Retrieval of Snow Albedo and Total Ozone Column from Single-View MSI/S-2 Spectral Reflectance Measurements over Antarctica

Round 1

Reviewer 1 Report

This report presents a snow albedo and ozone column retrieval algorithm by using Sentinel-2 MSI images. The validations show accurate retrievals. The high resolution snow albedo in Antarctic is new to the community. However, I have the llowing questions:

(1) Why only ozone is considered in the atmosphere transmmitance? Why rayleigh scattering, O2, CO2 and H2O absorption is neglected in this study? This should be justified further.

(2) Broadband albedo is the result of reflectance in wider spectral range than Sentinel-2 MSI sensor (900-2500 nm), the albedo at 900-2500 nm is determined by grain size in shallower layers than 400-900 nm (penetration and vertical inhomogenuous problem). I think this limitation should be mentioned.

Author Response

This report presents a snow albedo and ozone column retrieval algorithm by using Sentinel-2 MSI images. The validations show accurate retrievals. The high resolution snow albedo in Antarctic is new to the community. However, I have the next questions:

• Why only ozone is considered in the atmosphere transmittance? Why Rayleigh scattering, O2, CO2 and H2O absorption is neglected in this study? This should be justified further.

We select channels free of O2, CO2 and H2O absorption for our retrieval technique. The Rayleigh scattering is low for selected channels. In particular, the Rayleigh optical depth is smaller than 0.1-0.15 depending on the ground elevation for the MSI/S-2 channels. The aerosol load is also low. The paper has been modified (see Conclusions).

• Broadband albedo is the result of reflectance in wider spectral range than Sentinel-2 MSI sensor (900-2500 nm), the albedo at 900-2500 nm is determined by grain size in shallower layers than 400-900 nm (penetration and vertical inhomogeneity problem). I think this limitation should be mentioned.

We have mentioned this limitation in the paper (see Conclusions).

Reviewer 2 Report

Kindly find my detailed comments in the attached file.

Comments for author File: Comments.pdf

Author Response

 

Major comments

1. The negative sign is missing in the provided equation (Eq. 5). The equation for the value of L must be checked.

The misprint in the expression for K has been corrected. The expression for L given in the paper is correct. The absorption of light by ozone at 864.7nm is too small  to be accounted for.

2. Regarding snow grain size: If the ??? can be calculated using ??? = L⁄? and the value of ? is also assumed using previous research then how is this connected and can be derived using the algorithm proposed in the manuscript?

The relationship between the effective absorption length L and def is given by Kokhanovsky et al. (2019). We use this relationship to derive ??? from L.

3. Authors should provide TOC and albedo maps for the entire region first so that readers can get an idea regarding the applicability of the algorithm. Then authors can compare 1 or 2 observations with ground measurements to justify the algorithm.

The total ozone maps based on 300m OLCI/S-3 have been provided by Kokhanovsky (2021b). We do not expect major differences, if one uses 10m MSI/S-2 data due to relative horizontal homogeneity of the surface at Dome C site. The discussion of the maps is out of scope of this paper.

4. Page 2, Line 70, it is mentioned that L is the effective light absorption path. On page 5, line 141 manuscript states l is the effective absorption length. Kindly explain how they are different? Also, the letter l used to indicate the effective absorption length (Eq. 6 onwards) looks very similar to number 1. This is confusing with equations where numbers and letters are both used. I would suggest using a different letter for L the variable to avoid such confusion.

L is the effective light absorption path (ELAP). It is determined by Eq. (2). The effective absorption length (EAL)  l is determined by Eq. (6). The difference is due to the fact that ELAP depends on the geometry of observation and EAL does not depend on incidence and observation angles. They are related via Eq. (7). The details are given by Kokhanovsky (2019).

We propose to keep the variables as suggested in the paper.

 

 

Minor comments

 

1. Page 5, Line 142: Is it spherical or spectral albedo?

It is spherical spectral albedo (white sky spectral albedo).

2. 3: Legend mentions December 12, but the description says December 19.

Corrected.

3. Since the algorithm is applicable only over Antarctica, there seems to be minimal use of this algorithm in the future. Is there a way to extend this to other regions of the globe to improve its utility?

This is out of scope of this paper. Antarctica is twice the size of Australia and plays important role as far as climate change issues are of concern. Therefore, we assume that the algorithm will be used in the future. We are working towards the improved algorithm to be used over Greenland and other polar regions including cases of polluted snow and atmosphere.

 

Reviewer 3 Report

This work presents an interesting procedure to estimate narrowband albedo and ozone content at high spatial resolution using Sentinel 2 reflectance data. The procedure presented is clearly explained, making it easy to reproduce the results by other authors.

