Estimation of the Particulate Organic Carbon to Chlorophyll-a Ratio Using MODIS-Aqua in the East/Japan Sea, South Korea

Round 1

Reviewer 1 Report

I have reviewed the manuscript “Estimation of the particulate organic carbon to chlorophyll-a ratio using MODIS-Aqua in the East/Japan Sea, South Korea” by Lee et al. The authors developed a new regional satellite-POC algorithm based on their POC observations. Using their algorithm, they investigated the spatiotemporal variability of POC and POC to chl-a ratio based on satellite data. Considering that POC and POC to chl-a ratio are important in the marine biogeochemistry, many readers may be interested in this study. However, the authors may need to give more concrete results (insight discussions) and need to improve the writing before officially publish this manuscript. Below, I list some of the suggestions.

 

Major comments:

The methodology part can be detailed. For example, the authors may give more details on the calculation of POC to chl-a ratio, and the monthly distributions. When reading the title of the manuscript, I thought that the manuscript focuses on the estimation of POC to chl-a ratio. Of course, the authors explored the variability of POC to chl-a ratio at the end of the manuscript. However, I felt that this manuscript is more like a regional satellite-POC algorithm. If the authors prefer to emphasize the POC to chl-a ratio which I think is more interesting, they may further analyze the relationship between POC to chl-a ratio and environmental and physiological parameters.

 

Minor comments:

Line 42: change “are still not clear” to “remain unclear”?

Line 45: change “are variable depending on” to “depend on”?

Lines 55-57: Rephrase the sentence? For example, the sentence may be rephrased to “This study aims to: (1) develop a new satellite-POC algorithm; and (2) investigate spatiotemporal variability of the POC to chl-a ratio in the East/Japan Sea.”

Lines 90-92: We can determine the model parameter using linear region in the logarithmic space. Am I right?

Line 117: change “not found” to “found”?

 

Author Response

The methodology part can be detailed. For example, the authors may give more details on the calculation of POC to chl-a ratio, and the monthly distributions. When reading the title of the manuscript, I thought that the manuscript focuses on the estimation of POC to chl-a ratio. Of course, the authors explored the variability of POC to chl-a ratio at the end of the manuscript. However, I felt that this manuscript is more like a regional satellite-POC algorithm. If the authors prefer to emphasize the POC to chl-a ratio which I think is more interesting, they may further analyze the relationship between POC to chl-a ratio and environmental and physiological parameters.

First of all, thank you for your precious comments. In this study, we focused on the “accurate estimate” of POC and POC to chl-a ratio since the existing POC algorithm in the East/Japan Sea commonly showed a poor accuracy. Therefore, this study is a kind of ‘pilot study’ to derive a more accurate POC algorithm in the East/Japan Sea (also this is a technical note). We will prepare the detailed article about the POC to chl-a ratio in the East/Japan Sea. The methodology part was revised and added detailed explanations.

 Line 42: change “are still not clear” to “remain unclear”?

We revised the sentence in line 42

Line 45: change “are variable depending on” to “depend on”?

We revised the sentence in line 45

Lines 55-57: Rephrase the sentence? For example, the sentence may be rephrased to “This study aims to: (1) develop a new satellite-POC algorithm; and (2) investigate spatiotemporal variability of the POC to chl-a ratio in the East/Japan Sea.”

We revised sentences in lines 55-57

Lines 90-92: We can determine the model parameter using linear region in the logarithmic space. Am I right?

We didn’t tried, but maybe we can determine the model parameters in the logarithmic space with linear regression.

Line 117: change “not found” to “found”?

We revised the sentence in line 126

Reviewer 2 Report

To begin with, I commend the authors for presenting their work as an empirical regional algorithm for the East/Japan Sea. I have seen may reports where claims are made way beyond what the work or algorithm can actually do. It is an important result reported here that the P)OC/Chla algorithm derived from and applicd to the open ocean is probably not adequate for shallow, productive coastal seas. I will have a few comments on the presentation of what appears to be a sound report.

