Summary:

Somkuti et al. present OSSE simulations for XCH4 retrieval from the 2.3-micron band using an OCO-2-like observation pattern. The study realistically simulates clear-sky and aerosol-loaded conditions, identifying retrieval biases correlated with surface reflectance—significantly larger under aerosol influence. These biases are mostly mitigated when the true aerosol profile and optical properties are incorporated into retrieval.

The study aligns well with AMT’s scope and addresses a relevant topic. However, the manuscript requires improvement in a few areas. It does not sufficiently review prior literature on the interaction between surface reflectance and aerosols in GHG retrievals. Additionally, the final section assumes an idealized scenario where aerosol properties match the OSSE forward simulation, which may not reflect realistic uncertainties. The authors should assess the level of aerosol uncertainty that can be tolerated to reduce surface-dependent biases to acceptable levels. Addressing these issues would strengthen the manuscript and support its publication.

General Comments:

Section 2.1

• L58: Clarify how BRDF wavelength dependence is treated in the OSSE forward simulation, as it is a key difference between forward and inverse models.

Section 2.2

• L62: Do ISCCP cloud observations cover the simulated period? When combined with the CAMS there could be inconsistencies from where the chemical transport model simulates clouds. For instances where there are clouds in CAMS but not ISCCP, the AOD may be overestimated because hygroscopic growth is being accounted for. 

Section 4.1

• L202: Earlier in the section it was stated that Rayleigh scattering was insignificant, but is this still true for the small magnitude of the changes being considered in Fig. 7? For instance the negative bias over the dark surfaces seems consistent with atmospheric scattering, since photon paths from light scattered from the atmosphere will have a greater contribution to the total radiance at the sensor relative to the brighter surfaces. To first order this would be radiation from the solar beam directly scattered into the path of the sensor, which would effectively shorten the light path, making a forward model that does not account for it reduce the CH4 column, as is shown in the figure. Since the magnitude of the biases reduces between the non-sc and SS cases, this supports Rayleigh being potentially important for biases at this level.
  
  
  
  It should also be noted that a 2-5 ppb bias shown in the Figure may actually be significant depending on the application. E.g. A lot of diffuse agricultural sources produce enhancements around this magnitude. If anything the results show how close to perfect a retrieval would have to be to quantify these. 

Section 4.2

• The physical explanation for the curves in Fig. 9/10 should be discussed - e.g. see section 3.1 of Aben et al. (2007). https://doi.org/10.1016/j.jqsrt.2006.09.013
• Also these relationships between albedo/aerosol loading are well known. I would possibly reframe this section as a quantification of this type of bias for CH4 in the 2.35 micron band. At the moment it reads like somewhat of a discovery, but these effects have been written about since at least SCIAMACHY e.g. Houweling et al. (2005) https://doi.org/10.5194/acp-5-3003-2005

Section 5

• L299 - I suspect that the 1.65 micron band is susceptible to the same aerosol/albedo related biases as the 2.3 band when retrieving the absolute column because the physical light path shortening/lengthening effects are still at play. For retrievals based on the CO2-proxy method they may be lessened for moderate aerosol scenes (assuming that the albedo difference between the CO2 and CH4 bands is similar). 

Specific Comments:

Section 3

L141: Nit - “surface albedo” should be “Lambertian surface albedo” (assuming that is what is being fit). 

L145: What are the wavelength(s) corresponding to the radiances that are used to estimate the albedo 

L150: Where do the trace gas profiles come from?

Section 4.1

L186: SS and non-sc don’t need to be in brackets as they were defined in Section 4. 

L199: Since there are only two sets does it make more sense to say “non-sc and SS produce offsets of x and y ppb respectively” rather than “each set exhibits an offset between -6 and -3ppb. 

Somkuti et al. investigates the impact of surface reflectance biases on satellite-based methane retrievals in the 2.3um SWIR band. The study uses an observing system simulation experiment (OSSE) to analyze how these biases appear, particularly in the presence of aerosols. I hope that the team will continue to do research on this topic. Understanding and addressing such biases is crucial for improving the accuracy of methane retrievals and to prepare for future missions.

The manuscript is well written and structured. I greatly appreciate the fact that I couldn’t find any typos and the clarity of the narrative. I recommend publication after the comments below have been addressed.

General comments

I understand that interesting questions may be out of scope of this study, but I would encourage the team to pursue investigations on a couple of fronts: (1) retrievals in the 1.65um, where historically based on GOSAT, scattering biases have been deemed unsignificant. Given current and upcoming missions planning their retrievals from the 1.65um band, it would be incredibly valuable to have this analysis done with those missions in mind. (1) For scenes measured in sun-glint mode, given the relevance of methane measurements offshore, greatly affected by aerosol transport from onshore, to characterize methane emissions.

Given that a qualitative change in TROPOMI retrievals involved modifying the order of the polynomial that captures spectral dependence of the surface albedo, has the team considered analyzing this aspect? Additionally, it would be useful to discuss the findings in light of Jongaramrungruang et al. (2021) and the spectral differences between GeoCarb and TROPOMI, as mentioned in the discussion

Given all the aerosols scenarios available in the simulations, are there any other insights besides the aerosol extinction optical depth, maybe also considering spatial distribution of aerosols?

Has the team conducted any simulations for non-nadir viewing geometries? This may be particularly relevant for new instruments with higher spatial resolution, which focus on plume retrievals and also thinking on bidirectional reflectance distribution function (BRDF) effects. Exploring whether these factors impact retrieval biases could provide valuable insights.

It would be helpful if the authors provided more insights into why they believe the bias is already present in the absorption-only atmosphere. Additionally, the difference in the sign of the bias between Figure 6 and Figure 7 is not fully discussed. Some areas with a positive bias in Figure 6 appear as stronger negative biases in Figure 7. When averaging the maps into the curves that depict surface reflectance bias, some details about the underlying spatial distribution of the biases may be lost and are not explicitly discussed. Addressing these aspects would improve clarity and strengthen the conclusions.

Specific comments

One remark is to make the figures a little bit bigger, particularly Fig. 8, 9 for the final version of the manuscript.

What is the spatial resolution of the simulations?

Have you looked at scenes covered by snow?

Are low albedo values filtered in the analysis? Asking particularly after seeing the remanent bias for the lowest bin (green square, Fig. 11), and recalling the surface albedo threshold of 0.02 in the TROPOMI quality filtering or 0.05 used in many applications.

Can you, from this analysis, conclude on the usefulness of the oxygen A-band? Is the solution for TROPOMI to ingest “real” aerosol information in the retrieval? Can this be operationalized?