Steur et al. have presented a multi-year triple oxygen isotope time series of atmospheric CO₂ from two locations. Previous research has suggested that air CO₂ triple oxygen isotope data can help quantify atmospheric carbon cycle fluxes. However, there are only a few proof-of-concept papers on air CO₂ triple oxygen isotope measurements despite the importance of such research. This paper provides one of the first extended time series and, therefore, will be a valuable addition to scientific literature.

Specific comments:

Suggest that the authors change “λ“ in Equation 2 to “Θ “. Equation 2 refers to specific kinetic processes with unique isotope fractionation factors. Such physical variables are commonly designated as theta “Θ” values in the literature to distinguish them from the slope “λ_RL“ in Equation 3, which is an arbitrary number.

Strongly suggest the authors recalculate and report the ∆’¹⁷O values using a λ_RL = 0.528 instead of 0.5229. First, water triple oxygen papers use λ_RL = 0.528, and since the composition of air CO₂ is closely linked to water compositions, it is reasonable to use the same λ_RL. Second, the triple oxygen isotope community is now adopting λ_RL = 0.528 as a consensus value, independent of the field of study and materials analyzed; see Miller & Pack (2021). Using the consensus value of 0.528 will make comparing the presented data with existing and future literature easier.

In line 53, the authors state that O₂–CO₂ exchange currently provides the highest measurement precision triple O data. It may be worth noting that multiple papers in recent years have demonstrated sub-10 ppm precision for CO₂ measurements using laser spectroscopy (e.g., Bajnai et al., 2023; Hare et al., 2022; Perdue et al., 2022; Stoltmann et al., 2017).

To the paragraph starting with line 130: It may be worth noting that Perdue et al. (2022) doesn’t observe a shift in ∆’¹⁷O values related to the pCO₂ mismatch between the sample and reference (see their Fig. 8), whereas Bajnai et al. (2023) does, but they correct it by precisely matching the pCO₂ of the reference to the sample (see Fig. 4). It seems that the mismatch in pCO₂ between the sample and reference is the largest source of uncertainty in the presented data. As an outlook, could the authors discuss how to make their measurements more precise?

Bajnai et al. (2023) observed the dependence of the TILDAS-∆’¹⁷O data on the measurement temperature. While the reference bracketing method used in the presented dataset likely addressed such temperature variations, the authors could further increase the credibility of their data by discussing this effect in the paper.

In lines 267 and 336, the authors argue that CO₂-enriched signals are due to the contribution of fossil fuel emissions. In this case, one would expect to see correlations between δ¹³C, ∆’¹⁷O, and pCO₂. Can the authors underline their statements with data and possibly additional figures?

In the paragraph starting with line 310, the authors argue that they should be able to resolve a 130 ppm annual variation in ∆’¹⁷O, as observed by Hoffman et al. (2017). However, their argument that their uncertainty of ±100 ppm is lower than 130 ppm is misleading and needs to be revised. Instead, the authors should take into account the signal-to-noise ratio and the number of measurements to determine what cyclic signal can be resolved in their time series.

In line 315, the authors write that the amplitude of the seasonal ∆’¹⁷O signal in Göttingen is larger due to a stronger biosphere signal. This is an important statement in comparing the presented record with existing data and thus should be expanded upon. Would the 3-D model used in this paper be able to reproduce the 130 ppm signal observed by Hoffman et al. (2017)?

The following changes are suggested for Figures 4, 5, and 6: The range of the top and bottom plots should be the same, which will help the reader make visual comparisons easily. The measurement locations should be written above the curves and not on the vertical axis label. The coloring of the ∆pCO₂ should be changed to a diverging, color-blind-friendly color scale.

Please note the following suggestions for improving the visibility of Figure 7: The vertical year-markers should be made thinner so that they don't clash with the data and error bars. The red trend should be plotted accurately without any shift by 0.08‰ to avoid confusion. Moreover, the horizontal axis grids, similar to those in Excel-made figures, are unnecessary for any plots.

