Combining Cirrus and Aerosol Corrections for Improved Reflectance Retrievals over Turbid Waters from Visible Infrared Imaging Radiometer Suite Data

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The algorithm itself is described in Gao, B.C., Li, R.R.,2023 (The VIIRS cirrus reflectance algorithm. Sensors, 2023, 23, 2234)
The present paper is the results of the algorithm application and testing.

The problem approached by authors is vital and the value of the algorithm itself and data processing and testing in different regions is high.
The tests performed by authors seem extensive and appropriate.
In my opinion the paper lacks presentation specially taking into account rather general conclusions about the applicability of the present algorithm as opposed to the standard NASA one.
Paper presents case studies.
I would suggest authors to add some statistics describing the overall performance of the algorithm for most of the data. E.g. average discrepancies between standard Rrs and Rrs retrieved by presented ATREM.
Namely, for example, statistical descriptions for discrepancies as in Figure 6D but for representative  amounts of processed data. For example several dosens/hundreds of spectra in  coastal, areas of Bahamas, Arctic, East China Sea and same for open sea / non coastal waters. Perhaps subsetting every 100th or every 1000th pixel in overlapping pairs of Standard-ATREM spectra.
These above are just suggestions for statistics. Final point is that some kind of statistics are needed.

To highlight the quality of cirrus corrected parts of the images I suggest adding Chlorophyll or any type of turbidity product to the case studies presented in the paper.
Rrs seem to have minor artifacts of of removed Cirrus it is not obvious, how much it affects higher level products.
My concern is that Rrs retrieved by ATREM are spatially relatively smooth but this artifacts may be non-linearly amplified but consequently used algorithms and e.g. Chlorophyll concentration in cirrus-corrected areas may turn out unusable.
I understand that in present stage of research it may be unwise to demand by all means smooth results of Chlorophyll and similar level products. In that case I encourage authors to apply some simplest qualitative estimates of suspended matter based on band difference (M4-M7 or similar). If that is not possible for some reason then at lease I encourage authors to speculate on the question and give some comment.
The question in short is: spatial continuity and discontinuity of VIIRS data in transition between areas without Cirrus and areas here Cirrus was removed.

Figure 1
Could you please improve and/or correct the figure. Perhaps extend the caption to the figure.
It is not very clear what is shown: RGB shows coastal turbid waters and some show/ice. At the same time Kd490 shows highest values right next to the shore (filled red in the picture), but lower values of Kd490 are clipped out from (made transparent). Obviously Kd490 must be around 0.5-0.01 in most of the picture but authors leave  the RGB in these areas.
Maybe if appropriate, just place two panels A and B with RGB and Kd490 separately?

Figure 2
As I understand VIIRS VSNIR and SWIR bands (green ans violet bars) are shifted vertically dimply to avoid overlapping in the picture. If my understanding is correct then no changes needed.

Page 12
 "on the same MaBook pro computer" should be, probably ' on the same MacBook pro computer"
 
Figure 7
Cirrus feature that is located close to the top right corner of a red bounding square seems to affect the Chlorophyll distribution (Figure 7E).
Dark green patch located at the top right corner of a bounding red square has color similar to Chl. features color from panel Figure 7A. This seems a bit suspicious because to the north-east  and to the south-west the color is more "whitish" and not similar to chlorophyll patch. (attached illustration)
Would not it be safer to mask the Rrs data that is located below cirrus clouds that are thicker/stronger than some threshold? It would be nice to add Chl retrievals obtained by (some) the same algorithm from NGSFC and ATREM Rrs.
Please, indicate  on the maps the location of water pixel that was used for panel Figure 7F.
Please, Please add red bounding box to panels G and H.

Figure 8B
Please explain what the blue circle filled with blue color mean.

Firure 8A and C
On the left side panel a there are come cirrus clouds.  Wide patch is stretched from north to south and right next to it there is a more narrow patch of clouds that goes from Nort North-West to South South-East. This 'diagonal' patch crosses the water and it seems that this causes come artifacts in the RGB image. The "vertical (north to south) also seems to cause some artifacts, but for me it is more difficult to see partially due to dense cirrus clouds.
Question is: how severe these artifacts are. They do not seem strong on the RGB image but may be critical when retrieving Chlorophyll or other substances in the water. (attached Illustration)
It would be nice to add Chl or suspended sediments or Kd490 maps to see the result on a high level satellite data product.
Perhaps add a transect in Rrs across the cloud to show presence or absence of the artifact in the Rrs data.

Page 15
"By comparing Fig. 9C with Fig. 9B, it is seen that our retrievals has more processed pixel"
should probably be corrected to
"By comparing Fig. 9C with Fig. 9B, it is seen that our retrievals have more processed pixel" (retrievals has to retrievals have)

Page 16
Could authors, please ad some comment if it is always the case or may happen sometimes when for example, for some reason inappropriate aerosol model is selected during the NASA atmospheric correction procedure.  "NASA Rrs algorithm treats thin cirrus clouds as ‘white’ aerosols during the atmospheric correction process"

Comments for author File: Comments.pdf

Author Response

Dear Reviewer,

Please see the attached pdf file for our detailed responses.

Regards,

Bo-Cai Gao

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript “Combining Cirrus and Aerosol Corrections for Improved Reflectance Retrievals over Turbid Waters from VIIRS Data” presents new atmospheric correction algorithms for retrieval over water that handles both turbid, shallow, and rough water conditions. The paper provides a detailed review and context for atmospheric correction and demonstrations both how the algorithms were developed, how they were applied, issues with the current algorithsm and why these issues exist, and both the new methods along with how they address current limitations. The provide good discussions on limitations for the new approaches.

