Distinct iron cycling in a Southern Ocean eddy

Mesoscale eddies are ubiquitous in the iron-limited Southern Ocean, controlling ocean-atmosphere exchange processes, however their influence on phytoplankton productivity remains unknown. Here we probed the biogeochemical cycling of iron (Fe) in a cold-core eddy. In-eddy surface dissolved Fe (dFe) concentrations and phytoplankton productivity were exceedingly low relative to external waters. In-eddy phytoplankton Fe-to-carbon uptake ratios were elevated 2–6 fold, indicating upregulated intracellular Fe acquisition resulting in a dFe residence time of ~1 day. Heavy dFe isotope values were measured for in-eddy surface waters highlighting extensive trafficking of dFe by cells. Below the euphotic zone, dFe isotope values were lighter and coincident with peaks in recycled nutrients and cell abundance, indicating enhanced microbially-mediated Fe recycling. Our measurements show that the isolated nature of Southern Ocean eddies can produce distinctly different Fe biogeochemistry compared to surrounding waters with cells upregulating iron uptake and using recycling processes to sustain themselves.

In "Biogeochemical metrics reveal distinctive but unusual iron cycling in the Southern Ocean eddies" Ellwood et al discuss iron measurements taken in a Southern Ocean eddy. They find Fe cycling characteristics in the eddy, a cyclone, that are distinct from surrounding waters. They conclude that the isolation of waters due to the eddy's dynamics allow for those distinct characteristics to occur, and more generally, that these finding will help to better understand Southern Ocean eddy chl/biomass anomalies.
The study presents -to my knowledge-a unique assessment of natural iron cycling within a Southern Ocean eddy, comparing it to measurements taken in surrounding waters. The study contributes to explaining mechanisms causing biogeochemical anomalies of Southern Ocean eddies. Thus, it presents a valuable addition to the field. The paper is well written. I have only few comments (see below), mainly concerning referencing of previous literature and why the eddy features characteristic Fe cycling. I recommend the paper for publication.
Minor comments: *Title: How about a shorter the title, sth like "Distinct iron cycling in a Southern Ocean eddy"? Also, this is a case study about one eddy-that is, I suggest to stay with the singular in the title (see also comment below on how representative the eddy is for other cyclones). *General: I understand that Fe cycling in the eddy is distinct compared to surrounding waters because the eddy is isolating waters; though, it is not entirely clear to me if the iron conditions of the eddy are typical for conditions south of the SAF where the eddy is originating from? That is, is "mere" advection of water, including its material properties, the main player in setting the distinct eddy iron characteristics (as suggested, e.g., in L70/71/76)? Or do the iron characteristics of the eddy further evolve during the ~1 month after it's detachment from the SAF (as suggested, e.g., in L96)? Also, you highlight the high Fe-to-carbon ratio and efficient recycling/short residence times of Fe in the eddycan you comment on the larger-scale potential biogeochemical and/or ecological implications? *L20 "productivity … low" and L28 "persistent productivity": Appears contracting at first glance. I suggest to rephrase to make clearer. E.g,. ", even though low, sustained productivity"? *L34/35 "a crucial role in...": Please provide references for this statement (e.g., references you use later on, see References comment below). *L37 "typically have closed circulation leading to biogeochemical properties": I am somewhat hesitant with "typically" as the majority of eddies does not appear to be efficient in trapping waters, see also, e.g., Wang et al, 2015; the cyclone observed here rather appears to represent an extreme case in that it features a rather high temperature anomaly (see also comment below); can you comment on this? *L46 "supply of Fe... usually via": Do you mean to say that the supply usually is provided via ironenriched deeper waters? If so I suggest sth like "supply of Fe usually is provided via upwelling and upward mixing of deeper iron enriched waters..."; also, lateral advection of iron by eddies may play a role, too, may it not, see e.g., Xiu et al, 2011, as an example from the north Pacific? *L59 "about 2C lower": This appears to be a distinct/intense eddy; see "typical"/mean SST anomalies of eddies, e.g., in Haumann et al, 2012, of (well) below 1C; the large temperature anomaly suggests that the eddy rather is an extreme case, likely special also in terms of its biogeochemistry, and not so much an eddy representative for most eddies (see also comment above); I am happy to be convinced otherwise, though. If possible, could you comment on how many of such distinct eddies occur in the region versus the number of weaker eddies with less pronounced physical/biogeochemical anomalies (e.g,. based on satellite SLA and SST/Chl)? *L77 "unique": Unique compared to what, e.g., "unique in this region" or so. Or delete the second part of the sentence -is it necessary here? *Fig1: I suggest to add SSH or SLA contours, a typical proxy to identify eddies. This manuscript presents new data on Fe concentrations and Fe isotopes in the Southern Ocean, and interprets these data in the context of a 1D model of isotope cycling and other biogeochemical data.
