Establishment of Plot-Yield Prediction Models in Soybean Breeding Programs Using UAV-Based Hyperspectral Remote Sensing
Round 1
Reviewer 1 Report
Establishment of plot-yield prediction models in soybean breeding programs using UAV-based hyperspectral remote sensing
The study refers to the use of UAV-based hyperspectral remote sensing for developing plot-yield prediction models for soybean breeding lines. For achieving this, the authors used a hyperspectral spectrometer that was mounted on a UAV for collecting the spectral data of 1103 soybean breeding lines at different crop stages. In addition, the authors assessed different spectral vegetation indices for identifying the most suitable for the prediction models. The authors found that NDVI and RVI were the best combination of vegetation indices for developing their plot-yield prediction model and that the most suitable crop stage was the initial seed filling stage (R5) among with any other crop stage for a two-growth stage yield prediction model. Moreover, the best-fitted plot-yield prediction models achieved coincidence rate between 52.4 % and 58.5 %. Finally, the authors stated that stablished model construction elements and methods can be 36 used for the establishment of local models for pre-harvest yield-selection and post-harvest 37 integrated yield-selection in advanced breeding nurseries as well as yield potential prediction in 38 plant-derived-line nurseries in soybean breeding programs.
Comments
Abstract
L29. The authors should refer to the crop stage with text (e.g. the initial seed filling stage (R5)).
Introduction
L106-112. These sentences should be removed since they do not add any value to the manuscript.
Materials and Methods
L196. Table 1 title should be written to “The spectral vegetation indices used in the present study”.
L209-210. The authors should write in detail the crop growth stages e.g. full flowering stage (R2) etc.
L231-237. This sentence should be revised since the meaning is not so clear (e.g. “In the first method, all the models were evaluated...”. Accordingly “In the second method,...”).
L242-243. This sentence should be revised.
Results
L251. The authors need to write “12.81%and” as 12.81% and”
L252-258. Past tense should be used.
L277-279. “The dark red area presented the highest correlation zone...”.
L289-292. Past tense should be used.
L296-297. This information should be included in the Materials and Methods section (see L209-210).
L313-316. Past tense should be used.
L324-326. Past tense should be used.
L326-328. The sentence should be revised.
L328-330. Past tense should be used.
L331-334. Past tense should be used.
L339. This sentence should be removed.
L341-351. Past tense should be used.
L359-361. Past tense should be used.
L361-363. It is more appropriate to use third person in this sentence e.g. “Based on the aforementioned, a linear function with two vegetation indices (namely NDVI and RVI) at R5 stage was established for the second round of the yield-prediction models assessment.”
L366-368. Past tense should be used.
L368. “Linear models...” instead of “A linear model...”.
L368-369. “the established models were evaluated based on their precision,....” instead of “the established models are evaluated with their model precision,...”.
L374. ”the model MA1 presented the...”instead of ”the model MA1 is characterized with the...”.
L381-386. It is better to move this sentence in the discussion section.
L381-383. Past tense should be used.
L390-399. This information must be included in L389.
L401. Past tense should be used.
L412. Please use “significantly” instead of “obviously”.
L414-416. Past tense should be used.
L417-421. Please revise the sentence.
L421-422. Past tense should be used.
L439. It is better to use “Based on the results...” instead of “It is obvious...”.
L439-450. Past tense should be used.
L450-452. It is better to move these sentences in the discussion section.
L453-465. Past tense should be used.
L466. It is better to use “of best fitted” instead of “for superior”.
L467-478. The materials should be distinguished in data and breeding lines. Past tense should be used.
L480-484. Past tense should be used.
L481-484. The sentence should be revised.
L508-518. These sentences should be revised. It is recommended not to include the entire equations in the main text but provide a reference to the appropriate supplementary table.
L523-526. Past tense should be used.
Discussion
L529-621. The complete discussion section should be revised. The authors need to discuss on the results that they presented earlier in the manuscript and justify their outcomes based on scientific references. For example, the authors can discuss about the suitability of the spectral vegetation indices that they chose to use, the accuracy of the models (with single vegetation index and with combination of vegetation indices) that they developed, the suitable crop stages for assessing soybean breeding lines. The authors should check if their results are in accordance or not with the studies of other researchers and state their beliefs on what was the reason of this outcome.
