Machine-Learning-Based Multi-Site Corn Yield Prediction Integrating Agronomic and Meteorological Data

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript presents the use of ML to predict maize yield in three areas of Yunnan Province by integrating agronomic and meteorological data. The study compares three models (RF, SVM, and XGBoost) and uses PDP to enhance result interpretability. The results demonstrate the excellent performance of XGBoost and clearly identify key factors affecting yield. This manuscript would be significantly enriched if the authors could provide additional details on important aspects such as the rationale for site selection, the role of meteorological variables, and methodological transparency. The following suggestions will help improve the manuscript to meet publication standards.

Introduction
1. The authors should present a map of the study area and explain the rationale for selecting the three sites (Dali, Lijiang, and Zhaotong). This should include differences in climate, topography, and agroecological zones that make these areas suitable for comparative study, which would enhance the research reliability.

Materials and Methods
2. Consider adding a flowchart or framework diagram to illustrate the workflow from data collection, processing, and modeling to evaluation. This will help readers clearly understand and replicate the research process.

3. The authors should explain the rationale for selecting each meteorological variable (TEMP, DEWP, VISIB, WDSP, MAX, MIN, RH, and PRCP) and their theoretical relationship to maize growth and yield, as well as the temporal resolution, as these are key components of the research.

4. The authors mentioned using the caret package for hyperparameter optimization. It would be beneficial to clearly specify the final hyperparameters selected for each model, such as the number of trees for Random Forest, the learning rate and max depth for XGBoost, or the kernel type for SVM with related parameters. Providing these details will help readers understand the model configuration and accurately reproduce the results.

Results
5. The authors show that SVM performs poorly compared to other models. It would be beneficial to further explain why this model is not suitable for this dataset. What factors prevent SVM from capturing complex relationships between variables as effectively as XGBoost? This in-depth analysis will help readers better understand the limitations and advantages of each model in crop yield prediction contexts.

6. Table 1 shows that several variables are highly correlated, such as GP with HKW and GWPS, or PH with GWPS and HKW. It would be beneficial if the authors could further examine and discuss how these correlations among independent variables might affect model stability and interpretability, as well as how to address this issue to ensure the model remains reliable and can clearly explain each variable's influence. 

Discussion
7. To align with the manuscript's focus, the authors should discuss the importance of meteorological factors that, although not highly ranked in feature importance, have indirect influences on yield formation.

8. The authors should discuss practical implications for farmers and policymakers, including approaches for implementing the model in decision support systems, technology transfer methods, and trade-offs between accuracy and complexity in real-world applications.

Author Response

First of all, we greatly appreciate you for your careful reviewing and valuable comments. Accordingly, we revised the manuscript (MS) based on your comments and suggestions. The details of the revision are as follows:

Introduction:

Point 1: The authors should present a map of the study area and explain the rationale for selecting the three sites (Dali, Lijiang, and Zhaotong). This should include differences in climate, topography, and agroecological zones that make these areas suitable for comparative study, which would enhance the research reliability.

Response 1: We sincerely appreciate your constructive suggestions. In response, we have: (1) Supplemented the experimental analysis section with geographic coordinates of the study regions; (2) Added clear justification for site selection in the Materials and Methods section (Revised manuscript, Section 2.1, Lines 98-101)

Materials and Methods:

Point 2: Consider adding a flowchart or framework diagram to illustrate the workflow from data collection, processing, and modeling to evaluation. This will help readers clearly understand and replicate the research process.

Response 2: We sincerely appreciate this valuable suggestion. A comprehensive workflow diagram (Figure 1) has been added to Section 2, accompanied by a detailed caption and in-text reference (Lines 117–118).

Point 3: The authors should explain the rationale for selecting each meteorological variable (TEMP, DEWP, VISIB, WDSP, MAX, MIN, RH, and PRCP) and their theoretical relationship to maize growth and yield, as well as the temporal resolution, as these are key components of the research.

Response 3: We greatly value the reviewer's insightful suggestions about the meteorological variables. The selected data were obtained from GSODR as daily records and processed to monthly averages (March–November) to align with maize growing seasons. These variables were prioritized because they directly or indirectly influence yield .

Point 4: The authors mentioned using the caret package for hyperparameter optimization. It would be beneficial to clearly specify the final hyperparameters selected for each model, such as the number of trees for Random Forest, the learning rate and max depth for XGBoost, or the kernel type for SVM with related parameters. Providing these details will help readers understand the model configuration and accurately reproduce the results.

