General comments:

The article deals with retrievals of spectral AOD and surface properties from a AHI image on board of geostationary satellite HIMAWARI-8 using GARSP software. Generally detailed properties with high temporal resolution are of great demand for atmosphere modelling community and article is within the scope of AMT. It seems that GARSP have never been applied to geostationary satellite so from its point it is a rather novel topic, although it is not fully clear if any modifications or improvements to already existing software were made to apply perform the retrievals. Article is well written, properly illustrated and referenced with good logical organization and structure, validation results are quite impressive not only by results (as opposed to the existing JAXA product) but also by sheer number of points used to gather statistics.

I’d recommend this article for publication with minor revision, below I enlist some points that in my opinion authors should consider improving.

Major comments:

Part of AHI+MPL retrieval is rather novel, I can’t recall if anyone did anything similar, and I do understand the desire of authors to share these results, although they outstanding a little from the paper and not emphasized in the title. It looks more like a proof-of-concept study, although well described, it provides significantly lower amount of observation. I’m not completely against having it in the paper, but I’d recommend authors to provide a better introduction to this part to emphasize the points I mentioned.

In the Introduction, the only algorithms described and compared are GRASP and JAXA/AHI one, consider adding an overview of other GEO based algorithms, or ones applied to GEO observations, it feels like general overview of remote sensing algorithms for GEO observation will improve readers awareness of state-of-the art in the field, and why some algorithms are considered more “next-generation” than others. I can suggest Dark Target by  Remer et al., 2020  e.g. (https://dx.doi.org/10.3390/rs12182900) and MAIAC by Zhang et al., 2011 (https://acp.copernicus.org/articles/11/11977/2011/) for e.g., but I believe take that there are so many GEO satellites out there, they should have a more excessive algorithm reference list. And I strongly believe making an overview of comparison between different GEO approaches will make the paper stronger, at least it won’t make it feel that GARSP is the only algorithm that can be applied to both LEO and GEO observations.

Minor comments:

Eq.19: In regards of height that sometimes is retrieved as a exponent parameter or profile, it is not clear how this equation changes when profile is retrieved, and S_h are not mentioned anywhere in eq19.

Eq. 21: delta_f_i are not described, assuming it is the same as for one pixel, I may conclude uncertainties are set up differently for each pixel however nowhere in text how accuracy is estimated for different pixels, please clarify.

Line 76: it is not clear if there any community-wide recognized “generations” of the remote sensing algorithms, consider elaborating more what stands out it from the others.

Line 356: “AHI/MPL retrievals, since aerosol loading is typically very low above 5 km, a prior estimate of 1.0-6 is set for the normalized aerosol concentration at the top altitude layer.” It is not clear why this a priori is applied and how, eq. above do not have such explications. In general it seems that there are some differences introduced by the presence of the MPL in the retrieval consider to make it clearer how AHI/MPL and AHI differ.

Table 3: I’m a bit confused by 0 and – in the table, does 0 means the constraint is effectively not applied? How’s that different from – then?

Table 4: It seems – have different signification in different columns of the table, consider clarifying it for readers, it is already rather hard to grasp due to excessive math. Consider noting “unitless” for the units column or something different.

Line 431: “at least 5 valid AHI/GRASP retrieval pixels should be available”, please clarify are these spatial or temporal pixels? It is not clear for which group of pixels standard deviation is calculated”

Figure 4: It’s a bit hard to interpret this figure, can we have a supplementary table with the same parameters somewhere below?

Line 574: “Both products have been re-gridded to 0.2°x0.2° spatial resolution”, can authors elaborate more or justify why this resolution was chosen?

Figures 8 and 10: Are there any other AERONET sites that can be marked in these areas?

Section 3.4. The profile analysis for several cases are quite nice, but is it possible to have layer to layer comparisons for all the cases combined on one scatter plot for general overview?

Technical comments:

Line 173: I believe it is not final layout, but this one is particularly bad with huge spaces, same for line 301.

Line 186:  (http://www.eorc.jaxa.jp/ptree/index.html) consider providing the last access date

Section 2.4: Consider mentioning somehow the version of the code used in the study.

Eq.16: “For i-th pixel” there’s no _i the equation anywhere, and in eq. 18 i is not explicitly described.

Line 345: “S is the matrix” it seems there are multiple matrices S_i.

Figures 6, 7, 16, 23-27: Personally I’m not in favor of captions like this, consider copy pasting the full caption, it’s not convenient to scroll up and down all day.

Figure 8:  Found it hard to find a black circle, consider mentioning that it is “circled in black (on the left edge of the map)” same for figure 9.

This paper is highly innovative. As stated by the author, it not only realizes the first application of the GRASP algorithm in the retrieval of geostationary satellite data, but also synergetic retrieval between AHI/Himawari-8 and micro-pulse lidar (MPL). The significance of this article is prominent, as it not only improves the accuracy of aerosol product from geostationary satellites and provides surface parameters synchronously, but also confirms the reliability of the vertical profile of MPL for satellite retrieval accuracy. But there are still some works in this article that need to be improved and perfected:

1. This article is actually very rich in content, but the title of the paper is relatively weak to generalise all the contents. It is recommended to consider modifying the title of the article to more comprehensively display the content of this article.

2.The resolution of some Figures in the article is too low, for example, Figure 21. It is recommended to increase the resolution of these images.

3.To improve the readability of this article, the author can provide some explanations of professional terms. For example, starting from line 56 of this article, the author discusses the impact of the surface reflectance on aerosol retrieval. line 63-65 introduce the solar geometry condition influence surface albedo. Suggest the author briefly mention the correlation and difference between reflectivity and albedo to improve the understanding of graduate readers. 

4. There are many Figures in this article, and the author usually lists each case one by one (day by day, for example Figure 22, 24, 26). Can we consider reducing the number of pictures and integrating and displaying these cases.

5. line 838-839, "the new approach also improves upon the potential issues of non-physical negative values or the abrupt spikes that may occur with the Fernald method." Indeed, for Fernald and Mie scattering lidar (do not like Raman or HSRL), some hypothesis parameters may introduce significant errors. However, the signal emission frequency of MPL is very high, and have more time/opportunity to realize the synchronous observation with geostationary satellites. Did the author attempt to average multiple MPL profiles to smooth out these errors or remove erroneous signals to ensure the validity of MPL data? After all, ground-based observation equipment is usually the reference for satellite.