This is a very strong paper. It provides a very concise and understandable review of atmospheric correction over water. The paper is well written and the results should be published.

I found no issues with the science or the case studies, they are sound and well presented.

The only issue I found was a tpyo in Section 2.3.2 (without line numbers it is difficult to provide the location).  It is in the paragraph that begins with “Figure 6 shows examples…” and is the second sentence from the end. The next paragraph starts with “Around 2015,…”. The statement ”…0.86 um were staturate over…” should read ”…0.86 um were staturated over…”. E.g., change saturate to saturated.

I also suggest breaking the last paragraph in Section 2.3.3 into two paragraphs. It is very long and contains two ideas. I suggest breaking  with the new paragraph starting with the sentence “With consideration of these practical….”

This a very good paper and I commend the authors.

Author Response

Dear Reviewer,

 Please see the attached pdf file for our detailed responses!

Regards,

Bo-Cai Gao

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

1. Overall summary.

The presented article is devoted to the consideration of the features of obtaining satellite quantitative information products about ocean reflectance. The authors reveal many important features of calibration and atmospheric correction of satellite multispectral ocean images and propose original algorithms designed to improve existing methods of thematic data processing in the field of OCEAN COLOR. I believe that the article is very relevant and must be published. Some disadvantage of the article is the modest quality of the illustrations presented. There are also questions about the validity of some of the conclusions. These shortcomings do not affect the fundamental importance and contribution of the work. Perhaps my comments will help improve the final version of this work.

2. Comments

2.1. In order to maintain academic rigor, it is advisable to accompany the first few sentences in the Introduction section on page 1 with references to the literature, since the authors publish specific important quantitative characteristics.

2.2. The literature [1] is used to substantiate the current status of the problem, but it dates back to 2019 (more than 5 years ago). I think it would be correct to rephrase the thesis at the top of page 2, or provide more references.

2.3 Fig. 8c - it seemed to me that the new algorithm treated the edge of the cloud as a water surface (see above to the left of the elongated area of the lake). How can the authors comment on this? Isn't there too much connivance with pixel rejection in the new approach?

2.4. The figures in Section 3 compare the insets ‘Rrs(GSFC)’ and ‘Water Refl Image'. I think the following can be recommended:

2.4.1 place the insets ‘Rrs(GSFC)’ and ‘Water Refl Image’ side by side  – this will provide easier perception when trying to compare;

2.4.2 label raster insets as well as graphs (‘REFL (GSFC OC Rrs 3.14)’ and ‘REFL (ATREM (MATCH)) / (ATREM (SIMPLIFY))’.

2.5. Are the graphs in the figures in Section 3 formed by one selected pixel (if I understood correctly)? Did the authors try to conduct experiments comparing the values of a statistically significant number of pixels generated by the basic and new algorithms?

2.6. Are some of the conclusions of the article unfounded in view of comment 5?

2.7. Technical aspects of illustration design:

2.7.1 Unreadable menu elements of the ENVI interface in Fig. 5. (it is better to delete them, since technical control during publication does not allow unreadable elements of drawings)

2.7.2 There is a blue spot in Fig. 8d. Perhaps this is a computer interface element that accidentally got into the screenshot.

With respect and best wishes!

Author Response

Dear Reviewer,

Please see the attached .pdf file for our detailed responses!

Regards,

Bo-Cai Gao

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

The manuscript describes combining the cirrus and aerosol corrections for improved retrieval of water leaving reflectance from Suomi NPP VIIRS data over coastal waters. A broad historical background of the atmospheric correction for the Ocean Color (OC) products is also provided. The strong point of the presented work is the description and demonstration how VIIRS band M9 is used for the cirrus cloud correction, beyond the usual application for cloud detection (cloud mask). Several case studies are discussed to show that the proposed algorithm works better than the one currently applied in the NASA OC processing. The other presented point, application of the atmospheric correction based on the SWIR (shortwave infrared) bands is already better established: the NOAA OceanView website referenced in the manuscript allows for quick comparisons of the NOAA OC products derived with atmospheric correction based on both NIR (near infrared) and SWIR bands. While the emphasis in the manuscript is on the NASA products, duality of VIIRS data processing at NASA and NOAA exists for both Level-1b/SDR and Level-2/EDR (e.g., Ocean Color), with many commonalities between the algorithms.

An early obstacle in applying the SWIR-based methodology for OC atmospheric correction was the high noise level in the MODIS SWIR bands. VIIRS performance is significantly better: noise in the SWIR bands is much lower. The authors demonstrate that particularly for band M9, noise in VIIRS data meets the OC needs. Selection of the glint-affected scene for the demonstration is somewhat unexpected because the glint areas are usually excluded from the routine OC processing. For the glint area, the observed variability (striping) is not noise, but rather a BRDF (bidirectional reflectance) effect also mentioned by the authors as “specular reflection by ocean surface facets”. While this effect for the multi-detector scanners is not often discussed, it was described in the study for the VIIRS MWIR (midwave infrared) bands that detect not only the thermal emissive radiation but also the reflected solar one (DOI 10.1175/JTECH-D-13-00054.1). Polarization effects discussed by the authors are less likely to contribute for Suomi NPP in comparison to NOAA-20/JPSS-1 VIIRS (see also DOI 10.3390/rs17010074).

Author Response

Dear Reviewer,

  Please see our detailed responses in the attached .pdf file

Regards,

Bo-Cai Gao

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Authors,

At present I feel the paper should be published.

No further comments or corrections are needed.

Reviewer 3 Report

Comments and Suggestions for Authors

Dear authors!

Thank you for your reply!

Best regards!