Response 4: We sincerely appreciate your constructive suggestion. The specific code implementation has been added to Section 2.3 Data Analysis (Lines 133–140 in the revised manuscript).

Results :

Point 5: The authors show that SVM performs poorly compared to other models. It would be beneficial to further explain why this model is not suitable for this dataset. What factors prevent SVM from capturing complex relationships between variables as effectively as XGBoost? This in-depth analysis will help readers better understand the limitations and advantages of each model in crop yield prediction contexts.

Response 5: We sincerely appreciate your valuable suggestion. we have an explanation regarding the relatively poorer performance of SVM in the revised manuscript (Lines 235-236).

Point 6: Table 1 shows that several variables are highly correlated, such as GP with HKW and GWPS, or PH with GWPS and HKW. It would be beneficial if the authors could further examine and discuss how these correlations among independent variables might affect model stability and interpretability, as well as how to address this issue to ensure the model remains reliable and can clearly explain each variable's influence.

Response 6: Thank you for your valuable feedback. The traits influence each other, and during model analysis, we also considered the interaction between traits. In future research, we can delve into how the correlation between traits affects the model, which would be an excellent idea.

Discussions:

Point 7: To align with the manuscript's focus, the authors should discuss the importance of meteorological factors that, although not highly ranked in feature importance, have indirect influences on yield formation.

Response 7: Thank you for your valuable suggestions. We have already incorporated the indirect impact of climate factors on yield into the revised manuscript (Lines 260-262 in the revised version).

Point 8: The authors should discuss practical implications for farmers and policymakers, including approaches for implementing the model in decision support systems, technology transfer methods, and trade-offs between accuracy and complexity in real-world applications.

Response 8: Thank you for this important suggestion. We have enhanced the practical implications discussion in the revised manuscript (Lines 256-258).

Thank you again for taking time out of your busy schedule to review our paper, and thank you for your valuable suggestions. After revising the MS according to your suggestions, we all learned a lot from your nice comments.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Understanding of the paper:

The paper presents a machine learning-based approach for multi-site corn yield prediction by integrating agronomic and meteorological data from three Chinese regions (Dali, Lijiang, and Zhaotong) collected between 2019 and 2023. It evaluates three models—Random Forest, Support Vector Machine, and XGBoost—with XGBoost outperforming the others in terms of accuracy (R² = 0.99) and error minimization. The study identifies key yield determinants, including grain weight per spike, shelling percentage, growth period, and plant height, and uses partial dependence plots to explore their complex interactions. The research underscores the value of combining diverse data sources for accurate yield forecasting and supports the advancement of precision agriculture under climate change.

Comments for author File: Comments.pdf

Comments on the Quality of English Language

There are a lot of typos. 

Author Response

First of all, we greatly appreciate you for your careful reviewing and valuable comments. Accordingly, we revised the manuscript (MS) based on your comments and suggestions. The details of the revision are as follows:

Abstract:

Point 1: Lack of Context or Justification for Study Area. Example: “This study selected three regions, specifically Dali, Lijiang, and Zhaotong…” No explanation is provided for why these regions were chosen.

Response 1: We appreciate your suggestions have made the following adjustments: (1) Supplemented the geographic coordinates of the regions in the experimental analysis section; (2) Clarified the reasons for site selection in the Materials and Methods section (Revised manuscript, Section 2.1, Lines 98-101).

Point 2: Fragmented Sentence Structure. Example: "And three machine learning models, namely Random Forest (RF), Support Vector Machine (SVM), and GBoost, were employed." Starting a sentence with "And" is informal and not suitable for academic writing.

Response 2: We appreciate your suggestions and have made revisions accordingly (Revised manuscript, lines 18-19). We have also reviewed the entire manuscript to check for similar issues.

Point 3: Redundancy and Repetition. Example: “...developed an efficient data-driven methodology for predicting corn yield prediction …” Repetition of “predicting” and “prediction” is redundant.

Response 3: Thank you for your suggestion. We have refined the sentence to eliminate redundancy (Lines 28–30 in the revised manuscript).

Point 4: Inconsistent or Ambiguous Terminology. Example: “yield and two key factors: grain weight per spike (GWPS) and hundred-kernel weight (HKW).” It’s unclear if these two traits alone are being emphasized, while more traits are discussed later.