The manuscript does not present enough information about methods to judge whether the underlying data are correct. This is particularly worrisome because the information which the authors do present hints that some data may be incorrect. I urge the authors to engage in significant additional methods testing and/or present the results of any significant methods testing which has already been performed. My three most significant concerns regarding methods are below: 1) Blank for Fe isotopes. The authors state a methods blank of 0.4 ng for Fe isotopes, yet they use 200 uL of AG-MP1 resin for purification. AG-MP1 resin has quite a high resin blank, and multiple publications suggest resin blanks for 200 uL resin would be roughly an order of magnitude higher (e.g. Dauphas et al. Anal Chem 2004, 1 mL resin, 20 ng Fe blank; Conway et al Anal Chim Acta, 2013, 20 uL resin 0.3 ng Fe blank; and others). The process used here for determining blanks was not clearly explained, aside from a sentence about running a 50 mL sample, and then calculating "based on a 2 L sample". Wouldn't it have been more straightforward just to report the blank obtained for the 50 mL sample, as this represents at least a minimum blank for a 2L sample? In any case, if the authors are really able to provide a 10-fold reduction in blank compared to all previous studies using the same anion exchange resin, this should be highlighted and clearly described.
Even if the reported blank of 0.4 ng is correct, more information needs to be provided. For the lowest concentration samples, the blank would constitute about 30% of the entire Fe in the samples. Was any sort of blank correction done? In order to do such a blank correction, the authors would need to know not only the concentration of the blank (and it's variability) but also it's isotopic composition (and it's variability). How was this determined?
2)Fe uptake rates. Little information was provided about how Fe uptake rates were measured except that they were determined using 55Fe. However, such experiments are notoriously difficult to perform. Specifically, it is quite common in such experiments that some of the "uptake" just represents surface adsorption or precipitation, rather than true biological internalization. This would invalidate the author's conclusion about the recycling timescale of Fe beeing ~ 1 day. Was the added Fe bound to any sort of ligand? Do the authors have any way of knowing whether the Fe was actually translocated across the cell membrane, or just precipitated on the cell surface?
3)Fe concentrations. Surprisingly for a manuscript focused so heavily on Fe concentrations, the authors appear not to have described the methods they used to measure Fe concentrations. Typical blanks for Fe concentration by Seafast are 150 pM. Blanks for Fe concentration analysis based on large-volume stable isotope samples may be smaller, but they would suffer from the same lack of clear description as described for δ56Fe in point 1. What sort of ICPMS instrument was used to measure Fe concentrations? What was the blank for this method? How was blank and blank variability determined?
Reviewer #3 (Remarks to the Author): In this paper the authors present an analysis of the Fe chemistry of a Southern Ocean cold-core eddy. The analysis of Fe speciation and uptake rates is shown to infer that, while biomass and productivity in the eddy is low, the Fe uptake rate is high, suggesting rapid turnover of the Fe pool. The control of Fe availability on production in the S. Ocean has been extensively studied and this paper focusing on the role of mesoscale eddies adds to this story. The paper is well written and the data presented seems robust.
I felt however that the message of the paper was not well stated. The abstract suggests that a main finding is that fast Fe turnover sustains 'persistent production' even when Fe is low; however, the concluding remarks suggest that the main message is 'extreme Fe limitation is prevalent in summer and autumn'. This latter argument seems weaker than that suggested in the abstract (as we know Fe limits production in the S. Ocean). Can the authors relate the findings more to carbon and state/quantify the increase in production (albeit it low) as a result of fast recycling? This would strengthen the findings of the paper. Minor comments: A legend (similar to that on Figure 2) on figure 3 would help. *L20 "productivity … low" and L28 "persistent productivity": Appears contracting at first glance. I suggest to rephrase to make clearer. E.g,. ", even though low, sustained productivity"?
Rephrased to "with cells upregulating iron uptake and using recycling processes to sustain themselves" see line28 *L34/35 "a crucial role in...": Please provide references for this statement (e.g., references you use later on, see References comment below). *L37 "typically have closed circulation leading to biogeochemical properties": I am somewhat hesitant with "typically" as the majority of eddies does not appear to be efficient in trapping waters, see also, e.g., Wang et al, 2015; the cyclone observed here rather appears to represent an extreme case in that it features a rather high temperature anomaly (see also comment below); can you comment on this?
The system that we studied is a akin to what Frenger et al. (2018) classified as a trapping or monopole eddy. We have rephrased lines section to mention this in the text. See line 37 *L46 "supply of Fe... usually via": Do you mean to say that the supply usually is provided via ironenriched deeper waters? If so I suggest sth like "supply of Fe usually is provided via upwelling and upward mixing of deeper iron enriched waters..."; also, lateral advection of iron by eddies may play a role, too, may it not, see e.g., Xiu et al, 2011, as an example from the north Pacific?
Eddies may play a role in nutrient and iron vertical and lateral supply depending on where they form. The warm core Haida eddy referred to in the Xiu et al. (2011) study transported warmer, fresher, and nutrient-and iron-enriched water offshore from the British Columbian coast. For our cold core eddy we observed elevated nutrient levels relative to the surrounding waters, but these levels are lower than those in the waters where the eddy likely formed. The consumption of nutrients within the eddy during its transit may well explain why dissolved iron is lower than in the surrounding waters. We have mentioned nutrient consumption on lines 79 to 86 *L59 "about 2C lower": This appears to be a distinct/intense eddy; see "typical"/mean SST anomalies of eddies, e.g., in Haumann et al, 2012, of (well) below 1C; the large temperature anomaly suggests that the eddy rather is an extreme case, likely special also in terms of its biogeochemistry, and not so much an eddy representative for most eddies (see also comment above); I am happy to be convinced otherwise, though. If possible, could you comment on how many of such distinct eddies occur in the region versus the number of weaker eddies with less pronounced physical/biogeochemical anomalies (e.g,. based on satellite SLA and SST/Chl)?
