OC4-SO: A New Chlorophyll-a Algorithm for the Western Antarctic Peninsula Using Multi-Sensor Satellite Data

Round 1

Reviewer 1 Report

Dear Authors,

Line 442-445: Does The Importance term means the principal component or EOF? Must be changed for the principal component term, in order to understand the meaning.

Line 523-525: Must be included at end of the paragraph (Figure 10c). 

Line 544-545: Must be replaced "inadequate" per  "accuracy". The inadequate is not a statistical term.

Line 640 -644: The paragraph between  640-644 lines must be put in the 5. Final considerations.

Line 688-693: The lines between 688 and 693 must be eliminated, or put into the discussion section.

 

 

 

 

Author Response

[Reviewer]

Dear Authors,

Line 442-445: Does The Importance term means the principal component or EOF? Must be changed for the principal component term, in order to understand the meaning.

Line 523-525: Must be included at end of the paragraph (Figure 10c).

Line 544-545: Must be replaced "inadequate" per  "accuracy". The inadequate is not a statistical term.

Line 640 -644: The paragraph between  640-644 lines must be put in the 5. Final considerations.

Line 688-693: The lines between 688 and 693 must be eliminated, or put into the discussion section.

[Authors]

Thank you very much for your comments. We will now answer each comment point-by-point.

Lines 442-445: The importance term in random forest models is a measure of how much the model improves/worsens when a variable is either entirely excluded from the model or not selected a tree’s node, depending on the method used to calculate the importance term. In this case, we used drop-column importance, a method which first calculates the performance of a baseline model (with all predictors) and then compares the increase/decrease in performance when testing for each possible combination of predictors. The measure of performance used was the out-of-bag (OOB) error, a measure of accuracy widely used in random forest models (lines 280-282; Breiman, 2001). This information has now been added to the manuscript (see lines 307-308) and to Figure 7 (lines 467-469).

Lines 523-525: A reference to Figure 10c has now been included at the end of the paragraph (line 538).

Lines 544-545: We agree and have replaced term “inadequate”. The sentence was also rewritten to accommodate these changes (lines 557-558).

Lines 640-644: Thank you for the suggestion. This paragraph has now been added to section 5 (lines 696-702).

Lines 688-693: While we respect and understand the reviewer’s concerns, we believe this final paragraph is essential to quickly reinforce the need for and importance of accurate satellite measurements of phytoplankton under the context of climate change and potential effects on the Antarctic food web. If this paragraph was to be removed from the manuscript or moved to a less-visible section, we feel a less attentive reader could miss this important point.

We are grateful for the helpful comments and hope to have addressed your concerns.

Works cited:

• Breiman, L. Random Forests. Mach. Learn. 2001, 45, 5-32.

Reviewer 2 Report

Dear Authors I am happy to say that after a long time I am reviewing a manuscript didn't require any major comments. There are some minor typos that I believe you will address when you submit the final version of the paper. A very well written paper and deserves to be published in Remote Sensing.

Author Response

[Reviewer]

Dear Authors I am happy to say that after a long time I am reviewing a manuscript didn't require any major comments. There are some minor typos that I believe you will address when you submit the final version of the paper. A very well written paper and deserves to be published in Remote Sensing.

[Authors]

We are truly grateful for the kind comment and are glad that the Reviewer 2 appreciated the work. In the new version of the manuscript, we have corrected several typos and have added new relevant information as suggested by other reviewers. Again, thank you very much.

Reviewer 3 Report

Review Underestimation of chla in Arctic waters

 

This paper adresses the known problem of undestimation of chla in Arctic waters. The problem is well examined, statistics are excellent and show the underestimation at high chlorophyll concentration  It is of prime importance to estimate well Chla as it is a parameter of climate change impact that needs to be carefully followed over decades without biais. Using only HPLC measurements, they provide a new OC4 and use it to show that Chla exhibit a strong seasonal cycle, which is not shown with the CCI.

 

The hypotheses explaining this underestimation have been well exposed : 1) the fact they use the CCI 4.2 product which merges data from four sensors into daily measurements 2) not tight coincidence  with maximum difference of 12 hours between a given in-situ collection and a satellite overpass 3) low backscattering in arctic waters, 4) low fuco : Chla ratio. Moreover, as the data set of the study concerns mainly coastal data with only a few points offshore, it is evident that the error is higher for high concentrations as Chl a will vary much more at the coast than in the open ocean so the match ups at the coast are harder to obtain. Indeed, for low chla, it is not the case, the temporal and spatial variability is much less.

I have a few question regarding the data used. I understand that the match-ups concern only HPLC pigments. We have shown a 30% decrease when compared with fluorimetric data (at least in open ocean waters) for HPLC-Chla compared to fluorimetric Chla (Dupouy et al., 2018). In coastal waters, it may be more. Why consider only pure Chla ? and not consider the sum of Chla+ DV-chla  + pheopigments (if any) associated pigments to Chla all these pigments absorbing at 440 nm? While dv-chla proportion is probably nul in highly productive waters, it might influence the total absorption. In high latitudes, a reference read (Robinson et al., 2021) shows that at high latitude locations in the Southern Ocean, pheopigment concentrations can exceed mono-vinyl chlorophyll a concentrations, it may be the case in coastal Arctic waters. Have the authors try a correlation with the sum of all HPLC pigments linked to Chla and see how the correlation change and discuss the results ?

Their data confirm that, as previously suggested, the relationships between bio-optical properties and chlorophyll a in the Southern Ocean are different to other oceans

Nevertheless, the underestimation is strong and none of these bias (use of the sole Chla instead of sum of associated pigments to Chla as determined by HPLC, would not explain the strong underestimate of 50%.

 

 

The authors do not consider high CDOM as an explanation of the underestimation though I think it is a good explanation. As CDOM high concentration influences strongly the Blue channel and as CDOM maybe high in these coastal Arctic environments, it might lower the 440nm channel, causing artificially higher satellite chlorophyll and maybe explain the underestimate. The authors say that as no CDOM data are available, they can not evaluate the underestimation but could the authors make a litterature review of the subject to improve the discussion on that matter ?

Last, if the waters are essentially coastal water Type 13, evidently backscattering must be very strong and also biaise the result of the CCI algorithm. Could this be discussed also further ?

 

I would expect a brief description of the bathymetry of the region studied, missing in the Figure 1 of presentation. Could bottom reflectance influence the satellite signal ? This has also to be envisaged and discussed.

 

 

 

 

Author Response

Dear Reviewer 3,

Please see the attachment. Thank you.

Author Response File: Author Response.pdf