al., 2018Middag et al., , 2019Middag et al., , 2020. For example, Zn is used for protein repair, in the enzyme alkaline phosphatase, and for the uptake of CO 2 via the enzyme carbonic anhydrase (Morel et al., 1994;Buitenhuis et al., 1999;Hu et al., 2003;Shaked et al., 2006;Montsant et al., 2007). Low dissolved Zn concentrations can influence phytoplankton growth as illustrated by deformities in the shells of Emiliania huxleyi under Zn limitation (Crawford et al., 2003;Schulz et al., 2004). Cd also has a nutrient-type concentration profile and a tight relationship with the macro-nutrient PO 4 3− , but Cd is not considered an essential element for phytoplankton growth. However, it can replace Zn in carbonic anhydrase for some, but not all, phytoplankton species Lee and Morel, 1995;Sunda and Huntsman, 2000;Timmermans et al., 2001;Hendry et al., 2008). Although Cu, Co, and manganese (Mn) do not show typical nutrient-type profiles, they are essential micro-nutrients for phytoplankton growth. Enzymes based on the metal pairs Fe-Mn and Cu-Zn are required to protect the cell interior from damage by reactive oxygen species (De Baar and La Roche, 2003). Diatoms are known to be limited by low DCu, although as far as we know, this has not been encountered outside culture experiments (Morel and Price, 2003;Guo et al., 2012;Morel et al., 2014). Co is necessary for the biosynthesis of vitamin B12 by cyanobacteria and can, like Cd, substitute for Zn in carbonic anhydrase (Sunda and Huntsman, 1995;Rodionov et al., 2003;Zheng et al., 2019). The Ross Sea is prone to B12 limitation because there are no cyanobacterial populations (Caron et al., 2000;Marchant, 2005), although Bertrand et al. (2007) have shown that heterotrophic bacteria in the Ross Sea are able to produce B12. Thus, it is possible that low DCo limits the production of vitamin B12 (Gordon and Sullivan, 2008). The micro-nutrient Ni is usually not depleted in the surface ocean (Butler et al., 2013;Middag et al., 2020) and Morel et al. (2014) suggested this could be caused by slow uptake mechanisms.
Little is known in Antarctic waters about effects of trace metals other than DFe Alderkamp et al., 2019) on phytoplankton species distribution (Saito and Goepfert, 2008). However, in temperate and warmer waters, Zn addition stimulates growth of the coccolithophore E. huxleyi and Co and Ni stimulate Synechococcus and other cyanobacteria (Crawford et al., 2003;Saito et al., 2005;Dupont et al., 2008;Sunda, 2012). Despite the difference between temperate and high latitude systems, trace metals are likely to impact species composition in a similar way in Antarctic environments.
The metals Cu, Co, Fe, and Mn have a so-called hybrid type depth distribution, since the concentrations in the upper water column often increase with depth and, depending on their susceptibility to scavenging, decrease with depth deeper in the water column (Bruland et al., 2014). Oxidation and precipitation or co-precipitation can play a role in regulating trace metal concentrations here too, notably for Co and Mn (Middag et al., 2012;Sherrell et al., 2015;Noble et al., 2017;Saito et al., 2017).
Surface sources of these metals to the Southern Ocean consist of atmospheric input (Landing and Bruland, 1980;Baker et al., 2006;de Jong et al., 2013), ice and glacial melt (Frache et al., 2001;Lannuzel et al., 2014), while remineralization of organic matter, such as fecal pellets, can be an important process of retaining metals in the surface layer (Hendry et al., 2008;Saito et al., 2010;Croot et al., 2011;Wyatt et al., 2014;Laglera et al., 2017;Vance et al., 2017). Gerringa et al. (2012) concluded that the Pine Island Glacier formed the most important supply of DFe for the phytoplankton blooms in the Pine Island Polynya. However, the contribution of the glacial melt was overestimated due to an error in the computation of the horizontal flux gradient contributing to the biological uptake (Gerringa et al., 2020). Taking the distance over which the flux gradients were calculated into account, the contribution of the glacial melt is of the same order of magnitude as upwelling, sea ice melt and vertical transport from the sediment, which counted for approximately 10% of the assumed phytoplankton consumption. In retrospect, it appears that regeneration of Fe is vital to sustain the phytoplankton bloom within the Pine Island Polynya.
The metals titanium (Ti), lanthanum (La), yttrium (Y), and lead (Pb) have, as far as we know, no function in phytoplankton metabolism (Nozaki, 2001;De Baar and La Roche, 2003;Dammshäuser et al., 2011;Parker et al., 2016), but nevertheless, the dissolved concentrations are low in the surface layer and increase with depth, with the exception of Pb. Ti can enter the ocean by dust or lateral transport from terrestrial input and is used as a tracer for Patagonian dust input in Antarctic waters, as it is particle reactive and has a relatively short ocean residence time (Skabral, 2006;Dammshäuser et al., 2011). Over much of the global ocean, Pb concentrations used to be high in the surface oceans due to leaded automobile fuel emissions. In Antarctic waters, the decrease in Pb pollution resulting from use of lead-free fuels is not as noticeable due to residual Pb stored in glacial ice and released upon melt (McConnell et al., 2014;Ndungu et al., 2016). Dissolved Y and La have nutrient-type depth profiles (De Baar et al., 1985;Parker et al., 2016) although DLa hardly increases with depth since it is probably more susceptible to scavenging (Parker et al., 2016). This difference might be due to the smaller size of Y, probably causing it to be more strongly complexed with organic ligands and therefore less susceptible to scavenging compared to La (Shannon, 1976;Nozaki, 2001;Parker et al., 2016). Rivers and dust are known sources for both elements (Till et al., 2017), however, these sources play a minor role in the Southern Ocean.
Here we present the first densely sampled transects for a suite of trace metals (Zn, Cd, Co, Cu, Fe, Mn, Ni, Ti, La, Y, and Pb) from the surface to the bottom in the central Ross Sea. Data were collected during NBP13-10 (Phantastic 1) between 20 December 2013 and 5 January 2014, from two north-south transects and one east-west transect (Figure 1), mostly within the Ross Sea Polynya. Data for DFe concentrations from the present cruise have already been published, although analyzed with a different technique (Flow Injection Analysis, Gerringa et al., 2015) than discussed in this paper. Therefore, DFe is treated here briefly, only focusing on similarities and differences with other metals. We show relationships of trace metals with macro-nutrients and discuss sources and sinks of the metals to unravel the interactions between the marine ecosystem and metal cycling.

