Enhancing Crop Yield Predictions with PEnsemble 4: IoT and ML-Driven for Precision Agriculture

Round 1

Reviewer 1 Report

This is an interesting topic for the PEnsemble 4 model, a sophisticated machine learning framework that integrates IoT-based environmental data to accurately forecast maize yield. Also,the PEnsemble 4 model establishes a new standard in maize yield prediction, revolutionizing crop management and protection through the synergistic utilization of IoT and machine learning technologies. The author's overall analysis is good, however,some writing skills need to be strengthened.I think the following improvements could be made

Comments for author File: Comments.pdf

Minor editing of English language required

Author Response

Thank you very much for your compliment and comment. Your feedback is valuable to us as we continuously strive to improve the quality and appeal of our work. In response to your comment, please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper examined how the machine learning framework PEnsemble 4 model can be used and combined with environmental data from the Internet of Things (IoT) to accurately predict maize yields.The PEnsemble 4 model enables earlier estimation, advancing yield prediction from day 100 of the conventional R6 phase to day 79 of the R2 phase.The PEnsemble 4 model extends its benefits beyond yield prediction that facilitates the detection of moisture and crop stress, as well as disease surveillance in a broader agricultural setting. This research makes a significant contribution to precision agriculture by providing an efficient and sustainable alternative to traditional agricultural practices through accurate yield prediction. This study is a much-needed and timely study, but there still needs some revisions to improve the quality of this work.

Minor comments:

1. The abstract section does not accurately summarize the work in the full paper and the findings are not clear, please revise.

2. The introduction section lacks examples of existing relevant studies.

3. The introductory part is repetitive, please delete it.

4. Please give a brief description of the geographic location and climatic conditions of the study area in 2.1 Study area. the information in Figure 1A needs to be described in text form.

5. Please describe the maize varieties grown in the study area. It is recommended to list the varietal parameter descriptions.

6. Lines 398-406 are mainly about the evaluation methodology and it is recommended that they be formed into a new section.

7. The header in the fifth column of Tables 3 and 4 should be RMSE.

8. The results section is confusing, the results also contain a description of the methodology, it is suggested to reorganize the content. For example, lines 686-693 should be in the Materials and Methods section.

9. The figure in the Discussion section belongs in the Results section, please put it in the Results section.

Author Response

Thank you very much for your compliment and comment. Your feedback is valuable to us as we continuously strive to improve the quality and appeal of our work.

In response to your comment, please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

This paper proposes a machine learning framework integrating IoT-based environmental data to accurately forecast maize yield. It combines Huber and M estimates to analyze temporal patterns in vegetation indices that are used as indicators of canopy density and plant height.

The work is potentially valuable since it falls in a very interesting research area, although some aspects can be improved as detailed below.

1.- Please define ROI the first time it appears in the text.

2.- DJI phantom 4 multispectral (P4M) quadcopter. I guess it is a drone, please detail and add a photograph.

3.- CatBoost Regression, decision tree Regression, ElasticNet Regression, gradient boosting Regression, Huber Regression, K-nearest neighbors Regression (KNN), Lasso Regression, linear Regression, M estimators, passive-aggressive Regression, random forest (RF) Regression, Ridge Regression, support vector Regression (SVR), and XGBoost Regression. Why did the authors select these algorithms? There are no explanations of their theoretical basis.

4.- Table 2 relates the vegetation indexes analyzed in this work with formulas derived from the outputs of different sensors. Did the authors related the output values of these indexes to direct measurements of real parameters such as canopy density or plant height?

5.- I guess that the outcomes of the proposed system are highly dependent on the data acquired by the sensors. What happens under varying meteorological conditions? For example rainy versus sunny days.

The Reviewer encourages the authors to revise the work based on the suggestions above in order to improve its quality.

The authors are suggested to proofread the manuscript

Author Response

Thank you very much for your compliment and comment. Your feedback is valuable to us as we continuously strive to improve the quality and appeal of our work. In response to your comment, please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The author has carefully revised according to the reviewer's comments, and there are no other suggestions.

Author Response

Thank you for considering our submission and dedicating your time. We appreciate your feedback.

Reviewer 3 Report

Thank you very much for replying all my questions. There is still a minor question related to my previous Comment 5: I guess that the outcomes of the proposed system are highly dependent on the data acquired by the sensors. What happens under varying meteorological conditions? For example rainy versus sunny days. 

May be my question was not very precise. I refer to the data gathered by the sensors, since I guess it can depend on the meteorological conditions: bright days versus dark days, rainy days versus clear days, etc.

The authors are suggested to proofread the manuscript

Author Response

Thank you for your continued engagement and insightful comments on our work. We appreciate the opportunity to address your concerns regarding the potential influence of varying meteorological conditions on the data acquired by our sensors. 

While we acknowledge the importance of considering meteorological factors in the performance evaluation of our system, we would like to emphasize a few points that support the robustness and reliability of our approach, even in the face of changing environmental conditions: 

• 1) Comprehensive Data Collection: Our sensor network is designed to collect a wide range of environmental parameters, including but not limited to light intensity, humidity, temperature, and precipitation. By capturing these variables, we ensure that our dataset encompasses diverse meteorological conditions, ranging from bright and sunny days to overcast or rainy weather.
•  
• 2) Normalization and Calibration: To mitigate the impact of varying meteorological conditions on the sensor data, we employ rigorous normalization and calibration techniques. These processes enable us to standardize the collected data, accounting for fluctuations in environmental factors and ensuring consistency across different weather scenarios.
•  
• 3) Machine Learning Adaptation: Our system incorporates machine learning algorithms that are trained on diverse datasets, encompassing different meteorological conditions. Through this approach, our models can adapt and generalize effectively, even when operating in environments with varying levels of brightness, humidity, or precipitation.

• 4) Real-World Testing and Validation: Prior to deployment, our system underwent extensive real-world testing across a range of environmental settings, including diverse meteorological conditions. This validation process not only confirmed the reliability of our sensor data but also demonstrated the robustness of our system in practical scenarios.

With these considerations, we are confident that our system's outcomes are not overly dependent on specific meteorological conditions. Rather, our approach is designed to adapt and perform reliably in diverse environmental settings, ensuring its applicability and efficacy in real-world deployment scenarios.

We hope that this explanation addresses your concerns satisfactorily. 

We have revised our manuscript adding the corresponding description in one more paragraph in Subsection 2.4. Environmental Measurements and Optimal IoT Design for Soil Health Monitoring.

Author Response File: Author Response.pdf