Use of Phenomics in the Selection of UAV-Based Vegetation Indices and Prediction of Agronomic Traits in Soybean Subjected to Flooding

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This study is based on UAV multispectral vegetation index characterisation of soybean genotypes in response to flooding and prediction of agronomic traits, this study is interesting and I think it can be published in agriengineering journal after revision, below are my specific comments.

1. The description of the statistical methods is not clear enough, resulting in an incomplete presentation of the results, which needs to be improved. The purpose of the method description is to allow the reader to relate to the results presentation from the flow of the author's study.

2. It is suggested that the authors add subheadings in the results section to enhance the logic of the article.

3. The description of the conclusion part is too little, and it needs to highly summarise the patterns and conclusions obtained from the whole study.

4. It is suggested that the authors add a workflow diagram.

Comments on the Quality of English Language

Minor editing of English language required.

Author Response

REVIEWER 1

Quality of English Language

( ) I am not qualified to assess the quality of English in this paper.
( ) The English is very difficult to understand/incomprehensible.
( ) Extensive editing of English language required.
( ) Moderate editing of English language required.
(x) Minor editing of English language required.
( ) English language fine. No issues detected.

 AUTHORS: the manuscript has been carefully analyzed for English language.

 Comments and Suggestions for Authors

This study is based on UAV multispectral vegetation index characterisation of soybean genotypes in response to flooding and prediction of agronomic traits, this study is interesting and I think it can be published in agriengineering journal after revision, below are my specific comments.

 

REVIEWER: The description of the statistical methods is not clear enough, resulting in an incomplete presentation of the results, which needs to be improved. The purpose of the method description is to allow the reader to relate to the results presentation from the flow of the author's study.

AUTHORS: The description of the statistical methods (Section 2.4 - Statistical analysis) was rewritten and improved (Lines 206-231). Furthermore, we introduced one reference in this section (Line 230: Pradebon et al., 2024 – Reference 22) for further information about the methodology used in our study.

REVIEWER: It is suggested that the authors add subheadings in the results section to enhance the logic of the article.

AUTHORS: Subheadings were added in the results section

3.1 Analysis of variance components (Line 262)

3.2 Best linear unbiased prediction (BLUP) analysis (Line 317)

REVIEWER: The description of the conclusion part is too little, and it needs to highly summarise the patterns and conclusions obtained from the whole study.

AUTHORS: The conclusion section was improved with conclusions obtained from the whole study (Lines 507-509 and 512-518).

REVIEWER: It is suggested that the authors add a workflow diagram.

AUTHORS: A workflow diagram has been added to the manuscript by the insertion of the Figure 4 (Lines 199-203, Figure 4).

 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper presents a study on the application of phenomics for identifying vegetation indices (VIs) that can characterize the response of soybean genotypes to flooding stress. The research utilizes multispectral sensors mounted on unmanned aerial vehicles (UAVs) to collect spectral data, which is then analyzed to predict agronomic traits such as tolerance scores, relative maturity group, biomass, and grain yield. The study involved 48 soybean cultivars evaluated in both flooded and non-flooded environments, with data collected at various phenological stages. The results highlighted the effectiveness of certain VIs like MSAVI, NDVI, OSAVI, SAVI, VEG, MGRVI, EVI2, NDRE, GRVI, BNDVI, and RGB-index in selecting more flood-tolerant genotypes. However, the paper still has the following problems:

 

The paper could provide more details on the experimental setup and conditions to ensure reproducibility of the study. Could the authors provide additional details on the experimental setup, including soil types, water management practices, and any other environmental factors that might influence the results?

 

There is a need for further evaluation and ablation studies to understand the contribution of each vegetation index to the predictive models.

 

The theoretical analysis could be strengthened by including a broader range of genotypes and environmental conditions.

 

The paper might benefit from a more in-depth comparison with existing methods or baselines in the field of plant phenomics.