Response 4: We appreciate this careful observation. The text has been revised to specify that GWPS and HKW were primary focus traits among the multiple evaluated characteristics (Lines 19–21 in the revised manuscript).

Introduction:

Point 5: The actual research gap (lack of comparative analysis of multi-model integration strategies) is buried deep in the last paragraph (line 100+). This should appear earlier to anchor the narrative.

Response 5: We sincerely appreciate your constructive feedback. According to your comments, emphasized the research gap early on by clearly stating the lack of comparative multi-model analyses in the introduction (Lines 51-53).

Point 6: Several phrases and ideas are repeated or over-explained. For example, the benefits of RF are elaborated in multiple places, and interpretability is introduced both in the middle and at the end. Sentences like "These further underscore the promising capabilities..." could be more concise.

Response 6: We sincerely appreciate your valuable suggestions. Accordingly, we consolidated Random Forest applications, removed redundant interpretability explanations, aand deleted over-explaining transitional sentences. These revisions are in the Introduction's second and third paragraphs.

Point 7: Minor issues such as:

✓ "Research has showed" → should be "Research has shown"

✓ “Varied spatial variability” → redundancy ("varied" and "variability")

✓ "Enhance interpretability through the use of Partial dependence plot (PDP)" →

should be “Partial Dependence Plots (PDPs)” and "enhances"

✓ "These insights hold practical value for: optimize..." → should be "optimizing"

Response 7: We thank the reviewer for noting the grammatical and terminological issues. The suggested changes have been made  (Lines 41, 44, 86, 88).

Point 8: The final paragraph introduces the contribution but lacks a strong, clear thesis statement or objective sentence summarizing what exactly the study does and how it fills the gap.

Response 8: Thank you for your suggestion. Accordingly, we have improved structural clarity by using numbered lists to present our methodological contributions in the revised manuscript (Lines 84-88).

Point 9: Follow the best practices when making scientific writing: Don’t repeatedly expand

the full forms already introduced earlier. Also, expand some acronyms not introduced

(CNN, LSTM – line 58).

Response 9: We are grateful for this observation and have implemented the suggested modification (Line 56 in the revised version).

Materials and Methods:

Point 10: Revised grammatical errors such as:

✓ Line 117: “and was c composed” → typo, should be “and was composed”

✓ Line 117: “which spaced 0.8 meters apart” → revise for clarity

✓ Lines 137–138: The mention of the GSODR package is useful, but you could mention which meteorological variables it provided.

Response 10: We thank the reviewer for noting the grammatical errors, and have carefully revised and supplemented the corresponding sections (see Lines 105 and 124-128 in the revised manuscript).

Point 11: Data Analysis (Lines 128–142): Mention if feature selection, normalization, or data imputation was performed prior to modeling, especially given the machine learning context.

Response 11: We appreciate your comment on the data processing. To clarif, all data used in the analysis are mean values, without additional processing applied beforehand.

Point 12: Cross-validation Strategy (Line 140): Clarify whether stratified k-fold was used (especially important if class imbalance exists), and whether this strategy applied to all models or just RF and XGBoost.

Response 12: We sincerely appreciate your constructive feedback. In our methodology, we implemented 10-fold cross-validation for both RF and XGBoost algorithms, while employing the Radial Basis Function (RBF) kernel for SVM analysis (see Lines 133-140 in the revised manuscript).

Point 13: Software Versioning: You've mentioned R version, which is good—consider including

package versions if reproducibility is critical.

Include a workflow diagram for experimental and analytical flow.

Provide maps of the study regions (Dali, Lijiang, Zhaotong) to support spatial interpretation.

Response 13: Thank you for your suggestions. We confirm R packages work in version 4.3.3, and we added: (1)geographic coordinates of study sites,(2) justification for site selection, and (3) expanded methodological descriptions. These changes in the Materials and Methods section of the revised manuscript (Lines 98-101, 133-140).

Results And Discussions

Point 14: Grammatical and Stylistic Issues

• Several phrasing errors and awkward expressions reduce clarity and professionalism.
• Examples:

o “...which attributing its superiority…” → should be “attributing its superiority” or “which is attributed to”

o “...inefficiency in handing large-scale datasets...” → should be “handling”

o “...but also and aligns seamlessly...” → redundant, remove “and” →  Perform a careful grammatical edit to improve readability and flow. -

Response 14: We appreciate your suggestions and have implemented the revisions to improve grammar (Lines 234, 236 and 244-2436 in the revised manuscript).