Conclusions
L623-645. Past tense should be used.
L625-634. This sentence should be revised. The authors should present the results in a more simple way.
Supplementary Materials
Supplementary Figure 2. The figure is not presented in a clear way and should be revised because there are many overlaps among the lines.
General Comment
I recommend this manuscript to be accepted after minor revision. The study provides an innovative method to assess yield potential of multiple soybean breeding lines using hyperspectral data. Hyperspectral data are considered valuable in agriculture because they can provide information in a fast and non-destructive way, which this is not the case when applying destructive measurements. Moreover, the results of their study are very promising for adoption of their approach to the breeding programs of other arable crops of high importance. However, the manuscript needs to be revised in the discussion section because the authors need to support their findings based on the results that were found by other researchers. In addition, the authors need to clarify in their text the term “materials” since in some cases it refers to data and in other cases in breeding lines.
Author Response
RESPONSES TO THE COMMENTS ON THE MANUSCRIPT FROM THE EDITOR AND REVIEWERS
(Manuscript title: Establishment of plot-yield prediction models in soybean breeding programs using UAV-based hyperspectral remote sensing)
The authors thank the editor and reviewers very much for kindly spending much time in reviewing the manuscript and providing detailed comments to the manuscript for revision. According to the comments and suggestions, we have made a major revision on the manuscript. The following are the authors' responses to the comments.
RESPONSE TO THE COMMENTS FROM REVIEWER #1
Reviewer #1: Comments and Suggestions for Authors
Overview of the manuscript: The study refers to the use of UAV-based hyperspectral remote sensing for developing plot-yield prediction models for soybean breeding lines. For achieving this, the authors used a hyperspectral spectrometer that was mounted on a UAV for collecting the spectral data of 1103 soybean breeding lines at different crop stages. In addition, the authors assessed different spectral vegetation indices for identifying the most suitable for the prediction models. The authors found that NDVI and RVI were the best combination of vegetation indices for developing their plot-yield prediction model and that the most suitable crop stage was the initial seed filling stage (R5) among with any other crop stage for a two-growth stage yield prediction model. Moreover, the best-fitted plot-yield prediction models achieved coincidence rate between 52.4 % and 58.5 %. Finally, the authors stated that stablished model construction elements and methods can be used for the establishment of local models for pre-harvest yield-selection and post-harvest integrated yield-selection in advanced breeding nurseries as well as yield potential prediction in 38 plant-derived-line nurseries in soybean breeding programs.
General comment: I recommend this manuscript to be accepted after minor revision. The study provides an innovative method to assess yield potential of multiple soybean breeding lines using hyperspectral data. Hyperspectral data are considered valuable in agriculture because they can provide information in a fast and non-destructive way, which this is not the case when applying destructive measurements. Moreover, the results of their study are very promising for adoption of their approach to the breeding programs of other arable crops of high importance. However, the manuscript needs to be revised in the discussion section because the authors need to support their findings based on the results that were found by other researchers. In addition, the authors need to clarify in their text the term “materials” since in some cases it refers to data and in other cases in breeding lines.
Response: Thanks for the reviewer’s evaluation. We have revised the discussion section and clarified the meaning of materials as breeding lines in the text. In Discussion section, we have added a paragraph for comparisons of our results with those in the literature.
Abstract section:
Comment 1: L29. The authors should refer to the crop stage with text (e.g. the initial seed filling stage (R5)).
Response 1: According to the reviewer’s suggestion, the sentence ‘the initial seed filling stage (R5)’ has been added.
Introduction section:
Comment 2: L106-112. These sentences should be removed since they do not add any value to the manuscript.
Response 2: These sentences are introductory words to the aims of the study that the large scale yield-test program needs some auxiliary approach for final yield selection in addition to harvested real plot yield. According to the comment, the sentences have been improved to make a smooth connection with the followed sentences.
Materials and Methods section:
Comment 3: L196. Table 1 title should be written to “The spectral vegetation indices used in the present study”.
Response 3: According to the reviewer’s comment, the title has been changed.