 

Some results, particularly those related to the prediction of agronomic traits, could be clarified with additional statistical analysis or visual aids.

 

The exposition could be enhanced by providing a clearer explanation of the RELM-BLUP methodology and its application in the study.

 

The paper might consider discussing potential limitations in the application of the findings, such as scalability or adaptability to different regions.

 

There is a need for a more thorough discussion on the practical implications of the study for soybean breeding and cultivation under flood-prone conditions.

 

How were the specific UAV settings and sensor calibrations determined, and how do they impact the spectral data collected?

 

Can the authors elaborate on the selection criteria for the vegetation indices included in the study and their relevance to flood tolerance in soybeans?

 

What are the potential implications of the findings for soybean breeding programs, and how can these indices be integrated into existing selection strategies?

 

Are there any specific challenges or considerations for implementing the proposed phenomic analysis in large-scale agricultural settings?

Author Response

REVIEWER 2

Quality of English Language

( ) I am not qualified to assess the quality of English in this paper.
( ) The English is very difficult to understand/incomprehensible.
( ) Extensive editing of English language required.
( ) Moderate editing of English language required.
( ) Minor editing of English language required.
(x) English language fine. No issues detected.

Comments and Suggestions for Authors

The paper presents a study on the application of phenomics for identifying vegetation indices (VIs) that can characterize the response of soybean genotypes to flooding stress. The research utilizes multispectral sensors mounted on unmanned aerial vehicles (UAVs) to collect spectral data, which is then analyzed to predict agronomic traits such as tolerance scores, relative maturity group, biomass, and grain yield. The study involved 48 soybean cultivars evaluated in both flooded and non-flooded environments, with data collected at various phenological stages. The results highlighted the effectiveness of certain VIs like MSAVI, NDVI, OSAVI, SAVI, VEG, MGRVI, EVI2, NDRE, GRVI, BNDVI, and RGB-index in selecting more flood-tolerant genotypes. However, the paper still has the following problems:

REVIEWER: The paper could provide more details on the experimental setup and conditions to ensure reproducibility of the study. Could the authors provide additional details on the experimental setup, including soil types, water management practices, and any other environmental factors that might influence the results?

AUTHORS: Additional details on experimental setup were added to the “Materials and Methods” section, including soil type, soil physical and chemical characterization, water management practice and other setup description data related to the experimental site (Lines 90-93; 95; 118-120 and in other parts as shown by the sentences in red).

REVIEWER: There is a need for further evaluation and ablation studies to understand the contribution of each vegetation index to the predictive models.

AUTHORS: In our study, we evaluated the environmental effect (flooded and non-flooded), genetic effect (48 soybean genotypes) and their interaction on genetic heritability of each variable, which allowed us to select predictors with less effect from the environment and less influence from experimental error. These selected variables can be used to predict agronomic traits linked to the response of soybean genotypes to flooding with high accuracy. In the discussion section, the observation made by the reviewer was addressed (Lines 488-495) as well as in the conclusions (Lines 512-514).

REVIEWER: The theoretical analysis could be strengthened by including a broader range of genotypes and environmental conditions.

AUTHORS: In the present study, 48 commercial soybean genotypes from different breeding programs were included (as e.g. DonMario, Syngenta, Brasmax, TMG Tropical Melhoramento e Genética, Bayer Cropscience, Fundação Pró-Sementes and BASF), covering different genetic backgrounds, relative maturation groups, plant architecture and grain yield potential. We believe that we have covered a wide range of soybean genotypes, with a number of cultivars comparable to that used in other studies on this topic. Regarding environmental conditions, the objective of our study was to test two environments, that is, flooded and non-flooded. Another condition that, in fact, could be evaluated would be different locations. In any case, the methodology proposed here seeks to deduce such effects.

REVIEWER: The paper might benefit from a more in-depth comparison with existing methods or baselines in the field of plant phenomics.