Point 15:Unclear or Underdeveloped Insight

• Some points lack elaboration or are too brief to be meaningful.
• Examples:

o “Meteorological factors did not rank prominently... they indirectly influenced yield formation...” → This needs more detail. Which factors? How do they indirectly influence yield?

→  Provide a short, concrete example of how meteorological variables affect the growth period or yield.

Response 15: We sincerely appreciate your valuable suggestion. The corresponding section has been supplemented accordingly (see Lines 260–262 in the revised manuscript).

Conclusion:

Point 16: Inclusion: add a critical reflection on limitations, clarify how the findings can be practically applied, reduce repetition, and outline directions for future work. The conclusion does not mention any limitations of the study (e.g., dataset size, overfitting risks, regional specificity, data quality, etc.).

Response 16: We sincerely appreciate your constructive suggestion. We have now incorporated the study's limitations and future research directions in the Conclusion section (Lines 294-298 of the revised manuscript).

Point 17: Limited Generalizability and Scope - The study was conducted in only three regions in China, yet the conclusion doesn’t qualify whether the model would generalize well to other geographies or maize varieties.

Response 17: We sincerely appreciate your valuable suggestion and have supplemented in the Discussion section accordingly in the revised manuscript (Lines 294-298 of the revised manuscript).

Referrences:

Point 18: References should have the DOI where applicable.

Response 18: We sincerely appreciate your valuable suggestion. We have now supplemented all references with their corresponding DOI identifiers in the revised manuscript.

Thank you again for taking time out of your busy schedule to review our paper, and thank you for your valuable suggestions. After revising the MS according to your suggestions, we all learned a lot from your nice comments.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript presents findings that are valuable for precision agriculture and breeding applications. However, several sections need editing to improve clarity and conciseness. Specifically, long sentences in the abstract and introduction were shortened for clarity, and some repetitive descriptions in the results were condensed. These comments were provided in the file.

Comments for author File: Comments.pdf

Comments on the Quality of English Language

The manuscript is generally understandable, but it contains several long and complex sentences. It needs editing to make it easier for readers to understand.

Author Response

First of all, we greatly appreciate you for your careful reviewing and valuable comments. Accordingly, we revised the manuscript (MS) based on your comments and suggestions. The details of the revision are as follows:

Abstract:

Point 1: These can be summarized briefly without showing specific values.

Response 1: Thank you for your suggestion. I have now provided a brief summary of this section in the revised draft, see lines 22 to 23.

Point 2: Highlight the broader implications. How can this help farmers, policymakers, or breeding programs?

Response 2: Thank you for your question. The study yields three practical implications: First, farmers may enhance productivity through optimized management; second, policymakers could refine subsidy frameworks to prioritize drought-resilient cultivars; third, breeding initiatives stand to reduce experimental phases via targeted screening protocols.

Introduction:

Point 3: The introduction is informative but so much details. It could be more concise and simplified.

Response 3: Thank you for your suggestion. We have refined this section to enhance clarity and conciseness in the revised version.

Point 4:. Italic

Response 4: Thank you for your reminder. The relevant section has now been changed to italics (see Lines 36 in the revised manuscript).

Point 5: This paragraph lists many studies in one long flow. May be it can summarize into one strong statement.

Response 5: Thank you for your nice suggestion. We have summarized this part (the second paragraph of the Introduction in the revised manuscript).

Point 6: Why are multimodal models better than unimodal? How do these studies link to the current research gap?

Response 6: Thank you for your suggestion. Multimodal models enhance prediction robustness by integrating complementary data sources, outperforming unimodal approaches. However, existing studies focus on single architectures instead of systematically comparing integration strategies, which is the key gap our study aims to address through rigorous evaluation of RF/SVM/XGBoost combinations.

Materials and Methods:

Point 7: Why select these three locations? What are the significance of these areas? Are there a lot of corn crops and are there environmental differences?

Maybe give a short additional explanation.

Response 7: Thank you for your nice suggestion. Accordingly, we added: (1)geographic coordinates of study sites, and (2) justification for site selection (Revised manuscript, Section 2.1, Lines 98-101).

Point 8: The text doesn't specify what is a treatment.

Response 8: Thank you for your nice suggestion. The text has been revised to explicitly specify the experimental treatments (Lines 104 in the revised manuscript).

