Summary

This manuscript evaluates strategies to mitigate ozone (O₃) contamination during stratospheric sampling of carbonyl sulfide (OCS) using the AirCore balloon system. The primary finding is that O₃ does not appear to significantly affect OCS measurements across the atmospheric column: AirCore profiles with and without O₃ scrubbing agree well with independent datasets from ACE-FTS and SPIRALE.  Observed differences from these datasets fall within the QCLS measurement precision and the expected natural variability, particularly in high-latitude regions where seasonal OCS changes can reach ~150 ppt (https://gml.noaa.gov/hats/gases/OCS.html).

The authors tested three inlet O₃ filtration strategies: cotton, magnesium perchlorate [Mg(ClO₄)₂], and squalene. Laboratory results show that cotton is ineffective at removing O₃ (Figure S2), and squalene was not deployed during balloon flights. Thus, only Mg(ClO₄)₂-filtered and unfiltered (“free inlet”) profiles provide field-relevant comparisons. However, because the Mg(ClO₄)₂ dryer was not characterized in laboratory experiments, its O₃ removal efficiency in the field remains uncertain.

The motivation for O₃ mitigation stems from Engel et al. (1994), which reported up to 50% OCS loss when co-sampled with stratospheric O₃ in balloon-borne cryogenic samplers. Notably, in the nearly three decades since, no widely cited studies have confirmed such losses, suggesting that the effect may be specific to cryogenic sampling techniques. If so, the null result reported in this manuscript has important implications for interpreting historical in-situ stratospheric trace gas records.

Major Comments

1. Laboratory results vs. Engel (1994)

The laboratory results presented in this manuscript appear to contradict the findings of Engel et al. (1994). Specifically, Figure A2 shows a slight enhancement in OCS (~40 ppt) in the presence of stratospheric O₃ concentrations, whereas Engel reported up to a 50% loss of OCS under similar O₃ levels. Given the significance of these laboratory results to the paper’s conclusions, I recommend moving most of the information currently in Appendix A into the main body of the manuscript. However, it should be noted that the enhancement reported here is less than 10% of typical ambient OCS levels and approaches the instrument’s stated precision (~25 ppt), making it unclear whether the observed difference is statistically meaningful.

1. Appropriateness of multi-year dataset comparisons

Due to the substantial natural variability in tropospheric OCS abundance and in stratospheric OCS driven by transport processes, I do not consider it appropriate to perform a quantitative comparison of short-duration datasets separated by several years. I therefore recommend removing Figures 2, 6, and 7 as well as Tables 3 and 5. Table 2 could be moved to the Supplementary Information or omitted entirely, as its contents are already conveyed in Figure 1. Alternatively, a comparison with N₂O or other stratospherically photolytic tracers such as CFCs could allow for quantification of stratospheric transport, but this would require substantial additional analysis and constitute a major revision to the manuscript.

1. Concerns with LISA and BIG LISA “bag” data

The manuscript includes significant caveats regarding the LISA and BIG LISA MLF “bag” sampler data. In Section 4.1.2, the authors note that the 30229-U MLF bags used in BIG LISA have a manufacturer warning: “Although the deployed bags are indicated as suitable for sulfur compounds, they are not recommended for low-ppm volatile organic compounds due to background levels” (Sigma Aldrich, 2025). A similar warning appears on the Sigma Aldrich website for the 30228-U bags used in LISA: “Not recommended for low ppm VOCs due to background levels (we recommend the SupelInert PVDF Tedlar alternative film for VOCs)”. Considering that a common VOC such as CS₂ can oxidize to OCS with ~80% efficiency, even sub-ppm VOC contamination could significantly bias atmospheric OCS measurements (<1 ppb). In addition, a subset of the LISA OCS data appears to have been omitted by the authors without explanation (see Section 2.1.2). Given these concerns, I recommend removing the LISA dataset analysis from the manuscript.

Specific Comments:

Line 48: Remove: “Although the debate has not been fully resolved” 

Line 59: Provide a reference for the QCLS

Line 66:  Please explain why the impact of air samples of COS may be significant.  I can find no other examples aside from the Engel 1994 paper.

Line 72:  Add a reference to the Schmidt et al 2024:

Schmidt, Matthew, et al. "Trends in atmospheric composition between 2004–2023 using version 5 ACE-FTS data." Journal of Quantitative Spectroscopy and Radiative Transfer 325 (2024): 109088.

Line 83:  Explain how Aircore and Sulfinert specifically differs from the stratospheric Cryogenic air sampler apparatus.

Line 102:  Do you expect any Aqueous reactions with ozone to occur on the wetted cotton?