Sampling
Samples were taken during expedition NBP13-10 on the RVIB Nathaniel B. Palmer in the Ross Sea, entering from the northeast on 20 December 2013 and leaving to the northwest on 5 January 2014 (Figure 1) (for full details, see Gerringa et al., 2015Gerringa et al., , 2019Alderkamp et al., 2019).
A total of 350 samples were taken at 33 stations at depths between 2 and 714 m along three transects. The first transect was located in the eastern Ross Sea . A second longer transect formed a circle in the Ross Sea Polynya, starting 10 km from the Ross Ice Shelf along the 177.5 • E meridian going north, crossing troughs between the Ross and the Pennell Banks (St. 20-60), turning west from the Pennell Bank to the Joides Trough (St. 61-65), and bending south-west to south-east (via St. 75,(85)(86)(87)91,101,(111)(112)(113)(114), with station 114 revisiting station 31) (Figure 2). A third short transect was sampled from 9.5 km east of Franklin Island going eastwards (St. 88-90); these stations were covered by sea ice for 30-50% of the surface.
Modified 12 L Go-Flo (General Oceanics) samplers from the Royal NIOZ (the Netherlands) were attached to a trace metal clean frame from the United States Antarctic Program on a Kevlar cable. Temperature, depth, and salinity were measured with an SBE 9/11plus conductivity-temperature-depth (CTD) system (SeaBird Electronics). Transmission was obtained with a C-star transmissometer (WET Labs) and chlorophyll a (chl a) was estimated by fluorescence with a fluorometer (WET Labs). CTD data are expressed as conservative temperature ( in • C) and absolute salinity (S A in g kg −1 ) (transformed according to McDougall et al., 2009), as described in Gerringa et al. (2015).
After recovery of the trace metal clean frame, water was subsampled and filtered (Sartorius R , 0.2 µm; Sartobran 300) inside a trace metal clean van directly from the Go-Flo bottles. Samples were acidified with HCl (Seastar© baseline hydrochloric acid) to a final concentration of 0.024 M (pH 1.7-1.8) in a laminar flow bench in a plastic bubble with overpressure achieved by inflow of air through a HEPA filter. Before use, the sample bottles were rigorously cleaned in three steps according to GEOTRACES protocols.

Analysis of Dissolved Trace Metals by ICPMS
Trace metals were analyzed with a Thermo Finnigan HR-ICPMS element 2. A 30 mL sample was pipetted into an acid cleaned FEP vial. Peroxide (final concentration 10 µM) and an internal standard (indium and lutetium) was added at a final concentration of 5 nM prior to UV digestion after Middag et al. (2015). The samples were digested for 4 h in a custommade UV box after which they were pre-concentrated using a seaFAST pre-concentration system, using Nobias PA1 resin in the pre-concentration column (Lagerström et al., 2013;Bown et al., 2017). Using two loops of 10 mL each, a total of 20 mL sample was pre-concentrated into 500 µL elution acid, which gives a pre-concentration factor of 40. Samples were eluted using 1.5 M teflon distilled HNO 3 that contained rhodium as a standard to easily identify any sample introduction issues in a run (Middag et al., 2015). The system was calibrated via standard additions using a mixed stock solution made of 1000 ppm single spike standards. A typical calibration line consisted of seven standard additions where the maximum standard additions are comparable to maximum concentrations in the marine environment. The recovery was verified in each analytical run by comparing the slope of the seawater calibration curve (multi-element standard added to seawater) and the eluent calibration curve (multi-element standard added directly to the elution acid) after Biller and Bruland (2012). Blank contributions from sample handling, pre-concentration, and analyses steps were determined by analyzing acidified Milli-Q water (∼1.8 pH) as a sample and exposing the sample to similar procedures. The accuracy and precision of the measurements were determined by measuring GEOTRACES community consensus reference materials SAFE D1 and S, as well as in-house reference seawater samples ( Table 1).

Data Treatment of Trace Metals
Outliers were classified as described by Middag et al. (2015), where a linear regression was determined between the data deeper and shallower than a suspect data point; however, we used a rejection criterion of >20% deviation from the predicted value instead of 10%.
For Pb, contamination was evident for early stations , where the concentrations decreased and profiles became less irregular per station. The Pb data of these stations were removed from this study. It is possible that the Pb concentration FIGURE 2 | Conservative temperature ( • C) in the circle transect ( Figure 1). Black isolines indicate temperatures used for defining the water masses. White isolines show absolute salinity (g kg −1 ), with values used also for defining water masses. Some station numbers are given on top of the figure. The sequence of stations where trace metals were sampled and analyzed from right to left is: 44,45,43,114,31,30,20,46,47,48,49,59,60,61,62,63,64,65,75,85,86,87,101,91,111,112, are outside this transect (see Figure 1 for their positions). AASW = Antarctic Surface Water, WW = Winter Water, mCDW = modified Circumpolar Deep Water, MSW = Modified Shelf Water, HSSW = High Salinity Shelf Water, ISW = Ice Shelf Water. Figure made using ODV (Schlitzer, 2018). decreased from east to west in the eastern Ross Sea between stations 16 and 18, although a reduction of Pb contamination in the sampling system seems more likely. Ti and Pb data at St. 65 were eliminated because the concentrations were substantially higher than elsewhere and the vertical profiles consisted of more outliers than normal data. The software package R (R Development Core Team, 2011) was used for statistical calculations, to calculate Pearson and Tukey correlations. The median values with the variance in metal and nutrient concentrations are shown as boxplots (Figure 3). The actual dissolved metal concentrations in the circle transect ( Figure 4) were presented using Ocean Data View (ODV, Schlitzer, 2018).
The bottom nepheloid layer (BNL) was defined at a water depth of > 200 m by a reduction in transmission to below 98.5% where the DFe concentration was >0.5 nM, or by a reduction in transmission to between 99% and 98.5% where DFe was >1 nM (N = 44). The mixed layer depth (MLD) was defined as the depth where neutral density was 0.02 kg.m −3 greater that at the surface.
To test for significant differences in nutrient concentrations between water masses, a t-test and an ANOVA-Tukey test were applied using R, with a 95% confidence interval.
The median values are indicated by the thick horizontal line in the boxes, which contain the first and third quartiles, whereas the whiskers indicate the range excluding outliers. Circles are outliers (>1.5 interquartile range from the box, Teetor, 2011). The figures are made using the software package R.
The notation metal/P (e.g., Zn/P) is used for dissolved ratios and metal:P (e.g., Zn:P) for stoichiometric ratios derived from a slope of a linear regression.
Unless otherwise noted, data are expressed as the mean ± the standard deviation.
The datasets of FIA and ICPMS generally compare well, show the same trends (Supplementary Information (SI) Supplementary Figures S1, S2I,J) but differ slightly in absolute concentrations. The average offset for low DFe concentrations is 0.055 ± 0.021 nM higher when analyzed by ICPMS after the cruise (averaged for FIA DFe < 0.1, N = 146, the offset is 0.075 ± 0.026 nM averaged for FIA DFe < 0.5, N = 278).

