Evaluation of Satellite-Based Algorithms to Retrieve Chlorophyll-a Concentration in the Canadian Atlantic and Pacific Oceans
Round 1
Reviewer 1 Report
The authors provide the performance of two generic chlorophyll-a algorithms (a band ratio and semi-analytical one GSM) for two regions in the north area (Northeast Pacific (NEP) and Northwest Atlantic (NWA) during more 15 years (SeaWiFS, MODIS and VIIRS). They propose a new regionally-tuned versions of these two algorithms, which reduced the mean error (mg m3) of chlorophyll-a concentration modelled by OCx in the NWA from -0.40, -0.58 and -0.45 to 0.037, -0.087 and -0.018 for MODIS, SeaWIFS, and VIIRS respectively, and -0.34 and -0.36 to -0.0055 and -0.17 for SeaWiFS and VIIRS in the NEP.
This paper is of scientific relevance, well written and logically organized. They present several tables and figures including lot information published about the issue. To my opinion this work might be ready for publication. I have found few minor comments only (see below).
General comments
Language and grammar: generally, the manuscript is well written. Just a few sentences would need to be revised; there are some typos that would need to be corrected.
Title: the title reflects most of the author guidelines in the manuscript.
Abstract: the abstract presents a good summary of the manuscript. The context of the study is clearly defined. A suggestion could be to highlight the obtained results better.
Introduction: well written and exhaustive
Material and Methods: the methods are well written and the inputs dataset and data analyses are well detailed.
The authors present the original GSM algorithm in depth. This part could be briefer.
Results: The results chapter of results is clear. However, some minor comments should be addressed.
There are many tables of data that could lead to an appendix, since much information is given that is not discussed in the text
Discussion: in general, all the topics are well-treated and detailed.
Figures and tables: There are too many tables in the results section.
References: prior work are fully cited
Author Response
The authors provide the performance of two generic chlorophyll-a algorithms (a band ratio and semi-analytical one GSM) for two regions in the north area (Northeast Pacific (NEP) and Northwest Atlantic (NWA) during more 15 years (SeaWiFS, MODIS and VIIRS). They propose a new regionally-tuned versions of these two algorithms, which reduced the mean error (mg m3) of chlorophyll-a concentration modelled by OCx in the NWA from -0.40, -0.58 and -0.45 to 0.037, -0.087 and -0.018 for MODIS, SeaWIFS, and VIIRS respectively, and -0.34 and -0.36 to -0.0055 and -0.17 for SeaWiFS and VIIRS in the NEP.
This paper is of scientific relevance, well written and logically organized. They present several tables and figures including lot information published about the issue. To my opinion this work might be ready for publication. I have found few minor comments only (see below).
We thank the reviewer for the general comment about the manuscript.
General comments
Language and grammar: generally, the manuscript is well written. Just a few sentences would need to be revised; there are some typos that would need to be corrected.
We have read the manuscript multiple times to correct typos. As the reviewer did not point out the sentences that would need to be revised, we have left most of the manuscript in its original state except for modification made in relation to the reviewer’s comments.
Title: the title reflects most of the author guidelines in the manuscript.
Abstract: the abstract presents a good summary of the manuscript. The context of the study is clearly defined. A suggestion could be to highlight the obtained results better.
Introduction: well written and exhaustive
We thank the reviewer for this comment. We have taken special care to ensure that the abstract and introduction reflect the background knowledge, the objectives of the study and the main results.
Material and Methods: the methods are well written and the inputs dataset and data analyses are well detailed.
The authors present the original GSM algorithm in depth. This part could be briefer.
Again, we thank the reviewer for the positive comments. We think that the GSM algorithm requires some background information and the detailed description of the approach such that a non-expert can understand the method without referring to the original study. In addition, we have modified the approach to account for regional characteristics and performed a sensitivity study, for which the detailed description of the method is necessary to fully understand the changes we made. For that reason, we have decided to leave the description of the GSM approach as is.