Line 146:  What is the duration of the samples storage in: 1. MLF bags, 2. glass cylinders before analysis

Lines 149-151:  Without a specific reason to disregard a subset of samples, you must either present all of the data or none of it.

Lines 205-207:  Please present the data for the QCLS precision.

Lines 211-212: Present the data or references for the QCLS precision.

Line 215:  Which version of ACE-FTS are you using v5.2? v5.3?

Line 216:  Add Boone et al 2023. And Schmidt et al 2024:

Boone, C. D., P. F. Bernath, and M. Lecours. "Version 5 retrievals for ACE-FTS and ACE-imagers." Journal of Quantitative Spectroscopy and Radiative Transfer 310 (2023): 108749.

Line 228:  Do the 502 and 1681 profiles correspond to global samples at those latitude bands or localized profiles over TRN and KRN?

Line 244:  This analysis would be more quantitative if you compared with the Age of Air parameter or another photolytic species like N2O.

Line 263:  The MkIV spectrometer utilized by Toon et al is balloon borne into the stratosphere, but this is a long path solar FTIR measurements so it may be more appropriate to include with the remote sensing measurements.

Line 278:  Please provide more details for the “direct reaction of other gas species with O2.”

Line 284:  How does your observed tropospheric variability compare with NOAA GML flask network measurements for high latitudes?

Line 290:  How is the variable lapse rate an indicator of OCS convective transport?  Have you preformed trajectory analysis?

Line 299-301:  The low-ppm VOC warning is listed for both LISA and Big-LISA MLF bags.

Line 317:  The absolute differences in OCS above the tropopause are not useful.  You must compare with N2O to understand stratospheric transport and age of air factors, and correct for N2O increasing trend.

Line 323:  Include Schmidt et al 2024 to for stratospheric OCS trends.

Line 331:  On line 327 the authors state that no quantification of sample loss is available, but here they claim that “differences remain marginal.”  Please attempt to provide more quantification of the uncertainties

Line 339:  The changing relationship between Methane and N2O is interesting and either points to major instrument issues or more complex stratospheric dynamics and should be discussed more here.

Line 370:  The temporal differences really limit any kind of quantitative analysis.

Line 383:  How does the 8% difference compare with the decadal change in stratospheric OCS abundance?

Line 400:  This qualitative comparison is not useful, just refer to Figure 3

Line 413:  What version of ACE-FTS are you using?

Line 476:  Add discussion of Squalene-based scrubbers to the main body of the text.

Line 479:  This is a major result, and you should highlight the difference from Engel’s 1994 paper.

The manuscript “Balloon-borne Stratospheric Vertical Profiling of Carbonyl Sulfide and Evaluation of Ozone Scrubbers” by Alessandro Zanchetta et al. compare several new methods to measure carbonyl sulfide (COS) concentrations along vertical profiles into the stratosphere using balloon-based air sampling platforms. The authors compare measured COS profiles between the different sampling instruments, from three locations and assess their results in comparison with modelled COS profiles. The methods show good overall agreement and will undoubtedly be able to provide new and insightful information to the growing science community interested in COS.

The manuscript furthermore includes an assessment of the efficiency of ozone scrubbers, based on experiments undertaken on a purpose-built setup. The authors tested cotton, a material that has been suggested some 30 years ago and found that it is not as reliable in ozone elimination as it might have been expected. They tested an alternative material in laboratory experiments, which has not yet been tested on the balloon-based air samplers, but shows very promising results.

The manuscript is very well written, the experiments are clearly described, and the overall methods are sound. In my opinion, this is a fantastic study, and the manuscript is an excellent fit for publication in AMT.

There are, however, a few minor points that I suggest considering prior to publication to improve the clarity of the manuscript:

1) Instrument calibration. The observations span a period of four years, from June 2019 to August 2023. COS measurements are challenged by instability of COS in reference gas cylinders. While the authors provide information on the measurement precision, information on the long-term stability of the instrument is not presented. As this study potentially requires long-term stability of the instrument over four years, a demonstration or small assessment on the stability to underpin the robustness of the presented method would be desirable.

2) Figures. Many of the figures include a lot of data. Several figures are so crowded that I find it difficult to comprehend (for example 3 a). Figures often include very small symbols (for example 3 b and c), which are too small for me to tell apart by colour, blurring the figure and the story the data tell. Other examples include Figure 1 left panel, Figure 2, Figure 3 left panel, symbol size in Figure 3 middle and right, Figure 4, Figure 5. Can Figures 6 and 7 have larger symbols? Maybe split some of them and have Figures on single panels with full width (i.e., Figure 3 a)? Could log scales be useful to decompress the large numbers of symbols in the lower y-axis range? Or could panels b and c of Figure 3 be on top of each other as a second column in the Figure, allowing 3 a to be larger? I am not sure how to improve, but I think the present state doesn’t make the most out of the fantastic data and it would be well worth to improve their clarity.