AUTHORS: The aim of our work was to indicate the best variables to be used in the prediction of agronomic traits linked to the response of soybean to flooding, such as grain yield, flooding tolerance score and shoot biomass. Thus, high-throughput phenotyping using multispectral sensors embedded in unmanned aerial vehicles (UAVs) can be applied, through the use of vegetation indices with greater genetic heritability and lower environmental effect (as e.g. MSAVI, NDVI, SAVI, OSAVI, VEG and MGRVI), for a faster and more accurate selection of soybean genotypes tolerant to flooding. We inserted in the manuscript a discussion regarding this topic (Lines 488-496).

REVIEWER: Some results, particularly those related to the prediction of agronomic traits, could be clarified with additional statistical analysis or visual aids.

AUTHORS: We included the Figure 12 in the manuscript, showing the relationship between observed and predicted genetic data (RELM-BLUP analysis) using linear models for tolerance score (FTS), shoot biomass and grain yield, as suggested by the reviewer, improving the visual data presentation of the data shown in Table 4 (Lines 376-378 and 404-407 – Figure 12). Regarding the Figure 12, one paragraph was included in the discussion section (Lines 474-478).

Regarding additional statistical analyses, our work aimed to use the mixed model methodology (RELM-BLUP), which is widely used in plant breeding based on quantitative genetics, in the selection of spectral vegetation indices (VIs) obtained by UAV-based sensors with potential use in the indirect selection of soybean genotypes tolerant to flooding.

This procedure is still little explored in the area of ​​phenomics. On the other hand, such methodology can increase the selection gain based on spectral variables (IVs) by up to 20% (Oliveira et al., 2023 - Spectral variables as criteria for selection of soybean genotypes at different vegetative stage. Remote Sensing Applications: Society and Environment, v.32, p.1-8, 2023. DOI: https://doi.org/10.1016/j.rsase.2023.101026.)

REVIEWER: The exposition could be enhanced by providing a clearer explanation of the RELM-BLUP methodology and its application in the study.

AUTHORS: The description of the statistical methods (Section 2.4 - Statistical analysis) was rewritten and improved (Lines 206-231). Furthermore, we introduced one reference in this section (Line 230: Pradebon et al., 2024 – Reference 22) for further information about the methodology used in our study.

 REVIEWER: The paper might consider discussing potential limitations in the application of the findings, such as scalability or adaptability to different regions.

AUTHORS: One of the main limitations refers to the compilation of data on a larger scale, which could be minimized by creating a phenotyping package to automate data collection on a greater number of genotypes and, especially, different sites. To minimize the effect of the environment on the methodology proposed in the present study, it is crucial to standardize the procedures for acquiring and processing images obtained by a multispectral UAB-based sensor, such as radiometric correction by the irradiance sensor, geometric correction using ground control points (GCPs) and time of flight. This standardization was carried out in our study and must be observed in subsequent studies.

The above discussion has been inserted in the discussion section (Lines 497-504).

REVIEWER: There is a need for a more thorough discussion on the practical implications of the study for soybean breeding and cultivation under flood-prone conditions.

AUTHORS: The implications of the present study for soybean breeding were presented in the insertions made in the discussion. High-throughput phenotyping using sensors embedded in unmanned aerial vehicles (UAVs) can be used through the application of vegetation indices with greater genetic heritability and lower environmental effects (as e.g. MSAVI, NDVI, SAVI, OSA-VI, VEG and MGRVI), in the faster and more accurate selection of soybean genotypes tolerant to flooding. This discussion has been inserted into the text (Lines 488-496).

REVIEWER: How were the specific UAV settings and sensor calibrations determined, and how do they impact the spectral data collected?

AUTHORS: The UAV configurations are basically the same as those used and standardized in similar studies. The frontal as well as lateral overlap between successive images of at least 70% is essential to guarantee the necessary number of points to create an orthomosaic that represents the phenomenon observed in the field at the time of flight. Furthermore, image overlapping of 70-75% is crucial for creating digital surface models that can be used to estimate plant height and other spectral variables. The flight height was chosen to guarantee high spatial resolution (small pixel size) in the processed orthomosaics, with a ground sample distance (GSD – Pixel size) of 2 cm. This information has been added to the “material and methods” section (Lines 168-169).