Hydrography, Nutrients and Chlorophyll a
As described in previous papers on data from the same cruise (Gerringa et al., 2015(Gerringa et al., , 2019, four main water masses were present in the study area, based on definitions of Tomczak and Godfrey (2001) and Orsi and Wiederwohl (2009). Antarctic Surface Water (AASW, potential temperatures ( ) > −1.85 • C and neutral density < 28 kg m −3 ) covered the surface, below which, in the center of the polynya, remnants of cold Winter Water (WW) were present north of 76 • S (St. 48) (Figure 2). From the north, warm mCDW (34 < S < 35) flowed onto the shelf into the Ross Sea, whereas in the south, cold Shelf Water (SW) was formed near the Ross Ice Shelf ( < −1.85 • C) with variable salinity depending on salt rejection. The following subdivisions can be distinguished in SW: Ice Shelf Water (ISW), low salinity SW (LSSW), high salinity SW (HSSW), and modified SW (MSW) (Figure 2). The very cold ISW was only present close to the ice shelf ( < −1.95 • C, at St. 18,. The difference between LSSW and HSSW depends on the extent of salt rejection in the sea ice (for LSSW and HSSW, S < and > 34.62, respectively), where LSSW was only present at the stations in the eastern Ross Sea (St. 16, 17 and 18). MSW is formed through mixing of northward flowing SW with southward flowing AASW and mCDW, and has a broad range of salinities and temperatures (S > 34.5 and > −1.85 • C, Orsi and Wiederwohl, 2009). A BNL was observed above the bottom in the Ross Sea Polynya in HSSW and at two stations in MSW (Gerringa et al., 2015(Gerringa et al., , 2019. Relative to deeper water, mean macro-nutrient concentrations were relatively low in the mixed layer (ML), which varied between 8 and 99 m, averaging 36 ± 21 m, but were not low enough to limit phytoplankton growth (Alderkamp et al.,  In the BNL, Si(OH) 4 concentrations were similar to those in mCDW. However, the Si(OH) 4 concentrations showed more variance than NO 3 − and PO 4 3− in deeper water masses, notably in HSSW and in the BNL (Figures 3A-C). In HSSW, NO 3 − and PO 4 3− concentrations became almost constant with depth after a steep increase with depth in AASW, but Si(OH) 4 kept increasing slightly with depth. Figure 1 shows that the distribution of satellite obtained chl a has a distinct pattern, known to be reoccurring yearly with only slight variations in horizontal distribution. This was explained by Reddy and Arrigo (2006) to be due the interaction of the bathymetry and the currents, like the incoming mCDW in the Joides Trough.

Metals per Water Mass
In the following sections, the results are described per water mass from surface to bottom with a focus on the Ross Sea Polynya transect (Figures 3F-P, 4A-K, 5D-N, 6D-N) and with an additional focus on those metals that are correlated with macro-nutrient concentrations. The metals DZn, DCd, and DCo typically displayed a surface minimum and a steep increase with depth followed by more or less constant deep concentrations. The concentrations of DCu, DFe, and DMn also had a surface minimum and an increase with depth, but also a distinct increase toward the sediment. Last, the details for metals that often also featured a surface minimum, but with less distinct depth profiles, like DTi, DNi, DLa and DY will be presented. For DPb, there was no significant difference between water masses (average concentrations 17 ± 7 pM in the ML, elsewhere 16-19 ± 7pM, Tables 2, 3).

ML and AASW
All dissolved metals, except DPb, had significantly lower concentrations in the AASW and ML compared to deep water masses, similar to the macro-nutrients (Figure 3, Tables 2, 3). No differences existed between average dissolved metal concentrations in the ML and the overall AASW that includes the ML (Tables 2, 3). For DFe, the concentrations were not only extremely low in the AASW but in the whole upper 200 m (Figures 3J, 4E). For DCu, DNi, DTi, DY and DLa, the difference between surface and deeper water masses was relatively small (Tables 2, 3, Figures 3, 4, 6).
Stations sampled earlier in the season, in the eastern Ross Sea and stations close to the Ross Ice Shelf (St. 44, 45), had less pronounced surface minima for most metals, whereas even surface maxima existed here for DNi, DY and DLa. St.17 in the eastern Ross Sea exhibited constant DZn, DCd, and DCo concentrations with depth (Figures 4, 5).

WW
During the cruise, WW was present only in the center of the polynya 75,85 and 113). Like the macro-nutrients, DZn, DCd, and DCo increased steeply with depth in WW, with DZn increasing from 2.32 (±1.37) in ML to 5.23 ± 0.42 in WW ( Table 2, Figures 3-6). Other metals, such as DCu, DMn, DNi, and DTi, also increased, but not as steeply, resulting in non-significant differences in concentration between WW and AASW. However, DFe had equal concentrations and DY and DLa appear to have lower concentrations in WW compared to AASW (Figures 3N,O; DY: 0.18 nM in AASW and 0.176 nM in WW; DLa:33 pM vs. 31 pM in AASW and WW, respectively). For DFe, this is most probably due to severe depletion by phytoplankton uptake, continuing into WW whereas for DY and DLa the apparent WW minimum is related to spatial trends. The WW was present only in the center of the polynya 75,85,and 113), and here, average concentrations in AASW and WW were the same for both metals whereas concentrations outside the center of the polynya, where no WW was present, concentrations where relatively high in AASW. However, the concentrations of DY and DLa in AASW and WW together were significantly lower than in the deeper water masses.

