This manuscript by Ortega et al. presents an extensive evaluation of retrieval strategies for two major greenhouses gases, namely, methane and nitrous oxide, applied to high resolution solar absorption spectra obtained with Fourier transform infrared (FTIR) spectrometers.

For both targets, the authors carefully investigate the effect of the line parameters selection and of the adopted regularization on the retrieved geophysical products, using the SFIT4 algorithm. The parametrization is tuned such as to reach comparable information content (degree of freedom for signal or DOF of 2-2.5). Observations obtained at Boulder, CO, in the framework of the NDACC network are the focus here. Indeed, the authors capitalize on concurrent airborne (air core or aircraft) in situ measurements available in the vicinity of this site over several years.

This way, the numerous strategies investigated can be ranked not only on the basis of the smallest spectral residuals achieved, but also, and importantly, by minimizing the bias between the remote-sensing and in situ data.

The inter-technique bias for CH₄ and N₂O are determined considering large samples and found small or even not significant. Uncertainty budgets are also evaluated and presented considering the prescribed strategies.

Overall, the results are appropriately presented and discussed, and the manuscript reads well despite the large number of figures and tables.

Major comments

This work will undoubtedly be very useful for the NDACC FTIR community, likely leading to better harmonization and consistency across the stations while potentially minimizing biases with other techniques, consolidating the resulting data sets, making of them more reliable and relevant ensembles for satellite and atmospheric model validation.

Still, this manuscript remains very technical by nature and possibly of limited relevance for the broader scientific community. In my view, it could be more appropriate to consider this work as a “Technical note” and to include in a supplement the detailed descriptions of the selected and recommended retrieval strategies, including the regularization (especially for the optimal estimation approach). This way, this paper would be optimized for its most likely end users.

Finally, I found the section on the trend analysis quite irrelevant in the present context, considering that this study aims at the determination of optimum FTIR products as part of a network effort. Moreover, this section suffers from flaws and does not bring very useful information to the reader. More specifically:

• there is no information about the tool or method that is used to derive the trends, what is the approach used to estimate the trends and their associated uncertainties?
• the statistical uncertainties are extremely small: are they really representative and robust?
• there are some dissimilarities among the NH and SH trends in the stratosphere (see Table 13): this might be at least partly related to well-known stratospheric asymmetries that were the subject of earlier studies for other long-lived tracers (e.g., by Strahan et al., 10.1029/GL088567, 2020), but the authors do not discuss neither comment this feature
• furthermore, the tropospheric trends are also statistically different; can we really expect this for such a well-mixed greenhouse gas?
• the CH₄ trend is evaluated, but not the N₂O one; this is not explained nor justified

My suggestion would therefore be to just remove the trend section (section 4.4) and to investigate the CH₄ and N₂O trends in a follow-up paper, involving a larger number of FTIR stations after network-wide implementation of the recommended strategies.

Specific comments and suggestions

An originality of this manuscript is the use of concurrent in situ data to assess FTIR retrieval strategies. I would suggest mentioning that strength in the title, perhaps by adding “FTIR observations with the support of airborne in situ measurements”.

Line 36: its relatively short atmospheric…

Lines 88-89: it might be relevant to provide information about the filters 3 and 4? What are their characteristics and respective advantages or limitations? Or provide a reference?

Line 90: maximum optical path difference

Line 124: to a common altitude scheme…

Figure 2.a. why is the methane scale in ppm here, and in ppb previously? Perhaps harmonize?

Table 5 and similar: would bolding the best results helps the reader to identify the most relevant strategy?

Line 308: remove double opening parenthesis

Figure 11 (and 12): is it relevant to show the altitude range above 30-35 km, where there is no information available?

Line 331: can we consider that 13 years of data provide a “long-term” view for a geophysical parameter?

Figure 12: are the uncertainties similar for “filter 3” and filter 4”?

Figure 13 (and 14): I found counterintuitive to place the panel for the stratosphere below the one for the troposphere; also and of kept, they could be enlarged for readability

Line 395: remove the question mark after “Sweeney et al., 2015”?

In this manuscript, the authors present a broad evaluation of different retrieval strategies for methane (CH₄) and nitrous oxide (N₂O)—two major greenhouse gases that also affect the ozone layer—from solar absorption infrared spectra measured at the NDACC (Network for the Detection of Atmospheric Composition Change) station in Boulder, Colorado, USA.

The investigated retrieval strategies make use of different molecular absorption databases (including the latest releases), a priori profiles, spectral windows, and regularization methods, thus providing an exhaustive evaluation. The comparison is based on the assessment of fitting residuals (attributed to spectroscopic parameters) as well as the comparison of retrieved profiles and partial columns to a unique dataset from AirCore (CH₄ only) and aircraft measurements (both CH₄ and N₂O).

