Early Monitoring of Cotton Verticillium Wilt by Leaf Multiple “Symptom” Characteristics
Round 1
Reviewer 1 Report
The article is very interesting and has relevant aspects of research. All experimental data are statistically confirmed. The figures and tables are clear and concise. The link to the authors (References) is correct.
Author Response
General comment: The article is very interesting and has relevant aspects of research. All experimental data are statistically confirmed. The figures and tables are clear and concise. The link to the authors (References) is correct.
Response: We sincerely acknowledge the referee for her/his careful review of the manuscript, and we thank the reviewer very much for these positive and encouraging comments. Your positive comments are the greatest recognition of our research content.
Reviewer 2 Report
My comments regarding this study are as follows:
1- The data were collected by measuring five locations on each leaf. Why weren't imaging techniques used to monitor disease? This study did not use imaging spectroscopy, which is a promising approach for early detection.
2- The article uses the word "new" 13 times, but there is nothing new to be found in it. As an example, what is the novelty of using the Relief-F algorithm in calculating the wavelength feature weights?
3- It is necessary to develop regression models that predict physiological indicators using spectral data.
4- The purpose of this study is to categorize cotton disease into two levels. A regular camera can be used with an easier and cheaper methodology to accomplish the task. There should be a discussion of why the authors have chosen this strategy.
5- Can you explain the importance of using the Boruta algorithm to select the key physiological trait factors? It is possible to select the optimal wavelength for classification directly by using variable selection methods.
6- The Evaluation Indicator column should be removed from Table 6.
7- There are several parts of the text that are difficult to understand, so the paper needs to be revised for spelling and grammar.
Author Response
Point 1: The data were collected by measuring five locations on each leaf. Why weren't imaging techniques used to monitor disease? This study did not use imaging spectroscopy, which is a promising approach for early detection.
Response 1: We give special thanks to the referee for carefully reviewing the manuscript. In this study, we focused on the performance of the physiological symptoms of leaves on the spectral reflectance in the early stage of cotton Verticillium wilt. By integrating the spectral indices and spectral bands of key physiological characteristic factors, the early monitoring index of cotton verticillium wilt was established to realize the early monitoring of cotton verticillium wilt. This research can be achieved with ground-object spectrometers without the use of imaging spectroscopy. It is undeniable that imaging spectroscopy is very promising in early disease monitoring. As an information acquisition technology that integrates image processing and spectroscopy, imaging spectroscopy can not only allow us to visually observe the spatial information of the image, but also to obtain the spectral information of each pixel of the target [1]. In order to further improve the monitoring accuracy of disease degree, in future research we will use spectral imaging technology to integrate the spatial and spectral information of leaves under VW stress and enhance the ability to express the characteristic information of various "symptoms" caused by disease. We have added the discussions in the revised manuscript and presents the limitations of the study, in the revision Page 20 Line 506-509.
- Abdulridha, J.; Ampatzidis, Y.; Kakarla, S.C.; Roberts, P. Detection of target spot and bacterial spot diseases in tomato using UAV-based and benchtop-based hyperspectral imaging techniques. Precis. Agric. 2020, 21, 955-78.
Point 2: The article uses the word "new" 13 times, but there is nothing new to be found in it. As an example, what is the novelty of using the Relief-F algorithm in calculating the wavelength feature weights?
Response 2: We thank the reviewer for the useful suggestion. Our research goal is to achieve early monitoring of cotton Verticillium wilt through the "symptoms" characteristics of the leaves. The novelty of this article is that we have constructed relevant spectral indices for various physiological and functional traits that appear in the early stage of cotton Verticillium wilt, and established disease monitoring indicators that integrate multiple "symptom" characteristics. A large number of studies have proved that the vegetation indices are a fast and non-invasive method of identifying plant diseases [1, 2]. However, there are certain limitations of early monitoring of the use of vegetation index for diseases [3]. Remote sensing-based VIs are not disease-specific because they are related to different biophysical plant properties. Therefore, it is necessary to construct spectral indices of multiple sensitive physiological parameters under the coercion of specific diseases. Early monitoring indicators of the combination of multi-physiological indices are not only simple and effective, strongly explained, but also further improve the space adaptability and accuracy of the early detection model of the disease. In the revised manuscript we have removed some of the "new" expressions, On page 2 of the revised manuscript, lines 65, 75, page 1, line 317, page 17, line 378, and page 21, line 530.