Suggest adding Carlstad & Boering (2023) to the list of references in line 15.

The sentence in line 437, “A better precision…”, is without precedence in the text. The authors may consider either expanding on it or removing it.

Correct the spelling in line 87: “continues”.

References cited in the review:

Bajnai, D., Pack, A., Arduin Rode, F., Seefeld, M., Surma, J., & Di Rocco, T. (2023). A dual inlet system for laser spectroscopy of triple oxygen isotopes in carbonate-derived and air CO₂. Geochemistry, Geophysics, Geosystems, 24, e2023GC010976. https://doi.org/10.1029/2023GC010976

Carlstad, J. M., & Boering, K. A. (2023). Isotope Effects and the Atmosphere. Annual Review of Physical Chemistry, 74(1), 439–465. https://doi.org/10.1146/annurev-physchem-061020-053429

Hare, V. J., Dyroff, C., Nelson, D. D., & Yarian, D. A. (2022). High-precision triple oxygen isotope analysis of carbon dioxide by tunable infrared laser absorption spectroscopy. Analytical Chemistry, 94(46), 16023–16032. https://doi.org/10.1021/acs.analchem.2c03005

Hofmann, M. E. G., Horváth, B., Schneider, L., Peters, W., Schützenmeister, K., & Pack, A. (2017). Atmospheric measurements of ∆¹⁷O in CO₂ in Göttingen, Germany reveal a seasonal cycle driven by biospheric uptake. Geochimica et Cosmochimica Acta, 199, 143–163. https://doi.org/10.1016/j.gca.2016.11.019

Miller, M. F., & Pack, A. (2021). Why measure ¹⁷O? Historical perspective, triple-isotope systematics and selected applications. Reviews in Mineralogy and Geochemistry, 86(1), 1–34. https://doi.org/10.2138/rmg.2021.86.01

Perdue, N., Sharp, Z., Nelson, D., Wehr, R., & Dyroff, C. (2022). A rapid high‐precision analytical method for triple oxygen isotope analysis of CO₂ gas using tunable infrared laser direct absorption spectroscopy. Rapid Communications in Mass Spectrometry, 36(21), e9391. https://doi.org/10.1002/rcm.9391

Stoltmann, T., Casado, M., Daëron, M., Landais, A., & Kassi, S. (2017). Direct, precise measurements of isotopologue abundance ratios in CO₂ using molecular absorption spectroscopy: Application to ∆¹⁷O. Analytical Chemistry, 89(19), 10129–10132. https://doi.org/10.1021/acs.analchem.7b02853

Title: Interannual variations in the Δ(17O) signature of atmospheric CO2 at two mid-latitude sites suggest a close link to stratosphere-troposphere exchange

Author(s): Pharahilda M. Steur, Hubertus A. Scheeren, Gerbrand Koren, Getachew A. Adnew, Wouter Peters, and Harro A. J. Meijer

MS No.: egusphere-2023-2924

MS type: Research article

Iteration: Correction

The paper provides a set of D17O measurements of tropospheric CO2 at two mid-latitude sites, and the authors link the observed variation to cross-tropopause exchange. SICAS is potentially a good way for acquiring data in abundance, but its precision, stability, and limits require further and extensive examination and quantification before it can be fully applied. Overall, I found the paper difficult to follow and the results/interpretation not convincing. Major and some specific comments follow.

