In this work, Aydin et al. painstakingly recovered a 52,500-year record of atmospheric carbonyl sulfide from a South Pole ice core, with great care and attention devoted to correcting for the post-depositional COS production from sea salt aerosols and other artifacts during extraction. The resulting data set is a valuable contribution to the atmospheric history of carbonyl sulfide and will enable the climate and ecosystem modeling community to better understand biospheric changes since the last ice age.

Despite its scientific significance, the organization of the manuscript may hinder its key findings from being grasped by a broad audience. For this reason, I suggest clarifying the methods, especially the rationale behind every correction and sensitivity test, and streamlining the presentation of the main messages to strengthen the work. Below I list a few high-level issues followed by specific line-by-line comments.

My issues with the methods are:

• For the calculation of ssNa⁺ (Sect. 2.3), please check the denominator in the right hand side of Eq. (6). My derivation says it should be "1 – R_m/R_t" in the denominator (see attached slides). In other words, there is no reason for Eqs. (6) and (7) to have different denominators.
• For the regression between COS and ssNa⁺ (Sect. 2.4), it was not clear why errors in the response variable need to be scaled with a multiplier (α in Eq. 8). This multiplier did not appear in Eq. (5).
• It was also unclear how measurement errors in COS and ssNa⁺ (Sect. 2.5) were incorporated into the Bayesian errors-in-variables regression between COS and ssNa⁺ (Sect. 2.4).
• Different scenarios for correcting for ssNa⁺-produced excess COS (Table 1) should be clearly outlined in the methods section, with the rationale explained. It was not until I read the results did I get a hint of why these scenarios were needed and why they were set up like these. The choice of the averaging window size appears ad hoc.
• Too many scenarios overlaid on the same plot (e.g., Figs. 4b and 5) make it hard to discern the key information. Consider presenting only scenarios that are most robust and relevant to the interpretations.

The results section needs to provide answers to questions raised in the introduction. Currently it is a mix of methods, results, technical arguments, and discussion, making it challenging to navigate. The main issues are:

• Many paragraphs describe the nitty-gritty details of various corrections and make reference to figures in the appendices. Important as they are to ensuring data quality, these are probably not the high-level findings you want the readers to walk away with. Consider moving them to methods or the appendices.
• Descriptions of the correction for ssNa⁺-produced excess COS are scattered throughout the results section, making it difficult to follow. Consider consolidating the main points about this under a single subsection and offload nonessential details to supplementary materials.
• It was not clear how the "climate-driven" correlation between ssNa⁺ and COS was disentangled from the production artifact. Shouldn't the ssNa⁺ COS production correction be applied first before one can make any robust inference of a climate-driven relationship?
• West Antarctica and Taylor Dome ice cores (L233–243) need to be described in the methods.

With respect to the interpretations of the data, there are a few issues:

• The measurements presented here only show ssNa⁺ but say nothing about what the anions in sea salts are. Thus, it is unclear whether sea salt aerosols contain organic sulfur compounds and the analogy between COS dark production in the ocean and that in ice (L462–L476) does not seem to be empirically supported. Experimental evidence suggests that COS is produced from organic sulfur compounds (Modiri Gharehveran and Shah, 2018, https://doi.org/10.1021/acs.est.8b01618) such as cysteine, methionine, and chromophoric dissolved organic matter (CDOM). In any case, a reduced sulfur precursor is needed to produce COS in the absence of sulfate-reducing microbes.
• Given the large uncertainty in the dimethyl sulfide (DMS) contribution to the global ocean COS budget (53 to 680 GgS yr^–1; Jernigan et al., 2022, https://doi.org/10.1029/2021GL096838), it seems hasty to suggest that atmospheric COS is sensitive to changes in DMS emissions.
• Without paleoceanographic evidence, it seems speculative to single out changes in coastal upwelling zones and low-latitude oceans as likely contributors to the deglacial rise in COS. At best this represents one scenario among many possibilities.
• It was not clear why a climate-driven relationship between ssNa⁺ and COS has to be an anticorrelation. What are the biogeochemical or climatological reasons behind this?