The main concerns lie on the strong approximations imposed in the calculations:  no pollutants absortion by the snow, no atmospheric scattering and atmospheric absorption by ozone only. The areas on Earth where this model seems to fit is central Antarctica. The authors should include a brief discussion about the validity of the model, stating clearly where they expect the model to work, as well as the main source of error when applying the model to other areas.

I think that it would be a good idea to include a case study out of central Antarctica for comparison.

A minor coment: In figure 1, the authors use the word "measurements" to refer to Sentinel 2 data. This is a bit confusing, becasue "measurement" seem to indicate data collected in the field. The authors should refer to "Sentinel 2 TOA reflectance data" instead.

Author Response

This work presents an interesting procedure to estimate narrowband albedo and ozone content at high spatial resolution using Sentinel 2 reflectance data. The procedure presented is clearly explained, making it easy to reproduce the results by other authors.

The main concerns lie on the strong approximations imposed in the calculations:  no pollutants absorption by the snow, no atmospheric scattering and atmospheric absorption by ozone only. The areas on Earth where this model seems to fit is central Antarctica. The authors should include a brief discussion about the validity of the model, stating clearly where they expect the model to work, as well as the main source of error when applying the model to other areas.

We agree that the applicability of the technique decreases for areas with polluted snow and atmosphere. We include the discussion of this issue in the paper (see Conclusions).

I think that it would be a good idea to include a case study out of central Antarctica for comparison.

The measurements of snow albedo in Antarctica are rare. Also MSI/S-2 measurements over a given site occur only 5 times or so per month (including also the cases with clouds). In addition,  the intercomparison of ground and satellite measurements require the survey of a satellite ground pixel at many locations or horizontally homogeneous snow surfaces. Such datasets are not available to us. Although we do not expect major problems with the technique for the satellite observations over Antarctica.

A minor coment: In figure 1, the authors use the word "measurements" to refer to Sentinel 2 data. This is a bit confusing, becasue "measurement" seem to indicate data collected in the field. The authors should refer to "Sentinel 2 TOA reflectance data" instead.

The text has been changed.

 

 

Round 2

Reviewer 1 Report

The two concerns are addressed in the conclusion part, this report and can be published.

Author Response

Thanks for your comments.

Reviewer 2 Report

Most of the concerns raised earlier are clarified or corrected by the authors. However, I still have few issues which need clarification before I can recommend it for publishing:

1. The authors stated that "The relationship between the effective absorption length L and def is given by Kokhanovsky et al. (2019)." This was my point. If this has been already done in a previous paper then how does it add anything new in this paper? So stating specifically about snow grain size retrieval using this new algorithm is completely misleading. The way it is present in the abstract and the text indicate that this new algorithm is utilized to determine snow grain size.
2. Regarding maps: I understand and accept the response regarding TOC map. However, for surface albedo we need to see if the algorithm actually provides reasonable retrievals for other locations in Antarctica too? We cannot accept a new algorithm which works on only one data point at one location. If the generation of map is not possible then authors need to retrieve and compare surface albedo for multiple instances of the same location. Authors can get the surface albedo of Dome C for multiple days during the satellite pass over the region and then compare it with ground observation.

Author Response

Most of the concerns raised earlier are clarified or corrected by the authors. However, I still have few issues which need clarification before I can recommend it for publishing:

The authors stated that "The relationship between the effective absorption length L and def is given by Kokhanovsky et al. (2019)." This was my point. If this has been already done in a previous paper then how does it add anything new in this paper? So stating specifically about snow grain size retrieval using this new algorithm is completely misleading. The way it is present in the abstract and the text indicate that this new algorithm is utilized to determine snow grain size.

The text has been corrected (see line 151).

Regarding maps: I understand and accept the response regarding TOC map. However, for surface albedo we need to see if the algorithm actually provides reasonable retrievals for other locations in Antarctica too? We cannot accept a new algorithm which works on only one data point at one location. If the generation of map is not possible then authors need to retrieve and compare surface albedo for multiple instances of the same location. Authors can get the surface albedo of Dome C for multiple days during the satellite pass over the region and then compare it with ground observation.

The temporal and spatial collocation of 10m satellite pixels and ground data is not an easy task. Both satellite and ground data must be quality - assured and performed for the case of horizontally homogeneous terrain. This is a separate task, which we would like to explore in our future work. We have added 4 additional satellite retrievals of the spectral albedo (and grain size) at Dome C to this paper (see new Table 4 and Fig.4).

Reviewer 3 Report

The authors have responded to the questions posed in the revision.

Author Response

Thanks for your comments.