We have amended to the text to account for this point. Our was typical in size and life span but was more intense in terms of rotation speed and temperature gradients with respect surrounding waters. See lines 58-64.
*L77 "unique": Unique compared to what, e.g., "unique in this region" or so. Or delete the second part of the sentence -is it necessary here? Corrected.
*Fig1: I suggest to add SSH or SLA contours, a typical proxy to identify eddies.
We have included an extra panel to Figure 1 showing the sea level anomaly (SLA) for the region. The panel highlights that the eddy is characterised by lower SLA compared to surrounding waters. For decided to leave in the temperature contours for panels 1b and 1c as the nice define the edge of the eddy. Amended. Line 55 *L163-167: I suggest to add a sentence here, embedding these conclusions/hypotheses on light/iron limitation in eddies in previous works; do the statements/hypotheses (dis)agree with findings/hypotheses, e.g., of the observational works of Dawson  This manuscript presents new data on Fe concentrations and Fe isotopes in the Southern Ocean, and interprets these data in the context of a 1D model of isotope cycling and other biogeochemical data.
The manuscript does not present enough information about methods to judge whether the underlying data are correct. This is particularly worrisome because the information which the authors do present hints that some data may be incorrect. I urge the authors to engage in significant additional methods testing and/or present the results of any significant methods testing which has already been performed. My three most significant concerns regarding methods are below: Thank you for your constructive comments. We have addressed you concerns below 1) Blank for Fe isotopes. The authors state a methods blank of 0.4 ng for Fe isotopes, yet they use 200 uL of AG-MP1 resin for purification. AG-MP1 resin has quite a high resin blank, and multiple publications suggest resin blanks for 200 uL resin would be roughly an order of magnitude higher (e.g. Dauphas et al. Anal Chem 2004, 1 mL resin, 20 ng Fe blank; Conway et al Anal Chim Acta, 2013, 20 uL resin 0.3 ng Fe blank; and others). The process used here for determining blanks was not clearly explained, aside from a sentence about running a 50 mL sample, and then calculating "based on a 2 L sample". Wouldn't it have been more straightforward just to report the blank obtained for the 50 mL sample, as this represents at least a minimum blank for a 2L sample? In any case, if the authors are really able to provide a 10-fold reduction in blank compared to all previous studies using the same anion exchange resin, this should be highlighted and clearly described.
We have increased the detail within the relevant Section, and have spelled out the blanks we obtained for the anion-exchange separation component of the method, and for the full iron extraction and separation process. For these we get blanks of 0.39 ± 0.34 ng (n = 4) and 0.40 ± 0.32 ng (n = 5), respectively. We have also detailed how we precleaned our AG-MP1 resin and how the columns were stored between use. At face value, most of the blank can be ascribed to the anion exchange process. See lines 227 to 306 With respect to the earlier work of Dauphas et al. (2004), their blank results were for the AG1-X8 and AG50W-X4 resins, not the AG-MP1 resin.
Even if the reported blank of 0.4 ng is correct, more information needs to be provided. For the lowest concentration samples, the blank would constitute about 30% of the entire Fe in the samples. Was any sort of blank correction done? In order to do such a blank correction, the authors would need to know not only the concentration of the blank (and it's variability) but also it's isotopic composition (and it's variability). How was this determined?
We have included more details about how we dealt with blank contribution within samples. Because we were not able to determine the iron isotope composition of the blank we chose not to correct the results. For all our results, a blank correction of the isotope data would have resulted in values that are statistically indistinguishable from the results presented. For example, for the 70m sample (2 ng of Fe) from the cold core eddy, we obtained a dissolved Fe isotope value of 1.10 ± 0.43 ‰ (2.SE). Correcting this 70m value with a 20% blank contribution (and assuming a lithogenic isotope value for the blank of 0.1‰) produces a result of 1.35 ± 0.43 ‰. The difference between this and the measured value is not statistically significant. Likewise, if we take the dissolved iron isotope value for the 300 m depth sample (total of 45 ng Fe) from the cold core eddy, we obtain a blank corrected isotope value of 0.64 ± 0.05 ‰ which is indistinguishable from its measured isotope value of 0.64 ± 0.05 ‰.
Because we chose not to correct our iron isotope results we also chose not correct our concentration results. For the majority of the samples, this had no significant bearing on the final results. For a few this correction will be significant and thus they represent upper concentration bound.

See lines 301-306 and 326-327
2)Fe uptake rates. Little information was provided about how Fe uptake rates were measured except that they were determined using 55Fe. However, such experiments are notoriously difficult to perform. Specifically, it is quite common in such experiments that some of the "uptake" just represents surface adsorption or precipitation, rather than true biological internalization. This would invalidate the author's conclusion about the recycling timescale of Fe beeing ~ 1 day. Was the added Fe bound to any sort of ligand? Do the authors have any way of knowing whether the Fe was actually translocated across the cell membrane, or just precipitated on the cell surface?
Earlier field studies used 55 Fe concentrations one to two orders of magnitude higher (i.e., 2 -20 nmol L -1 ) than we used in this study Ti(III) reagent has shown to be an efficient washing agent as it relies on both redox potential and strong binding affinity for Fe(III) to first reduce 55 Fe oxyhydroxides that may precipitate on the cell surface, and then subsequently chelates the 55 Fe liberated into the dissolved phase. The extracellular 55 Fe is then removed from the filters with multiple seawater rinses.