Page 2, lines 65-66. It is a common misconception that the “pore-size” of Whatman glass microfiber filters has a Diameter of 0.7 μm. The structure of glass microfiber filters does not allow for a pore-size to be estimated and what is reported is actually a mean pore size determined from a range of pore sizes possible for this filter type, glass fibres randomly positioned one on top of another. Chavez (1995) has reported quantitative recovery of phytoplankton down to 0.2 μm in the north Pacific with GF/F filters.

Page 2, lines 67-68. Since this is a preliminary report on the effort to determine an algorithm for POC/Chla for the East/Japan Sea, I would recommend the authors calibrate the algorithm from Chla determined by HPLC. Fluorometry is chancy and empirical and not nearly as certain as an HPLC determination of Chla. I say this only because the effort expended here is to calibrate the remote sensing algorithm and this requires the most reliable determination possible of Chla.

Page 2, lines 70-71. The authors need to expand on the rationale of the stable isotope method for determining POC.

Page 6, lines 134-137. First of all, the authors are to be commended for a correct plot of modeled variable (algorithm POC on x-axis) and in-situ variable (measured POC on y-axis). The majority of studies I have reviewed plot it in the opposite fashion. Thus the plot in Fig. 7 correctly shows the ability of the model to predict the in-situ measurement. Indeed, the argument that the algorithm proposed here appears visually to have a slope in this plot that approaches 1.0 is a good one. However, I think the argument would be strengthened by calculating the regression slope for the coastal algorithm and using standard regression tests to show that the slope of the plot does not differ significantly from 1.0. I suspect that the same effort for the open ocean algorithm would show that the slope does differ significantly from 1.0. It is interesting that both algorithms appear to have the same y-intercept when the plots are extended until they cross the y-axis.

Page 6, lines 144-146. Again, I recommend that the authors use chlorophyll concentration derived from HPLC analysis to further validate and calibrate the algorithm proposed here.

Reference:

Chavez, F. 1995. On the chlorophyll a retention properties of glass-fiber GF/F filters. Limnol. Oceanogr., 40(2): 428-433

 

Author Response

Page 2, lines 65-66. It is a common misconception that the “pore-size” of Whatman glass microfiber filters has a Diameter of 0.7 μm. The structure of glass microfiber filters does not allow for a pore-size to be estimated and what is reported is actually a mean pore size determined from a range of pore sizes possible for this filter type, glass fibres randomly positioned one on top of another. Chavez (1995) has reported quantitative recovery of phytoplankton down to 0.2 μm in the north Pacific with GF/F filters.

A description about pore size was removed in line 67.

Page 2, lines 67-68. Since this is a preliminary report on the effort to determine an algorithm for POC/Chla for the East/Japan Sea, I would recommend the authors calibrate the algorithm from Chla determined by HPLC. Fluorometry is chancy and empirical and not nearly as certain as an HPLC determination of Chla. I say this only because the effort expended here is to calibrate the remote sensing algorithm and this requires the most reliable determination possible of Chla.

Page 6, lines 144-146. Again, I recommend that the authors use chlorophyll concentration derived from HPLC analysis to further validate and calibrate the algorithm proposed here.

Thank you for your good suggestion. We know the Cal/Val of the Chl-a algorithm with the HPLC is essential to investigate the POC to chl-a However, it is hard to meet the deadline if we try to derive a new chl-a algorithm from HPLC data which is not currently available in the East/Japan Sea. To do so, we have started to collect seasonal HPLC samples in the East/Japan Sea since last year. Our next study will be proceeded with calibrated chl-a algorithm with in-situ HPLC datasets.

Page 2, lines 70-71. The authors need to expand on the rationale of the stable isotope method for determining POC.

POC samples were not determined by a stable isotope method. POC samples were analyzed along with carbon and nitrogen stable isotopes in a parallel study using a mass spectrometric method. Stable isotope is only a part of the lab name.

