Improving Agricultural Efficiency of Dry Farmlands by Integrating Unmanned Aerial Vehicle Monitoring Data and Deep Learning
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis study proposes a method for predicting soil moisture and yield of dryland wheat based on unmanned aerial vehicle (UAV) multispectral imagery, combining the Modified Pedotransfer Function-based Drought Index (MPDI) and Gradient Boosting Regression (GBR) model. The research topic is of practical significance, the method is innovative to some extent, and the experimental design is generally rational. However, there may be some problems with the manuscript in its present form, and some suggestions are as follows.
Point 1: The introduction falls short in elaborating on the research background and current status. Although the authors have touched upon the irrigation needs of agriculture in arid regions and the significance of soil moisture monitoring, the overall content remains superficial, lacking depth and comprehensiveness. It is recommended that the authors supplement the introduction with specific data and case studies to more clearly illustrate the impact of climate change on agriculture in arid regions. Moreover, the authors' discussion of soil moisture monitoring technologies is rather one - sided, focusing only on a few selected techniques without systematically summarizing the advantages and disadvantages of other relevant technologies. It is suggested that the authors further expand and elaborate on these aspects to enhance the completeness and persuasiveness of the introduction.
Point 2: There are deficiencies in model validation. The authors only used data from a single plot for training and validation in the paper, which may affect the generalizability of the model. It is recommended that the authors add datasets from different soil types and climatic zones for validation to more comprehensively evaluate the model.
Point 3: There are certain inconsistencies in the method of sample division. The authors mention in the paper that the training and testing sets are randomly divided in a 7:3 ratio. Random division of samples may lead to instability in the model results. It is recommended that the authors clearly state in the paper whether the random seed is fixed to ensure the reproducibility of the results. In addition, it is suggested to consider multiple repeated random divisions and take the average result to evaluate the stability and reliability of the model.
Point 4: The rationality of the soil sampling strategy is open to discussion. The paper mentions that 12 sampling points were set up in the study plot for soil moisture sampling, but no detailed spatial data of the sampling points is provided. It is recommended that the authors supplement the relevant data to intuitively show the layout of the sampling points.
Point 5: Lack of comparative study of the model. The paper uses the GBR model to model the research data, but no comparative experiments or quantitative comparisons with similar models are provided. It is recommended that the authors add comparative experiments and literature references in the conclusion section to demonstrate the superiority of the GBR model's performance.
Author Response
Thank you for your valuable comments and guidance. We have revised the manuscript according to your suggestions. Please refer to the attached document for detailed information.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsGENERAL COMMENTS
This study enhances dryland agriculture resilience by monitoring soil moisture using UAV-mounted hyperspectral sensors. As such, this paper falls within the scope of the journal “lands”.
The Modified Perpendicular Drought Index (MPDI) and soil samples trained a Gradient Boosting Regression (GBR) model to estimate moisture, optimising irrigation. A second GBR model, combining MPDI and wheat spike data, predicted yields with >90% accuracy in subsequent years. Integrating remote sensing and machine learning enables precise, data-driven water management, improving crop productivity under climate stress. The approach supports sustainable agriculture through timely, spatially targeted resource allocation.
The approach presented in this study is of general interest and worthy of being spread within the scientific community.
The article is well conceptualised, its figures are instructive, and most of the ideas are clearly explained. However, the Introduction is weak and introduces elements more appropriate for Section 2; some specific descriptions regarding soil properties should be improved, and the References contain many mistakes. While the English language expression is adequate, the article contains significant shortcomings.
Consequently, I cannot recommend this paper for publication at this time.
MAIN CONCERNS AND QUESTIONS
The authors of this article have used AI Model (the GBR Model). The updated MDPI guidelines recognise that “these tools do not meet authorship criteria and thus cannot be listed as authors on manuscripts. While AI can contribute intellectually to the writing process, it is now widely accepted that it cannot take responsibility for the content it produces”. In this regard, the reviewer accepts the Editor’s final decision on this matter.
1-The study does not clarify the rationale for selecting the 0-10 cm soil layer. While this layer is easily accessible for sampling and facilitates straightforward image acquisition -both clear advantages- a significant drawback is that the moisture in this shallow layer dissipates quickly due to exposure to wind and solar radiation, as well as the natural movement of water to deeper layers, which reduces its retention time. These factors may compromise the reliability of moisture measurements at this specific soil depth.
2-On the other hand, the study does not consider root depth. The crops mentioned (sorghum, wheat, corn, and soybeans) are characterized by relatively shallow root system, and their influence on soil water dynamics should be taken into account.
3-The Introduction section requires improvement. I offer the following suggestions:
i)Ls47-48: The final statement is overly general. Therefore, I recommend revising it to something like “…comprehensive data; in contrast, satellite monitoring lacks sufficient detail to represent water status at the plot level. Thus …”. In other words, the methodology investigated by the authors plays a significant role for end users.
ii)Please define the acronym (UAV) “unmanned aerial vehicle” (in Ls48-49).
iii)L52-66: These lines describe the research site and should be moved to Section 2.
iv)At this point, the Introduction is very brief. It would be benefit to add the current state of the art, including topics such as those mentioned in lines 110 and 111, the role of the AI model (Gradient Boosting Regression), a clear statement of the article's objectives, and the necessary approaches to address the research questions. These enhancements should be supported by up-to date references.