Shelf Waters ISW, LSSW, HSSW, and MSW
A BNL was present predominantly in the HSSW, as shown by the salinity (Figure 3D), and only at two stations (St. 49 and 59) in MSW. Because of the high metal concentrations in the BNL, especially for DFe and DMn, the BNL samples were treated as a separate "water type" in order to distinguish the specific BNL metal source. Dissolved metal concentrations in ISW were not significantly different from those in other water masses with the exception of the BNL, which is probably due to the large variation within this water mass and the fact that it was sampled only four times.
DZn, DCu, DFe, DMn, DNi, DTi, DLa, and DY had elevated concentrations in the shelf water masses, with generally lower concentrations in MSW relative to the other shelf water masses, probably due to mixing with overlying AASW and WW. DZn showed higher concentrations in ISW and LSSW compared to the other shelf water masses; notably, DZn was significantly higher in LSSW than in HSSW, but it must be noted that this trend is based only on 14 samples in LSSW vs. 127 samples in HSSW ( Figure 3F, Table 3). Moreover, the DZn concentrations were higher and the variation was larger in the eastern northsouth part of the Ross Sea circle transect (St. 20-60) compared to the western north-south transect (St. 75-113) (Figure 4A at 0-500 km). Akin to DZn, variation in DCu along the eastern part of the transect was larger than along the western part of the transect. The variations within a water mass for DCu and DZn, were relatively large when compared to DCo or DCd (Figures 3F-I). Concentrations of DFe, DMn, DLa, and DY were relatively high in LSSW (St. 16-18) and HSSW (mean DMn 1.40 ± 0.78 and 1.24 ± 0.50, respectively) ( Table 2, Figures 3J,K,M,N) compared to MSW and, except for DFe, to mCDW. High DFe (>1.5 nM) probably showed influence of the BNL (Figure 3J), akin to the elevations in DCu and DZn. As already discussed for DFe (Gerringa et al., 2015), DMn decreased with increasing distance from the island along the Franklin Island transect at depths greater than AASW, indicating the island was a source of DMn that was laterally transported away from it. For DNi and DTi, the influence of shallow banks and a BNL as a source is less clear.

mCDW and BNL
The inflow of mCDW onto the shelf, characterized by higher temperatures and low oxygen (Figures 2, 3D), is seen in elevated Si(OH) 4 and in relatively low DMn, DY, and DLa concentrations (Figures 3C,K,N,O; Tables 2, 3) when compared with HSSW. A slight increase in the DCd concentrations with depth is visible in Figure 4B at St. 46-64, where the mCDW was encountered, up to 0.78 ± 0.02 nM in the otherwise constant DCd in HSSW and MSW (0.72 nM, Table 2). DZn is slightly higher in mCDW than in HSSW, but DCo, DCu, DNi, and DTi have concentrations comparable to those in HSSW (Figures 3, 4).  The significance levels are indicated as: Bold: p < 0.005; underlined: p < 0.01; italic: p < 0.05.
The highest DCo, DCu, DFe, DMn, DNi, DTi, DY, and DLa concentrations were observed near the sediments in the BNL (Figures 3, 4). At most BNL stations, DCo increased toward the sediment, but only in samples taken closest to the sediment. For DFe and DMn, elevated concentrations were found near the bottom as well as further above the sediments, but always connected to the presence of a BNL.

Revisiting St. 31
The location of St. 31 was revisited again after 14 days as St. 114. During these 14 days, the temperature at a depth of 25 m (AASW) increased from −1.11 to −0.18 • C and macronutrient concentrations decreased (NO 3 − from 24.2 to 17.5 µM), showing the progressive warming and associated nutrient uptake by phytoplankton. In AASW, the decrease in NO 3 − , PO 4 3− , and Si(OH) 4 was associated with a decrease in DZn, DCd, DCo, DCu, DMn, and DNi (Supplementary Figure S2). In the bottom layer of this particular station, the transmission decreased from 98.8% to 98.3% and the accompanying increase in thickness of the BNL was associated with elevated DFe, DMn, and DCu. For most metals, the concentration at mid depths was very similar between the two occupations, with the exception of DPb. This could indicate contamination for Pb, complicating interpretation of the Pb data.

The Importance of Organic Complexation
Apart from direct sources and sinks, the competition between complexation by dissolved organic ligands and scavenging can be important in determining dissolved metal concentrations, notably DFe concentrations. When the Fe-binding dissolved organic ligands are saturated, the excess DFe concentration is not stable due to the low solubility product of Fe-oxyhydroxides (Liu and Millero, 2002) and Fe will be scavenged or precipitated, as has been demonstrated in the Ross Sea for this cruise data (Gerringa et al., 2019).
Although it is known that DZn, DNi, DCo and DCu form organic complexes (Bruland, 1989;Achterberg and Van Den Berg, 1997;Saito et al., 2004Saito et al., , 2010 no organic complexation data is available for the samples of this cruise. The concentrations and solubilities of DZn and DNi are probably not depending on organic complexation (Nimmo et al., 1989;Ellwood and Van den Berg, 2000;Saito et al., 2004;Baars and Croot, 2011) because the concentration and the strength of the ligand is relatively low. Saito et al. (2010) studied organic speciation of Co in the Ross Sea during spring and concluded that both the strong Co(III) and weaker, more labile, Co(II) organic complexes were present over the whole water column. These complexes were present in more or less constant ratios and did not change with depth, indicating that scavenging is probably not important in the Ross Sea, explaining the more or less constant deep DCo concentrations in our research, as also observed by Saito et al. (2010).
Although we have no information on the organic speciation of Cu in our samples, the relatively large variation in DCu in the shelf waters in our research might be explained by changes in organic complexation. Frache et al. (2001) observed a release of particulate Cu from melting pack ice in the Ross Sea, which was followed by an increase in the dissolved phase 10 days later. The production of Cu binding organic ligands like thiols by algal blooms contribute to seasonal differences in DCu and might increase the solubility (Whitby et al., 2018) and thus aid in the partitioning from the particulate to the dissolved phase. It seems plausible that after opening up of the Ross Sea Polynya, the partitioning of Cu between the dissolved and particulate phase released from sea ice and the Ross Ice shelf face, was not yet at a steady state since primary productivity and thus ligand production were still increasing. Given that the eastern southnorth part of our Ross Sea Polynya transect was sampled early, it can be assumed that steady state had not yet been reached. The western north-south part of the Ross Sea Polynya was sampled later and these data showed considerably less variation and smooth mixed-type profiles, i.e., low concentrations at the surface and an increase near the bottom (Figures 4D, 5G, 6G), implying steady state had been reached by the time of sampling.