Results:

Point 9: This is reiterated in several sentences. Based on these results, consider that XGBoost is a strong model in terms of various traits. Combine the results to determine which traits are advantageous and how they differ from other models.

Response 9: Thank you for your nice suggestion. Indeed, as noted in our Discussion (Lines 232–234), which may explain the divergence in results compared to Random Forest.

Point 10: This is visually clear from the figure. Can be shortened.

Response 10: We sincerely appreciate your constructive suggestion. The text has been revised (see Lines 176–177 in the revised manuscript).

Discussions:

Point 11: Needs an interpretation of how this information can improve the management of breeding.

Response 11: Thank you for this valuable suggestion. The revised text now explicitly integrates feature importance outcomes with actionable breeding strategies, outlining clear selection priorities and optimized utilization of phenotypic markers (Lines 256-258). This linkage facilitates targeted genetic improvement and resource allocation for future breeding programs.

Conclusions:

Point 12: Including recommendations for practitioners and researchers (e.g., Future studies should integrate soil data, satellite remote sensing, and multi-year datasets to improve generalizability).

Response 12: Thank you for your nice suggestion. We have now incorporated the study's limitations and future research directions in the Conclusion section (Lines 294-298 of the revised manuscript).

Thank you again for taking time out of your busy schedule to review our paper, and thank you for your valuable suggestions. After revising the MS according to your suggestions, we all learned a lot from your nice comments.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

This is a solid and relevant application of machine learning for yield prediction across multiple sites. The methodological setup is appropriate, and the conclusions are valid. However, the novelty is moderate since many similar studies exist. Figures are informative but need better visual clarity. I recommend minor revision to address issues with clarity, justification of methodological choices, and slight English polishing. No ethical or scientific misconduct detected.

The manuscript presents a relevant application of machine learning methods (RF, SVM, XGBoost) for maize yield prediction using agronomic and meteorological data across multiple sites in Yunnan Province. The results are generally well-supported, and the integration of partial dependence plots (PDPs) enhances interpretability to some extent.

However, several aspects need clarification or improvement:

1. Model Development Transparency: Please provide additional detail regarding model training, particularly the hyperparameter tuning process for each algorithm. Were default settings used, or was there a systematic grid/random search?
2. Figures and Visualization: Figures 1 and 3 are dense and somewhat hard to interpret. Consider splitting the panels across multiple figures or simplifying the layout for better readability. Also, some axis labels and legends are too small.
3. Explainability Tools: While PDPs offer insight into marginal effects, stronger interpretability tools such as SHAP could add more clarity and robustness. If PDPs were chosen intentionally over other methods, this choice should be justified.
4. Language and Clarity: The English is understandable but could benefit from minor editing throughout to improve clarity, especially in the Introduction and Discussion sections.

Overall, the study is scientifically sound and contributes to the literature on ML-based agricultural prediction. A minor revision is recommended to address the issues above and enhance clarity, reproducibility, and presentation.

Comments on the Quality of English Language

Language and Clarity: The English is understandable but could benefit from minor editing throughout to improve clarity, especially in the Introduction and Discussion sections.

Author Response

First of all, we greatly appreciate you for your careful reviewing and valuable comments. Accordingly, we revised the manuscript (MS) based on your comments and suggestions. The details of the revision are as follows:

Point 1: Model Development Transparency: Please provide additional detail regarding model training, particularly the hyperparameter tuning process for each algorithm. Were default settings used, or was there a systematic grid/random search?

Response 1: We sincerely appreciate your valuable suggestion. In response, we have supplemented the model training details in the revised manuscript (see Line 133-140).

Point 2: Figures and Visualization: Figures 1 and 3 are dense and somewhat hard to interpret. Consider splitting the panels across multiple figures or simplifying the layout for better readability. Also, some axis labels and legends are too small.

Response 2: We sincerely appreciate your constructive feedback on the figures. In response to your suggestions, we have thoroughly optimized Figures 1 and 3 to improve clarity and readability.

Point 3: Explainability Tools: While PDPs offer insight into marginal effects, stronger interpretability tools such as SHAP could add more clarity and robustness. If PDPs were chosen intentionally over other methods, this choice should be justified.

Response 3: We sincerely appreciate your insightful suggestion regarding explainability tools. The requested discussion has been added in the Introduction section (Lines 64–66 of the revised manuscript).

Point 4:. Language and Clarity: The English is understandable but could benefit from minor editing throughout to improve clarity, especially in the Introduction and Discussion sections.