Specific comments

• L24–25: Seems speculative. The results may allow us to infer a likely increase in total COS emissions, but not in the emissions of each precursor.
• L33: "COS and CS₂ are produced primarily by photochemical reactions" - COS and CS₂ are also produced in the dark. See Lennartz et al. (2019) Ocean Sci. (https://doi.org/10.5194/os-15-1071-2019) and Modiri Gharehveran and Shah (2018) ES&T (https://doi.org/10.1021/acs.est.8b01618).
• L35: "Warmer waters can act as a seasonal sink due to temperature dependent loss to hydrolysis" - But both COS and CS₂ are less soluble in warmer waters (De Bruyn et al., 1995, https://doi.org/10.1029/95JD00217), and wouldn't this lead to more outgassing from the ocean?
• L52: It would be helpful to add a brief note on how the bubble–clathrate transition zone (BCTZ) affects the preservation of ancient air in ice cores for those who are not familiar with the BCTZ.
• L80: How were the extraction efficiencies of the wet and dry methods characterized? If these methods have been previously examined, a citation would help.
• L112, Eq. (1): Is α the ratio of aqueous concentration divided by gaseous concentration?
• L165, Eq. (8): Is α here the same as that in Eq. (1)? If not, please use a different symbol.
• L191–192: The numbers of effective sample sizes should be reported properly as 4900, 2500, etc.
• L197: "The 2σ uncertainty ranges shown in the figures represent" - Which figures?
• L215–L292: This section seems to belong to the supplementary material or methods. I would distill a few key messages only to put in the results section.
• L215: "COS was measured at the over the length of ..." - Check typos here.
• L220: Fig. A1a shows the amount of gas extracted, not gas extraction efficiency per se.
• L222–232: This paragraph left me wondering: are the spikes real or not?
• L255: "This reversal occurs at a depth where ice impurity concentrations are increasing steeply (Fig. 1b)." - I believe here you intended to reference Fig. 1c.
• L256–264: This paragraph seems to belong to the methods.
• L265: "significant" -> "statistically significant"?
• L266–267: "The slope is stronger ... with ssNa" - If the mechanism by which COS is produced from non-sea-salt aerosols differs from sea salt aerosol-caused COS production, can the slopes be compared on the same scale?
• L269–271: Skimming these sentences the first time, I was confused how an R² value of 0.04 could serve as the basis for the correction. The next time I realized that this was not the correlation between excess COS (the noise, which is unknown) and ssNa⁺, but that between the total COS (signal + noise) and ssNa⁺. You might want to add a brief note somewhere to get this point across more effectively.
• L276: "In fact, the presence of spikes is a contributing factor to the low R² of the correlations despite the high significance" - You might want to point the readers to scenario G2 here to show that the issue of spikes has been taken into account.
• L293–304: This seems to belong to the methods.
• L303–304: It was not obvious to me how these analysis scenarios supported the idea that "the relationship between ssNa and COS is driven primarily by millennial scale variability."
• L317–318: Why was it necessary to scale up the errors? Section 2.4 asks me to go to section 3.1, but section 3.1 refers me back to section 2.4.
• L334–342: Why are measurements of the glacial period and the Holocene corrected separately? Do you expect the relationship between ssNa⁺ and excess COS to differ between the last glacial period and the Holocene?
• L345: "150 ppt" and "90% lower" - Check the numbers. It's not like that the Holocene has a COS level at 1500 ppt. Also the referenced figure does not seem to tell this information.
• L370–376: The point of these three analyses is lost on me. How do the regressions of smoothed ssNa⁺ onto unsmoothed ssNa⁺ support the validity of climate-driven same-age anticorrelation between ssNa⁺ and COS?
• L454: "COS production in the firn is approximated by an advective-diffusive model of the South Pole firn" - Has this correction been applied to the present study?
• L487: "It is possible atmospheric COS is sensitive to changes in ocean DMS emissions modulated by winter sea ice" - According to Lana et al. (2011) climatology, DMS emission hotspots seem to lie beyond the sea-ice covered regions of the Southern Ocean.
• Fig. 1: For wet extraction results, maybe show only the solubility-corrected values (red line in 1a)? Captions seem misplaced for panels b and c.
• Fig. 2: I don't see the "black error bars" described in the caption.
• Fig. 4a: Isn't this panel a repeat of Fig. 1a, but with error bars? It may be removed because it adds little new information.
• Fig. 5: Could be combined with Fig. 4 for the comparison.
• Fig. 6: Seems like a supplementary figure to me.