Major comments:

1. Need to have a paragraph summarizing the errors/biases of SICAS and possible sampling/storage biases. The SICAS D17O measurements/results are suspicious. Detail analysis of IRMS D17O, though limited, is not available. See specific comments below.
2. Keeling binary-mixing analysis (and Keeling plots) is suggested to be made, to understand the endmembers, if any, controlling the variations of the isotope data. Color-coded diagram is hard to see. Scatter plots of D17O vs. d13C and D17O vs. conc(CO2) can be used to understand how much the variation in D17O is due to anthropogenic (e.g., see Liang et al., AAQR, 2017). Anthropogenic contribution (or even stratospheric influence) can also be assessed by comparing CO2 (including its isotopologues) and CO. This exercise is essential to tell whether the CO2 isotope data contain useful information, or just noise/errors from the measurements.
3. Need a more detail discussion on the modeling. Are the changes mainly in the D17O value in the downwelling flux or the changes are due mainly to the enhanced flux with D17O value little changed? For Eq(11), please elaborate it further. How much contribution is from the newly added 100 mbar temperature term? Is the term the anomaly from the climatology temperature? Please define “anomaly.” Please compare with PV and/O3 at 100 mbar. What is the source of 0.08 per mil mentioned in Fig 7 caption? If it’s from the newly added term, does it mean that the D17O from the model stratosphere is biased too high?
4. Figure 7: mid-year peak in most of the years except 2020, due to enhanced STE in spring, mentioned in the text. What is the cause of the missing peak in this particular year? Also what is source mechanism causing D17O less than 0? If I understand correctly, one has to subtract 0.08 per mil from the modified model, inconsistent with the statement -0.061-0.056 per mil variation range mentioned in Line 400. Does this mean the model was not appropriately made?

Other comments:

1. The CO2-O2 exchange method for D17O measurements was first developed by Mahata et al., not Adnew et al. Please acknowledge the previous effort.
2. make needed correction/clarification to small delta and big Delta in the presentation in the Introduction section.
3. Line 54: rephrase/elaborate 10 per meg for reference gas measurements. Do you mean 10 ppm is achieved for “reference” gas only?
4. Line 70: Please include Liang et al. (2023, Scientific Reports) who reported an updated data set that also include new data from Palos Verdes peninsula, CA.
5. Line 105-106: Rephrase/elaborate “the stability of trace gas amount fractions.” It is not clear whether you referred to CRDS instrumental precision/stability or the concentrations of the gases of interest in the atmosphere/flasks.
6. Trace gas concentration and isotope measurements: are the measurements made for the same flasks collected?
7. Line 181-186: Are the D(17O) D(17O) or d(17O)?
8. Section 2.4, first paragraph. I believed you meant to compare SICAS with DI-IRMS. The first sentence seemed to say that you compared SICAS at CIO with that at IMAU. Please rephrase and make needed correction/clarification.
9. Section 2.4. Figure 2 caption: how is the “combined” uncertainty defined and source of errors? Is the length of the error bar 1-sigma or +/- 1-sigma? Please define it clearly. Are the errors in the difference mainly from CIO? Why the extraction at IMAU is more variable?
10. Figure 7: Is 0.08 per mil from the model? What’s the source/cause of this?
11. Section 3.1 last paragraph. From Figure 3, I don’t see “clearly” the drought points mentioned. Normally I’d expect drought would reduce biospheric uptake and thus cause CO2 increase. Here it said the opposite that the decrease in May-June 2018 was due to the drought. I would suggest to have a separate figure showing the deviation from the average (the background) and discuss the cause of the deviation, such as droughts, in more detail. Also there are two NOAA points next to the referred Lut(CO2) reduction, and that can be used to support the reduction.
12. Line 328: D17O is affected little by transpiration. It’s mainly due to evaporation, or evapotranspiration.
13. Line 395: better agreement “than” the …
14. Line 473: I believe here you meant d17O, not D17O.
15. Appendix B: Are the results experimental results or from model simulation? If they are experimental, please provide measurement errors? What’s the D17O value of the water? With that, is the change in D17O reflected in d18O? That is, is the co-variation of d18O and D17O following water-CO2 equilibration line?
16. Figure C1 and App C: Other than the two lowest SICAS points, there is no correlation between SICAS D17O and IRMS D17O. IRMS higher precision measurements show a factor of ~3 more variation than SICAS. IRMS as claimed has higher precision. One has to discuss whether the large variation in D17O is also seen in and supported by other data, such as CO2 (conc, d13, d18O) and CO.