Combining data from all filter fractions and locations for the present study, Fe uptake rates and Fe:C uptake ratios were 2.65 ± 0.14 (±SE, n = 54) times higher, and C uptake rates 1.07 ± 0.03 times higher in unwashed samples compared to Ti(III) reagent-washed samples, indicating that: 1) intracellular Fe accounted for ~38% of total Fe 'uptake'; and 2) the Ti(III) EDTA -citrate reagent did not damage cells appreciably, as indicated by the good agreement in the C uptake rates between washed and unwashed filters.
We note also that the authors have a proven, >15 year, track record conducting these dual-label radiotracer experiments in the field, and have published extensively on how 55 Fe concentrations, 55 Fe chelation, and the use of the Ti reagent influence Fe:C uptake rates

See lines 223 to 251
3)Fe concentrations. Surprisingly for a manuscript focused so heavily on Fe concentrations, the authors appear not to have described the methods they used to measure Fe concentrations. Typical blanks for Fe concentration by Seafast are 150 pM. Blanks for Fe concentration analysis based on large-volume stable isotope samples may be smaller, but they would suffer from the same lack of clear description as described for δ56Fe in point 1. What sort of ICPMS instrument was used to measure Fe concentrations? What was the blank for this method? How was blank and blank variability determined?
In the Methods section, we now present the blanks in two forms: the blank associated with anion-exchange separation component of the method, and for the full iron extraction and separation from each sample. To determine, the dissolved iron concentration we made use of the amount of double spike used to determine the iron isotope composition of each sample. Specifically, we state that "Dissolved iron concentration for each sample is calculated from sample volume and the amount double spike added to the sample. This calculation is based on isotope dilution using the known proportion of 58Fe in the 57Fe-58Fe double spike. Note that the dFe concentrations presented here were not blank corrected, thus, they represent an upper concentration bound." The instrument used to make iron isotope measurements was a ThermoScientific NeptunePlus which is mentioned in the methods section.

See lines 324 to 327
Reviewer #3 (Remarks to the Author): In this paper the authors present an analysis of the Fe chemistry of a Southern Ocean cold-core eddy. The analysis of Fe speciation and uptake rates is shown to infer that, while biomass and productivity in the eddy is low, the Fe uptake rate is high, suggesting rapid turnover of the Fe pool. The control of Fe availability on production in the S. Ocean has been extensively studied and this paper focusing on the role of mesoscale eddies adds to this story. The paper is well written and the data presented seems robust.
Thank you for your positive and constructive comments.
I felt however that the message of the paper was not well stated. The abstract suggests that a main finding is that fast Fe turnover sustains 'persistent production' even when Fe is low; however, the concluding remarks suggest that the main message is 'extreme Fe limitation is prevalent in summer and autumn'. This latter argument seems weaker than that suggested in the abstract (as we know Fe limits production in the S. Ocean). Can the authors relate the findings more to carbon and state/quantify the increase in production (albeit it low) as a result of fast recycling? This would strengthen the findings of the paper. 210 pmol kg -1 , thus these low in-eddy dFe values appear to result from in-eddy processes (Table 1). 117 Outside the eddy, dFe concentrations were 35 to 68 pmol kg -1 and 57 to 210 pmol kg -1 for the SAZ 118 and SOTS stations, respectively (Figure 1), consistent with measurements for this region and the 119 Southern Ocean generally (Table 1)  Our results raise the question: How can dFe be consumed to such low concentrations by the in-eddy 125 phytoplankton community, compared to external waters? We explored this question by measuring 126 the rate of Fe uptake by cells inside and outside of the eddy. In-eddy Fe uptake rates were higher at 127 15 m (80% incident irradiance) compared to rates outside the eddy at the two reference sites ( Figure  128 2 and S2). With depth, in-eddy Fe uptake rates decreased, but they were always higher than rates 129 measured for the SAZ site ( Figure 2). In-eddy carbon-normalised Fe uptake rates (expressed as a Fe:C 130 ratio) for the resident phytoplankton community were approximately 4-fold higher compared to the 131 two references sites and historical measurements for the region (Table 1) This short residence time indicates that Fe is being heavily trafficked within the euphotic zone 155 between the dissolved pool and the microbial community. The increased importance of iron 156 recycling favours smaller phytoplankton cells, which is reflected in the cell abundances, the size-157 fractionated iron uptake and the Fe:C ratio datasets ( Figures S3 and S4). 158 The elevated in-eddy Fe:C uptake ratios also raise the question as to how phytoplankton are 159 enhancing dFe uptake. Enhanced dFe uptake can occur through a combination of processes 14 , 160 including increased production of Fe transporters on the surface of cells, a reduction in cell size, the 161 production of Fe binding ligands, and the use of Fe(III) reductase proteins to enhance Fe(II) 162 production and hence the acquisition of Fe from organic complexes. We used the isotopic 163 composition of dFe and pFe to probe Fe uptake within the eddy. (Hitachi-Hitec, Japan) at a flow rate of 2 mL min -1 . Samples were rinsed with 4 mL of ammonium 368 acetate buffer solution (1% w w -1 ) followed by elution with 4 mL of 1 mol L -1 nitric acid. Samples 369 were evaporated to dryness and redissolved with 0.5 mL of 6 mol L -1 hydrochloric acid containing 370 H2O2. Samples were further purified using an anion exchange procedure similar to that described by 371 Poitrasson and Freydier 40 . Precleaning of the AG-MP1 resin involved rinsing with