Sources and Sinks Related to Primary Productivity
All metals except DPb, had their lowest concentrations in AASW and concentrations increased with depth. Remineralization of organic material was probably an important source responsible for the increase (Orians et al., 1990;Bau et al., 1995;Hendry et al., 2008;Bown et al., 2011;Croot et al., 2011;Middag et al., 2013Middag et al., , 2020Bruland et al., 2014). The highest concentrations for all nutrients in the AASW and ML, both macro and micro, were measured at the stations sampled earlier in the season (St. [16][17][18] and those close to the Ross Ice shelf (St. 44,45) where primary productivity was low and indeed was just starting. Specifically, at these stations the nutrient-type trace metals Zn, Cd, and Co failed to show the classic nutrient type distribution, although DFe showed a slight surface minimum at these stations from early season depletion (Sedwick et al., 2011;Gerringa et al., 2015; Figure 5H). Our second visit to the position of St. 31 two weeks later (St. 114 for second occupation) showed the development of a nutrient type profile, with relatively depleted surface concentrations. The revisit also showed a surface water decrease for DTi, suggesting that Ti was most likely passively adsorbed onto settling organic particles (Dammshäuser et al., 2011). Although DY and DLa showed a nutrient type depth distribution (Figures 3N,O, 6L,M), they are not known to be micro-nutrients for primary producers and the nutrient type profile shape has been explained by passive adsorption (Nozaki et al., 1997;Nozaki, 2001). Also at St. 31, a nutrient type profile was observed in DLa and the surface DLa concentration was further depleted at the time of the revisit, indicating that adsorption onto organic matter indeed plays an important role in the cycling of La. In contrast, the revisit showed almost no change in DY, which is less susceptible to scavenging due to its smaller size compared to La (Shannon, 1976;Nozaki, 2001).
In addition to being limited by light, primary production in the Ross Sea is limited by DFe and possibly DMn availability (Wu et al., 2019). Most bio-active metals are not fully depleted in the Ross Sea, but notably, Ni is relatively high in AASW (average 6.71 nM vs. 7.19 in HSSW), consistent with other observations in the global oceans (Morel et al., 2014;Middag et al., 2020). However, macro-nutrients and other trace elements such as Zn, Co, Cd and Cu can play a role in influencing phytoplankton growth or determining phytoplankton community composition (Sunda and Huntsman, 2000;Bertrand et al., 2007;Hoppema et al., 2007;Alderkamp et al., 2019). We observed significant relationships between the three macro-nutrients and DZn, DCo, DCd and DCu in the Ross Sea (Supplementary Table S1). The relationships in Table 4 are restricted to these four trace metals because their biogeochemistry is not strongly influenced by environmental processes other than uptake and release upon mineralization. Indeed DZn, DCd, DCu, and DCo, show linear relationships with the macro-nutrients, reflecting most probably phytoplankton uptake (Supplementary Table S1, Figure 8). Even though the role of Cd in biological processes is still unclear (Lane and Morel, 2000;Xu and Morel, 2013), this element strongly correlates with the macro-nutrients. For example, DCd decreased in experiments containing natural phytoplankton while this did not occur in the same setup with only the dissolved seawater (Mellett et al., 2018). In our study, good correlations existed between Cd and both NO 3 − and PO 4 3− . When nutrient concentrations are influenced by phytoplankton growth, we expect the changes in concentrations to reflect phytoplankton elemental stoichiometry (Redfield et al., 1963;Sunda and Huntsman, 2000;Saito et al., 2010;Twining and Baines, 2013). However, mixing is also known to lead to relationships between nutrients that can be linear (Middag et al., 2018(Middag et al., , 2020. As observed in other ocean basins, the macro-nutrients NO 3 − and PO 4 3− in the Ross Sea correlate better with each other than with Si(OH) 4 (Supplementary Table S1, Figure 7). All macro-nutrient-relationships showed a  Löscher (1999); Ellwood (2008), Lane et al. (2009);Bown et al. (2011. *** Twining et al., 2004. Frontiers in Marine Science | www.frontiersin.org  Table S1), although the change in slope in the Cd-P relationship was small ( Figure 8B). All slopes, except in the Cd-P relationship, increased from AASW to deeper water masses, indicating processes affecting nutrient concentrations in the AASW, most probably biological uptake and release by decomposition, are different from those that affect the concentrations in the deeper water masses (Frew and Hunter, 1992;Xie et al., 2015;Saito et al., 2017;Middag et al., 2018Middag et al., , 2019Vance et al., 2019). In our results, DZn correlated best with PO 4 3− and NO 3 − , and although still significant, less so with Si(OH) 4 (Pearson correlations r = 0.88, 0.87, and 0.83; linear regressions R 2 = 0.75, 0.76, and 0.69, respectively) (Supplementary Table S1, Figure  8A). Also, in the relationships of Zn with the macro-nutrients, a clearly visible discontinuity at DZn concentrations above ∼4 nM exists, where the macro-nutrients remain relatively constant (between AASW and other water masses, at 3-4 nM DZn, see also above) but DZn and Si(OH) 4 increased. The discontinuity in the DZn and Si(OH) 4 relationship is less obvious (Supplementary Table S1). The upper part of the nutrient like profile of DZn is thus caused by biological uptake and release by degradation taking place in the AASW, deeper DZn in influenced by other processes.
Samples in the ML were assessed separately from the overall AASW as well (N = 73 for ML, N = 102 for AASW, Table 2 and Supplementary Table S1). DCo, DCd and DZn correlated well with all macro-nutrients, but best with PO 4 3− , with only DCu correlating best with Si(OH) 4 (Supplementary Table S1, Figure 8E). That is why for the relation with primary productivity in the following paragraphs, in both the overall AASW as well as the ML, we focus on uptake ratios relative to phosphorus, but for DCu also with respect to Si(OH) 4 (Supplementary Table S1).
Overall, the presence of discontinuities in the relationships between macro-nutrients and trace metals implies that local uptake and remineralization are not the only processes, determining the slopes of the relationships in the ML and AASW. Therefore, we used three approaches to assess the stoichiometry of phytoplankton uptake: (1) The slope of the linear regression in the ML and AASW for all stations.
(2) For the stations from which we have information on phytoplankton community composition (St. 20, 91, 101, and 114), we calculated deficits for each nutrient by subtracting the average ML concentrations from deep water average concentrations . For the latter, we used the average concentrations in HSSW, the deep water present at these stations ( Table 2) and assumed that the deficits were due to consumption. We divided deficits for each micro-nutrient by the deficit calculated for PO 4 3− , or in case of DCu, also by the deficit in Si(OH) 4 . Station 31 was added to this approach for a better comparison with approach 3, even though the phytoplankton composition was not known.
(3) For the re-occupied station (St. 31 and 114), by calculating the decrease in average nutrient concentrations within the ML over the two weeks between the successive occupations, and dividing the change in metal concentrations by the change in PO 4 3− or Si(OH) 4 concentration.
With the exception of DCo, the derived uptake ratios of the first two approaches compared relatively well with P-normalized stoichiometries reported previously (Saito et al., 2010;Bown et al., 2011); indicating that the approaches mostly give representative results ( Table 4). The correlations and average values in the ML did not differ significantly from those in the AASW. With the exception of St. 20, the second approach for specific stations resulted in uptake ratios that compare reasonably well with the results of approach 1 (based on all stations), only Zn:P was higher at St. 31 (approximately 7 mM M −1 close to 7.7 mM M −1 at St. 20 versus 2.9-4.4 mM M −1 of approach 1 and St. 91, 101, and 114, Table 4). The uptake ratios at St. 20, where the community consisted of only 26% diatoms, are distinctly different from the other stations and approach 1. Station 20 was  Table S1) between the parameters, in black for samples from AASW, in brown for samples outside AASW.
sampled in an earlier stage of the bloom compared to the other stations, in date of sampling but also in days after being ice free, and had lower POC (maximum POC concentrations = 47 µM at St. 20 versus 47, 58, and 52 µM at the other stations 31-114). Uptake ratios should not be affected by changes in biomass unless there is a coinciding change in community composition. Uptake ratios compensate for changes in biomass. For St. 20, the lower Si:P uptake ratio together with the higher Cu:Si, correspond with little Si(OH) 4 consumption due to the low contribution of diatoms to the overall community. In the Ross Sea, the N:P uptake ratio of diatoms is 9.52, whereas for P. antarctica the ratio is 19 (Arrigo et al., 2000, illustrating the influence of the phytoplankton population on macro-nutrients and this influence is now also evident for trace metal uptake ratios in the study region. The results of the third approach, using the change over time at the position of St. 31 and 114, compared less favorably with those of approaches 1 and 2. Except for Cd, the uptake ratios were lower when using approach 3 compared to both the other approaches and the literature values. Possibly, a local surface source of Zn, Co, and Cu contributed to the surface concentrations in between visits, with sea ice melting being the most probable source, since sea ice melting is not a Cd source (see section 4.2.3) or lateral transport.
Nevertheless, overall all three approaches compared reasonably well within the variation known from other studies and these current results provide an opportunity for additional insight into uptake ratios related to the phytoplankton community composition. According to Hendry et al. (2008), the plankton community influences the Cd:P ratio. Indeed, we do see that our Cd:P uptake ratio decreased significantly (Pearson correlation, p < 0.05) from 0.65 to 0.54 mM M −1 with increasing % diatoms (Table 4). Saito et al. (2010) suggested that diatoms take up more Co than P. antarctica, but in contrast, we find that the derived Co:P uptake ratio was higher at P. antarcticadominated St. 20 and lower at the diatom dominated St. 101. This discrepancy might be explained by differences in Fe stress that also influence metal:P uptake (Sunda and Huntsman, 2000;Cullen et al., 2003;Lane et al., 2009). Additionally, there is a large range in Co:P uptake ratios reported for various regions ranging between 23 µM M −1 from subantarctic waters reported by Dulaquais et al. (2014a), 38 µM M −1 in the Ross Sea, (Saito et al., 2010) to 44-49 µM M −1 north of the polar front (Bown et al., 2011) Our Co:P uptake ratio was 18 µM M −1 in the ML using approach 1 and 18-29 µM M −1 using approach 2, with generally lower ratios in diatom dominated stations (St. 91 and 101) consistent with the conclusion that diatoms need less Co to grow (Sunda and Huntsman, 1995), but no significant relationship between Co:P and the % diatoms existed.
As well established in other ocean regions (Bruland et al., 1991;Morel et al., 1994;Arrigo et al., 2000;Sunda and Huntsman, 2000;Hendry et al., 2008;Bown et al., 2017;Saito et al., 2017) and also in the Ross Sea, the Zn, Co and Cu cycles are influenced by phytoplankton. This is apparent from changes in the slopes and uptake ratios, which were possibly influenced by the phytoplankton composition, as shown by the significant relationship between the Cd:P uptake ratio and the % of diatoms present in the Ross Sea Polynya and the results from St. 20, the station with mostly P. antarctica. In AASW, Fe or Zn limitation is known to influence relative uptake ratios between the macro-and micro-nutrients (Sunda and Huntsman, 2000;Cullen et al., 2003;Cullen, 2006;Lane et al., 2009). DFe was shown to be limiting in the Ross Sea during our cruise , but DZn concentrations of 2.32 ± 1.37 nM in AASW (N = 100, DZn < 1 nM for N = 7) were probably not limiting phytoplankton growth. It is likely that the influence of Fe limitation on uptake ratios is different between P. antarctica and diatoms and that Fe limitation is more severe later in the season.