Comment 4: L209-210. The authors should write in detail the crop growth stages e.g. full flowering stage (R2) etc.
Response 4: According to the reviewer’s suggestion, the growth stages have been completed as full flowering stage (R2), the full podding stage (R4), the initial seed filling stage (R5), and the full seed filling stage (R6). Please see the revised manuscript.
Comment 5: L231-237. This sentence should be revised since the meaning is not so clear (e.g. “In the first method, all the models were evaluated...”. Accordingly “In the second method,...”).?
Response 5: According to the reviewer’s comment, the sentence has been re-written and added. Please see the revised manuscript.
Comment 6: L242-243. This sentence should be revised.
Response 6: According to the reviewer’s comment, the sentence has been re-written. Please see the revised manuscript.
Results section:
Comment 7: L251. The authors need to write “12.81%and” as 12.81% and”.
Response 7: According to the reviewer’s comment, it has been changed.
Comment 8: L252-258. Past tense should be used.
Response 8: According to the reviewer’s comment, the sentence has been re-written in past tense.
Comment 9: L277-279. “The dark red area presented the highest correlation zone...”.
Response 9: According to the reviewer’s comment, the sentence has been re-written.
Comment 10: L289-292. Past tense should be used.
Response 10: According to the reviewer’s comment, the sentence has been re-written in past tense.
Comment 11: L296-297. This information should be included in the Materials and Methods section (see L209-210).
Response 11: According to the reviewer’s comment, this information has been moved to 2.5. Establishment and verification of the yield prediction models of Materials and Methods.
Comment 12: L313-316. Past tense should be used.
Response 12: According to the reviewer’s comment, the sentence has been re-written in past tense.
Comment 13: L324-326. Past tense should be used.
Response 13: According to the reviewer’s comment, the sentence has been re-written in past tense.
Comment 14: L326-328. The sentence should be revised.
Response 14: According to the reviewer’s comment, the sentence has been re-written: “However, all the results showed only slight differences of CV among band values and vegetation indices under 21 different sampling areas.”
Comment 15: L328-330. Past tense should be used.
Response 15: According to the reviewer’s comment, the sentence has been re-written in past tense.
Comment 16: L331-334. Past tense should be used.
Response 16: According to the reviewer’s comment, the sentence has been re-written in past tense.
Comment 17: L339. This sentence should be removed.
Response 17: According to the reviewer’s comment, the sentence has been removed with the material notes included in the parentheses.
Comment 18: L341-351. Past tense should be used.
Response 18: According to the reviewer’s comment, the sentence has been re-written in past tense.
Comment 19: L359-361. Past tense should be used.
Response 19: According to the reviewer’s comment, the sentence has been re-written in past tense.
Comment 20: L361-363. It is more appropriate to use third person in this sentence e.g. “Based on the aforementioned, a linear function with two vegetation indices (namely NDVI and RVI) at R5 stage was established for the second round of the yield-prediction models assessment.”
Response 20: According to the reviewer’s comment, the sentence has been changed as the reviewer revised.
Comment 21: L366-368. Past tense should be used.
Response 21: According to the reviewer’s comment, the sentence has been re-written in past tense.
Comment 22: L368. “Linear models...” instead of “A linear model...”.
Response 22: According to the reviewer’s comment, the expression has been changed.
Comment 23: L368-369. “the established models were evaluated based on their precision,....” instead of “the established models are evaluated with their model precision,...”.
Response 23: According to the reviewer’s comment, the sentence has been re-written as the reviewer indicated.
Comment 24: L374. “the model MA1 presented the...”instead of “the model MA1 is characterized with the...”.
Response 24: According to the reviewer’s comment, the sentence has been re-written as the reviewer indicated.
Comment 25: L381-386. It is better to move this sentence in the discussion section.
Response 25: According to the reviewer’s suggestion, the sentence “Thus, it was recognized that the model precision depends on their source materials, those from a same test was usually better than those from different tests even the sample size (number of total lines) increased, such as MA1+B1 and MA2+B2 (but not MA3+B3) were better than MA4+B4, and different material tests may provide different model precision, such as MA1+B1 is better than MA3+B3, and MA1 and MB1 better than MA2 and MB2, and different year (environment) may cause different model precision even for a same set of materials, such as MA6 better than MB6. ” (L381-386) has been moved to Discussion section (Line 600-606).