Page 6, lines 134-137. First of all, the authors are to be commended for a correct plot of modeled variable (algorithm POC on x-axis) and in-situ variable (measured POC on y-axis). The majority of studies I have reviewed plot it in the opposite fashion. Thus the plot in Fig. 7 correctly shows the ability of the model to predict the in-situ measurement. Indeed, the argument that the algorithm proposed here appears visually to have a slope in this plot that approaches 1.0 is a good one. However, I think the argument would be strengthened by calculating the regression slope for the coastal algorithm and using standard regression tests to show that the slope of the plot does not differ significantly from 1.0. I suspect that the same effort for the open ocean algorithm would show that the slope does differ significantly from 1.0. It is interesting that both algorithms appear to have the same y-intercept when the plots are extended until they cross the y-axis.

Thank you for your comments. The slope for the new POC algorithm in this study was 1.23 while Stramski et al. (2008) was 1.78. The slope of our algorithm is much closer to 1.0 compared to Stramski et al (2008). Also, y-intercepts of both algorithms look similar, but they are not exactly the same (-39.407 and -41.484 for Stramski et al (2008) and This study, respectively). We added the regression equations in Figure 7.

Reviewer 3 Report

General comments

This article proposes the use of the ratio POC/Chla to try to explain changes in primary production rates in the East/Japan Sea. To do so they propose two objectives: (1) to derive a new regional POC concentration deriving algorithm and (2) to investigate the spatial and temporal variations of the POC to chl-a ratio in the East/Japan Sea. Although the proposal is interesting and many work was done by the authors, the organization of the manuscript is confusing, specially the methodology and results, and it was hard to understand the steps and procedures applied. For this reason, it is hard to be sure about the final results regarding temporal variability of POC in the study region. Please, see my specific comments below. In general I recommend and extensive revision of the manuscript before it can be accepted for publication.

  

Specific comments

Abstract. The coefficient of determination has as symbol R² instead of R2. Please change.

Line 36. Many previous studies reported that the East/Japan Sea is productive ... please change to "is a productive region"

Lines 47 to 54. Please review this paragraph. You are using many times the same word (e.g. conditions) and expressions (ecological conditions and physiological states). The paragraph reads confusing and ideas are repeated many times. Please, see pdf with marks. There are many others mistakes in English marked in the document. You have to send the manuscript for en English edition.

Methodology section. Satellite dataset. I do not understand how you are using the algorithm of Stramski but at the same time you are incorporating your in situ observations. This has to be better explained. If you are deriving your own algorithm this is part of your results. Besides, in the power low equation (1) please indicate the constants (a and b). In the equation you indicate Rrs between 547 and 565, but in the text you mention 547. Please explain. It is not really clear how you proceeded here and this is very important to understand the following sections... After beginning the results section I understand what you were trying to explain in methodology. However, you will have to change this explanation because as it is, is confusing. I recommend to rewrite this section of the methodology.

Results

You have here two things to consider. First, you are modelling, i.e., you have your Rrs data and you compared it to in situ POC. From this step, you derive an equation (2), and you have to indicate how your in situ data is explaiend by this model (or equation). You will do that using R², which is an statistics that indicates how much of the in situ variability is explained by the model. Here, the RMSE is not really relevant. Instead, you can use the Spearman correlation coefficient, that will describe the degree of correlation between variables (when the relationship is not linear). Second, after proposing the model (equation 2), you will compare it with your in situ data. In this comparison (your figure 2b) you will see the errors of the model. Here is recommended the use of RMSE and also the BIAS to indicate how well the model is reproducing your in situ data and if it underestimates or overestimates (BIAS) in situ data.   

Fig. 2. (b) remove the regression equation because here the only important thing to know is the degree of correlation between the modeled data and the in situ data. Please include RMSE and BIAS.

In figure 3 I suppose you are applying your algorithm? Please explain.

How did you calculate the POC/Chla ratio shown in figure 4. This is also using satellite Chla I suppose. Please, explain.

Line 117. You mention that there are no differences between time series. However, did you check if the differences (or no differences) are statistically significant?

Fig. 6. Please, indicate what is the meaning of the vertical lines. Do they indicate minimum and maximum? In text, please explain from which region/station/point are you deriving these values. Are they an average of all pixels from the image?

Fig. 7. What is the difference between this figure and figure 2b?

