This paper deals with exploring the capabilities in aerosol optical properties retrieval with the Flexible Combined Imager (FCI) onboard the Metosoat Third Generation satellite. The core of the study is to explore the improved retrievals in AOD using the VIS04 channel in FCI and compare the potential of these new retrievals with those in SEVIRI MSG. Authors also explore the capabilities of determining aerosol fine mode fraction using NIR22 channel. The topics are of interest to the scientific community and deserve publication in AMT. However, there are some concerns that need to be addressed before: First, I believe there is essentially nothing new in the retrieval algorithm and the title should be re-though. But most importantly, in my opinion the paper is too lengthy, and it is difficult to catch the main message. I believe that the manuscript can be shortened significantly but keeping the essence of the study.  But I understand that the style of the authors must be respected.

             Apart of my previous concerns related to the shape and structure of the manuscript, I believe that there is not enough discussion of the sensitivity of the retrieval to errors in the input optical data – or at least it as not been clear for me -. For example, I see many diurnal patterns in TOL, AOD… that seems to follow the pattern in SZA and I wonder if there is an artifact in the retrievals that amplifies errors. I also believe that the assumptions in surface and aerosol models are too strong when going to the real world. This should be at least discussed, although it is true that the information content for aerosols is low in the FCI when compared with more advanced imagers/polarimeters.

MINOR REVISONS

• Line 36: It is not true that you try to address particle size distribution. You are trying to estimate fine mode fraction. Please correct
• Line 48: FCI is not constantly covering the entire Earth, just part of it. Please correct
• Figure 1: It is not discussed in the text. What is the point of including Figure 1.
• Figure 2: Biomass-burning and dessert dust are illustrated as aerosol types. But in the real world, there are large variability and mixtures of biomass-burning and dessert dust. Assuming all aerosols within these types have the same properties is too simple, although it helps in satellite retrievals. I recommend being careful in the discussions.
• Line 139: I do not understand the expression ‘satellite synoptic’.
• Section 2.3 Simulation of Synthetic Data: I strongly recommend using a box-chart to help the reader better understand the different steps.
• Line 147: What is TOL? Please define the acronyms at least the first time they appear. Similar happen to others (i.e. MSA, DFS, SZA, RAA). Please, be careful.
• Section 3.1 Sensitivity Study: To me, a sensitivity study is to study the impact of random and systematic errors in the retrievals. Maybe it is necessary to rephrase this section.
• Line 259: I do not see clearly how Table 2 confirms NIR22 potential to distinguish between fine and coarse mode. Can you specify ?
• Table 3: What is w.r.t VIS06?
• Table 5: Can you explain better the first column? What is w?
• Section 4.5.1: I wonder how you are going to apply your inversion approach to real FCI data. Can you explain something?

This manuscript explores the additional information content gained from the new channels for the Flexible Combined Imager (FCI) on the Meteosat Third Generation-Imager compared to its predecessor, the Spinning Enhanced Visible and Infrared Imager (SEVIRI) on the Meteosat Second Generation platform. It uses synthetic top of layer reflectances generated using a combination of aerosol models using known exact atmospheric and solar and sensor geometric parameters as well as surface reflectance data derived from monthly GRASP/POLDER surface reflectance data which have been curve fit to the satellite spectral bands and then calculated at 15-minute intervals for two challenging case study regions, one in a desert dust event and another in a biomass burning event. An information content analysis is performed on retrievals for both AOD-only (to compare FCI with SERVIR) and AOD+Fine Mode Fraction (to highlight the potential for FCI to further characterize aerosol properties with the new channels).

General Comments:

The topic itself is incredibly relevant and the scientific community can stand to gain a lot from it, but I feel that this manuscript is not finished in terms of characterizing the uncertainties and utilizing them in their sensitivity analyses for retrievals using the different channels, particularly over the course of the diel cycle and at different AOD concentrations. The assumptions of constant error covariance matrices for both the input satellite observations as well as the a priori inputs to the aerosol models seems overly simplistic, and I worry about their impacts on the final results, especially as these are very complex aerosol case studies. I'd like to see if both the final retrievals and the sensitivity analysis significantly changes due to a dynamically-varying uncertainty at the very least for:

• the trace gas correction as a function of solar and sensor angles,
• the polynomial fit function to interpolate the GRASP/POLDER spectra to the FCI channels,
• a systemic bias due to the magnitude of the AOD retrieval itself as an a priori,
• the instrument uncertainty itself as a function of wavelength (~5% for the VIS and ~10% for the NIR), and
• a spectrally-invariant BRDF shape as a function of solar angle

This might be outside the scope of this manuscript as it stands, but while I do really like the evaluation of the information content changes for extreme aerosol events like desert dust and biomass burning, I was left wanting to see it also in the context of more "typical" retrievals like standard atmospheres over land and/or ocean to give a larger perspective on how much more information content we can gain from FCI in a more global context. 