Comment 26: L381-383. Past tense should be used.
Response 26: According to the reviewer’s comment, the sentence has been re-written in past tense (It has been moved to Discussion).
Comment 27: L390-399. This information must be included in L389.
Response 27: According to the reviewer’s comment, this information has been moved to table notes.
Comment 28: L401. Past tense should be used.
Response 28: According to the reviewer’s comment, the sentence has been re-written in past tense.
Comment 29: L412. Please use “significantly” instead of “obviously”.
Response 29: According to the reviewer’s comment, “obviously” has been changed to “significantly”.
Comment 30: L414-416. Past tense should be used.
Response 30: According to the reviewer’s comment, the sentence has been re-written in past tense.
Comment 31: L417-421. Please revise the sentence.
Response 31: According to the reviewer’s comment, the sentence has been re-written as “Based on the above results, the third round yield-prediction models for the 17 material sets with two growth stages (R5+R4 for each material set and R5+R2, R5+R6 and R5+R4 for A1 and A6 material sets) and two vegetation indices (NDVI and RVI), in a total of 21 yield-prediction models were established using the MATLAB procedure and then evaluated further. As indicated before, half set of breeding lines was used for modelling and half set for validation. The results were summarized in Table 5 (the model equations listed in Supplementary Table 8).”
Comment 32: L421-422. Past tense should be used.
Response 32: According to the reviewer’s comment, the sentence has been re-written in past tense.
Comment 33: L439. It is better to use “Based on the results...” instead of “It is obvious...”.
Response 33: According to the reviewer’s comment, the sentence has been re-written as the reviewer indicated.
Comment 34: L439-450. Past tense should be used.
Response 34: According to the reviewer’s comment, the sentence has been re-written in past tense.
Comment 35: L450-452. It is better to move these sentences in the discussion section.
Response 35: According to the reviewer’s suggestion, these sentences have been moved to the discussion section Line 597-599.
Comment 36: L453-465. Past tense should be used.
Response 36: According to the reviewer’s comment, these sentences have been re-written in past tense.
Comment 37: L466. It is better to use “of best fitted” instead of “for superior”.
Response 37: According to the reviewer’s comment, the expression has been changed as the reviewer indicated.
Comment 38: L467-478. The materials should be distinguished in data and breeding lines. Past tense should be used.
Response 38: According to the reviewer’s comment, the “materials” and “breeding lines” have been distinguished throughout the manuscript. In addition, the English tense has been checked throughout the manuscript.
Comment 39: L480-484. Past tense should be used.
Response 39: According to the reviewer’s comment, these sentences have been re-written in past tense.
Comment 40: L481-484. The sentence should be revised.
Response 40: According to the reviewer’s comment, the sentence has been re-written: ‘After a further comparison comprehensively, the models of MA1-2, MA6-2 and MA4-2 were good in coincidence rates for all the selection categories (eliminated, reserved and promoted) in all the populations and were chosen for utilization in plot-yield prediction in yield breeding programs (See Table 7 and its notes for details).’
Comment 41: L508-518. These sentences should be revised. It is recommended not to include the entire equations in the main text but provide a reference to the appropriate supplementary table.
Response 41: According to the reviewer’s comment, these equations have been moved to Supplementary Table 8 and cited in the main text. Please see the revised manuscript.
Comment 42: L523-526. Past tense should be used.
Response 42: According to the reviewer’s comment, the sentence has been re-written and the past tense were used. Please see the revised manuscript.
Discussion section:
Comment 43: L529-621. The complete discussion section should be revised. The authors need to discuss on the results that they presented earlier in the manuscript and justify their outcomes based on scientific references. For example, the authors can discuss about the suitability of the spectral vegetation indices that they chose to use, the accuracy of the models (with single vegetation index and with combination of vegetation indices) that they developed, the suitable crop stages for assessing soybean breeding lines. The authors should check if their results are in accordance or not with the studies of other researchers and state their beliefs on what was the reason of this outcome.