These investigations ultimately lead to recommended retrieval strategies for the Infrared Working Group (IRWG) of NDACC, aiming to harmonize high-quality global datasets of these critical trace gases.

Given that this manuscript presents novel methods for producing high-quality datasets of two major greenhouse gases and leverages a unique dataset of airborne in situ measurements to validate these methods, I recommend its publication in Atmospheric Measurement Techniques once my comments have been addressed.

Major comments

• N₂O retrieval strategy

Based on the results presented in the manuscript, the N₂O retrieval strategy that fits both filter 3 and filter 4 spectra simultaneously is claimed to be the best among the tested strategies and is therefore recommended by the authors. I have several comments regarding this conclusion:

          - Fitting multiple spectra simultaneously is not a common approach within the IRWG and raises methodological questions. The manuscript does not provide sufficient details on how this retrieval strategy is implemented in practice. For example, were any time co-location criteria applied when selecting the spectra? FTIR high-resolution spectra have a long integration time (several minutes), and it is common to record several spectra for the same filter before switching to another one. If filter 3 is not immediately followed by filter 4 in the measurement protocol, then the selected spectra could span a relatively long time. What, then, is the impact of varying solar zenith angles and other parameters? Additionally, is the assessed error budget comparable to other strategies, such as r6? Since this strategy is recommended for implementation, more details should be included in the manuscript to help network members adopt it in the future.

          - Looking at Figure 9 and Table 9, one could argue that the retrieval strategy r6 yields comparable or even better results than r8 (the recommended strategy). Since r6 is essentially an updated version of the existing IRWG strategy (incorporating HITRAN 2020 instead of 2008, a new version of the WACCM profiles, and a shift to Tikhonov L1 regularization), wouldn't it be simpler for the network to implement r6 instead?

• Trend analysis section

I understand that the purpose of the trend section is not to conduct a full trend analysis but rather to provide a foundation for future studies. However, I would argue that the residual resampling bootstrap method is not well-suited for geophysical time series. This method is known to be sensitive to the strong autocorrelation typically observed in such datasets, often leading to underestimated confidence intervals for the slope parameter. This issue should at least be acknowledged in the manuscript.

As currently presented, this section does not contribute significantly to the manuscript and may be outside its primary scope and objectives. The authors might consider either improving the methodology—potentially by comparing explicitly their trends with those derived from independent datasets—or removing this section altogether.

Minor comments/suggestions/typos

Page 10: The manuscript demonstrates that including HI00 for CH₄ spectroscopic parameters results in the lowest residuals in window 4 and, to a lesser extent, in window 5. However, as shown later in Section 4.2, the r8 and r9 (ATM20) strategies provide the best comparison to the in situ data. What is causing better results with higher residuals is probably not known, but could this be discussed or acknowledged somewhere in the paper? Perhaps in Section 4.2 or the conclusion section?

Page 10, the sentence in lines 219 to 220 is not clear. There is a typo with the molecular absorption databases: HIT04CH4 should be HIT00CH4. For the statement: “micro-window 4 shows superior residuals”, I interpret it as larger residuals, while it is the opposite.

P2, L37: add the actual GWP for CH4

P3, L79: define semi-co-located?

P3, L89: define spectral range of filters

Section 2.2, this is only a suggestion: add subsections for AirCore (2.2.1) and aircraft (2.2.2)?

P4, L116: Maybe add that those years (2018 to 2022) were selected because they are spanning the common period with aircraft data?

Section 4.2, again this is only a suggestion: To improve readability and help the reader navigate the results more easily, this section could be further divided into two subsections, one for CH₄ and one for N₂O. The first paragraph (lines 235–247) could remain as it is, followed by the introduction of Section 4.2.1 at line 248 (dedicated to CH₄) and Section 4.2.2 at line 289 (dedicated to N₂O). This structuring might make it easier to follow the discussion of each gas separately.

P12, L244: remove the brackets for “(and N2O)”

P15, L279: I am not sure to understand how the partial columns were estimated for the in situ data. This sentence could be clarified.

P16, L283-285: I am not sure to understand this sentence. What exactly is being compared here? These numbers are not included in any tables.

P17: there is a typo for the bias, it is 0.18 instead of 0.10 (see Table 9)

P18, L302. Even if I understand that the authors are referring to the AirCore technique, one could argue that long-term satellite datasets exist (e.g., ACE-FTS). I would re-phrase or remove this statement.

In the first or second sentence of the conclusion, I would suggest highlighting that this study leverages a unique airborne in situ dataset to perform a profile comparison—an approach that is rarely undertaken within the IRWG. This aspect significantly strengthens the study’s findings and underscores its contribution to improving retrieval strategies.

P26, L388-389: this is a strong conclusion, but it unfortunately lacks supporting references. As the trend section is currently conducted and written, it does not fully support this statement. I suggest rephrasing it to make it less definitive, ensuring that this conclusion aligns with the evidence presented in the manuscript.