4-The authors state that the region does not have uniform soil types (L110). However, the soil textural classes could be more homogeneous and could provide a useful range of textural classes for this analysis. This factor is crucial for determining the soil’s water retention capacity. Only general aspects are mentioned in Ls56-57, and the textural class “clayey sand” is not incorporated into the text, at least not in the Discussion section.
5-There is little information about the method used in this study to determine soil moisture, even though it is an essential variable (Sections 2.4, 3.2 and Table 2). This procedure is standardised: it must be measured in the soil’s active fraction (<2 mm) by weighing the sample before and after oven-drying. Failure to do so would invalidate the results.
6-How were the soil samples collected? (Ls133-134)
7-Ls92-93. The authors should specify the range of wavelengths used.
8-The final reference list contains 25 entries. However, only ten references are cited in the text, with the last citation appearing at L146. It is customary for Section 4 (Discussion) to include comparisons with other studies or to use data from other authors to support the assessments made. In this manuscript, however, the authors omit this important aspect. Please review and improve this section accordingly.
9-As I have indicated previously, I believe the following point could be addressed in Section 4.4: “Soils are three-dimensional entities and only the surficial layer is studied in this study. Many crops have deeper root systems that interact with subsurface soil layers; therefore, the model used in this research may require different modifications and adjustments than those obtained in this research.”
SPECIFIC COMMENTS
For the sake of rigor and readability, the following general issues should be addressed throughout the text:
The variable “fv” should be consistently formatted as “fv” in formulas and text (e.g. Ls209,301,…)
The regression coefficient “R” should be written as “R2” (as in Tables, e.g. Ls215,269,…) and should not be italicised.
Title:
I suggest the following Title, which is more synthetic: “Enhance dryland farming efficiency with UAV data and deep learning”
Abstract:
Ok.
Keywords:
Please sort alphabetically.
1.Introduction:
See “Main Concerns” section.
2.Materials and Methods:
L93. Why are the dates written in red?
L163. Typing error: This table should be “Table 3”. Please verify.
3.Results:
L280. Regarding the reported prediction accuracy of 95.33%: The reader notes that Table 8 contains two values that exceed this figure, specifically for Jan.25 and Feb.11. Are additional reasons supporting the selection of this particular value?
4.Discussion:
Ls285-287. Please clarify this sentence.
L292. Since this point has already been mentioned above (Section 2.3), I believe you might also consider referencing this section here.
L329. The text reads “displayed”; please use “showed” instead.
Ls339-340. This sentence appears repetitive with the content in Ls278-282. However, it would be more appropriate to include this observation here, as it pertains to the Discussion. Please revise for clarity and conciseness.
L349. I believe that this table should be Table 8. Please verify.
Ls352-354. These lines seem repetitive with Ls340-342. Please revise to avoid redundancy.
References:
As commented above in the “Main Concerns” section, References 11 to 25 are not cited in the text. This section should be significantly improved.
Tables:
Tables 1 and 6. I recommend including a bar scale and the North arrow in each image to facilitate calculations.
Table 2. The label “SP1.” in the first row should be “SP1”.
Table 3. No issues.
Tables 4 and 5. The images labels should be expanded to improve readability. The Y-axis labels should be arranged in ascending order.
Tables 7 and 8. The “R2” label at the header should be written in regular font, not italics.
Figures:
Figure 1. Please include scale bars and the North arrow.
Figure 2a. Why do the authors use greyscale images in the panels, while the tables use coloured multispectral images?
Figure 3. No issues.
Figure 4. No issues.
Comments on the Quality of English LanguageThe English in this article is acceptable, making for a fluent read. However, it could benefit from more thorough revision. I have identified only a few shortcomings, although I did not focus on these aspects.
Author Response
Thank you for your valuable comments and guidance. We have revised the manuscript according to your suggestions. Please refer to the attached document for detailed information.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThe paper is interesting; however, it is recommended to clarify some doubts about the document.
Comments for author File: Comments.pdf
Author Response
Thank you for your valuable comments and guidance. We have revised the manuscript according to your suggestions. Please refer to the attached document for detailed information.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsFINAL COMMENT
I have read with interest the new manuscript, which incorporates all the guidelines outlined in the previous revision, and all these corrections have been made appropriately.
I have only two additional formal suggestions for improvement:
Ls89,90,91: The text currently reads “21 degrees”. To avoid misunderstandings, I recommend specifying “21ºC”, as Fahrenheit degrees are commonly used in many countries.
L187. I suggest adding the following sentence: “Soil samples were obtained using a handheld sampler.”
Comments on the Quality of English LanguageWhile the English in this article is acceptable, it could be enhanced by a more thorough revision.
Author Response
Thank you very much for your kind and constructive feedback.
We sincerely appreciate your careful review and valuable suggestions. In response, we have revised the manuscript as follows:
-
At Lines 89–91, we have clarified the temperature unit by changing “21 degrees” to “21ºC” to avoid any potential misunderstanding.
-
At Line 187, we have added the suggested sentence: “Soil samples were obtained using a handheld sampler.”
We hope these adjustments meet your expectations. Thank you again for your insightful comments, which have helped us further improve the manuscript.
Reviewer 3 Report
Comments and Suggestions for AuthorsThe changes made by the authors have been clarified, substantially improving the paper.
Author Response
Thank you very much for taking the time to review our manuscript. We are pleased to know that the revised version meets your expectations. Your support and recognition are greatly appreciated and encouraging to us.