Response 43: According to the reviewer’s suggestion, a brief description of the study and comparisons of the present results to those in the literature as the reviewer suggested have been added in Discussion section. Please see the revised version.
Conclusions section:
Comment 44: L623-645. Past tense should be used.
Response 44: According to the reviewer’s comment, the sentence has been re-written in past tense.
Comment 45: L625-634. This sentence should be revised. The authors should present the results in a more simple way.
Response 45: According to the reviewer’s comment, the sentences have been re-written and simplified. Please see the revised version.
Supplementary Materials section:
Comment 46: Supplementary Figure 2. The figure is not presented in a clear way and should be revised because there are many overlaps among the lines.
Response 46: In Supplementary Figure 2, the 21 sampling-unit curves are close to each other, even seriously overlapped. We tried to enlarge the figure for separating the curves, but not obviously changed. It indicates that all the curves (average spectral reflectance vs. wavelength) among the sampling unit size are very similar with the smaller unit size (pink color) of a little higher reflectance, the medium unit size (light green color) of medium reflectance and larger unit size of a little lower reflectance. This figure was put in the Supplementary Materials section rather than in main text only for showing the small difference of average spectral reflectance vs. wavelength among the sampling-unit size.
Author Response File: 
Author Response.docx
Reviewer 2 Report
Dear Authors,
I revised the manuscript “Establishment of plot-yield prediction model in soybean breeding programs using UAV-based hyperspectral remote sensing” submitted to the Remote Sensing Journal. The paper is very interesting, well written and presented plot-yield prediction models for soybean using UAV based hyperspectral remote sensing for yield-selection in large scale soybean breeding programs. However, I have some concerns, which need to be conferred before final publication.
Major Comments:
· In Introduction chapter, last paragraph, you discussed some of your methods. Please take that into the Methodology section. The last paragraph should be mostly on the objective of your study.
· I would suggest to separate Figure 1 into two and discuss your UAV in Figure 2.
· Please include the Supplementary Figure 1 and Figure 2 in the main manuscript.
Minor Comments:
· Line 95, GNDVI is built with the combination of Green and Near Infrared. Please correct the sentence.
· Line 98, please elaborate different vegetation indices, such as SAVI, EVI etc. at the first instance.
· Line 102, please use correlation of determination before (R2).
· Line 103, please delete the year after the author name in citation. Don’t need to use (2001) after Ma et al. Same for whole manuscript.
· Line 116, please explain UVA. It should be UAV.
· Line 132, please explain NJRIKY in the first instance.
· Line 144, please use plants·ha-1, instead of plants/ha, to follow the style of journal. Same for whole manuscript.
· Line 146, please use high resolution image for Figure 1 (A).
· Line 204, please follow the journal guideline for equation. The equation should be in the middle.
· Please increase the size of Figure 4.
Author Response
RESPONSES TO THE COMMENTS ON THE MANUSCRIPT FROM THE EDITOR AND REVIEWERS
(Manuscript title: Establishment of plot-yield prediction models in soybean breeding programs using UAV-based hyperspectral remote sensing)
The authors thank the editor and reviewers very much for kindly spending much time in reviewing the manuscript and providing detailed comments to the manuscript for revision. According to the comments and suggestions, we have made a major revision on the manuscript. The following are the authors' responses to the comments.
RESPONSE TO THE COMMENTS FROM REVIEWER #2
Reviewer #2: Comments and Suggestions for Authors
General comment: I revised the manuscript “Establishment of plot-yield prediction model in soybean breeding programs using UAV-based hyperspectral remote sensing” submitted to the Remote Sensing Journal. The paper is very interesting, well written and presented plot-yield prediction models for soybean using UAV based hyperspectral remote sensing for yield-selection in large scale soybean breeding programs. However, I have some concerns, which need to be conferred before final publication.
Response: Thanks for the reviewer’s evaluation. We have revised the manuscript according to the reviewer’s comments.
Comment 1: In Introduction chapter, last paragraph, you discussed some of your methods. Please take that into the Methodology section. The last paragraph should be mostly on the objective of your study.