Response 4: We sincerely appreciate your constructive suggestions. In response, we have conducted a comprehensive revision of the manuscript, with particular attention to refining both the Introduction and Discussion sections to enhance clarity and coherence.

Thank you again for taking time out of your busy schedule to review our paper, and thank you for your valuable suggestions. After revising the MS according to your suggestions, we all learned a lot from your nice comments.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Quick Fixes needed:

Abstract.

1. Briefly state how many samples, traits, or years of data were collected. Example: “Data comprised 2,500 plot-level observations and 12 meteorological variables.” "Agronomic traits and meteorological data" is vague. [Line 17].
2. Mentions “lowest RMSE” and “good R²” without numeric values. Please include the quantitative metrics in numbers. Like (RMSE = X.X, R² = X.Xxxxx). [Line 22]
3. You state “providing scientific support for precision agricultural management,” but lack concrete examples of real-world application. Add examples: “These predictions can guide fertilizer optimization and irrigation scheduling at the farm level.” [Line 29]

General.

• Ensure that the full form of any acronym is introduced once, for example, RF, SVM, ...
• The formula should be referenced (Numbered).
• Revise the manuscript for grammatical errors. 

Author Response

First of all, we greatly appreciate you for your careful reviewing and valuable comments. Accordingly, we revised the manuscript (MS) based on your comments and suggestions. The details of the revision are as follows:

Abstract:

Point 1: Briefly state how many samples, traits, or years of data were collected. Example: “Data comprised 2,500 plot-level observations and 12 meteorological variables.” "Agronomic traits and meteorological data" is vague. [Line 17].

Response 1: Thank you for your suggestion. We have added the number of samples and measured traits, and the year of data (Line 17-18). Detailed information were presented in the Materials and Methods section.

Point 2: Mentions “lowest RMSE” and “good R²” without numeric values. Please include the quantitative metrics in numbers. Like (RMSE = X.X, R² = X.Xxxxx). [Line 22]

Response 2: We appreciate this precise suggestion. The requested quantitative metrics have been added in Line 22-23 of the revised manuscript.

Point 3: You state “providing scientific support for precision agricultural management,” but lack concrete examples of real-world application. Add examples: “These predictions can guide fertilizer optimization and irrigation scheduling at the farm level.” [Line 29]

Response 3: We sincerely appreciate your valuable suggestion. We have accordingly updated the revised manuscript by incorporating the relevant content on lines 29-32.

General

Point 4: Ensure that the full form of any acronym is introduced once, for example, RF, SVM, ...

Response 4: Thank you for your suggestion. We have carefully reviewed the entire document to ensure that the full name was used only once at the beginning, and abbreviations were used thereafter. (Lines 128, 137, 159, 201, 210, 223, 228, 247, 296).

Point 5: The formula should be referenced (Numbered).

Response 5: Thank you for your suggestion. Thank you for your suggestion. We have numbered the formulas in the manuscript.

Point 6: Revise the manuscript for grammatical errors.

Response 6: We sincerely appreciate your careful reading of our manuscript. Following your suggestion, we have performed an extensive grammatical assessment and made significant enhancements to the overall text.

Thank you again for taking time out of your busy schedule to review our paper, and thank you for your valuable suggestions. After revising the MS according to your suggestions, we all learned a lot from your nice comments.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Thank you for editing this article. The revised version is clearer and more concise. This article has the potential to be developed as a tool for agricultural decision-making related to climate change in the future.

However, it would be clear if the cultivar names and the number of cultivars used in the test were provided (in line 104).

Author Response

First of all, we greatly appreciate you for your careful reviewing and valuable comments. Accordingly, we revised the manuscript (MS) based on your comments and suggestions. The details of the revision are as follows:

Point 1: Thank you for editing this article. The revised version is clearer and more concise. This article has the potential to be developed as a tool for agricultural decision-making related to climate change in the future.

However, it would be clear if the cultivar names and the number of cultivars used in the test were provided (in line 104).

Response 1: We sincerely appreciate your valuable suggestion. Accordingly, we have reorganized the materials information into supplementary tables, the corresponding description has been presented in line 110-111 of the updated manuscript..

Thank you again for taking time out of your busy schedule to review our paper, and thank you for your valuable suggestions. After revising the MS according to your suggestions, we all learned a lot from your nice comments.

Author Response File: Author Response.pdf