methanol, multiple 372 washes with 6 mol L -1 HCl and 0.5 mol L -1 HNO3 before storage in dilute HNO3. When required ~200 373 µL columns filled with the precleaned anion exchange resin AG-MP1 (Bio-Rad), conditioned by 374 washing with 0.5 mol L -1 HCl, 0.5 mol L -1 HNO3, Milli-Q water and finally 6 mol L -1 HCl before use. 375 Between use, columns were stored filled in 2% w w -1 HNO3. Columns were typically used 5-8 times 376 before being refilled with new precleaned resin. After sampling loading, salts and other elements not 377 of interest were eluted from the column by passing 3x 1 mL of 6 mol L -1 hydrochloric acid. Iron was 378 eluted with 3x 1 mL of 0.5 mol L -1 hydrochloric acid and evaporated to dryness. Samples were 379 redissolved in either 0.30 or 0.35 mL of 2% (w w -1 ) nitric acid. The blank associated with the anion 380 exchange separation was 0.39 ± 0.34 ng (n = 4). Procedural concentration blanks for the whole 381 process were determined by passing small volumes (~50 mL) of an in-house seawater standard with 382 a concentration 0.78 ± 0.08 nmol kg -1 over the Nobias PA Chelate PA1L resin and then through the 383 whole elemental and Fe isotope separation procedure. The dissolved iron concentration for this 384 smaller seawater was then scaled to 2L thus allowing us to estimate the blank associated with the 385 buffering of the sample, passing it over the Nobias PA Chelate PA1L resin and then over the anion 386 exchange columns. The blank associated with this test was determined to be 0.40 ± 0.32 ng (n = 5). 387 Note that we were not able to determine the isotope composition of the blank associated with the 388 extraction and processing procedure, so the isotope values presented in have not been blank 389 corrected. For dissolved samples, the total amount of Fe analysed ranged between 2 and 63 ng, thus 390 the contribution of the blank to the lowest concentration samples could have been between as 391 much 20 ± 17% of the lowest δ 56 Fediss signal, i.e. for samples collected from the upper water column 392 within the CCE. 393 Iron isotopes were determined using a Neptune Plus multi-collector ICPMS (ThermoScientific) with 394 an APEX-IR introduction system (ESI, USA) and with X-type skimmer cones. Samples were measured 395 in high-resolution mode with 54 Cr interference correction on 54 Fe and 58 Ni interference correction on 396 58 Fe. Fe isotope ratios ( 56 Fe/ 54 Fe) ratios are reported in delta notation (‰) relative to the IRMM-014 397 The potential processes that influence the distribution and isotope fractionation of dFe and pFe 452 were explored using a 1D model ( Figure S7). The rationale for using this 1D model is to explore the 453 relative influence (and interplay) of processes such as phytoplankton utilisation of Fe, its 454 complexation to natural organic ligands, its regeneration from sinking organic matter and the role of   I can't shake the feeling that the Fe isotopes data could possibly be incorrect. Seawater Fe isotopes are quite challenging to measure, and the authors are presenting d56Fe for some of the lowest-Fe (and thus most challenging) waters ever reported. Also, they present a blank for purification of Fe on AG-MP1 resin which is lower than ever reported previously, by a factor of something like 5-10 by my rough estimation, without any obvious reason why their blank should be so much lower than other studies. Also, their reported blank of 0.4 +/-0.32 ng Fe suggests great variability in this value, and thus uncertainty in d56Fe.
The key "gold standard" experiment which would allay my analytical concerns would be to remove Fe from seawater, then dope back in a small amount of non-crustal Fe standard (e.g. 50 pM Fe which is either +2 permil or -2 permil), then extract that new standard Fe and show that the correct d56Fe value is obtained. Successfully performing such an experiment would be very convincing, and indeed such experiments are often presented by labs which are developing new seawater metal isotope methods. However, I recognize that, depending on whether they are still set up to measure Fe isotopes routinely, doing such experiments could require a fair amount of work, and it seems unfair to hold up publication of a paper until a whole new round of methods-development can be undertaken.
Alternatively, perhaps the authors could more honestly discuss the incredible challenges of measuring Fe at such low concentrations, and acknowledge that they have not yet done the "gold standard" experiment of doping an isotope standard back into Fe-free seawater (as described above). They could also provide even more information about issues such as the variability in their blank. Then they could discuss the reasons why they still believe that their data are sound, and reasons why any expected small errors in their measurement should not undermine their basic conclusions.
I appreciate that the authors have responded in a detailed fashion to my earlier concerns about methods. They have made a convincing argument that their Fe radioisotope uptake experiments and Fe concentration data are valid. And if the reported blanks and other methods details are taken at face value, then their d56Fe data should also be correct. The interpretation of the data is reasonable and the conclusions are significant.
I support eventual publication of this manuscript, and I don't wish to ask for an insurmountable amount of additional effort. If the authors are able to perform the "gold standard" experiment, that would greatly strengthen this manuscript, and it would clearly establish them as a lab capable of measuring Fe isotopes even in low-Fe Southern Ocean waters, which in turn would open up exciting new areas of research. Alternatively, I hope that an extraordinarily thorough and honest discussion of the strengths, and possible weaknesses, of their d56Fe methods can be included, as a step towards the eventual goal of establishing unequivocally how Fe isotopes cycle in the Southern Ocean. Best,

Seth John
Reviewer #3 (Remarks to the Author): I am happy the authors have addressed my comments.

