The article titled "Spring and summertime aerosol optical depth retrieval over the Arctic cryosphere by using satellite observations" provides an analysis of AEROSNOW-retrieved satellite aerosol optical depth (AOD) statistics and validation using AERONET data over the Arctic region from 2003 to 2011. However, the main objective of the study is not clearly defined. The introduction highlights the importance of AOD retrieval using passive sensors, which suggests an algorithm development focus, which is suitable for the AMT journal. Nonetheless, the majority of the article discusses the distribution of retrieved AOD over the Arctic region and comparisons against AERONET data, with a greater emphasis on understanding the AOD distribution during spring and summer. The content of the article would be better suited for journals with a stronger emphasis on scientific aspects rather than technical aspects, as found in AMT. If the article intends to evaluate the accuracy and uncertainty of the retrieved AEROSNOW AOD, further investigation into uncertainties and comparisons with field campaign data would be necessary.

Several significant sources of uncertainty are mentioned in the article but not adequately addressed. Firstly, cloud contamination poses a major uncertainty source, requiring further examination. Additionally, the assumption of a fixed snow surface parameterization is mentioned but not sufficiently analyzed. The article should discuss the uncertainties associated with these assumptions and explain why they hold true for the study region.

In line 94-95, the author claims that the cloud identification algorithm meets the requirements for high-latitude AOD studies, but this statement requires a citation to support it. Additionally, the phrase "a given sampling period" in line 96 needs to be clarified since a larger time window could introduce risks to this assumption. Line 172 introduces the use of cloud fraction as a parameter in the quality flag, yet the uncertainty associated with cloud fraction over the Arctic region is not discussed. It would be valuable to explore whether the limited impact of cloud fraction is due to the large uncertainty associated with this parameter.

In line 118-119, the article mentions a fixed snow surface parameterization, the uncertainties related to this assumption should be analyzed and discussed. Specifically, it is important to explain why this assumption is valid for the study region, considering the Arctic's limited precipitation. Additionally, line 166 mentioned mixed snow regions, but its impact on AOD retrieval is not mentioned.

Some additional points to address include providing information about the basic aerosol properties of the models used (e.g., single scattering albedo and asymmetry factor) in line 137. Furthermore, the impact of a solar zenith angle cutoff of 75 degrees mentioned in line 160 should be discussed in terms of its impact on data sampling, particularly if there are specific times within a season when aerosol retrieval is not possible. Additionally, all monthly plots should display the variation in data for both AERONET and AEROSNOW datasets, including the number of retrieved data points aggregated into the monthly data. Lastly, in line 233, Figure 8 is introduced before Figure 7, which should be corrected.

Overall, the article would benefit from addressing these points to enhance clarity, provide a more comprehensive analysis of uncertainties, and align the content with the appropriate journal's focus.

The manuscript "Spring and summertime aerosol optical depth retrieval over the Arctic cryosphere by using satellite observations" carries out an analysis of satellite AOD retrievals over ice and snow. Development of satellite aerosol retrievals over ice and snow is a highly important topic and would require some further development to improve our understanding of, for example, the aerosols over the Arctic regions.

The title of the manuscript indicates some kind of development work regarding AOD retrievals over cryosphere. However, the manuscript is mainly about an analysis of AEROSNOW algorithm produced AOD from AATSR. The manuscript is basically lacking the development work that the title is suggesting. This makes me think if AMT is a suitable journal for this type of work.

The analysis shown in the manuscript is based on AEROSNOW algorithm. The authors state that the AEROSNOW algorithm is based on Istomina et al. (2011) that has been further developed by the authors of this manuscript. There are no citations to any work that would describe the further developments by the authors neither any detailed descriptions is given in this manuscript. If this manuscript describes the further development, it is very much lacking the details necessary even to basic understanding on how the algorithm works. There are no details to reproduce the results used in AEROSNOW retrievals based on this manuscript or at least it would require an extensive literature search and details from multiple cited papers. This is a big issue in this manuscript.