- Zheng, Q.; Huang, W.; Cui, X.; Dong, Y.; Shi, Y.; Ma, H.; Liu, L. Identification of wheat yellow rust using optimal three-band spectral indices in different growth stages. Sensors-Basel. 2018, 19, 35.
- Chen, T.; Zhang, J.; Chen, Y.; Wan, S.; Zhang, L. Detection of peanut leaf spots disease using canopy hyperspectral reflectance. Comput. Electron. Agr. 2019, 156, 677-83.
- Meena, S.V.; Dhaka, V.S.; Sinwar, D. Exploring the Role of Vegetation Indices in Plant Diseases Identification.:IEEE, 2020,372-7.
Point 3: It is necessary to develop regression models that predict physiological indicators using spectral data.
Response 3: We thank the reviewer for this useful comment and followed the suggestion. And used the dataset of diseased leaves to construct linear models of relevant physiological parameters for the three newly constructed spectral indices and the existing classical vegetation indices. the performance of EWT1500 (R2 = 0.70, RMSE = 0.02) and EWT2100 (R2 = 0.70, RMSE = 0.01) in evaluating EWT is better than that of DSWI1 (R2 = 0.70, RMSE = 0.10) and DSWI2(R2 = 0.46, RMSE = 0.23), and the performance of Chla701 (R2 = 0.42, RMSE = 0.11) in evaluating Chla is better than PSSRa (R2 = 0.09, RMSE = 0.71). These results indicated that the newly constructed physiological spectral index could more accurately represent the variation of physiological characteristics of cotton leaves under VW stress. In addition, the leaf temperature sensitive wavelength of 1610 nm screened by the Relief-F algorithm also has a good quantitative relationship with the leaf temperature under VW stress (R2=0.35, RMSE=0.23), which can reflect the temperature change of cotton leaves under VW stress (Figure 8-G). These have been added on pages 15-16, lines 339-359 of the revised manuscript.
Figure 8. Quantitative relationship of physiological characteristics of cotton verticillium wilt under early stress based on spectrum. Where (A), (B), (D), (E) are the quantitative relationship between spectral indexes DSWI1, DSWI2, EWT2100, EWT1500 and equivalent water thickness (EWT); (C) and (F) are the quantitative relationship between PSSRa, Chla701 and chlorophyll a (Chla); (G) is the quantitative relationship between the reflectivity with the spectral wavelength of 1610 nm and the leaf temperature.
Point 4: The purpose of this study is to categorize cotton disease into two levels. A regular camera can be used with an easier and cheaper methodology to accomplish the task. There should be a discussion of why the authors have chosen this strategy.
Response 4: We appreciate this suggestion. We have made a detailed comparison of the advantages and limitations of hyperspectral technology and RGB images in early disease detection, in lines 482-510 on pages 20-21 of the revised manuscript. Image detection has good performance in the identification of lesions, the latest research has used RGB images to automatically detect plant diseases [1] and can detect the diseases that have just appeared in the lesions, but research is not considering the disease type, which means that it cannot be determined whether it is caused by a specific biotic or abiotic stress. compared with the image detection technology of RGB cameras, hyperspectral data has the characteristics of high resolution and strong continuity [2], which can detect the physiological changes of cotton leaves caused by Verticillium wilt in a more targeted manner, which helps us to better early and more accurately detect cotton leaf infestation with VW. RGB imaging technology has rich spatial information and has certain potential in early disease monitoring. In future research work, we will combine imaging technology to achieve more accurate cotton yellow disease early detection.
- Palma, D.; Blanchini, F.; Montessoro, P.L. A system-theoretic approach for image-based infectious plant disease severity estimation. Plos One. 2022, 17, e272002.
- Abdulridha, J.; Ampatzidis, Y.; Ehsani, R.; de Castro, A.I. Evaluating the performance of spectral features and multivariate analysis tools to detect laurel wilt disease and nutritional deficiency in avocado. Electron. Agr. 2018, 155, 203-11.
Point 5: Can you explain the importance of using the Boruta algorithm to select the key physiological trait factors? It is possible to select the optimal wavelength for classification directly by using variable selection methods.
Response 5: Through the method of variable selection, the optimal wavelength can be directly selected for classification, but it is not enough to select variables from the perspective of data mining, and the final result of variable selection must be interpretable. The goal of Boruta is to select all feature sets related to the dependent variable rather than select the feature set that can minimize the model cost function for a specific model. The significance of the Boruta algorithm is that it can help us more comprehensively understand the influencing factors of dependent variables so as to perform better and more efficient feature selection. [1] Our purpose is to explore which physiological traits are the most critical under the early stress of Verticillium wilt, and to select characteristic wavelengths more pertinently, so as to accurately judge the infection of cotton leaves by Verticillium wilt. We used the Boruta algorithm to determine the key physiological traits under the early stress of Verticillium wilt and select the characteristic wavelengths in a more targeted manner, which is beneficial to our constructed early monitoring index that is more reliable and has strong interpretability.