Dear Dr Frischkorn,
We are grateful to all three reviewers for their comments on our manuscript. Below is our response to points raised by the reviewres.

Reviewers' comments:
Reviewer #1 (Remarks to the Author): Thank you for addressing my comments. From my (non-iron but eddy-focused) perspective, I am fine with the publication of the manuscript as it is.
Grammar: * L37ff: The sentence "These eddies can have closed circulation thus 'trapping' the biogeochemical properties of these water such as nutrients, chlorophyll and particle concentrations differ compared to external waters. " needs correction.
Response: We have revised this sentence so it now reads "These eddies can have closed circulation thus 'trapping' ref 9 the biogeochemical properties of these features such that nutrient, chlorophyll and particle concentrations can be distinct relative to those in the surrounding waters2,9-11" Reviewer #2 (Remarks to the Author): I can't shake the feeling that the Fe isotopes data could possibly be incorrect. Seawater Fe isotopes are quite challenging to measure, and the authors are presenting d56Fe for some of the lowest-Fe (and thus most challenging) waters ever reported. Also, they present a blank for purification of Fe on AG-MP1 resin which is lower than ever reported previously, by a factor of something like 5-10 by my rough estimation, without any obvious reason why their blank should be so much lower than other studies. Also, their reported blank of 0.4 +/-0.32 ng Fe suggests great variability in this value, and thus uncertainty in d56Fe.
Response: In 2019 I (Michael Ellwood) published a paper describing iron isotope transformations in Lake Cadagno Switzerland . In that study, the anion exchange resin AG-MP1 was also utilised to separate iron from other ions for a freshwater study for a Swiss mountain lake. This work was undertaken in the labs at ETH Zurich and utilised their chemicals and their resin. The overall dissolved iron procedural blank of that study was 0.6 ± 0.5 ng , which included evaporating samples to dryness and passing the samples over the AG-MP1 resin. The blank from that study is comparable to the blanks results obtained in the present study (0.40 ± 0.32 ng) where column separation of iron was undertaken at the ANU using a different batch on AG-MP1 resin to that used at ETH. Subsequent to the analysis of the 2016 samples (i.e., results featured in our manuscript), we have managed to obtain the same, and in some instances lower, blank values for the AG-MP1 resin for processing other iron isotope samples. Thus, in the present study we present a blank that is comparable to that reported previously, and which has subsequently been repeatable (and surpassed) at the ANU.
The key "gold standard" experiment which would allay my analytical concerns would be to remove Fe from seawater, then dope back in a small amount of non-crustal Fe standard (e.g. 50 pM Fe which is either +2 permil or -2 permil), then extract that new standard Fe and show that the correct d56Fe value is obtained. Successfully performing such an experiment would be very convincing, and indeed such experiments are often presented by labs which are developing new seawater metal isotope methods. However, I recognize that, depending on whether they are still set up to measure Fe isotopes routinely, doing such experiments could require a fair amount of work, and it seems unfair to hold up publication of a paper until a whole new round of methods-development can be undertaken.
Response: Prior to the analysis of the samples from the 2016 voyage , we did indeed test that the ANU method was producing high-quality results. This was done by undertaking an inter-calibration exercise (perhaps the "platinum standard" as each analysis is completely independent) with Tim Conway and Matthies Sieber at ETH. In this inter-calibration exercise, samples from the GEOTRACES crossover station at 32.5 o S, 170 o W for the GP13 were analysed by the ETH group. The ANU results were also compared to iron isotope results generated by the ETH for samples collected GP19 voyage at 32.5 o S, 170 o W crossover station. The results from the two groups showed that were no systematic differences between methods used by the two groups -see the figure below. The results from this inter-calibration exercise were also presented at the 2018 Ocean Sciences meeting (Conway et al., 2018), and Conway et al., are in the process of preparing a manuscript detailing the results from this inter-calibration exercise.
Prior to developing the double spike method at the ANU, the ANU group also participated in the GEOTRACES inter-calibration exercise where groups were able to analyse samples collected from the BATS site near Bermuda. The results produced by the ANU group using the standard-samplestandard bracketing technique were comparable to the other groups undertaking iron isotope analysis (see the table below). This work was published by Boyle et al. (2012).
Overall, the ANU double spike method produces iron isotope results that are comparable to other field-leading interantional groups working in the space. Table 2   . The iron isotope results are for samples collected and analysed by the ANU group, samples the ANU group shared with the ETH group and samples collected on GP19 and analysed by the ETH group. One sample is highlighted as an outlier. The trends seen in the GP13 and GP19 profiles are consistent with each other and there is no systematic offset in the datasets, highlighting the quality of both the ANU and ETH methods.
Alternatively, perhaps the authors could more honestly discuss the incredible challenges of measuring Fe at such low concentrations, and acknowledge that they have not yet done the "gold standard" experiment of doping an isotope standard back into Fe-free seawater (as described above). They could also provide even more information about issues such as the variability in their blank. Then they could discuss the reasons why they still believe that their data are sound, and reasons why any expected small errors in their measurement should not undermine their basic conclusions.
