Inversion and Analysis of Global Ocean Chlorophyll-a Concentration Based on Temperature Zoning

Round 1

Reviewer 1 Report (Previous Reviewer 1)

Comments and Suggestions for Authors

This is my second time reviewing the manuscript, now entitled: “Inversion and Analysis of Global Ocean Chlorophyll-a Concentration Based on Temperature Zoning”. I am glad by the series attitude of the authors, addressing my previous criticisms about the very limited data applied in the ‘inversion’ experiment. In the current version, after fitting the OC-CCI data in Oct 2018 for four different temperature ranges (<10, 10-20, 20-25, 25-30 degree Celsius), four OC3V-style retrieval models were constructed. The authors then tested the generalization capabilities of the four models to (1) daily retrieval for Oct 2018 and (2) other months (i.e., Jan, Apr, Jul, and Oct for 2017, 2018, and 2019). Moreover, the authors also added the comparison with in-situ data. In all cases, the zoning retrieval method presents considerable improvement compared with the standard OC3V scheme.

 

Putting all these together, now the study is scientifically sound. However, I still have some presentational problems, which inhibit the acceptance of the manuscript in its current form. I apologize for not mentioning these problems in the first revision, and raising these concerns only to improve the scientific value of this study. See follows. Given the efforts required, my recommendation is major revision.

1.     The discussion part is not well organized. Three aspects of the retrieval were discussed, including spatial autocorrelation, so-called clustering and outlier analysis, and hotspot analysis. The problem here is that these analyses were conducted for no reason, especially lacking any reference to justify them. If I were the authors, I would remove these analyses entirely. If the authors really like these analyses, then at least several examples using these methods in ocean sciences (especially for the ocean temperature or chl-a) should be provided, compared with, and discussed along with the results here. In addition, the analysis should be simplified (to the length of 1/3). The Introduction, Material and Methods should present a corresponding background and detailed description of the methods, while the Discussion part should focus on the main finding and the linkage with other existing studies.

 

2.     In my opinion, a more important question is the retrieve Chl-a results over space and time. In general, such a zoning retrieval often suffers from spatial discontinuity (also temporal discontinuity for the daily case). In this case, it would be problematic when using the product. Therefore, the authors should also show this is not the case for their method. This should also be covered in the discussion.

Comments on the Quality of English Language

The organization needs to be improved. The grammar is OK.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report (New Reviewer)

Comments and Suggestions for Authors

lns.148-9

"...temperature zoning can significantly improve the inversion accuracy of chlorophyll-a concentrations"

Question:

Usually, when accessing water quality, water temperature an Chl-a are actually measured in conjunction...

Is it possible to access to water monitoring data in order to confirm this correlation?

 

Abstract & text

> please explain the meaning of "SST"

 

ln.177

> excess of significant figures?

measuring temperatures with 1 decimal and estimating mean with 2 decimals?

 

2.2.2. In-situ Data

> try to access other countries "in-situ" data?

(collaborative studies to verify and generalize this relation in other waters)

 

2.2.3. Data Preprocessing > till the end of section 2

No reference to used software / programming languages / algorithms / figures...

> please add more information in respect to data processing!

(very nice figures!)

 

2.3.1. Temperature Zoning Idea

> I suspect that the relation "Chl-a - temperature" is not necessarily accurate!

What is the value of R² for a significant amount of data?

 

2.3.2. OC3V Inversion Algorithm Based on Temperature Zoning

> I was expecting to see some dependencies with temperature in eq.(1)... expressing the relationship "Chl-a - Temp."

Why do they not appear?

 

eq(1)

This equation is "non-linear" in respect to parameters...

a) Please add in text information on used algorithm for parameter estimation 

(what parameter optimization method was here used?)

b) have you estimated parameter errors?  (to evaluate respective significance and significant figures)

 

lns.260-262

"...the transition from a single-term polynomial with two coefficients to a cubic polynomial with five coefficients represents an improvement in accuracy."

> this may be due to an "overparameterized model" 

Have you statistically tested your model?

Have you validated your model?

 

Table 1

> "lg" is very similar to "Ig" -> preferable to use "Log" instead?

(if accepted this suggestion... please verify all Text!)

 

Figure 4. Accuracy Verification Plots

> observing these figures...

a) with OC3 model it seems that data become more fuzzy (higher dispersion and a curved tendency?)

b) for me OC2 model is more accurate - it seems to need some SHIFT to be totally aligned with the first quadrant bisector

Can you test this model?

 

lns.346-349

> you concluded that OC3 is better than OC2 based on FITTING performance indicators...

Preferable to chose in accordance to PREDICTED errors!

