Estimation Method of Chlorophyll Concentration Distribution Based on UAV Aerial Images Considering Turbid Water Distribution in a Reservoir

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

 

The paper deals with a very important topic and thematically is well-targeted and appropriate to the journal scope. The structure is clear and logical. I would recommend a minor revision as I feel some improvements in the following directions can be accommodated in the paper before it is published:

·         Abstract: please underline much more clearly the innovative contribution of this work make it a bit more quantitative in the results and underline the key implications of your study findings. Also, would be good to make the abstract shorter in length.

·         Introduction: perhaps make a bit punchier and underline stronger the innovative aspect of this work. Also, add some more key papers from the current literature particularly so in reference to your study objectives just before those are introduced.

·         Discussion: How the results of the study can be generalised, if they can? What else needs to be done if the interest is in this direction? Comment on the methodology reproducibility…?

·         Conclusions Make it clear why this study is important and what it offers. The same also should had been done in the introduction of the paper.

·         Figure and table captions: ensure that those provide sufficiently the description of the table or figure on which each refers to

·         Figures: consider ways to perhaps help you reduce the figures number in the manuscript.

·         Improve the writing style in overall, as it is sloppy in some areas in terms of English writing usage.

Author Response

Dear Reviewer

Thanks for your helpful comments and advises. I revised the manuscript and reply to your comments as below

• Abstract: please underline much more clearly the innovative contribution of this work make it a bit more quantitative in the results and underline the key implications of your study findings. Also, would be good to make the abstract shorter in length.

Thank you for the advice. The results are presented more quantitatively in the new Abstract, and the contributions made are specified at the end. The number of words is limited to 200 words as defined by the submission guideline.

 

• Introduction: perhaps make a bit punchier and underline stronger the innovative aspect of this work. Also, add some more key papers from the current literature particularly so in reference to your study objectives just before those are introduced.

Thank you for the advice. We have added some recent important literature to the Introduction. Specifically, 12, 15, 19, 30, 31, 32, 35, 42, 52, 53, 60-63 on the new manuscripts. In addition, the new manuscript describes the research goal starting from line 144, and is structured to identify the issues that need to be overcome to achieve this goal while listing references.

 

 

• Discussion: How the results of the study can be generalised, if they can? What else needs to be done if the interest is in this direction? Comment on the methodology reproducibility…?

The other reviewers pointed out the influence of CDOM. The reservoirs targeted in this study have short retention time and almost no CDOM, but this must be taken into account when generalizing. The results of this study were obtained under an environment where CDOM can be ignored, so while we were able to propose a calibration method specific to the relationship between turbidity and chlorophyll, we took the influence of CDOM into account when generalizing. There are two possible directions: (1) Differenciating and eliminating the influence of CDOM on satellite images or UAV aerial images, and (2) including many images with high CDOM in the training data for machine learning.

We have described these points in the Discussion and Conclusion.

 

• Conclusions Make it clear why this study is important and what it offers. The same also should had been done in the introduction of the paper.

I added such a note at the beginning of the Conclusion. Similarly in the introduction.

 

• Figure and table captions: ensure that those provide sufficiently the description of the table or figure on which each refers to

I added description on some figures, Thanks for your pointing out

 

 

• Figures: consider ways to perhaps help you reduce the figures number in the manuscript.

We have carefully selected the minimum necessary figures from the first draft stage, so we do not want to reduce them any further. Please understand.

 

 

 

• Improve the writing style in overall, as it is sloppy in some areas in terms of English writing usage.

In particular, there were many places in the Introduction where I listed several references all at once and neglected to write short introductions for each one, so I focused on making corrections to those parts.

 

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

Excellent paper.

Author Response

Dear Reviewer

Thanks for your comments. I revised the manuscript following to the other reviewer's comments. see attache revised manuscript

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript presents a very good combination of classic scientific methods with machine learning techniques. 

In the conclusions section I would recommend to insert some suggestions regarding future work for further refinement of this methodology aiming in the adoption of the research results and its integration into the standard operation procedures applied by water quality monitoring organisations and local authorities.

Author Response

Dear Reviewer

Thanks for your comments. I added some description about the limitations and issues for generalization of this method to any reservoirs in the discussion and conclusion. please check the revised manuscript

Author Response File: Author Response.docx

Reviewer 4 Report

Comments and Suggestions for Authors

This study aimed to conduct observations using photographs taken from a UAV and to be able to conduct spatially high-resolution analysis without restrictions due to photographing date or cloud cover. After calibrating the reflectance of the images based on insolation, the authors designed a method for estimating Chl-a concentration using machine learning based on the calibrated reflectance values and turbidity. This is an interesting study. However, from the perspective of the optical transmission principle of water, this method still has many small problems.

