Estimation of Anthocyanins in Leaves of Trees with Apple Mosaic Disease Based on Hyperspectral Data
Round 1
Reviewer 1 Report
Comments and suggestions：
The authors utilized different nonparametric fitting methods, including PLSR, BP, SVM, RF and SSA-RF, for estimation of anthocyanin in apple leaves with mosaic disease with hyperspectral data. Abundant work has been involved in the manuscript, and key information on data analysis have been presented. Nevertheless, novelty of the research seemed to be not enough, since the manuscript just focused on improvement of estimation accuracies with comparison of different nonparametric fitting methods, limited content was presented on explanation of anthocyanin retrieval mechanism, and key problems that haven’t been solved for anthocyanin estimation should be considered and paid with more attention. Moreover, here is a query: Generally, target variables (here in the manuscript is anthocyanin) were truth values, and these truth values were either measured with units using laboratory analysis methods, or converted to values with units through modification equations. Anthocyanin used in the manuscript were just Dualex 4 values without units, hyperspectral data measured were also values with dimensionless unit. The models established between these two variables without units couldn’t be convincing, errors of the models couldn’t be quantified and these models could not be suitable for authentic anthocyanin content estimation. These problems must be addressed. Moreover, I believe the manuscript should clarify several crucial issues which are presented in the general comments below, and some minor specific comments also should be addressed for the improvement of the manuscript.
General comments
1. Rare introduction were presented, especially for literature on anthocyanin assessment. These contents were critical, since readers could know abundant previous work on anthocyanin estimation, and thus could evaluate the novelty and contribution of the manuscript for anthocyanin retrieval. This issue must be addressed.
2. in line 96 “Multiple reflectance curves of anthocyanin position were measured”, here detailed information on reflectance measurement should be provided, such as the light? how the leaves were placed in measuring area? The height of SVC sensor? Fov of the sensor? And so on.
3. in Table 2, the criteria to determine specific band, such as 477, 634, should be explained. In fact, band selection methods like this were affected by the data used. Majority of the features couldn’t be explained by detection mechanism of anthocyanin. Such as DSI1, that used BAND836 and BAND894. this spectral index was so weird, since these two bands were so close in the NIR band, and NIR bands had no correlation with anthocyanin according to the detection mechanism of anthocyanin. These spectral indices selection process lacked of convincing evidence.
Specific comments
1. in line 176 “as shown in Figure 2”, here should be Figure 3.
2. no information and results on Table 5 were shown in the manuscript. And since BP, SVM, RF were used in the manuscript, details information on how to set parameters for these models should be explained clearly.
3. in line 17 “concisability”, don’t know the meaning of this word.
4. problems concerning grammatical expression should be carefully checked throughout the whole manuscript.
Comments for author File: 
Comments.pdf
Author Response
Dear reviewer:
Thank you for your letter and the reviewers’ comments on our manuscript entitled " Estimation of Anthocyanins in Leaves of Trees with Apple Mosaic Disease based on Hyperspectral Data " (ID：remotesensing-2233782). Those comments are very helpful for revising and improving our paper, as well as the important guiding significance to other research. We have studied the comments carefully and made corrections which we hope meet with approval. The main corrections are in the manuscript and the responds to the reviewers’ comments are as follows (the replies are highlighted in red ).
Point 1: Generally, target variables (here in the manuscript is anthocyanin) were truth values, and these truth values were either measured with units using laboratory analysis methods, or converted to values with units through modification equations. Anthocyanin used in the manuscript were just Dualex 4 values without units, hyperspectral data measured were also values with dimensionless unit. The models established between these two variables without units couldn’t be convincing, errors of the models couldn’t be quantified and these models could not be suitable for authentic anthocyanin content estimation. These problems must be addressed.
Response 1: We are glad that you have noticed this problem. We discussed with the professor and reached the following conclusion:
Dualex 4 is a portable handheld polyphenol analyzer ,it is mainly based on the principle of plant chlorophyll fluorescence remote sensing, and uses the dual-excited chlorophyll fluorescence to detect the ultraviolet light (375nm) absorption rate of the leaf epidermis to reflect the content of anthocyanin in the leaves. Dualex 4 is calibrated at the factory and incorporates a GPS receiver to ensure a linear response between the measured value and the true value. So,it can be useful for non-destructive estimation of leaf Ahth contents for ecophysiological research and ground truthing of remote sensing of vegetation.