Response 1: According to the comment, the last paragraph in Introduction has been adjusted to mainly on the objectives of this study. Those contents related to methods have been moved to Materials and Methods section. For easy to understand the complete procedure, an overview of the research methodology was added to the beginning of Materials and Methods section.
Comment 2: I would suggest to separate Figure 1 into two and discuss your UAV in Figure 2.
Response 2: According to the reviewer’s comment, the Figure 1 has been separated into Figure 1 and Figure 2.
Comment 3: Please include the Supplementary Figure 1 and Figure 2 in the main manuscript.
Response 3: We are worrying about too many figures and tables in the manuscript, so we add an overview of the whole research process at the beginning of Materials and Methods section as in the response to comment 1 to save Supplementary Figure 1, and since we have put another figure (Figure 4) for hyperspectral reflectance sampling in main text so Supplementary Figure 2 was still saved. Thus we would still to keep them as Supplementary Figures. We are happy if the reviewer and editor prefer to move them in the main text.
Comment 4: Line 95, GNDVI is built with the combination of Green and Near Infrared. Please correct the sentence.
Response 4: According to the reviewer’s comment, the sentence has been re-written and added. Please see the revised manuscript.
Comment 5: Line 98, please elaborate different vegetation indices, such as SAVI, EVI etc. at the first instance.
Response 5: According to the reviewer’s comment, these different vegetation indices have been elaborated with reference cited. Please see the revised manuscript.
Comment 6: Line 102, please use correlation of determination before (R2).
Response 6: According to the reviewer’s comment, the coefficient of determination has been added before R2.
Comment 7: Line 103, please delete the year after the author name in citation. Don’t need to use (2001) after Ma et al. Same for whole manuscript.
Response 7: According to the reviewer’s comment, all the citations have been checked and corrected with the year deleted.
Comment 8: Line 116, please explain UVA. It should be UAV.
Response 8: Yes, it should be UAV, thanks.
Comment 9: Line 132, please explain NJRIKY in the first instance.
Response 9: According to the reviewer’s comment, NJRIKY was explained as the name of RIL (recombinant inbred line) population in the text.
Comment 10: Line 144, please use plants·ha-1, instead of plants/ha, to follow the style of journal. Same for whole manuscript.
Response 10: According to the reviewer’s comment, all the similar expressions have been replaced with the suggested ones.
Comment 11: Line 146, please use high resolution image for Figure 1 (A).
Response 11: According to the reviewer’s comment, Figure 1 (A) has been replaced with high resolution image, in addition, Figure 1 has been separated into two figures according to reviewer #1’s comment.
Comment 12: Line 204, please follow the journal guideline for equation. The equation should be in the middle.
Response 12: According to the reviewer’s comment, the equation has been placed in the middle place.
Comment 13: Please increase the size of Figure 4.
Response 13: According to the reviewer’s comment, the size of Figure 4 has been increased. Please see the revised manuscript.
Author Response File: Author Response.docx
Reviewer 3 Report
The objectives of the study, and the first part of the discussion point out at relevant problems to be solved.
The paper presents a significant amount of data and a lot of results that are presented in quite a confusing manner. There are too many tables, too many models, too many and complicated abbreviations (for example those referring to the experiments), that makes the manuscript very difficult to read and understand.
The analysis methods are voluminous but they end on empirical relationships that seem significant because of the high number of datapoints. However there is no justification about why some models are better than others. The authors recognize that the model to be used depend on the source materials, and that different years and environments could generate different relationships that make the results presented useless even for themselves in years to come.
One of the problems stated in the introduction is the soil heterogeneity in enlarged field trials (page 2, line 52), that apparently is solved with the UAV-based hyperspectral reflectance (page 17, line 617), but the result is that the plot could be smaller. It does not really solve the soil heterogeneity problem.
I consider that there is a positive result from this study, which is the estimation of the plot size that should be discarded due to border effects (figure 4).
Other minor comments:
I don't understand why the authors use the expression "blocks-in-replication" if they finally used randomized complete block instead (page 3 line 137-138).