The performance of the AEROSNOW show in the manuscript is in principle very good. I, however, feel there is a major issue in the analysis. The authors derive a quality flag (QF) parameter. The derivation and use of this parameter is not clear in the manuscript. As far as I understand both AERONET and AEROSNOW AOD are used in computing these results. Furthermore, the QF parameter is later applied to post-process filtering of the AEROSNOW data. This makes the AEROSNOW data dependent on AERONET and it is a major issue as no independent validation data is available. Then all the analysis is based on the dataset that has a dependency on AERONET data. This may result in overoptimistically good results. As mentioned, due to unclear description of the QF parameter I may have misunderstood this as well. In any case, the derivation and use of QF needs to be clarified and analysis of the results needs to be carried out using independent AERONET data.

The English language itself used in the manuscript is mainly good and understandable. However, the contents of the manuscript are often confusing and may even be misleading. In my opinion, the manuscript would require some re-organizing.

More specific comments:

• l.8 "The AOD is retrieved assuming that the surface reflectance observed by the satellite can be well-parametrized by a bidirectional snow reflectance distribution function, BRDF." More detailed analysis on the effects of this assumption needs to be carried out.
• l.19 As Arctic is changing rapidly this almost 20 year old study already gives outdated information. More recent studies indicate AA of about 4 (e.g. Rantanen et al., "The Arctic has warmed nearly four times faster than the globe since 1979", Communications Earth & Environment, 3:168, 2022).
• l.104 "...an iterative procedure to obtain the AOD over Arctic snow and ice regions at 555 nm." No real details of the retrieval are given, e.g. what is the cost function in the iterative procedure.
• l.118 "We fixed the free parameters for the entire time series of AATSR, which involved a
  
  fixed snow grain size and snow impurity assumptions." How do these assumptions affect the retrievals?
• l.161 Please define the NDSI.
• l.166 To better understand the effect of post-processing with NDSI, it would be nice to know how much of the data was filtered out and how much the metrics changed because of the filtering.
• l.171 As mentioned earlier. The derivation and use of QF parameter needs to be clarified. Also independent validation data to validate AEROSNOW needs to be taken into account. Filtering cannot depend on AERONET data.
• l.195 "Uncertainties due to both space and time sampling differences are minimized." No details are given on how this is done and if the uncertainties really are minimized.
• l.202 "...the latter does not take into account the relationships between geophysical variables." I do not believe this statement is true. Ordinary Least Squares regression is used exactly to determine the linear relationship between variables. Ordinary least squares does not take into account the uncertainties in x variable at all and, for example, therefore leads to biased slope estimates. Maybe this is what the authors meant here.
• l.203 "Therefore, it is recommended that the use of reduced major axis (RMA) regression replace the use of standard linear regression" RMA regression is not the only one tackling the issues of OLS. There are many other methods as well. This statement written by the authors may give the reader a wrong impression here.
• l.218 "This difference can be explained by using an optimal chemical transport model by separating the total AOD to aerosol components." This is a vague statement not explained well enough and not supported by any results shown in this manuscript.
• l.246 The whole paragraph is a bit confusing. The paragraph is in the results section and no comparison between AEROSNOW and model data has been carried out.
• l.254 "The AEROSNOW algorithm uses the dual-viewing capability of the AATSR instrument to minimize retrieval uncertainties" It was not shown that the use of dual view instrument really minimizes the retrieval uncertainties.
• l.255 "It showed good agreement with ground-based AERONET observations, with a correlation coefficient R = 0.86 and a low systematic bias." It may be totally misleading to give retrieval metrics without stating any details. Use of different time averaging for example hugely affects the metrics.
• l.256 "The high anthropogenic aerosol loading (Arctic haze events) due to long-range transport over Arctic snow and ice is captured by the AOD determined by AEROSNOW." There were no well justified experiments in the manuscript that would confirm this exact statement about long-range transport of anthropogenic aerosols. The authors themselves listed this type of experiment as future work.
• l.263 "The promising AOD results obtained with AEROSNOW indicate that these can be used to evaluate and improve aerosol predictions for various chemical transport models." In previous parts of the manuscript, the authors say that in the future models can be used to test and validate the AEROSNOW retrievals. Here in the conclusions, the authors say that AEROSNOW can be used to evaluate and improve the models. This is confusing.