All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.
Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.
Thank you for the marvelous contribution. Many thanks for the wonderful suggestions to the reviewers.
[# PeerJ Staff Note - this decision was reviewed and approved by Xiangjie Kong, a PeerJ Section Editor covering this Section #]
Everything is OK
yes
Impact and novelty assessed
Please address the remaining minor issues identified during the peer review process.
**PeerJ Staff Note:** It is PeerJ policy that additional references suggested during the peer-review process should only be included if the authors agree that they are relevant and useful.
no comment
no comment
Discuss your novelty compared with recent similar works.
Use and cite:
https://doi.org/10.1016/j.compag.2025.110317
https://doi.org/10.1088/2515-7620/ad2d02
http://dx.doi.org/10.3390/rs12081232
**PeerJ Staff Note:** It is PeerJ policy that additional references suggested during the peer-review process should only be included if the authors agree that they are relevant and useful.
Please follow the requests, criticisms, and suggestions of the reviewers assiduously.
**PeerJ Staff Note:** Please ensure that all review, editorial, and staff comments are addressed in a response letter and that any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate.
- Line 40: “broad-scale crop yield prediction” would be clearer as “large-scale crop yield prediction”.
- Line 46: “Remote sensing data provides…” should be “Remote sensing data provide…”
- Line 86: “also incorporated” → rephrase to “was also incorporated”.
- The literature section (Lines 110–252) is overly long, repetitive, and sometimes reads like a summary of unrelated works. It also lacks important recent citations that use the hybrid approaches and hybrid machine learning, including automated ML. Please clarify your work novelty beyond these works.
Hence, this section should be condensed and streamlined to focus on:
1.Gaps in current wheat yield prediction models.
2. Justification for stacking ensemble methods and multi-source data fusion.
3. Direct comparison with benchmarks in similar agroecological zones.
- Suggest consolidating and reorganizing by themes (e.g., ML methods, data types used, region-specific applications).
- Some headings are missing or inconsistent (e.g., the model architecture is discussed in the middle of the Results section).
- Figures are relevant and appropriately labeled, but some need clearer captions and integration into the discussion (e.g., Figure 5 on residuals is mentioned briefly without interpretation).
-Raw data are stated to be available upon request. However, for full transparency and in line with PeerJ’s data sharing policy, the authors should provide open access to all preprocessed datasets and code used in model training, preferably through a permanent repository (e.g., Zenodo, Figshare, or GitHub with DOI linkage).
- Clarify the novelty of the model design and how it fills the stated research gap.
- Justify model choices, especially the ensemble configuration. Why were only FFNNs used as base learners in the ensemble? Why not CNNs, LSTMs, or regression trees to increase model diversity? Why is Random Forest used as a meta-learner instead of a more typical linear regressor in stacking? Please justify this choice with a performance comparison or references.
- Clearly describe how training, validation, and testing were performed. Was cross-validation used? Any temporal or spatial considerations?
- Key details are missing (e.g., hyperparameters, training settings). Include them or provide code in a public repository.
- Explain how data from different sources were aligned and fused.
- The algorithm claims to “synchronize” heterogeneous data into a unified feature vector, but the methodology (step 13 in Algorithm 1) lacks detail. How was spatio-temporal alignment handled across different modalities with varying resolutions and sampling frequencies?
- The data splitting strategy (train/test/validation) is not explicitly described. Was cross-validation used? Were samples randomly shuffled? If the dataset has temporal or spatial structure, how was this handled?
- The reported performance metrics (e.g., R² = 0.81, MAE = 59.07 kg/ha) are strong, but the values may be too optimistic considering the complexity and variability of wheat yield data. You should clarify whether data leakage was ruled out, what validation strategy was used (e.g., k-fold cross-validation, temporal holdout), and if hyperparameter tuning was done only on the training set.
- Figure 5 (residuals) suggests heteroscedasticity, particularly for high-yield observations. This should be interpreted in the text with possible explanations (e.g., outliers, missing variables, changing agronomic practices).
- Conclusion should be sharpened and focused to reflect only the supported findings — avoid overstating model superiority without broader benchmarking or validation.
1. Why were three feedforward neural networks chosen for the base learners? Was this empirically validated?
2. The Pearson correlation threshold r > 0.3 is arbitrary; explain why it was chosen!.
3. What method was used to prevent multicollinearity among the selected variables?
4. The study lacks a discussion of temporal trends or seasonality. Did you consider using LSTM or attention-based models to leverage time series characteristics?
5. Were the weather variables monthly, seasonal, or annual averages? Please clarify.
6. An MAE of 59 kg/ha seems unrealistically low given the mean yield of ~18,000 kg/ha and reported MAEs in comparable literature. Double-check for data leakage or overfitting.
7. Provide the standard deviation of prediction errors or confidence intervals.
8. In the Title, the term “Stacked Ensemble Neural Network” (SENet) may confuse readers, as SENet is also the name of a popular CNN architecture (Squeeze-and-Excitation Networks). Choose a more specific acronym (e.g., MS-YieldStackNet) to avoid ambiguity.
9. Add a spatial map of the study area with sampling points.
• Rewrite the abstract. In the abstract, include the importance of the work, what was done, and the main results of the work.
• Rewrite the Introduction. In the Introduction, write the research gap.
• The introduction should summarize the relevant literature so that the reader can understand why the topic is important and worthy of investigation.
• The study does not compare the performance of the models with baseline models (e.g., linear regression, etc.). Include baseline models and cross-validation.
• There is no discussion on feature importance or interpretability techniques, which is crucial for model transparency
• The study does not consider the feasibility of implementing the model in low-resource environments, which is important for real-world use.
• The possibility of adaptation of the model to other regions, crops, or agro-ecological zones has not been explored.
• In the result and discussion, include comparative analysis with other machine learning and deep learning models.
• Include sensitivity analysis to enhance interpretability.
• Provide details on data preprocessing and handling of missing data
• Include limitations, challenges, and future directions in the manuscript.
• The typographical errors in the manuscript need to be corrected meticulously.
• Read the sentence properly and correct the whole manuscript.
• Do all the corrections as suggested in the manuscript.
The manuscript needs major revisions for clarity, accuracy, and originality. Its contribution lies in application. A clearer focus, writing, and stronger experimental validation are required for publication.
.
.
.
All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.