Section 4.2. You are supposing many things here that you cannot be sure (phytoplankton composition, size, physiological changes). Maybe you should concentrate in model implementation and validation, giving some insights about temporal variability. Some differences you are mentioning between south and north were not statistically proved.

 

 

 

 

 

Comments for author File: Comments.pdf

Author Response

Abstract. The coefficient of determination has as symbol R² instead of R2. Please change.

We revised the symbol in line 26

Line 36. Many previous studies reported that the East/Japan Sea is productive ... please change to "is a productive region"

We revised the sentence in line 36

Lines 47 to 54. Please review this paragraph. You are using many times the same word (e.g. conditions) and expressions (ecological conditions and physiological states). The paragraph reads confusing and ideas are repeated many times. Please, see pdf with marks. There are many others mistakes in English marked in the document. You have to send the manuscript for en English edition.

Repeated words and expressions in lines 44-54 were revised. Also, several mistakes in English in manuscript were modified.

Methodology section. Satellite dataset. I do not understand how you are using the algorithm of Stramski but at the same time you are incorporating your in situ observations. This has to be better explained. If you are deriving your own algorithm this is part of your results. Besides, in the power low equation (1) please indicate the constants (a and b). In the equation you indicate Rrs between 547 and 565, but in the text you mention 547. Please explain.

Rrs between 547 and 565 means any available band between the range can be used as a green band. A detailed explanation was added in lines 85-91.

Results

You have here two things to consider.

First, you are modelling, i.e., you have your Rrs data and you compared it to in situ POC. From this step, you derive an equation (2), and you have to indicate how your in situ data is explaiend by this model (or equation). You will do that using R², which is an statistics that indicates how much of the in situ variability is explained by the model. Here, the RMSE is not really relevant. Instead, you can use the Spearman correlation coefficient, that will describe the degree of correlation between variables (when the relationship is not linear). Second, after proposing the model (equation 2), you will compare it with your in situ data. In this comparison (your figure 2b) you will see the errors of the model. Here is recommended the use of RMSE and also the BIAS to indicate how well the model is reproducing your in situ data and if it underestimates or overestimates (BIAS) in situ data.   

As suggested, Spearman’s correlation coefficient was added in figure 2a and RMSE and BIAS were added in figure 2b.

Fig. 2. (b) remove the regression equation because here the only important thing to know is the degree of correlation between the modeled data and the in situ data. Please include RMSE and BIAS.

We revised Figure 2. We added RMSE and BIAS were added in figure 2b.

In figure 3 I suppose you are applying your algorithm? Please explain.

Yes, POC concentrations in Figure 3 were derived by using our regional algorithm described in equation (2). A brief explanation was added in line 108.

How did you calculate the POC/Chla ratio shown in figure 4. This is also using satellite Chla I suppose. Please, explain.

POC/Chl-a ratio was calculated by dividing POC by Chl-a. We used POC derived by our model and Chl-a obtained from MODIS-Aqua. A description was added in lines 89-90.

Line 117. You mention that there are no differences between time series. However, did you check if the differences (or no differences) are statistically significant?

Yes, statistical tests for the differences were conducted. There were no differences between time series of northern and southern parts (t-test, p>0.05).

Fig. 6. Please, indicate what is the meaning of the vertical lines. Do they indicate minimum and maximum? In text, please explain from which region/station/point are you deriving these values. Are they an average of all pixels from the image?

Vertical lines in Figure 6 indicate standard deviations of climatological monthly mean POC in each month. Explanation was added in the caption for Figure 6. Domains for Northern and Southern parts of the East/Japan Sea were shown in Figure 1. Related explanations were added in lines 60-62 and the caption in Figure 1.

Fig. 7. What is the difference between this figure and figure 2b?

In Figure 7, we compared the accuracy of our POC models with existing POC model (Stramski et al., 2008; NASA standard) while Figure 2 just shows our validation result during the derivation of a new regional model.

Section 4.2. You are supposing many things here that you cannot be sure (phytoplankton composition, size, physiological changes). Maybe you should concentrate in model implementation and validation, giving some insights about temporal variability. Some differences you are mentioning between south and north were not statistically proved.