Ice Melt
In our data, indications for sea ice melt as a possible source were present for all metals, even for DPb, although it was not always apparent. The water column was well mixed at the stations sampled earlier in the season in the eastern Ross Sea and at stations near the Ross Ice Shelf, especially at St. 17, 44, and 45, obscuring whether sources were ice melt or the sediment. Salinity and dissolved metal concentrations did not co-vary, except for DCu, DY and DLa. These latter metals showed non-significant, positive relationships (data not shown), similar to salinity versus the macro-nutrients. If melt were a significant source, a negative relationship with salinity would exist, indicating that any evidence of sea ice melt release is probably hidden by processes of phytoplankton growth and mixing of surface and deep water masses.
For both DY and DLa, a higher concentration was observed in AASW outside the center of the polynya, where no WW was present (Figure 3). Since DY and DLa are not used by phytoplankton, a release from melting sea ice and/or the Ross Ice Shelf could explain the small surface maxima in AASW compared to the lower WW concentrations, as release from melt would occur predominantly in spring and early summer, our sampling period. Thereafter, the concentrations would decrease during summer and early fall by passive adsorption and sedimentation, thus causing the lower concentrations conserved in WW, as the WW concentrations should be more representative of concentrations in those seasons.
We do see indications for the melting Ross Ice Shelf as a source of Pb, as higher concentrations occurred at the southern end of the two transects. However, given that the concentrations were only slightly elevated and concentrations overall were close to the detection limit, no estimates of the magnitude of this source were attempted. Moreover, the DPb data from the first stations had to be discarded due to suspected Pb contamination as a decrease in concentrations was observed with every station sampled and no probable biogeochemical process for such a decrease could be identified. Most likely the Go-Flo samplers became cleaner with every deployment. However, since at the final stations of the expedition in the Ross Sea Polynya, concentrations increased again with decreasing distance from the ice shelf, we decided to incorporate the Pb data in this paper and suggest that in the Ross Sea, ice shelf melt indeed is a source for DPb. This source would be a combination of both naturally sourced Pb as well as anthropogenic Pb deposited on the ice sheet. Although concentrations of DPb were relatively low, this does not automatically mean that the DPb was solely derived from natural sources. Flegal et al. (1993) concluded, based on the isotopic composition of DPb, that DPb near the Antarctic Peninsula was for a significant part derived from industrial sources despite decreases in contamination of Antarctic ice over the last decades. According to Ndungu et al. (2016), 60-95% of Pb in the Amundsen Sea has a natural source. As far as we know, no information exists on the origin of DPb in the Ross Sea, but it seems reasonable to assume a similar contribution of a natural source for the Ross Sea.
The identification of sources of metals into AASW and quantification of such fluxes, for example the opening of the polynya, the timing of brine release, melting of the Ross Ice Shelf, or dust input, can be difficult because of immediate uptake by micro-organisms or sorption onto (biogenic) particles. An indirect indication of input from ice melt can be the disturbance or deviation from the relationship between metals with other nutrients, as shown schematically by Noble et al. (2008); their Figure 8), assuming the ice melt source supplies the nutrients in a different ratio than the biological stoichiometry. The correlations of DZn, DCd, and DCo with PO 4 3− (Figures 8A-C), reveal data points in AASW deviating from the regression where DZn, DCd, and DCo were high relative to PO 4 3− , and these were all from the stations 16-18 with very low phytoplankton growth. The highly deviating metal concentrations were probably due to input from melting ice ( Figure 8A as example: DZn > 4 nM where PO 4 3− = 1.5-2 µM). Additionally, samples from St. 59 in the polynya also deviated from the general regressions. Here, an explanation is not directly apparent, but the elevated concentrations might be due to release from detached floating sea ice. Deviations in the DCu:PO 4 3− correlation occurred also in St. 43 to 45, near the Ross Ice Shelf probably due to slow reaction kinetics of Cu. Frache et al. (2001) showed the importance of melting pack ice on DCu concentrations. As discussed above (section 4.2.1), the lack of steady state or equilibrium between release of Cu and Cu-binding by organic ligands or scavenging probably explains variation in DCu.
Another indication of ice melt as a source of metals are elevated concentrations in shelf water masses. DFe, DMn, DZn, DCo, DY, and DLa concentrations were elevated in LSSW and DCu, DY, and DLa concentrations were relatively elevated in ISW. But as noted earlier, the number of samples in these water masses was low with a rather large variability in concentrations, and therefore the concentration differences from other water masses were not significant ( Table 3). Concentrations of DFe, DMn, DNi, DY, and DLa were higher in HSSW.
Increased concentrations in HSSW may be influenced by the sediment. Overall, ice melt is a likely source of dissolved metals in the Ross Sea Polynya and isotopic compositions or other proxies would be a great help in future studies to constrain meteoric input.