Specific Comments:

• Line 84 (and throughout the manuscript): I think it might be better to call the channels by their center wavelengths (e.g. "FCI 444 nm channel") throughout the text, or at the very least have a table with the center bands and the FWHM (but I'd much prefer having it in the text to avoid the reader having to keep referencing it).
• Figure 1: This figure by itself seems like it could be removed for brevity, especially as the FCI data were still pre-operational. More attention then could be devoted to Figure 2 to highlight the FCI additions.
• Figure 2: Could we change the SEVIRI bands to grey (or something slightly different from the gridlines) and the FCI to blue to highlight them as the newer wavelengths? Also, the "Sand" color and the "Desert Dust" color are blending to my eyes (especially with the right legend overlaid so closely to the desert dust line). Could this be swapped to another color, and the spectral reflectance lines given a different weight or dotted/dash style to discriminate them better? This figure could also be partnered with a table going over the wavelength comparisons between FCI and SEVIRI (showing channel names, band centers, FWHM, etc).
• Section 3.1: This section might be better titled "Information Content Analysis" since it primarily analyzes the DFS which isn't quite the same as the uncertainties of the retrievals.
• Line 225: I wonder if this is an oversimplification to be set as constants when it could vary spatially and temporally depending on the input models (such as the polynomial fit uncertainty for the spectra, or the uncertainty from trace gases especially at high angles, the biases in retrievals due to AOD magnitude, etc). How would the impacts from these affect the information content retrievals performed here?
• Table 2: Instead of simply calling it case study 1 and 2, for clarity it might be better to call it dust and biomass burning.
• Figure 4: All three orange-red markers as well as the light green markers are very difficult to distinguish here.
• Line 338: Would a systemic bias in TOL reflectance from this influence the final results?
• Line 464: What were the values for these? Are they based on the Georgeot et al 2024 paper?
• Line 500: A 50x multiplier seems very arbitrary.What would different Sa values do for final retrievals?
• Line 585: Could you add a bit more detail about this correction in the text here? Just enough to cover how it's done for the full diurnal timeframe. As they use SMAC to correct for molecular effects, what are the minimum and maximum ranges of solar and sensor angles for these stations? How are the increased uncertainties here at sunrise and sunset accounted for in the sensitivity analysis? Also, there are two Ceamanos et al 2023, so there might need to be an "a" and "b" discriminant depending on how this journal does it.
• Line 595: Is there any random or systemic uncertainty that needs to be accounted for using this polynomial fit, in particular for the FCI 510nm and FCI 640/SEVIRI 635 channels, and, if so, how is it being accounted for? From Figure A1 it seems like there might be some systemic biases being introduced here.
• Line 602-603: How bad of an assumption would a spectrally invariant BRDF shape be? Is there a quick thought or reference you can put in here to give an estimate on how it might impact the retrieval capabilities, particularly at larger angles?
• Line 625: For clarification, this was used to estimate the surface reflectance in Figure A1b?
• 639: view of both solar and viewing zenith angles for the mu terms?
• Line 660: To me, Figure C2 looks to have a pretty similar error, if not more in some day/variable combinations, as C1. Could these be better visualized as a scatter plot or a Tukey mean-difference plot?

The paper presents a sensitivity study for the retrieval of AOD using FCI. The topic is important for the new sensor; however, the overall structure of the paper is difficult to follow. The authors divide the study into several sections based on different settings, channels, and methods, but the logic behind the sensitivity study is not clearly articulated. The authors are encouraged to reconsider the organization of the paper. For instance, the one-channel retrieval could be presented as a subsection alongside the two-channel retrieval, clearly indicating the added value of including more channels. Similarly, the distinction between the idealized cases and real-world cases should be better structured, with a clear conclusion on how ignoring real-world conditions affects retrieval accuracy.

Some more specific comments are listed below:

• P155–P160: The radiative transfer model used for the simulations is critical. More detailed information about the strengths and limitations of the aerosol and surface simulations within the model is needed. For example, is the two-layer assumption sufficiently accurate? What are the potential impacts of the aerosol vertical distribution on the results?
• Using only two scenarios to represent real-world conditions is insufficient, particularly for assessing anthropogenic contributions. Additional cases representing various anthropogenic aerosol types are recommended.
• Clarification is needed on how the AERONET-based aerosol models from MAIAC are implemented in the simulations. Specifically, which parameters are used as inputs? Models 6 and 7 are mentioned—are these fixed models, or are they updated using real-time AERONET data?
• Regarding surface reflectance, the authors state that POLDER products were used and fitted to the FCI wavelengths. A thorough validation of this fitting procedure is necessary. Why not use a well-established BRDF model for the simulation instead?
• In the Gaussian noise section, is the wavelength-dependent behavior of noise considered? If so, what are the impacts of this dependence?
• Degrees of Freedom for Signal (DFS) is a useful theoretical metric for understanding information content, but I strongly recommend that the authors also evaluate retrieval uncertainty by comparison with real measurements—for example, comparing simulated and observed surface reflectance, TOA reflectance, and retrieved AOD.
• P218: How are the observation and prior covariance matrices incorporated into the DFS calculation? This needs further clarification.
• The sensitivity study sections are largely qualitative. A more comprehensive and quantitative analysis is strongly recommended.
• Section 3.2.1: In the ideal case, why does such a large retrieval error occur? The authors briefly mention potential sources such as noise and the inversion method, but this requires more detailed investigation and explanation.
• Why is a one-channel retrieval necessary given that the sensor has multiple channels? Will both one-channel and two-channel retrievals be used in practice when the satellite becomes operational?
• All similar comments regarding Section 3 also apply to Section 4.