Response: We have included the following sentences in the Methods section of the manuscript to highlight the quality of the iron isotope method. "The performance of the iron isotope method was also assessed through an inter-calibration exercise for samples from the GEOTRACES GP13 and GP19 campaigns at a crossover station located at 30°S; 170°W. The iron isotope results from the exercise were comparable -i.e., the trends seen in the GP13 and GP19 profiles are consistent with each other 43 ." I appreciate that the authors have responded in a detailed fashion to my earlier concerns about methods. They have made a convincing argument that their Fe radioisotope uptake experiments and Fe concentration data are valid. And if the reported blanks and other methods details are taken at face value, then their d56Fe data should also be correct. The interpretation of the data is reasonable and the conclusions are significant.
Response: Thank you I support eventual publication of this manuscript, and I don't wish to ask for an insurmountable amount of additional effort. If the authors are able to perform the "gold standard" experiment, that would greatly strengthen this manuscript, and it would clearly establish them as a lab capable of measuring Fe isotopes even in low-Fe Southern Ocean waters, which in turn would open up exciting new areas of research. Alternatively, I hope that an extraordinarily thorough and honest discussion of the strengths, and possible weaknesses, of their d56Fe methods can be included, as a step towards the eventual goal of establishing unequivocally how Fe isotopes cycle in the Southern Ocean.
We have added the following sentence to the manuscript and point the reader to additional figure (S10) in the support materials highlighting the quality of the data. "The performance of the Fe isotope method was also assessed through an intercalibration exercise for samples from the GP13 and GP19 campiagns at a crossover station located at 30°S; 170°W. The iron isotope results from the exercise were comparable -i.e., the trends seen in the GP13 and GP19 profiles are consistent with each other ( Figure S10

Thank you
For my last review, I was encouraged by the editor to present a complete and honest description of my concerns about the potential for inaccurate data when pushing Fe isotopes methods to report data on lower-Fe samples than previously reported. I suggested that if the authors didn't want to perform more methods development, then they might: "honestly discuss the incredible challenges of measuring Fe at such low concentrations, and acknowledge that they have not yet done the "gold standard" experiment of doping an isotope standard back into Fe-free seawater (as described above). They could also provide even more information about issues such as the variability in their blank. Then they could discuss the reasons why they still believe that their data are sound, and reasons why any expected small errors in their measurement should not undermine their basic conclusions." In response they added two sentences stating that their methods perform well in intercalibration. In not adding more detail, I think the authors are missing an opportunity to move the field forwards in terms of understanding how far we can push Fe isotope measurements for low-Fe waters, which is extraordinarily important considering that such HNLC waters are exactly where Fe is a limiting nutrient. I don't mean to imply that the measurements presented here are somehow lagging behind the data quality of other labs (indeed their responses demonstrate that their data matches that of other top labs, see below), but just to point out that they are reporting d56Fe for waters with roughly an order of magnitude less Fe than anything else previously reported. In my opinion, discussing the challenges and successes of such measurements is an important and useful part of such a boundarypushing effort.
While the authors have provided a great deal of additional information, which I appreciate, it doesn't quite get to the heart of my concern. The authors have demonstrated that they can perform Fe isotope analyses with skill just as high as other international labs. And they have demonstrated through intercalibration that they are able to accurately measure d56Fe on relatively highconcentration (>~0.2 nM) samples. But I'm still not totally convinced that they (or anybody!) can accurately measure d56Fe on samples with tens-of-picomolar concentrations of Fe. All of the intercomparison exercises and plots they present deal with samples containing several hundred picomolar Fe. I would argue that the "platinum standard" intercomparison exercises are appropriate for showing that their methods are useful for hundreds-of-picomolar Fe, but that the "gold standard" re-doping experiments are really the only way to show whether or not it is possible to accurately measure d56Fe on tens-of-picomolar Fe. Of course, realizing the difficulty of such experiments, I also suggested that they might provide more detail in the manuscript about their method's possible strengths and weaknesses. With a little more detail, I think this manuscript will not only contribute the main points of their paper, but also contribute to an understanding of these meethods.
All that said, I do realize that we are now discussing tiny details about the discussion of methods in this manuscript. This is not central to the overall message and value of the work, which I appreciate, and I do honestly hope it can be published soon.

Seth
Dear Dr Frischkorn, Below in blue text is our rejoinder to comments raised by reviewer 2.

Regards Michael
Reviewer #2 (Remarks to the Author): For my last review, I was encouraged by the editor to present a complete and honest description of my concerns about the potential for inaccurate data when pushing Fe isotopes methods to report data on lower-Fe samples than previously reported. I suggested that if the authors didn't want to perform more methods development, then they might: "honestly discuss the incredible challenges of measuring Fe at such low concentrations, and acknowledge that they have not yet done the "gold standard" experiment of doping an isotope standard back into Fe-free seawater (as described above). They could also provide even more information about issues such as the variability in their blank. Then they could discuss the reasons why they still believe that their data are sound, and reasons why any expected small errors in their measurement should not undermine their basic conclusions." In response they added two sentences stating that their methods perform well in intercalibration. In not adding more detail, I think the authors are missing an opportunity to move the field forwards in terms of understanding how far we can push Fe isotope measurements for low-Fe waters, which is extraordinarily important considering that such HNLC waters are exactly where Fe is a limiting nutrient. I don't mean to imply that the measurements presented here are somehow lagging behind the data quality of other labs (indeed their responses demonstrate that their data matches that of other top labs, see below), but just to point out that they are reporting d56Fe for waters with roughly an order of magnitude less Fe than anything else previously reported. In my opinion, discussing the challenges and successes of such measurements is an important and useful part of such a boundary-pushing effort.