Lateral and Vertical Input and Losses Due to Scavenging
Along the Ross Sea Polynya transect, St. 60-65 were sampled from the top of Pennell Bank (St. 60) and into Joides Trough (St. 61-65) to study lateral fluxes from banks as shown by Sherrell et al. (2015) for DMn (Supplementary Figure S3). We did not detect trends in concentrations that suggested lateral input from banks. However, DMn did decrease from 0.7 to 0.13 nM over a distance of 50 km from Franklin Island at depths between 100 and 400 m ( Figure 6I), indicating lateral transport of elevated concentrations of Mn from a point source. Here also lateral transport of DFe was estimated ( Figure 6H and Gerringa et al., 2015). Assuming lateral diffusion and subsequent dilution from a point source, as done for DFe, the decrease should be exponential. An exponential decrease occurred in the depth layer around 400 m (0.7 nM/50 km), although the fit was modest (R 2 = 0.36). For DFe, the exponential equation with distance best fit the data at approximately 300 m, whereas for DMn, an exponential decrease best fit the data at 400 m depth, where the scale length, a measure to express the influence of a point source, was 132 km. This implies that DMn decreased to 37% of the source concentration at 132 km distance from the source (Okubo, 1971). This results in an estimated flux at 100 km from Franklin Island of 2.3 * 10 −4 mol DMn per m 2 per day, compared to 3.4 * 10 −5 mol DFe per m 2 per day at the same depth (the best fit of the exponential decrease in DFe at 300 m attributed to a flux of 2.6 * 10 −5 mol DFe per m 2 per day; Gerringa et al., 2015). Apparently, the flux of Mn was about ten times higher than for Fe, but note that this is a coarse estimate as the assumptions are apparently not completely fulfilled as shown by the modest fit. Probably other processes disturbed the assumption of dilution from a point source, such as losses through scavenging, vertical fluxes from the sediment, or the BNL in deep waters and phytoplankton uptake in the surface layer.
In the relatively shallow waters of the Ross Sea, we saw no indications of scavenging given that depth profiles of dissolved metals below the MLD and/or AASW were either constant or increased with depth, as was also observed by Saito et al. (2010) for DCo. Scavenging determines the decreasing depth profiles of DMn in the deep oceans (Bruland et al., 2014), but here in the relatively shallow Ross Sea, DMn increased with depth (Figures 4F, 6I). Even DTi, which is very reactive and sensitive to scavenging, and therefore has a short residence time in seawater (Orians et al., 1990;Skabral, 2006;Dammshäuser et al., 2011), increased slightly with increasing depth with a slope of about 10 pM/100 m (Figures 4H, 6K). This increase was small and not always visible in depth profiles for individual stations. Moreover, there is little grazing in the Ross Sea and thus relatively few fecal pellets (Caron et al., 2000;Smith et al., 2000;Tagliabue and Arrigo, 2003). Probably the low scavenging is due to a combination of the shallow depth, few fecal pellets and rapid recycling (Martin et al., 2013;Rijkenberg et al., 2014;Laglera et al., 2017Laglera et al., , 2019, whereas the sediment source further obscures evidence of scavenging by increasing deep dissolved metal concentrations. The sediment as a source for metals is most evident in the BNL, where especially DFe and DMn, but also DCu, DCo, DNi, DTi, DY, DLa and the macro-nutrient Si(OH) 4 had significantly higher concentrations compared to HSSW and other deep water masses (Tables 2, 3, Figures 3, 4). Even a particle reactive element like Ti (Skabral, 2006;Dammshäuser et al., 2011) had higher dissolved concentrations in the BNL, although DTi remained the same in the BNL of St. 31 and 114 (Supplementary Figure S2). For DFe, the main supply to the ML proved to be the flux from the sediment, especially at shallow water depths with a BNL (Gerringa et al., 2015). Gerringa et al. (2015) calculated average vertical diffusive fluxes of 3.3 10 −8 mol m −2 day −1 for DFe, being positively correlated to DFe in the BNL and negatively correlated with the water depth. Since the average absolute difference in nM for DMn between ML and BNL and ML and HSSW is slightly higher (1.4 nM DMn vs. 1 nM DFe and 0.6 nM DMn vs. 0.3 nM DFe, respectively), the average upward flux of DMn will amount an average of 5 to 6.6 10 −8 mol m −2 day −1 . A rough estimate of the Fe flux necessary to sustain the bloom was 4-91 10 −8 mol m −2 day −1 (Gerringa et al., 2015) and since phytoplankton need less Mn per cell than Fe (approximately 20% according to Twining and Baines, 2013), the upward vertical diffusive flux of DMn would approximately meet the demand of the phytoplankton bloom observed during the cruise.
Relatively high concentrations of the macro-nutrients NO 3 − , PO 4 3− , and Si(OH) 4 , as well as DCd indicate the incoming mCDW as a source in the Ross Sea, but for most the concentrations are not significantly higher than in the shelf water masses (Figure 3, Table 3). It is possible that winter mixing of the water masses over the Ross Ice Shelf obscures the input of other metals by the mCDW. However, the CTD oxygen content, DMn, DY, and DLa were significantly lower in mCDW than in the LSSW and HSSW and thus mCDW was probably not a source for these metals. The position of our cruise track with only a few stations crossing the inflow of mCDW, together with the small differences in concentrations of most metals between mCDW and the shelf water masses, hardly allows an interpretation of the consequences of the interplay between the currents and bathymetry (Reddy and Arrigo, 2006) for the distribution of dissolved trace metals.

