Retrieval of Phytoplankton Pigments from Underway Spectrophotometry in the Fram Strait

Round 1

Reviewer 1 Report

Nice paper

i have only two observation..

- in the lines 165-171 have a lot information may be you can express all these in a table?

some figures are so small ...may be you can considerate made one figure by page

Author Response

We thank the reviewer for the compliment to our work. We give you the individual responses to each comment in blue.

 

i have only two observation..

 

- in the lines 165-171 have a lot information may be you can express all these in a table?

We have moved this information to Table 1.

 

some figures are so small ...may be you can considerate made one figure by page

We have resized Figure 8 that contains many small figures into one page.

Reviewer 2 Report

General Comments

 

The work reports adaptations and comparisons among procedures for retrieving pigment concentration from the coefficients of light absorption by phytoplankton (aph). The idea is to take advantage of a high-resolution dataset of aph, estimated using a underway system.

The work is well written and very clear.

 

Some suggestions:

 

-      Introduction could have a paragraph about the mean spatial resolution of pigment data available in the ocean 

-      Finding the potential number of pigments retrieved may be an objective of this work

-      Figure 2b may be better in B&W

-      As many of the equations are already available in the original publications, perhaps it would be interesting to prepare a flow chart diagram with all the procedures being compared and their respective “products”. There are a lot of abbreviation through the text and an illustration can be helpful.

-      Discussion and results could be merged. 

 

Few questions:

 

-      What was the final spatial scale of the estimates in the end?

-      Was it really 10% NaClO solution used? The methods usually call for a much weaker solution (Line 181). 

-      Lines 187-188- The use of the median suggests that some extreme values were observed, please explain. 

-      Would a fixed value for S (slope for the light absorption coefficient by NAP) interact with the measured Pigment Package Effect?

-      Why are Figure 3 so much bigger than Figure 6?

 

 

Minor comments

 

Line 44 – “their abundance”- their relative abundance 

Line 46 “In the current environment of climate change”- maybe rephrase.

Line 126 – “altering light conditions and nutrient transport.” – perhaps altering light and nutrient regimes. 

Lines 260-262 – Missing a reference to Moisan et al, 2011.

Line 350 -  can “High levels” be quantified? 

Line 354- p estimate significance not R2 

Author Response

We thank the reviewer for the comments that helped us to improve the understandability of the manuscript and the precision of the description. Please find our responses to each comment in blue in the attached file.

Author Response File: Author Response.docx

Reviewer 3 Report

Measurement of oceanic phytoplankton pigments plays a crucial role in understanding biological response to global climate change. This study assessed the performances of two approaches, Gaussian Decomposition (GD) and Matrix Inversion Technique (MIT), to estimate the pigment concentrations from phytoplankton absorption (aph) in the Fram strait. It is an interesting and important study to improve our understanding on the existing remote sensing methods. However, I still have several concerns need to be addressed before the consideration of its publication.

 

First, what is the rationale for selecting the DG and MIT methods, rather than others, like principle component analysis, and artificial neural networks stated in your introduction? That is, the advantages of the DG and MIT over other methods, especially correspondingly to your study, should be clearly presented.

 

Second, both DG and MIT have significant limitations. For example, it is unclear how to determine the number of Gaussian functions in the DG method. The authors used five in their analysis. Is it dependent on the characteristics of the study areas? How will the performance of DG be influenced if we used an inappropriate number of Gaussian functions? How can we determine a proper number if we have no ideas on a certain study area?

 

As for the MIT, in addition to the number of pigments discussed in the current manuscript, the number of spectral bands used for retrieval would also significantly influence its performance. Therefore, please clarify which bands have been used in this study, and how will its performance be influenced by using different wavelengths. Moreover, the MIT is a two-step approach, in which the HPLC measurements are needed to estimate the pigment-specific absorption spectra. Does it mean that the MIT cannot be applied using the sole observations of AC-S?

 

According the above limitations of DG and MIT, how can we expect the two methods to be operationally applied to the spectra data derived AC-S observations? Especially, when the prior-information, like HPLC data set, is not available. If we should not expect the DG and MIT to be used as operational methods to obtain pigment concentrations from AC-S data, what is the meaning for evaluating these methods?

 

Third, more details are needed on data collection in section 2.1. For example, the sampling intervals, depths for collecting water samples, velocity of the ship, should be provided. You should not send your readers to another paper to get the necessary information.

 

Last, do you assume that aNAP is not changeable for all the measurements of AC-S? In this way, you get the aph for each site by subtracting a constant aNAP from the total absorption, ap, derived from the AC-S? If so, what is the theoretic basis for this assumption? To what extent it will result in uncertainties in the aph measurements?

 

Specific comments:

Line 5, please add the abbreviation HPLC.

 

Line 23, “continuous” is very vague here. My first read impressed me that it was a time series of in situ observation, but practically the AC-S instrument cannot be fixed in the open oceans.

 

Line 141, why local algorithms are needed?

 

Line 156, “continuous” is not a proper expression here. In the real world, every kind of measurement is discrete, but the difference is sampling interval. I guess you mean “continuous” is that the AC-S data had been collected using a very high sampling frequency. But no matter how high it is, it should not be theoretically continuous.

 

Lines 203-204, why is the 20-minute average ap?

 

Lines 279-280, these should not belong to methods.

 

Line 323, why the leave-one-out cross-validation is used, even if you have 298 match-up of HPLC pigments and AC-S ap? May be you can try to simply separate the data set as calibration and validation groups.

 

Line 344, the maximum TChl-a is 3.87 in Table 1.

 

Lines 350-351, how this information can be used to improve the application of MIT method?

 

Line 365, why the leave-one-out cross-validation is need for the GD method?

 

Line 374, the regression coefficient B is slope? If so, revise it as slope.

 

Line 434, is it possible of clarify the minimum threshold of pigment concentration of the applicability of MIT?

 

Line 501, it is a time series of pigment concentrations, but the location of measurements also varied with the moving of ship. Please clarify it.

Author Response

We thank the reviewer for the comments.  They greatly helped us improve the manuscript. We provide the individual responses to each comment in blue in the attached file. Besides, we also worked on English language and style, as highlighted in the revised manuscript.

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

The athors have addresed my concerns properly. I agree to accept it as its current version.