Summary:

Aydin et al present a new 52kyr record of COS from the South Pole ice core (SPC14). The record was generated using both a dry-extraction (most of the samples) and a wet-extraction technique. An empirical solubility correction is applied to the wet-extracted samples based on comparison with dry-extraction data over the Holocene, where analytical artifacts are thought to be insignificant for the dry technique. Prior work has shown that COS appears to slowly degrade in glacial ice, with the hypothesized mechanism being hydrolysis. The authors argue that SPC14 is too cold for this process to be significant, providing an important advantage in terms of COS preservation. However, the authors find an unrealistically large amount of COS variability in the part of the record from the last glacial period. They observe a very weak but statistically significant correlation between COS and sea-salt sodium (ssNa) concentrations. The authors propose that there is a mechanism in the firn (and firn only) that somehow produces excess COS of non-atmospheric origin. On the basis of the observed correlation with ssNa, the authors develop and apply a correction for excess COS for SPC14. They also apply a version of this correction to COS data from two other ice cores: WAIS Divide and Taylor Dome. The resulting 52kyr record suggests very large changes in atmospheric COS, with 2 – 4 times greater COS in the Holocene than during the last glacial maximum (LGM). This large change is interpreted as an increase in the oceanic sources of COS during the deglaciation.

Major comments:

COS is present in air in very low concentrations (ppt level), and thus is extremely challenging to measure in glacial ice. The authors are therefore to be commended on the very demanding analytical effort involved in producing this record. This is an interesting record, and I think should be published if the authors are able to address the points below, but I find this record very challenging to interpret and have some concerns and recommendations, as follows.

I am not convinced by the authors’ arguments for a ssNa-based COS correction. The argument that excess COS is produced only in the firn does not make sense to me. The authors effectively propose a mechanism that “burns out” fast (so that there is nothing below the firn zone). Such a “fast burn” mechanism would be expected to be most intense near the surface, but here ventilation to atmosphere would remove any excess COS that is produced. Trapped air below the firn layer would be much more sensitive to any excess production because there is less air and because this air can’t exchange with overlying air.

The ssNa-based correction is very large (up to ≈70% of the measured values for some samples) and results in a record that shows 2 – 4 times lower COS at LGM than at the Holocene, and the authors argue that this must be mainly source-driven. Sources are mainly linked to ocean microbiota, and I find it difficult to accept that oceanic sources could have declined this much (the surface ocean was quite productive during the LGM). The fairly good (although far from perfect) agreement among three ice cores after the ssNa correction is encouraging, but the ssNa correction may or may not be transferrable between sites.

Considering the above, I would recommend presenting the ssNa correction and the resulting temporal trend as speculative and one of possible scenarios (the other obvious scenario being no ssNa correction), and more clearly emphasizing how uncertain the interpretation of the measurements is, both in terms of the resulting reconstruction and implications for COS budget, which Table 1 shows to be very uncertain even today.

Minor comments:

I think the manuscript could benefit from a stronger explanation for the motivation for this study, which clearly involved a great deal of effort. It seems that direct radiative forcing due to COS is negligibly small. The authors mention its link with DMS (which has a larger forcing), but DMS appears to be a relatively smaller source of COS. COS is removed by terrestrial plant uptake – could this be a stronger motivation for the record, as a possible proxy for terrestrial biospheric productivity?

Section 2.2:

The temperature-dependent expression for COS solubility should be given, and the actual solubility value used should be stated, and compared to the value for air. What fraction of COS is typically in the meltwater?

Line 149: “above” and “below”  “shallower than” and “deeper than” would be less ambiguous here. “above” could mean “greater than”

Line 215: “COS was measured over the length of the SPC14 ice core…”

Line 252: Do you mean “ice from the last glacial period”? All of the ice core is “glacial ice”. Please edit to clarify.

Line 254: Fig 1c does not show COS during time interval being discussed