CONCLUSION
In the Ross Sea Polynya, the cycles of Zn, Co, and Cu are shaped by phytoplankton demand, as shown by depthdependent changes in correlation slopes and nutrient uptake ratios, which are probably influenced by phytoplankton community composition. Although the stoichiometric ratios of Zn, Cd, and Co relative to P were higher at a station dominated by P. antarctica, it is only for Cd, the element without a clear biological function, where a significant correlation between nutrient stoichiometry and the percentage of diatoms was observed.
The incoming mCDW had slightly higher concentrations of macro-nutrients and DCd compared to the shelf water masses and forms a source of these nutrients for the Ross Sea.
The stations sampled earlier in the season in the eastern Ross Sea, as well as some stations very close to the edge of the Ross Ice Shelf, were completely mixed from surface to bottom and did not show nutrient type profiles of the macro-and micronutrients. The input of Zn, Cd, Co, and Cu from ice melting was visible in elevated dissolved metal concentrations that deviated from correlations with the macro-nutrients. Sea ice melt is also a source of DLa and DY, possibly via dust deposition on the ice. Most likely, DLa and DY concentrations slowly decreased during summer due to passive adsorption, as reflected by the lower concentration in WW compared to AASW in spring and early summer.
Variable DCu concentrations near the Ross Ice Shelf point to a lack of steady state or equilibrium between release of DCu from melting sea ice and scavenging or binding by organic ligands. In the Ross Sea Polynya, however, scavenging does not influence dissolved metal concentrations noticeably, since no decrease with depth was observed for even the most scavenging-prone elements Mn and Ti.
As for DFe, sea ice melt releases DMn as can be concluded from elevated concentrations in shelf water masses compared to the mCDW, but the sediments are an additional important source for both. Concentrations of DFe and DMn, as well as DCo, DNi, DTi, DY and DLa were elevated in the BNL demonstrating the influence of the sediments. This led to elevated metal concentrations in HSSW that may be transported over the shelf into the circumpolar current, making the Ross Sea a source of these metals to the open ocean.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repositories and accession numbers can be found below: https://dataportal.nioz.nl/doi/10. 25850/nioz/7b.b.r and https://doi.org/10.25850/nioz/7b.b.g.