1. Suspended solids, chlorophyll-a and CDOM are three important substances that affect the spectral characteristics of water bodies. These three substances can be considered independent of each other. The WTI described in the study is mainly affected by suspended solids. Therefore, estimating chlorophyll-a concentration based on WIT, temperature, etc. may not be suitable in principle, although in some regions they do have a strong correlation.

2.The data source of this study is UAV optical images, which only have three visible light bands: red, green, and blue bands. The lens itself has very limited capability in chlorophyll-a estimation. Although the author has designed a method to estimate chlorophyll-a concentration based on the images. How accurate is this method? There are still questions about whether it can be applied in other areas.

3.Also, is the temperature data estimated from the red, green and blue bands too? If so, the model is very dependent on the red band, potentially causing significant uncertainty in the model's results.

Author Response

Dear Reviewer

Thanks for your helpful comments and advises. I revised the manuscript and reply to your comments as below

1. Suspended solids, chlorophyll-a and CDOM are three important substances that affect the spectral characteristics of water bodies. These three substances can be considered independent of each other. The WTI described in the study is mainly affected by suspended solids. Therefore, estimating chlorophyll-a concentration based on WIT, temperature, etc. may not be suitable in principle, although in some regions they do have a strong correlation.

 

Thanks for your comment. I think you are referring to the fact that three characteristics appear at different wavelengths regarding the absorbance obtained by irradiating a small area in water body. We do not think that the three characteristics can be interpreted as independent in satellite images or UAV aerial images where the average reflection intensity within the area is obtained from a satellite at a macroscopic size (several meters to tens of meters). The references listed in 60-63 of the new manuscript also states that the three characteristics interfere with each other. This point was also stated in the discussion in this paper.

 

 

2.The data source of this study is UAV optical images, which only have three visible light bands: red, green, and blue bands. The lens itself has very limited capability in chlorophyll-a estimation. Although the author has designed a method to estimate chlorophyll-a concentration based on the images. How accurate is this method? There are still questions about whether it can be applied in other areas.

According to Fig.11, we can say the result is accurate for this reservoir. However, as you worried, it is just for the reservoir. Success or failure of Machine learning depends on training. In this study, our observation was carried out only in the reservoir with low CDOM. Perhaps that network cannot estimate the Chl-a conc. of high CDOM lake. Training with the images of that lake will be required for that application. In the new manuscript, I described that point in the discussion and conclusion. Thanks for your comments.

 

 

3.Also, is the temperature data estimated from the red, green and blue bands too? If so, the model is very dependent on the red band, potentially causing significant uncertainty in the model's results.

I never tried to estimate temp from satellite images in this study. As mentioned, the reservoir is small so that the water surface temp is almost uniform(Fig.4) and the measured temp at the point is applicable to the whole area.

I introduced water temp. as an explanatory parameter in order to reflect the seasonal fluctuation of some influences including unknown factors. The purpose of this study is not to explain the relationship between water quality and reflectance scientifically but to develop an easy estimation method of Chl-a. From that point of view, one of the important points is feeling in the field. The water surface I saw in summer looks different from that in winter, I felt that. water temp is indicating that and easy to measure.

Even if you plan to this method to a large lake with surface temp difference, the water temp measured at one fixed site can be the explanatory parameter to estimate Chl-a distribution.

 

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I ma satisfied with the revision done on the paper.

Comments on the Quality of English Language

check spelling mistakes here and there, still exist...

Author Response

Dear the reviewer

Thanks for your checking my manuscript for spelling mistakes and bad word order in detail.

Finally, I would like to thank you for your helpful advice throughout.

Reviewer 4 Report

Comments and Suggestions for Authors

The authors have made detailed revisions to the paper. I agree with the existence of the relationship between UAV images and Chl-a concentration obtained in this study. However, my concern is still whether this method can be applied and used in other waters. If the applicability is low, the method is not meaningful for other researchers or other studies. This is just my personal opinion. It is up to the editor to decide whether the paper could be accepted or not.

Author Response

Thank you for your comment.
In machine learning, the quality and quantity of training data are the keys to success or failure.
Regarding whether the method proposed here can be applied to other reservoirs, I think it is possible to evaluate chlorophyll concentration if the classifer is trained with the lake surface images of that reservoir. It is difficult to evaluate the reservoirs with the classifier trained only with the lake surface images of the research site targeted in this study. This point was already indicated in the manuscript.

Finally, I would like to thank you for your helpful advice throughout.