Of course, we can’t sure about what we have suggested in section 4.2 because the results were not obtained by field observation. However, we tried to explain spatial and temporal variability of POC to chl-a ratio based on general physiological characteristics of phytoplankton as well as environmental features of the East/Japan Sea reported by several previous studies. As you mentioned, although the difference between south and north was not statistically proved, monthly distribution of the POC to chl-a ratio shows the difference.

Reviewer 4 Report

Review MDPI remote sensing: Estimation of the particulate organic carbon to chlorophyll-a ratio using MODIS-Aqua in the East Japan Sea, South Korea. Lee et al.

 

POC and chlorophyll a of surface samples obtained during several years of the East Japan Sea were used to create a regional algorithm for remote sensed POC. The newly derived algorithm was used to investigate the POC: chl a ratio during 2003-2018. POC:chl a showed a clear seasonal pattern, but showed no clear inter annual trend.

From an eco-physiological point of view changes in POC:chl a are interesting. However, there are several limitations to remotely sensed POC: chl a ratios that are insufficiently addressed in the manuscript.

The study of the POC:chl a ratio is introduced as an indicator to investigate recent changes in productivity in this sea. However, the usefulness of this parameter is not discussed for this purpose in the discussion. Before expanding this research, I suggest to evaluate the usefulness of an improved algorithm for POC:chl a.

POC in the marine environment is not only derived from photosynthetic phytoplankton, but also from its grazers. This makes the POC: chl a ratio less useful when measured in marine samples. POC : chl a varies with irradiance acclimation, and with nutrient availability, and taxonomic composition (i.e. large and small phytoplankton cells). In addition, it also reflects the organic carbon of non-photosynthetic organisms, which can also show considerable seasonal variability. A discussion of the usefulness of the POC:chl a ratio for assessing changes in ocean productivity should be added to the discussion.

Seasonal decreases of the POC: chl a ratio were attributed to the prevalence of diatoms. However, no evidence for this is provided. Furthermore, changes in POC:chla during summer can be due to changes in experienced mixed layer irradiance, in addition to changes phytoplankton composition.

 

Overall I suggest to accept this paper for publication after major revision of the discussion.

Author Response

The study of the POC:chl a ratio is introduced as an indicator to investigate recent changes in productivity in this sea. However, the usefulness of this parameter is not discussed for this purpose in the discussion. Before expanding this research, I suggest to evaluate the usefulness of an improved algorithm for POC:chl a.

POC in the marine environment is not only derived from photosynthetic phytoplankton, but also from its grazers. This makes the POC: chl a ratio less useful when measured in marine samples. POC : chl a varies with irradiance acclimation, and with nutrient availability, and taxonomic composition (i.e. large and small phytoplankton cells). In addition, it also reflects the organic carbon of non-photosynthetic organisms, which can also show considerable seasonal variability. A discussion of the usefulness of the POC:chl a ratio for assessing changes in ocean productivity should be added to the discussion.

A paragraph explaining the usefulness of the POC to chl-a ratio for investigating recent changes in primary production in the East/Japan Sea was added in line 175-178.

Seasonal decreases of the POC: chl a ratio were attributed to the prevalence of diatoms. However, no evidence for this is provided. Furthermore, changes in POC:chla during summer can be due to changes in experienced mixed layer irradiance, in addition to changes phytoplankton composition.

Although we didn’t provide some evidences for the changes of diatom biomass, several previous studies observed diatom-dominated bloom in spring. Thus, we can only suggest a kind of hypothesis. We added an explanation of the limitation of this study in lines 177-180. Also, we agree with your opinion about the changes in POC to chl-a ratio during summer. A deeper study will be needed to understand temporal variation of the POC to chl-a ratio in the East/Japan Sea.

Round 2

Reviewer 1 Report

Thank you for the significant revisions!

Author Response

Thank you for your help.

Reviewer 3 Report

Although the authors answered some of my concerns from previous version of the manuscript, it still lacks of clarity and organization and many mistakes in English are still found. I recommend major revisions. Please see below specific comments and in the manuscript attached some observations are also included.