There are non-scientific expression all over the manuscript such as "vary obviously" (page 3 line 139), "the ten most popular vegetation indices" (page 5 line 192), "their experiment errors not very large" (page 7, line 253), "but not obviously" (page 11 line 412). These are only examples of subjective expressions that should be replaced with objective statements.
Some tables end with the expression "The same is true for later tables". Each table should be a complete piece of information, and the reader should not be asked to read all the tables for a complete understanding.
In table 1 there are two indices (NDVI and RVI) that are defined "in this paper" but are well known in the literature. Those indices are constructed with "anyone band" (as stated by the authors), but one of them should be in the visible range while the other should be in the infra-red part of the spectrum. If this is not true, then the names of the indices should be changed, and it would be true that this manuscript is the first to formulate them.
As a summary, due to the lack of explanation about why one model is better than the other, and the lack of reproducibility of the results, I consider that the paper is not suitable for publication in remote sensing.
Author Response
RESPONSES TO THE COMMENTS ON THE MANUSCRIPT FROM THE EDITOR AND REVIEWERS
(Manuscript title: Establishment of plot-yield prediction models in soybean breeding programs using UAV-based hyperspectral remote sensing)
The authors thank the editor and reviewers very much for kindly spending much time in reviewing the manuscript and providing detailed comments to the manuscript for revision. According to the comments and suggestions, we have made a major revision on the manuscript. The following are the authors' responses to the comments.
RESPONSE TO THE COMMENTS FROM REVIEWER #3
Reviewer #3: Comments and Suggestions for Authors
Comment 1: The objectives of the study, and the first part of the discussion point out at relevant problems to be solved.
Response 1: According to the comment, we have emphasized the objective of the present study in Introduction section that “the present study aimed at to explore how to establish prediction models for plot-yields in breeding programs for soybeans using UAV-based hyperspectral remote sensing, to establish, validate and select optimal plot-yield prediction models, and then to demonstrate their efficiency and effectiveness in real breeding programs”. Then in the first part of discussion, the outlines of the realized objectives (results) were summarized briefly.
Comment 2: The paper presents a significant amount of data and a lot of results that are presented in quite a confusing manner. There are too many tables, too many models, too many and complicated abbreviations (for example those referring to the experiments), that makes the manuscript very difficult to read and understand.
Response 2: Yes, we understand that the manuscript may be difficult to read if not well-organized because it involves a large number of data on plot-yield and UAV-remote-sensing, optimization of each of the four model construction elements (spectral reflectance sample size, vegetation index, growth for remote sensing and regression types, as well as their combinations) and five linked analytical steps from data collection to model construction and validation. Therefore, according to the comments, we introduce the objectives in details in Introduction section, describe the overview procedures at the beginning in Materials and Methods section, and introduce the results along with the five linked analytical steps. In addition, the numbers and sizes of tables and figures were reduced already. We expect that the revised version is better than the original version.
Comment 3: The analysis methods are voluminous but they end on empirical relationships that seem significant because of the high number of datapoints. However there is no justification about why some models are better than others. The authors recognize that the model to be used depend on the source materials, and that different years and environments could generate different relationships that make the results presented useless even for themselves in years to come.
Response 3: Yes, we agree with the reviewer’s comments. The objective of the present study aimed at to find plot-yield prediction models as an auxiliary tool for yield selection in advanced breeding nurseries. The materials used were breeding lines in real soybean breeding programs for that the established prediction models can fit the local breeding program and so for the data, rather than those designed material sets with built-in differences. Various models were established and compared with their RMSE to choose better ones and then validated using another half set of materials (half set for model construction and half set for model validation). The models were further validated with the total material sets and then compared with the real selection results of the breeding program. These are the justification process in the present study. The model establishment and selection process is impersonal with subjective factors avoided, even with some empirical property. In this situation there is no additional justification for the selected models except the reasons explored in the four model construction elements and five linked analytical steps. Thus, we prefer to choose several models rather than only a best one and use them jointly to make plot-yield prediction since there is no critical difference among the selected ones and it does not cause any more expense except some more computer calculation. Accordingly, in our conclusion we indicated: “Seventeen prediction models composed of NDVI and RVI vegetation indices at R5 growth stage and then 21 prediction models composed of the two vegetation indices at two growth stages (R5 plus another one) were established. In choosing the best models, the modelling RM2 and modelling RMSEM, verification RV2 and verification RMSEV, and their sums RS2 and RMSES were evaluated and compared. Integrated with the coincidence rate between the model-predicted results and the real selection results, the models of MA1-2, MA6-2 and MA4-2 were chosen for utilization in practical breeding programs…….The established model construction elements and methods could be used in the establishment of local models for pre-harvest yield-selection and post-harvest integrated yield-selection in advanced breeding nurseries as well as yield potential prediction in plant-derived-line nurseries in soybean breeding programs.”