- Hornero, A.; Zarco-Tejada, P.J.; Quero, J.L.; North, P.R.J.; Ruiz-Gómez, F.J.; Sánchez-Cuesta, R.; Hernandez-Clemente, R. Modelling hyperspectral- and thermal-based plant traits for the early detection of Phytophthora-induced symptoms in oak decline. Remote Sens. Environ. 2021, 263, 112570.
Point 6: The Evaluation Indicator column should be removed from Table 6.
Response 6: Thanks for carefully reviewing the manuscript. We followed this suggestion, and adjusted Table 5 (The revised draft has adjusted the table numbers, and Table 6 has been revised to Table 5.) in the revision Page 18, Line 409-410.
Table 5. Accuracy evaluation of the cotton VW early detection model.
Input |
Evaluation Indicator |
Training Set |
Validation Set |
DSWI1 |
accuracy kappa |
79% |
83% |
0.19 |
0.50 |
||
R2 RMSE |
0.31 |
0.35 |
|
12.50% |
20.20% |
||
HI |
accuracy kappa |
66% |
63% |
0.19 |
0.34 |
||
R2 RMSE |
0.43 |
0.47 |
|
13.41% |
8.33% |
||
EWT1500 |
accuracy kappa |
97% |
85% |
0.91 |
0.67 |
||
R2 RMSE |
0.26 |
0.38 |
|
13.41% |
9.07% |
||
Chla701 |
accuracy kappa |
94% |
85% |
0.84 |
0.66 |
||
R2 RMSE |
0.58 |
0.59 |
|
9.63% |
7.17% |
||
CVWEI |
accuracy kappa |
99% |
95% |
0.97 |
0.89 |
||
R2 RMSE |
0.65 |
0.73 |
|
8.17% |
3.15% |
Point 7: There are several parts of the text that are difficult to understand, so the paper needs to be revised for spelling and grammar.
Response 7: Thank you for finding our misspellings. Our manuscript has got English proofreading by a native speaker, and revised the misspellings and grammars. So we hope it can meet the journal’s standard. Thanks so much for your useful comments.
Author Response File: Author Response.docx
Reviewer 3 Report
rs-1924682 “Early Monitoring of Cotton Verticillium Wilt by Leaf Multiple “Symptom” Characteristics” Mi Yang, Changping Huang, Xiaoyan Kang, Shizhe Qin, Lulu Ma, Jin Wang, Xiaoting Zhou, Xin Lv, and Ze Zhang
This article describes the design and testing of a spectral index for detecting the disease severity of VW in cotton plants. It does a good job of describing the method, the resulting combination equation, and superior performance compared to existing vegetation indices.
Table 3 needs an actual description of the upper- and lower-case letters indicating significance: what is “A” compared to “a”, or “AB” compared to “ab”, what is the difference between “A”, “B” and “C”?
In Table 6 the RMSE of Validation Set for EWT1500 appears to be an incorrect entry, as its magnitude is very much smaller than all others and it is not expressed as a percentage. Is it correct?
Figure 8 does not label all panels, although I understand that the top panels are for the Training Set (n=100) and the lower panels are for the Validation Set (n=41), and it should be noted as such in the caption for completeness. I assume that the red shaded area around each line is the 95% confidence interval of the regression, as it also is not discussed. Figure 8 is barely mentioned in the text and its contents are subsumed numerically with Table 6 that follows immediately. Unless the scatter observed in Figure 8 is useful to note and discuss somewhere then this figure can be omitted.
I would like to see the Discussion talk about the end goal for this work. The work presented shows that an index designed specifically for cotton VW is superior to other general plant disease indices, so the only competition for this procedure is traditional visual inspection. If using a hand-held custom instrument, as suggested by Mahlein et al [4], it still requires manual operation but without a high degree of experience. Further if VW can progress in severity over days to weeks, then frequency of inspection is similar so you cannot save on labour time. If there is an advantage of this “unskilled” method over “skilled” visual inspection that (accurately and reliably) discriminates when cotton leaves enter SL2 and this is an important management trigger, that would be a supporting argument for its use.