Line 83. Did you use the Stramski equation but you used your empirically derived constants a and b? This is not clear. English is still confusing. 

Fig. 2. Your value for RMSE is extremely high !!! I am not sure if you did the correct calculation. Please, verify and show which equation did you use.

Line 104. You mention that R2 is 0.8017 and RMSE is 85.95, but for what? In figure 2 numbers are different? It is not clear what are you talking about. 

Line 116-117. How the ratio POC:Chla was calculated should be explained in the methodology section. In addition, I suggest the use of the abbreviation POC:Chla instead of ¨POC to Chla ratio¨.

Figure 5 and text. I suggest the use of the term ¨region¨instead of ¨part¨.

Lines 123-135. You are comparing tendencies but differences were not statistically verified. For example, in figure 6, you mention that ratios during spring and autumn were lower than summer and winter. However, the variability is strong and standard deviations (vertical bars) overlap. You cannot be sure of such differences. 

Figure 7. I do not understand these results. I do not also understand why this figure is here and not in the result section. You mention that ¨orange crosses¨ represent the correlations between POC modeled by Stramski and that blue circles are for your data. You have to check english ... what you are saying is not what you are showing in the figure. In addition, you are here ¨validating¨these models and comparing their perfomrance. You should use again RMSE and BIAS. 

I do not agree with your discussion about temporal variability. You should first discuss you model. See also for example comments for figure 6. 

 

Comments for author File: Comments.pdf

Author Response

Line 83. Did you use the Stramski equation but you used your empirically derived constants a and b? This is not clear. English is still confusing.

Yes, you are right. We used the equation reported by Stramski et al. (2008), and two constants; a and b were empirically derived. We revised the sentence in lines 83-85.

Fig. 2. Your value for RMSE is extremely high !!! I am not sure if you did the correct calculation. Please, verify and show which equation did you use.

We verified that our RMSE was correct as below: The high RMSE values could be due to the wide range of in-situ POC values ranging from 84.1 to 713.7 in our study area and our algorithms still need to be improved.

Line 104. You mention that R2 is 0.8017 and RMSE is 85.95, but for what? In figure 2 numbers are different? It is not clear what are you talking about.

We revised the sentence in lines 105-106.

Line 116-117. How the ratio POC:Chla was calculated should be explained in the methodology section. In addition, I suggest the use of the abbreviation POC:Chla instead of ¨POC to Chla ratio¨.

The explanation about calculating POC:Chla was described in lines 90-91. Also, we revised the abbreviation throughout the manuscript.

Figure 5 and text. I suggest the use of the term ¨region¨instead of ¨part¨.

We revised the term throughout the manuscript.

Lines 123-135. You are comparing tendencies but differences were not statistically verified. For example, in figure 6, you mention that ratios during spring and autumn were lower than summer and winter. However, the variability is strong and standard deviations (vertical bars) overlap. You cannot be sure of such differences.

We added statistical analysis results in lines 134 and 141-142. Actually, there is a statistically difference in POC:Chla between the spring-autumn and summer-winter.

Figure 7. I do not understand these results. I do not also understand why this figure is here and not in the result section. You mention that ¨orange crosses¨ represent the correlations between POC modeled by Stramski and that blue circles are for your data. You have to check english ... what you are saying is not what you are showing in the figure. In addition, you are here ¨validating¨these models and comparing their perfomrance. You should use again RMSE and BIAS. 

We moved Figure 7 into the result section (Figure 3), and we revised the figure caption in lines 113-116. RMSE and bias were also added in Figure 3.

 I do not agree with your discussion about temporal variability. You should first discuss you model. See also for example comments for figure 6. 

As we mentioned previously, there is a statistically difference in POC:Chla in the East/Japan Sea between the spring-autumn and summer-winter (lines 147-148). However, as we suggested in lines 192-194, a further research with field observations is needed to understand the spatial and temporal variations of the POC:Chla in the East/Japan Sea.

Author Response File: Author Response.docx