In recent years, DX4 has been widely used in the field of agricultural remote sensing. The number of articles directly using measured values and spectral data to build models is also increasing. It has been proposed that the unit of Dualex 4 is ug/cm2 , but this issue is still controversial. Therefore, the anthocyanin content in this study was unitless.To better illustrate this issue, we had reviewed the following articles and attached a link. I hope we can explain clearly.
A new optical leaf-clip meter for simultaneous non-destructive assessment of leaf chlorophyll and epidermal flavonoids https://doi.org/10.1111/j.1399-3054.2012.01639.x
Estimating Plant Nitrogen Concentration of Maize Using a Leaf Fluorescence Sensor across Growth Stages https://doi.org/10.3390/rs12071139
Assessment of Portable Chlorophyll Meters for Measuring Crop Leaf Chlorophyll Concentration https://doi.org/10.3390/rs11222706
Point 2: Rare introduction were presented, especially for literature on anthocyanin assessment. These contents were critical, since readers could know abundant previous work on anthocyanin estimation, and thus could evaluate the novelty and contribution of the manuscript for anthocyanin retrieval.
Response 2: We sincerely appreciate the valuable comments. We have checked the literature carefully and added more references on anthocyanin assessment into the INTRODUCTION part in the revised manuscript.Please see the attachment.
Point 3: In line 96 “Multiple reflectance curves of anthocyanin position were measured”, here detailed information on reflectance measurement should be provided, such as the light? how the leaves were placed in measuring area? The height of SVC sensor? Fov of the sensor? And so on.
Response 3: Thank you for your valuable comments. We have added relevant content to Data acquisition and preprocessing. We are sorry that our instruments cannot be taken again after being sent back to the factory for maintenance.Please see the attachment.
Point 4: In Table 2, the criteria to determine specific band, such as 477, 634, should be explained. In fact, band selection methods like this were affected by the data used. Majority of the features couldn’t be explained by detection mechanism of anthocyanin. Such as DSI1, that used BAND836 and BAND894. this spectral index was so weird, since these two bands were so close in the NIR band, and NIR bands had no correlation with anthocyanin according to the detection mechanism of anthocyanin. These spectral indices selection process lacked of convincing evidence.
Response 4: We feel sorry for our carelessness. In our resubmitted manuscript, the Band is revised. (DSI1 Band 424 and 656) We checked and calculated the data. Fortunately, errors only occurred in writing, and the calculation process of the data was not affected. As shown in the figure below, we added the reason for band selection. After consulting the data, we added references on the feasibility of estimating anthocyanins using the NIR band. Thanks for your correction. Please see the attachment.
Point 5: In line 176 “as shown in Figure 2”, here should be Figure 3
Response 5: We were really sorry for our careless mistakes. Thank you for your reminder.
Point 6: No information and results on Table 5 were shown in the manuscript. And since BP, SVM, RF were used in the manuscript, details information on how to set parameters for these models should be explained clearly.
Response 6: Table 5 was deleted due to operational errors during the modification process. We have uploaded it again in the new version. In addition, we have added the relevant parameter information of the model. Please check.
Point 7: In line 17 “concisability”, don’t know the meaning of this word.
Response 7: Concisability means using minor independent variables to obtain the best estimation accuracy. We added an explanation on line 86.
Point 8: Problems concerning grammatical expression should be carefully checked throughout the whole manuscript.
Response 8: Thanks for your suggestion. However, we do invite a friend of us who is a native English speaker from the USA to help polish our article. And we hope the revised manuscript could be acceptable for you.
Once again, thank you very much for your constructive comments and suggestions which would help us both in English and in depth to improve the quality of the paper.
Kind regards,
Zijuan Zhang
E-mail: [email protected]
Corresponding author: Qingrui Chang
E-mail address: [email protected]
Author Response File: 
Author Response.docx
Reviewer 2 Report
I have read the manuscript. In all, the structure of the manuscript is well, and the dataset is reliable. I have some suggestions for the manuscript.
The significance of the study is not very clear. The author claimed thet ‘most of the current studies have focused on healthy leaves, and few studies have estimated the anthocyanin content in diseased leaves’, but why this study focused on the diseased leaves? How does the anthocyanin content in diseased leaves benefits us? I think the anthocyanin content in normal or early infected leaves is more important.