Results: The results chapter of results is clear. However, some minor comments should be addressed.
There are many tables of data that could lead to an appendix, since much information is given that is not discussed in the text
We assumed that the reviewer was referring to tables 3, 4 and 5. We think that tables 3 and 4 contain pertinent information that is discussed in the results section (page 12 to 14). Some of the values reported for the different metrics (e.g., r2, MLE, MMLE) are used in the computation of the overall score that rank the performance of the algorithm, and therefore we think that it is important that these metrics are reported. The other ones, such as the merror are mentioned in the results and discussion sections (e.g., Page 18, Line 400).
We agree with the reviewer that Table 5 is rather long, which is due to the nature of our study that compares several methods and sensors. We have made an effort to move several tables to the Appendix and we feel that shortening the tables or moving some to the Appendix would remove important information that can be directly accessed by readers.
Discussion: in general, all the topics are well-treated and detailed.
We thank the reviewer for this positive comment. We have taken great care to discuss our results in detail and go beyond a simple analysis of performance by trying to explain some of the sources of uncertainties associated with the algorithms that we have tested. We hope that the readers will find the results interesting and will apply some of the finding to their own region of interest.
Figures and tables: There are too many tables in the results section.
As stated above, we think that the tables are providing pertinent information in support of the results and discussion sections. The main text contains 7 tables, which is on the higher end of the number of tables in a manuscript, but all the data is relevant and we have already added 10 extra tables in the Annex.
References: prior work are fully cited
We thank the reviewer for acknowledging our inclusion of previous work on the matter.
Reviewer 2 Report
remotesensing-620896
Title: Evaluation of satellite-based algorithms to retrieve chlorophyll-a concentration in the Canadian Atlantic and Pacific Oceans
This article reports a satellite product validation exercise (i.e. Chlorophyll-a product) for the Canadian waters, focusing the NASA products. Using an extensive in-situ dataset, they compare the performance of standard products against its regionalised-versions. The authors also investigate long-term variations associated to these datasets and try to evaluate potential impacts of phytoplankton communities in products accuracy retrieval.
The work presented is well written and organized and is very relevant considering the perspective of end-users such as the Fisheries and Ocean Canada. Although the in-situ sampling has not been conducted for satellite validation purposes, a large number of match-ups was identified allowing for a robust statistical analysis. This analysis however is not of the algorithm performance itself, as no radiometric data was available. Authors base their evaluation of the algorithms on the radiometric data provided by the sensors, i.e. they assume that atmospheric correction performs equally at all stations/sampling conditions. Considering that there are some very coastal stations this may not be the case but I believe authors have addressed this in section 3.4.2 by exploring the effect of latitude and bathymetry in the results obtained.
My only concern regarding this work refers to the evaluation of the regionalised versions of the algorithms. I was expecting that the regionalization coefficients would have been obtained with a random part of the dataset and tested with the other part. If I understood correctly, the coefficients were obtained by adjusting the algorithms to the whole dataset and tested against the same dataset. Considering the number of matchups obtained, I think the authors should try to make an independent testing.
Minor comments or suggestions.
Line 75, Figure 1 – I would suggest that a map with the sampling stations would be included. In alternative in Figure 1, a different colour (maybe black) could indicate the stations and this way the figure would show all stations and not just the ones with matchups.
Line 93-95 Note that liquid Nitrogen is not -80C, please correct.
Line 98 – Could you please specify what was considered to determine Chla concentration, was it just mv-Chla, epimer and allomer also included, or dv-Chla and any degradation products. Being a validation exercise I think this would be important to be stated.
Figure 3 – caption – It is not explained in the caption the meaning of the red and green colours (as it has been done in Figure 4)
Table 3 and 4 – caption- the difference between N and n should be stated in the caption.
Author Response
Dear Reviewers,
See attached PDF file for response to reviewer #2,
Best Regards,
E. Devred
Author Response File: 
Author Response.pdf