Comment 4: One of the problems stated in the introduction is the soil heterogeneity in enlarged field trials (page 2, line 52), that apparently is solved with the UAV-based. hyperspectral reflectance (page 17, line 617), but the result is that the plot could be smaller. It does not really solve the soil heterogeneity problem.
Response 4: The variation of soil heterogeneity is different from land piece to land piece, the breeders usually can find a piece of land fitting there testing scale (certain number of breeding lines) with a relative smaller soil heterogeneity. If the testing scale enlarged, the experiment has to extend to another piece of or neighboring land, which will introduce extra-heterogeneity causing more experiment errors. But if using a same piece of land with plot size reduced to fit the enlarged testing scale, the experiment error may be not increased even decreased, thus there is a balance between plot size and experiment error on a same piece of land. Utilizing the UAV-based hyperspectral reflectance does not solve the soil heterogeneity problem itself, but if utilizing UAV-based hyperspectral reflectance for plot-yield prediction, the plot size can be reduced to certain size providing the border area influence excluded, and that even the replication number can be reduced if a single plot can be of representativeness because our results showed that the sampling unit-size for hyperspectral reflectance can be reduced to keep a similar precision between 20% to 80% of a plot size.
Comment 5: I consider that there is a positive result from this study, which is the estimation of the plot size that should be discarded due to border effects (figure 4).
Response 5: Yes, as indicated in the last comment response, the results showed that when the proportion of the sampling unit-size was between about 20% to 80% of a plot, the obtained canopy reflectance data could be used for plot-yield prediction (especially, when the border effect removed).
Comment 6: I don't understand why the authors use the expression "blocks-in-replication" if they finally used randomized complete block instead (page 3 line 137-138).
Response 6: To make a precise yield evaluation and selection, we took incomplete block design (Blocks in Replication design) with similar maturity accessions and additional check in a same incomplete block while all blocks and replications in a same piece of land. It is convenient for field observation and management, especially in real breeding programs. Since the ANOVA showed not very large block in replication variation, to avoid use adjusted means in the model prediction process, the experiments were analyzed using randomized complete block design analytical approach. This is a usual way in using Blocks in Replication design.
Comment 7: There are non-scientific expression all over the manuscript such as "vary obviously" (page 3 line 139), "the ten most popular vegetation indices" (page 5 line 192), "their experiment errors not very large" (page 7, line 253), "but not obviously" (page 11 line 412). These are only examples of subjective expressions that should be replaced with objective statements.
Response 7: According to the reviewer’s comment, the manuscript has been checked for the non-scientific expressions which were replaced with scientific expressions. Please see the revised manuscript.
Comment 8: Some tables end with the expression "The same is true for later tables". Each table should be a complete piece of information, and the reader should not be asked to read all the tables for a complete understanding.
Response 8: According to the reviewer’s comment, the notes of tables have been completed with "The same is true for later tables" deleted.
Comment 9: In table 1 there are two indices (NDVI and RVI) that are defined "in this paper" but are well known in the literature. Those indices are constructed with "anyone band" (as stated by the authors), but one of them should be in the visible range while the other should be in the infra-red part of the spectrum. If this is not true, then the names of the indices should be changed, and it would be true that this manuscript is the first to formulate them.
Response 9: Yes, we should distinguish NDVI and RVI from the others. According to the comment, literature citations were put in Table 1, and “near infrared and visible red region” were put in the table note for NDVI and RVI.
Author Response File: Author Response.docx
Round 2
Reviewer 3 Report
The paper has improved from the last version and it is now suitable for publication.
