Modeling and Prediction of Environmental Factors and Chlorophyll a Abundance by Machine Learning Based on Tara Oceans Data
Round 1
Reviewer 1 Report
Attached Separately.
Comments for author File: 
Comments.pdf
Author Response
Dear editor and reviewer,
On behalf of my co-authors, we are very grateful for your positive and constructive comments and suggestions on our manuscript entitled “Modeling and Prediction of Ocean Environmental Factors for Tara Ocean Based on Machine Learning”. The comments are valuable and very helpful. The authors appreciate the reviewer’s constructive comments, which helped us improve the manuscript significantly.
We have read through the comments carefully and tried our best to revise our manuscript according to the comments.
We are very sorry for the bug in our original program code, which led to the wrong MRE values in the original manuscript. We updated the program code and recalculated the value of MRE. MRE results have been corrected in Tables 4-7.
In order to get more reliable results and reduce the random effects of machine learning, we conducted every experiment ten times. Deeper optimization for modeling and predicting has been done, and a large of experiment results have been added to this manuscript.
The following are the responses and revisions.
Response to comments
- The manuscript is too long and often confuses the reader. Authors may enhance the novelty and the intent of this research to bolster the entire manuscript.
Response:
We thank the reviewer for his effort in reading our manuscript and providing us with useful comments.
We have removed some redundant descriptions and reorganized the revised manuscript.
We have labeled the Hierarchical titles as below,
- Introduction
- Materials and Methods
2.1 Data Sources
2.2 Data Cleaning
- Comprehensive Correlation Model
3.1 Character Importance
3.2 Pearson Correlation
3.3 Comprehensive Correlation Evaluation
- Integrated Prediction Model
4.1 eXtreme Gradient Boosting Regression
4.2 LOOCV Validation
4.3 Objective Function
- Results and Discussion
5.1 Data Processing
5.2 Characteristic Importance Evaluation
5.3 Marine Environmental Factors Modeling
5.4 Chl-a Prediction
5.5 Results Comparison and Model Discussion
- Conclusions
The main changes includes:
The abstract has been updated.
Chapter “1. Introduction” has been updated. More previous works and references relevant to the study have been introduced in the revised manuscript.
Table 3 and relative presentations have been removed from section 5.5 to section 5.2.
Some explanations about Figur6 have been reduced in section 5.3.
Based on more detailed experiments, the original “Figure 7. Different by different machine learning methods”, “Figure 8. Average relative errors with different numbers of environmental factors”, “Table 9. Results for some prediction models”, and related presentations have been removed in section 5.4. “Figure 7. MRE based on different methods” was added in the revised manuscripts.
Deep learning requires a large number of samples to conduct sufficient model training, which is not suited for our work in section 5.5. The expressions and results comparing(In Table 9) about deep learning have been removed from in revised manuscript.
More experiments were conducted about MRE variation with different Max_depths and n_estimators have been done in section 5.5. Based on the detailed results, Figure 9 was updated to “Figure 8 Variation of MRE with different Max_depths and n_estimators” in the revised manuscript, and the related presentations have been updated.
- The entire writing should be in a more scholarly sense.
Response:
We thank the reviewer’s useful comments.
We have worked on both language and readability. The abstract, introduction, integrated model, conclusions, and other expressions in the revised manuscript have been updated in the revised manuscript.
- More discussion is needed for previous works. Summary and comment are needed for others’ works. More discussions about the parameters and their observation on them may be provided.
Response:
We are grateful for the suggestion.
More previous works relevant to this research have been introduced in the revised manuscript.
- The quality of the figures must be improved.
Response:
We apologize for the poor quality of the images.
We have replaced all the poor-quality images in the original manuscript with improved images.
- Almost all of the abbreviations are not defined.
Response:
We are grateful for the comments.
We have checked the expressions and defined the abbreviations in the revised manuscript.
- The contribution of the authors in terms of the developed algorithm needs to be described elaborately.
Response:
We are grateful for the comments.
The relations of environmental factors are strongly nonlinear and the samples are insufficient, which makes the prediction of Chl-a very difficult. In order to get more reliable results, the proposed marine environment factors prediction model integrates XGBoost regression, LOOCV validation, MRE optimization, and other strategies.
- The repeatability of the method and the results are important to be discussed in the manuscript with supporting arguments.
Response:
We are grateful for the suggestion.
In order to get more reliable results and reduce the random effects of machine learning, we conducted every experiment ten times. Deeper hyper-parameter optimization for modeling and predicting has been done and the results have been updated. The parameters for the model training based on XGBoost are listed in Table 8, and the URL of the raw data was presented in the manuscript.
- The over-enthusiastic claim of authors that their prediction is much better than any other models with only a very few investigation-specific ocean environmental factors is not justified in any part of the manuscript.