 

Figure 6

> please add mor information to figure legend

 

Table 3 / Table 4 / Table 5 / Table 6

> excess of significant figures in "MRE"

e.g.: 19.817% -> 19.8%

 

Table 7

> excess of significant figures in "z-score"

e.g.: 955.778 -> 955.8 (or 956)

 

5. Conclusions

I totally agree with your final comment...

"Future work will take into account a detailed study of temperature zoning and incorporate the temperature

variable directly as a parameter in the inversion algorithm."

 

Last question:

I do not understood the term "Inversion" in estimating Chl-a using remote sensing...

Is it due to the fact that...

- you are "transferring" "in-situ" to "remote sensing" evauation?

- you are "converting" Chl-a concentration to "Log(Chl-a)"?

- there is a relation Chl-a to temperature?  (this was not in fact demonstrated in your work)

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report (Previous Reviewer 1)

Comments and Suggestions for Authors

The authors addressed all my comments. Now the manuscript is acceptable. Congratulations!

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript entitled: “Inversion and Analysis of Global Ocean Chlorophyll-a Concentration Based on Temperature Zoning” shows a preliminary work of surface chlorophyll-a concentration retrieval scheme. By using a mere snapshot of Jan 2023, the authors used a temperature zoning technique to ‘improve’ chl retrieval from the OC3 scheme.

 

While the authors well reviewed existing literature on chl retrieval, I was surprisingly disappointed by the small data set used to test the idea. Only one snapshot (Jan 2023) was used. In this case, any crude subsetting of the data can improve the retrieval, not to mention the good relationship between SST and CHL. Since CHL retrieval is a very classic problem, I would suggest that the authors repeat the test over a global, long-term daily data set (or at least a regional data set with daily resolution) to systematically show the superiority of the proposed approach. Otherwise, perfect CHL retrieval for Jan 2023 was proposed, so what? We can still not well retrieve CHL for another time.

 

Moreover, the quality of English presentations should be improved to a large extent. For example, Lines 297-303 (fishing nets?) make no sense.

 

Based on the two major flaws, I recommend rejection of this manuscript.

Comments on the Quality of English Language

The English writing need to be improved to very large extent to be more precise and accurate.

Reviewer 2 Report

Comments and Suggestions for Authors

This is an interesting idea in that the authors propose to incorporate temperature as a variable in statistical chlorophyll algorithms applied to VIIRS. While in theory this makes sense, I have several issues with the way this was conducted:

1)Analyzing a single month is probably not sufficient, since you've automatically selected winter conditions in the northern hemisphere and summer conditions in the southern hemisphere, which probably introduces some phenological biases. 

2) My major concern is that VIIRS is already part of OC-CCI, and OC-CCI is not "true" data, it's merged satellite data with known errors. You should really complete this analysis with in situ data, which is also readily available (for example the SeaBASS NOMAD dataset) to show that the method provides improved statistics when using truly independent ("true") data. So really you are testing whether your version of VIIRS is biased compared to VIIRS (and other sensors) that OC-CCI processed. 

3) As implemented, you will get discontinuities whenever crossing temperature regions. A way around this would be to incorporate SST directly into the algorithm so that the response function is continuous. You could easily do that with VIIRS since it has colocated SST and Rrs at each pixel. This would also simplify your algorithm if SST is really important, since it would be a direct term in your modified polynomial (and that would be quite interesting), while also making for a continuous dataset. You could then assess accuracy (RMSE etc) for your temperature bands on that algorithm, to see if there are biases that remain corresponding to your large regions (although again, you would want to use more than a month of data because you introduce biases in other ways--ice melt might introduce more CDOM in some regions and less in others, for example, which would show up as "error" in the algorithm but would really be identifying mesoscale features). 

I really think you need to redo the analysis or at least add an analysis with NOMAD or similar datasets, since those are the ones that have been used to validate the underlying sensors (e.g. VIIRS) and composite products (OC-CCI). 

A few minor points:

You talk about "red tides" and high chlorophyll as being basically the same thing but a global analysis of chlorophyll will include red tides, diatom blooms, etc. so unless you are running some sort of PFT algorithm I would avoid emphasizing red tides so much. 

Line 113: if you are going to define the acronyms you should do so for all sensors, the first time it is used (which for MERIS and others was well before line 113)

 

Line 171-172: are those real limits, or just where the algorithm cuts off? 

 

Section 3.1. I’m not sure what fishing nets is referring to in relation to this analysis

 

Comments on the Quality of English Language

English is mostly fine but there are some typos and inconsistencies. Minor copyediting would be fine.