Furthermore, sensor calibration (radiometric correction) was carried out in a standardized way in our study, using the drone's own irradiance sensor (sunshine sensor), which records the amount of incident radiant energy for each photo recorded during the flight. Subsequently, when processing the images in the software, these image metadata are considered to perform the radiometric correction. This procedure is mandatory for the analysis and comparison of images acquired at different crop growth stages and under changing lighting conditions. These observations are shown in the description of the methodology used (2.3 Image acquisition using UAV) (Lines 171-172).

REVIEWER: Can the authors elaborate on the selection criteria for the vegetation indices included in the study and their relevance to flood tolerance in soybeans?

AUTHORS: The vegetation indices (VIs) used in the present study were chosen because they are widely used in high-throughput plant phenotyping studies and, in general, in remote sensing studies, as cited by Gao et al. (2023) and Lozada et al. (2020).

Furthermore, the VIs included in our study includes the most important bands of the electromagnetic spectrum related to vegetation studies, such as from the visible bands (blue, green and red) as well as from the infrared bands, as e.g. red-edge (around 700 nm) and the near-infrared (NIR). In this sense, the objective of our study was, in fact, select and recommend the most relevant and related VIs to flooding tolerance in soybean genotypes.

This observation was included in the material and methods section (2.3 Image acquisition using UAV) (Lines 186-191).

References:

Gao, S. et al. Evaluating the saturation effect of vegetation indices in forests using 3D radiative transfer simulations and satellite observations. Remote Sensing of Environment, v.295, p. 1-18 (113665), 2023. DOI: https://doi.org/10.1016/j.rse.2023.113665

Lozada, D.N. et al. Genomic prediction and indirect selection for grain yield in us pacific northwestwinter wheat using spectral reflectance indices from high-throughput phenotyping. International Journal of Molecular Sciences, v.21, p.1-18, 2020. DOI: https://doi.org/10.3390/ijms21010165 

REVIEWER: What are the potential implications of the findings for soybean breeding programs, and how can these indices be integrated into existing selection strategies?

AUTHORS: The results of the present study indicate the vegetation indices (VIs) with the greatest relationship between spectral data and genetic characteristics, which allows the use of VIs as an indirect selection variable in soybean breeding programs for flooding tolerance. These results can be used in high-throughput phenotyping using UAV and remote sensing, aiming to facilitate and, above all, accelerate the selection of genotypes with greater tolerance, since the methods traditionally used in this selection, such as tolerance score and sampling destructive biomass, are time-consuming and laborious. These tools shown here can be used in a complementary way in a plant breeding program, both for predicting variables as well as for genotype selection (Lines 488-496).

Understanding the association between variables obtained by multispectral imaging, such as vegetation indices (VIs), and agronomic traits of interest for soybean is crucial for breeding programs. This approach can identify easier-to-measure variables to be used in the indirect selection of genotypes for more expensive-to-measure traits, resulting in a faster, labor-saving, and large-scale selection process. (Paragraph included in the Introduction section, Lines 73-78).

REVIEWER: Are there any specific challenges or considerations for implementing the proposed phenomic analysis in large-scale agricultural settings?

AUTHORS: The implementation challenges are linked to the need for specific knowledge of the proposed methodologies, such as statistics based on mixed models and BLUP, and the need for knowledge in the areas of UAV, sensors and remote sensing. Additionally, knowledge of the programming language in the R software becomes essential. However, there are already packages, such as “Metan” and “Plimanshine”, which already have ready-to-use phenomics functions in an automated manner. Considerations about this were inserted in the discussion section (Lines 497-504).

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The author responded to the reviewer's comments by clicking on the email, and I have no further comments.