Retrieving Aerosol Optical Depth over Land from Landsat-8 Satellite Images with the Aid of Cloud Shadows

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript presents a new method using the cloud shadow to retrieve aerosol optical depth. Usually, the retrieval of AOD is highly influenced by the bright surface. Cloud shadow and adjacent bright pixel can be combined to remove the influence of surface. The Landsat-8 satellite data is used to do the retrieval and the results are validated using the MODIS and the sunphotometer data.  Also, the results are further validated under  the 3D cloud effect and the BRDF surface. Overall, this is a great idea and could be an important supplement to previous AOD retrievals. I recommend to accept with minor revisions. I have two minor comments.  

1. The equation (4) is missed and should be in the position of line 172.

2. About using equations (1) and (2) to give equation (3), it looks like to me that E_f and S in the bright and cloud shadow pixels should not be the same because with and without clouds should have large differences to the reflection and transmission processes.  

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

Remote sensing of aerosol optical properties over land has long been a challenge due to the relative strong and unknown surface contributions. By taking the advantages of high spatial resolution satellite images, the authors of the draft proposed a new way to derive the aerosol optical depth by using the radiance differences over cloud shadows and adjacent bright pixels, and the surface information is not neccessarirly to be known. This work is sure brand new and reasonable for aerosol remote sensing.

There are some English language defects in the draft.

1. Line 15, "satellite measurements over shadowed and nearby illumniated" need to add "cloud" before "shadowed".

2. line 16, "Building on this principle" be replaced by "Based on the assumption".

3. Beginning of Line 19, the word "results" be "AOD".

4. line 40, "(Giles et al., 2019)" need to be deleted. 

5. line 64, "To evaluate its performance" need to be deleted.

6. line 82, "associated" to be deleted.

7. line 187, "using the aerosol three..." to be changed to "and the aerosol three...".

8. line 188, "...as input are used" changes to "...are used as input..."

9. line 192, "...includes 72 pressure levels" to delet "includes"

10. line 204, "the dots" be changed to "filled circles"

11. line 244, "rendering the plane-parallel atmosphere assumption invalid" be changed to "and the plane-parallel atmosphere assumption is not applicable"

12. line 295, "eliminated in the difference making process" be changed to "eliminated in the radiance difference algorithm."

13. line 303, "Leveraging" to be changed to "By taking the advantages of..."

14. line 306, "components" to "step"

15. line 313 need to be rewrited to make it clear.

16. line 323, "usable data" be changed to "reasonable".

 

 

Comments on the Quality of English Language

The English text may be slightly modified for better understanding. Specific comments listed in the previous comments. 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

RS-3363534 proposed a approach for retrieval of AOD using satellite image over land. Based on comparison with MODIS and sun-photometer, it shows a good availability and providing a option for calculation of AOD, especially in cloudy regions. Authors should consider and address blow questions before publication.

 

1. Line 36: aerosol optical properties are not just AOD, why only AOD was selected in this work?

 

2. Line 37：which instruments could observe or retrieve AOD at ground? Please describe briefly

 

3. Line 53: what means of 0.05±0.20*AOD?

 

4. Line 58：please explain the valuable of your work.

 

5. Line 71:give link to Landsat.

 

6. Line 91: why this two sites were selected for study in your paper? Do they have special characteristic or meaning?

 

7. Line 103: give link to AERONET

 

8. Line 118: give link to MODIS

 

9. Line 184: please elaborate the method of calculating AOD.

 

10. Line 197： will different cloud types influence results?

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 4 Report

Comments and Suggestions for Authors

The idea of ​​using cloud shadows to better identify AOD is interesting, but there are several potentially critical issues in the developed method (see below).