The introduction is too short, and lots of information was not covered. Retrieval of the parameters from the plants with hyperspectral data and machine learning should be added in detail.
The infected leaves will lose chlorophyll, so I wonder if the presented SSA-RF model can discriminate the mosaic leaves from other infected leaves or even the dead leaves.
Line 149, ‘Random Forest algorithm (RF) [33] is a classification regression’, should be ‘random forest is a supervised machine learning algorithm that is used for classification or regression problems’.
The fitting analysis result of RF should be given in the discussion, as shown in Figure 7.
Author Response
Dear reviewer:
Thank you for your letter and the reviewers’ comments on our manuscript entitled " Estimation of Anthocyanins in Leaves of Trees with Apple Mosaic Disease based on Hyperspectral Data " (ID：remotesensing-2233782). Those comments are very helpful for revising and improving our paper, as well as the important guiding significance to other research. We have studied the comments carefully and made corrections which we hope meet with approval. The main corrections are in the manuscript and the responds to the reviewers’ comments are as follows (the replies are highlighted in red ).
Replies to the reviewer’ comments:
Point 1: The significance of the study is not very clear. The author claimed thet ‘most of the current studies have focused on healthy leaves, and few studies have estimated the anthocyanin content in diseased leaves’, but why this study focused on the diseased leaves? How does the anthocyanin content in diseased leaves benefits us? I think the anthocyanin content in normal or early infected leaves is more important.The introduction is too short, and lots of information was not covered. Retrieval of the parameters from the plants with hyperspectral data and machine learning should be added in detail.
Response 1: We agree with you that the anthocyanin content in normal or early infected leaves is more important. Our research sample contains leaves with various disease degrees, so that we can obtain the change rule of anthocyanins in the whole disease period. We established the estimation model of anthocyanins by the relationship between anthocyanins and spectral data during the onset of disease. This model allows us to monitor the content of anthocyanins through remote sensing technology, so that we can timely detect and treat fruit trees at the early stage of disease.
As suggested by the reviewer, we have added more references on the parameters from the plants with hyperspectral data and machine learning. Please see the attachment.
Point 2: The infected leaves will lose chlorophyll, so I wonder if the presented SSA-RF model can discriminate the mosaic leaves from other infected leaves or even the dead leaves.
Response 2: In this study, the estimation of the anthocyanin concentration of apples was based on the relationship between the variation of anthocyanin concentration caused by mosaic disease and the change of spectral characteristics. The changes in vegetation characteristics caused by different conditions are different. In theory, the method proposed in this paper can be used to estimate the biochemical parameters of various diseases and even dead leaves, as long as the change rules of the biochemical parameters and spectral characteristics of vegetation can be found. However, the estimation accuracy of the model needs further study. And we will collect vegetation samples infected with different diseases to solve this problem. Thank you for your suggestions.
Point 3: Line 149, ‘Random Forest algorithm (RF) [33] is a classification regression’, should be ‘random forest is a supervised machine learning algorithm that is used for classification or regression problems’.
Response 3: We were really sorry for our careless mistakes. Thank you for your reminder.
Point 4: The fitting analysis result of RF should be given in the discussion, as shown in Figure 7.
Response 4: We thought this is a good suggestion and had added relevant pictures to the article.
Once again, thank you very much for your constructive comments and suggestions which would help us both in English and in depth to improve the quality of the paper.
Kind regards,
Zijuan Zhang
E-mail: [email protected]
Corresponding author: Qingrui Chang
E-mail address: [email protected]
Author Response File: Author Response.docx
Reviewer 3 Report
Disease monitoring is one of the important contents of agricultural research. The manuscript aims to assess the anthocyanin values of apple tree leaves with mosaic disease using leaf hyperspectral data. The response of spectral characteristics to anthocyanin content was analysed. Based on the optimal spectral variables, anthocyanin content was estimated by SSA-RF regression model. The proposed estimation model in this paper obtained a high estimation accuracy, which could provide a theoretical basis for the use of remote sensing technology to monitor the health status of apple in a large area. The topic is of interest to the journal, the manuscript is generally easy to follow. My major comments include, (1) I find it difficult to understand the experimental design of this work. the description was too concise to obtain useful information from it. Also, there were few explanations of the rationale for the study design. Please give more details about the study area and experiment design.