Some multi-panel Figures have letters as identifiers (i.e., Figure 3 a-c), others (i.e., Figure 1) don’t, while some are referred to as second/third (p16, l377). Furthermore, the text refers to the Figures in a sequence that is different from the appearance and enumeration of the Figures, i.e., the text first refers to Figure 3, then Figure 5, then Figure 1. This might be a matter of personal taste, but greater consistency on those points would make the manuscript more accessible to me as the reader. It can be difficult to follow at times, especially when the text is jumping between different figures across different pages with focus on different events in those figures, i.e., P17, l386, where it might be useful to have markers in the figures (arrows?) to mark these events.

3) Structure. The main part of the manuscript includes an Appendix, which talks about the setup to assess different O3 scrubber materials, as well as the results of experiments made using gas from one cylinder with 0 ppt COS. There are also supplementary information to the manuscript, which include more test results on O3 scrubbers using a different cylinder with a different test gas, containing some 750 ppt COS. It is not clear to me why the O3 scrubber assessment is separated between appendix and supplementary information? It would seem to me that they could be merged into one assessment in either appendix OR supplementary information, which can then be referred to in the main text. As the assessment of the scrubbers in the appendix is made on a gas containing 0 ppt COS, I believe this doesn’t allow to assess whether the scrubbers could potentially reduce COS in the sample, as it is already at 0 ppt. To me, this is an important point that could potentially have affected the outcome of the study, if tests hadn’t been made with the 750 ppt COS gas as well. Here, the combined information from tests made with the 0 ppt COS and 750 ppt COS provide the robustness of the results that is needed. Therefore, the separation of those sections into Appendix and Supplements seems confusing to me and I’d suggest keep together in the manuscript. This would enable presenting the analysis in a more robust and compact way, which I think would improve the clarity of the manuscript.

Specific comments:

Title: “Ozone Scrubber Materials” instead of “Ozone Scrubbers”?

P2, l42: “, with a mole fraction range of 350-520 parts per trillion (ppt) in the unpolluted free troposphere (Berry 2013, Remaud 2023).”

P2, l47: “to CO2 and sulfur dioxide (SO2), a precursor…”

P5, l142: What is the COS mole fraction in that gas?

P6, l 169: Are these valves used for heating?

P6, l181: Is the number of 1800 s correct, i.e., 30 minutes?

P9, l217: Should this be section number 2.3.1 SPIRALE, instead of 2.4.1 SPIRALE?

P10, l223: Should this be section number 2.3.2 ACE-FTS, instead of 2.4.2 ACE-FTS?

P12, l272: Can you provide details of the differential pressure sensors? Could be quite useful to know for interested audience, in terms of materials to avoid.

P14, l327: Is the “self-consistent” statement fully applicable when comparing systems that sample either during ascend or during descend?

P14, l331: I’d suggest making the statement in that sentence more clear and spell out the conclusion with more clarity, for example: “However, we assume that differences due to instrumental effects remain marginal, while we believe that the day-to-day variability and long-term trends in COS mole fractions are the most important cause for the observed differences.”

P14, Table caption: Should this be expanded a bit? It is not clear to me what this exactly means. Are there data gaps at specific heights, and are these suggested to be due to contamination from glue, surface effects, and differential pressure sensors? If so, how could this only affect short sections/a small number of the sample, but not all samples?

P15, l334: Maybe add “as” to “and as we have no O3 measurements”?

P15, l358: Can you be more specific on “the variability” you refer to? Could you please explain what difference have in mind here?

P17, l386: Example where a marker/arrow within the data figure to identify/highlight that event could be helpful to better understand what the discussion is about.

P20, Figure captions: Here and in other figures where relevant, can the time interval represented by the modelled COS profiles be included into the caption as well?

P22, Figure 6: Here and in other figures where relevant, can the regression functions be included into the figures? This would be much easier to follow than having them in a separate table. If they are needed in a table, I suggest showing in both forms.

P24, l470: Say which laboratory.

P25, l480: Change to “different sorts of inlets with different O3 scrubber materials.”

P25, l482: “… variability, a hypothesis…”

Supplements, Figures 7-15:

• Where indicated in the figure legend, I am unable to tell ascent data and descent data apart; all I see is one colour.
• Include in captions that the tropopause height is indicated by the green line.