 

Major comments:

The developed method may contribute to satellite remote sensing of aerosols, but there is doubt whether the accuracy is as high as declared by the authors. Both the diffuse radiation component (E_f) and I_path depend on aerosol microphysical parameters, which are generally unknown but necessary for calculating E_f and I_path. AOD, which is part of the aerosol microphysical parameters, is determined by the method developed by the authors. This makes the procedure for AOD retrieval difficult. The microphysical parameters depend on the shape and composition of the aerosol particles and the way in which individual materials are aggregated. The influence of these effects on AOD, as well as on the aerosol asymmetry parameter (essential for modeling E_f), is clearly documented in the paper JQSRT 109, 2108–2123, 2008, which indicates AOD can change over wide range depending on how the particle core and shell are formed. The authors should provide a quantitative or at least qualitative assessment of the errors in obtaining AOD, taking into account the limits of the developed method (i.e., particles are considered to be homogeneous spheres).

 

Other comments:

The developed method may be a contribution to satellite remote sensing of aerosols, but there is a doubt whether the accuracy is not much lower than declared by the authors. Both the diffuse radiation component (E_f) and I_path depend on the aerosol microphysical parameters, which are generally unknown, but are necessary for calculating E_f and I_path. AOD is also a part of the aerosol microphysical parameters, which is determined by the method itself. The above-mentioned microphysical parameters are also strongly dependent on the shape and composition of the particles and the way in which the individual materials are aggregated. The influence of these effects on AOD, but also on the aerosol asymmetry parameter (which is essential for modeling E_f) is clearly documented in the paper JQSRT 109, 2108–2123, 2008, which I am a co-author of. I am not aware of any other work that would directly focus on the analysis of these effects. In the review, I drew attention to this work and its conclusions without any recommendation for the authors of the manuscript under review.

 

[lines 43-45] Satellite remote sensing does not cover the entire Earth's surface, so ground-based measurements are still an important part of AOD monitoring. Many measurements are still made using portable radiometers, which allow data to be obtained in different locations and with higher accuracy than satellites allow.

 

[line 56] This value is probably overestimated. Moreover, in the work [22] it is stated somethinf different, specifically "the global average fraction of ice clouds relative to all clouds range from 20 to 70%". The total amount of cloud cover over land is estimated to be more like 55% (IEEE Trans. Geosci. Remote Sensing 51, 3826–3852)

 

[Fig. 3] What do the individual colors mean?

 

[Tab. 2] I cannot identify the spectral range from the text in the right column.

 

[line 129] If I understand correctly, Fmask is a software product and you are using version 4.6.

 

[line 156] IIpath should be Ipath

 

[Eq. 2] the values of Ipath in Eq. (1) and Eq. (2) do not equal in case the clouds are present in the atmosphere

 

[Eq. 5] Eq. (5) should be on line 173.

 

[line 182] terrain orography must be known to make the calculate of Rayleigh optical depth possible

 

[line 184] Both Ipath and Ef depend on tau and many other factors, such as aerosol asymmetry parameter, single scattering albedo ... These parameters are unknown. So how do you perform numerical simulations?

 

[Fig. 6] It is not entirely clear from the color indicators how shadow retrievals differ from sun-photometer measurements. For example, in the right plot, the blue ranges from about 0.3 to almost 0.8, which is a huge span compared to the accuracy of the method indicated by the authors.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 4 Report

Comments and Suggestions for Authors

I am still not satisfied with the corrections made to the manuscript. For example, the authors did not sufficiently address the question regarding the different values of Ipath in Eq. (1) and Eq. (2). These values will differ more significantly with an increase in cloud fraction. Despite the presence of cloud windows, which still allow for remote sensing of the Earth's surface, Ipath may not account for the scattering of light along the entire trajectory from the Earth's surface to the satellite. In the case of cloud masking, this trajectory will be shorter and will only include scattering in the atmosphere above the cloud layer. These differences in Ipath values can be significant, especially in polluted areas, because aerosol concentration usually peaks near the ground. The lower atmosphere then contributes significantly to the scattered signal, but need not be included in Ipath in case of cloud arrays. Therefore, the question remains: what are the limits of the method developed? Was the error ever estimated by the authors?

Author Response

Please see the attachment

Author Response File: Author Response.pdf