Overall the work is well presented and requires only minor changes.
Author Response
General comment: This article describes the design and testing of a spectral index for detecting the disease severity of VW in cotton plants. It does a good job of describing the method, the resulting combination equation, and superior performance compared to existing vegetation indices.
Response:We thank the reviewer very much for these positive and encouraging comments. In particular, we really appreciate the referee for her/his suggestions. Please see our responses to the detailed comments below.
Point 1: Table 3 needs an actual description of the upper- and lower-case letters indicating significance: what is “A” compared to “a”, or “AB” compared to “ab”, what is the difference between “A”, “B” and “C”?
Response 1: Thanks for carefully reviewing the manuscript. Since Table 3 duplicates what is expressed in Figure 4, we have deleted Table 3 in the revised draft and added the difference analysis results in Figure 4. In the title of Figure 4, we explain the actual descriptions of uppercase and lowercase letters' meanings, in the revision Page 11-12 Line 282-285.
Point 2: In Table 6 the RMSE of Validation Set for EWT1500 appears to be an incorrect entry, as its magnitude is very much smaller than all others and it is not expressed as a percentage. Is it correct?
Response 2: We are very sorry for the mistake. As the reviewer suggested, because of our carelessness, we filled in the wrong response. We have revised the RMSE of the EWT1500 to 9.07%.and corrected Table 5 captions in the revision Page 18, Line 408-409. (The revised draft has adjusted the table numbers, and Table 6 has been revised to Table 5.)
Point 3: Figure 8 does not label all panels, although I understand that the top panels are for the Training Set (n=100) and the lower panels are for the Validation Set (n=41), and it should be noted as such in the caption for completeness. I assume that the red shaded area around each line is the 95% confidence interval of the regression, as it also is not discussed. Figure 8 is barely mentioned in the text and its contents are subsumed numerically with Table 6 that follows immediately. Unless the scatter observed in Figure 8 is useful to note and discuss somewhere then this figure can be omitted.
Response 3: We thank the reviewer for the useful suggestion. We improved the annotation of the content in Figure 9. We also agree with the referee regarding that it is necessary to Analysis results of the scatter. We focused on comparing CVWEI based on the fusion of multiple ‘symptom’ features, and the quantitative relationship between the single spectral index Chla701 with the best VW correlation and the disease degree (Figure 9). The results confirmed the excel-lent performance of CVWEI in evaluating the early VW. Compared with Chla701, CVWEI has a more obvious boundary between health and VW stress at 0.2 (Figure 9). We have made adjustments in lines 389-399 on page 18 of the revised manuscript.
Point 4: I would like to see the Discussion talk about the end goal for this work. The work presented shows that an index designed specifically for cotton VW is superior to other general plant disease indices, so the only competition for this procedure is traditional visual inspection. If using a hand-held custom instrument, as suggested by Mahlein et al [4], it still requires manual operation but without a high degree of experience. Further if VW can progress in severity over days to weeks, then frequency of inspection is similar so you cannot save on labour time. If there is an advantage of this “unskilled” method over “skilled” visual inspection that (accurately and reliably) discriminates when cotton leaves enter SL2 and this is an important management trigger, that would be a supporting argument for its use.
Response 4: We are grateful to the reviewers for pointing out inadequacies in the discussion. We have added the discussion in this section, in lines 477-510 on pages 20-21 of the revised manuscript. The traditional method of visual inspection is to determine if disease symptoms are present on leaves, stems, or other plant parts, which is simple and quick. However, this method also relies on continuous monitoring of plants by experienced professionals and is prone to a considerable risk of error [1]. In this study, hyperspectral technology was used to invert various "symptoms" characteristics of early leaves of cotton by Verticillium wilt, so as to accurately judge the infection of cotton leaves by Verticillium wilt. In contrast to traditional visual inspection, this method does not require experienced professionals to detect large-scale disease. At the same time, CVWEI can effectively detect VW in early cotton, help managers prevent large-scale diseases in advance, and reduce economic losses caused by VW. In order to be better used in the field, it is necessary to develop portable tools to meet this requirement, which is more convenient and practical for work in the field.
- Li, X.; Zhang, Y.N.; Ding, C.; Xu, W.; Wang, X. Temporal patterns of cotton Fusarium and Verticillium wilt in Jiangsu coastal areas of China. Rep.-Uk. 2017, 7, 1-8.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
The authors have modified the paper and addressed the comments in the appropriate manner so that the paper can be accepted.