(2) Please highlight the novelties of this study. What is the research gap? Moreover, the language should be improved by native speaker.
(3) What is the author's rationale for using the sparrow Search Algorithm to construct the optimal random forest model, have other methods been tried, and why does the method yield the optimal model?
(4) I noticed that Anth in Figure 3 is missing units of measurement, please add.
(5) Please check the names of Figure 4 for errors, as well as the other figures in the article. Before submitting a revision be sure that your material is properly prepared and formatted.
Author Response
Dear reviewer:
Thank you for your letter and the reviewers’ comments on our manuscript entitled " Estimation of Anthocyanins in Leaves of Trees with Apple Mosaic Disease based on Hyperspectral Data " (ID：remotesensing-2233782). Those comments are very helpful for revising and improving our paper, as well as the important guiding significance to other research. We have studied the comments carefully and made corrections which we hope meet with approval. The main corrections are in the manuscript and the responds to the reviewers’ comments are as follows (the replies are highlighted in red ).
Replies to the reviewer’ comments:
Point 1: I find it difficult to understand the experimental design of this work. the description was too concise to obtain useful information from it. Also, there were few explanations of the rationale for the study design. Please give more details about the study area and experiment design.
Response 1: Thank you for your comments. We have described the experimental design in more detail in the revised version. Please see the attachment.
Point 2: Please highlight the novelties of this study. What is the research gap? Moreover, the language should be improved by native speaker.
Response 2: We have listened to your comments and made large-scale changes to the introduction. We hope the results are satisfactory to you. In addition, we do invite a friend of us who is a native English speaker from the USA to help polish our article. And we hope the revised manuscript could be acceptable for you.
Point 3: What is the author's rationale for using the sparrow Search Algorithm to construct the optimal random forest model, have other methods been tried, and why does the method yield the optimal model?
Response 3: In fact, we also tried to use sparrow search algorithm to improve the SVM model, and the results are as follows. SSA-SVM can improve the estimation accuracy of SVM in the modeling set, but the estimation accuracy of SSA-SVM model in the validation set is very poor. In general, SSA can improve RF better, so SSA-RF is used to estimate anthocyanin content in this paper. Please see the attachment.
Point 4: I noticed that Anth in Figure 3 is missing units of measurement, please add.
Response 4: We are glad you found this problem. Allow me to explain. The anthocyanin value obtained by Dualex 4 instrument is a relative value, which has a linear response to the real value. Therefore, most people will use the measured value as the real value in the field of agricultural remote sensing. In recent years, it has been proposed that the unit of Dualex 4 is ug/cm2, but this issue is still controversial. Therefore, the anthocyanin content in this paper is unitless.
Point 5: Please check the names of Figure 4 for errors, as well as the other figures in the article. Before submitting a revision be sure that your material is properly prepared and formatted.
Response 5: Thank you for your suggestion. We have conducted a comprehensive inspection of the article and hope that the result is satisfactory to you.
Once again, thank you very much for your constructive comments and suggestions which would help us both in English and in depth to improve the quality of the paper.
Kind regards,
Zijuan Zhang
E-mail: [email protected]
Corresponding author: Qingrui Chang
E-mail address: [email protected]
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
Thanks to the authors for considering the comments and suggestions. In the present version of the manuscript, the authors have explained and answered general comments accordingly. Nevertheless, grammar of the manuscript could be further improved throughout the whole manuscript, and figures in the manuscript, such as figure1,2,3, and so on, are combination of different figures, these figures should be numbered, i.e., figure 2(a), 2(b), to make these figures more clearly to readers. I believe these minor problems should be addressed before manuscript acceptance.
Author Response
Thank you very much for your recognition and for your serious responsibility in reviewing. We have modified it according to your comments, please check.
Reviewer 2 Report
The revision has improved a lot. It is time to consider the publication of the manuscript.
Lines 69-72, 'With the improvement, ..., redundant data', should be removed. You have a minimal dataset, you do not need GPU, and this is not relative to the SSA.
Author Response
Thank you very much for your recognition and for your serious responsibility in reviewing. We have modified it according to your comments, please check.
Reviewer 3 Report
No more comments.
Author Response
Thank you very much for your recognition and for your serious responsibility in reviewing.