In response, we have added a paragraph to the manuscript highlighting the challenges to making iron isotope measurements on low concentration samples. see lines 335 to 365.
"As with all open ocean seawater work, during the collection and processing of samples contamination can hinder the production of accurate and meaningful data. The added challenge for Fe isotope studies, particularly for low concentration systems such as the Southern Ocean, is obtaining enough material for isotope analysis. For the result presented here, the dFe processing blank associated represents as much 20 ± 17% of the concentration and the isotope signal. While concentration uncertainties are highest for shallow samples collected in the CCE, the structure of the dFe concentration versus depth profile for this station, and indeed the other two stations, are oceanographically consistent, i.e. they have low surface water concentrations that increase with depth45. In a companion study, dissolved zinc concentration and zinc isotope results obtained from the same samples showed no indication of trace metal contamination associated with sample collection and processing46. For the dFe isotope results, there is also the added challenge of obtaining enough material for isotope analysis. Here we optimised the isotopic measurement of dFe by reducing the volume of each sample presented for analysis (0.3 to 0.35 mL) thereby upping its concentration to reduce errors associated with instrument noise41,47. We also utilised a samplespike ratio of ~1.6 (spike 57Fe-58Fe ratio = 1.05) such that counting errors are minimised for 56Fe, 57Fe, and 58Fe. Even with these steps, the influence of instrument noise increased for low concentration Fe samples ( Figure S6). While the uncertainty window around these measurements is larger than that for samples with a higher dFe concentration, the upper water column variations for δ56Fediss between 15 and 150 m are statistically distinct and oceanographically consistent. The enrichment of δ56Fediss within the euphotic zone is consistent with measurements made at 32.5°S, 150°W ( Figure S10) and other recent measurements for low dFe concentration waters of the Southern Ocean48. Likewise, the decline in δ56Fediss values below the euphotic zone is consistent with measurements made at 32.5°S, 150°W ( Figure S10), although one should be mindful that this station is outside of the Southern Ocean such that the biological community leading to variation in δ56Fediss is likely to be different." While the authors have provided a great deal of additional information, which I appreciate, it doesn't quite get to the heart of my concern. The authors have demonstrated that they can perform Fe isotope analyses with skill just as high as other international labs. And they have demonstrated through intercalibration that they are able to accurately measure d56Fe on relatively highconcentration (>~0.2 nM) samples. But I'm still not totally convinced that they (or anybody!) can accurately measure d56Fe on samples with tens-of-picomolar concentrations of Fe. All of the intercomparison exercises and plots they present deal with samples containing several hundred picomolar Fe. I would argue that the "platinum standard" intercomparison exercises are appropriate for showing that their methods are useful for hundreds-of-picomolar Fe, but that the "gold standard" re-doping experiments are really the only way to show whether or not it is possible to accurately measure d56Fe on tens-of-picomolar Fe. Of course, realizing the difficulty of such experiments, I also suggested that they might provide more detail in the manuscript about their method's possible strengths and weaknesses. With a little more detail, I think this manuscript will not only contribute the main points of their paper, but also contribute to an understanding of these meethods.
We would like to point out that the intercalibration study actually covered the following concentration range 0.017 and 0.72 nmol kg -1 . Indeed for samples shallower than 150 m were below 100 pmol kg -1 with the lowest being 17 pmol kg -1 . We have pointed this out too. That said, in the added paragraph (line 335 onwards), we highlight the challenges with the method in overcoming instrument noise which increases isotopic errors. We have also justified why we think our dissolve iron isotope dataset is reasonable.
All that said, I do realize that we are now discussing tiny details about the discussion of methods in this manuscript. This is not central to the overall message and value of the work, which I appreciate, and I do honestly hope it can be published soon. waters with cells upregulating iron uptake and using recycling processes to sustain themselves. 29 Recognising these eddy properties is essential to understanding how they contribute to the 30 biophysicochemical structure of the Southern Ocean. 31 Words = 187 32 33

Main text 34
Mesoscale eddies are ubiquitous in the ocean 1,2 and play a crucial role in the transfer of heat, carbon 35 and nutrients between the deeper ocean, surface waters and the atmosphere 3-6 . Cold-core eddies in 36 the Southern Ocean are defined by strong clockwise rotation, cooler temperatures and negative sea-37 surface height anomalies 7,8 . These eddies can have closed circulation thus 'trapping' ref 9 the 38 biogeochemical properties of these features such that nutrient, chlorophyll and particle 39 concentrations can be distinct relative to those in the surrounding waters 2,9-11 . They can transport 40 these biogeochemical properties vast distances 12 thus they are important from an oceanographic 41 point of view, especially if they cross water mass boundaries such as the Polar Front or the 42 Subantarctic Front 7,11,13 . 43 The concentration of dissolved Fe (dFe) in remote Southern Ocean surface waters, away from 44 continental and island input sources, is typically sub-nanomolar (60-200 pmol kg -1 ) ref 14,15 . The lower 45 limit for this dFe range is thought to be controlled by organic complexation and atmospheric supply. 46 In the Southern Ocean, atmospheric inputs are very low and the supply of Fe usually is provided via