Response:
We appreciate it very much for this good suggestion.
We have conducted more experiments and carried out deeper optimization in the modeling and predicting. Many new experiment results have been added to the revised manuscript. The abstract, conclusions, models, and related expressions have been updated in the revised manuscript.
- What are investigation-specific ocean environmental factors? What is not must be listed with analysis and results.
Response:
We are grateful for the comments.
Based on more experiment results, we have updated the relevant expressions
- Too many claims in the conclusions are not supported by results.
Response:
We are grateful for the comments.
The claims and conclusions have been updated in the revised manuscript.
Author Response File: 
Author Response.docx
Reviewer 2 Report
The authors have done modeling and prediction of ocean environmental factors for Tara Ocean based on machine learning. This work is interesting for the environmental scientists. However, there are various issues in this work.
1. The quality of images are poor. For example figure 1, 3, 6, 7, 9.
2. It is difficult to go through because of the poor organization and structure of this manuscript.
3. Equations 10, 13, 16, 17, 21 an so on are not embedded well into the texts.
4. Font size of most of the images are too small. Please increase font size for better readability.
5. The conclusions are more detailed. It can be bulleted. Only important ones should be mentioned.
6. Additionally, English should be polished well.
Author Response
Dear editor and reviewer,
On behalf of my co-authors, we are very grateful for your positive and constructive comments and suggestions on our manuscript entitled “Modeling and Prediction of Ocean Environmental Factors for Tara Ocean Based on Machine Learning”. The comments are valuable and very helpful. The authors appreciate the reviewer’s constructive comments, which helped us improve the manuscript significantly.
We have read through the comments carefully and tried our best to revise our manuscript according to the comments.
We are very sorry for the bug in our original program code, which led to the wrong MRE values in the original manuscript. We updated the program code and recalculated the value of MRE. MRE results have been corrected in Tables 4-7.
In order to get more reliable results and reduce the random effects of machine learning, we conducted every experiment ten times. Deeper optimization for modeling and predicting has been done, and a large of experiment results have been added to this manuscript.
The following are the responses and revisions.
Responds to the comments
- The quality of images are poor. For example figure 1, 3, 6, 7, 9.
Response:
We apologize for the poor quality of the images.
We have replaced them with improved images in the revised manuscript.
- It is difficult to go through because of the poor organization and structure of this manuscript.
Response:
We thank the reviewer for his effort in reading our manuscript and providing us with useful comments.
We have removed some redundant descriptions and reorganized the revised manuscript.
We have labeled the Hierarchical titles as below,
- Introduction
- Materials and Methods
2.1 Data Sources
2.2 Data Cleaning
- Comprehensive Correlation Model
3.1 Character Importance
3.2 Pearson Correlation
3.3 Comprehensive Correlation Evaluation
- Integrated Prediction Model
4.1 eXtreme Gradient Boosting Regression
4.2 LOOCV Validation
4.3 Objective Function
- Results and Discussion
5.1 Data Processing
5.2 Characteristic Importance Evaluation
5.3 Marine Environmental Factors Modeling
5.4 Chl-a Prediction
5.5 Results Comparison and Model Discussion
- Conclusions
The main changes includes:
The abstract has been updated.
Chapter “1. Introduction” has been updated. More previous works and references relevant to the study have been introduced in the revised manuscript.
Table 3 and relative presentations have been removed from section 5.5 to section 5.2.
Some explanations about Figur6 have been reduced in section 5.3.
Based on more detailed experiments, the original “Figure 7. Different by different machine learning methods”, “Figure 8. Average relative errors with different numbers of environmental factors”, “Table 9. Results for some prediction models”, and related presentations have been removed in section 5.4. “Figure 7. MRE based on different methods” was added in the revised manuscripts.
Deep learning requires a large number of samples to conduct sufficient model training, which is not suited for our work in section 5.5. The expressions and results comparing(In Table 9) about deep learning have been removed from in revised manuscript.
More experiments were conducted about MRE variation with different Max_depths and n_estimators have been done in section 5.5. Based on the detailed results, Figure 9 was updated to “Figure 8 Variation of MRE with different Max_depths and n_estimators” in the revised manuscript, and the related presentations have been updated.
- Equations 10, 13, 16, 17, 21 an so on are not embedded well into the texts.
Response:
We are grateful for the suggestion.
We have checked all the equations and embedded them into revised manuscripts.
- Font size of most of the images are too small. Please increase font size for better readability.
Response
We thank the reviewer for this comment.
We have increased the font size in the figures for better readability.
- The conclusions are more detailed. It can be bulleted. Only important ones should be mentioned.
Response
We thank the reviewer for this comment.
We have conducted more experiments and carried out deeper optimization in the modeling and predicting. According to the detailed results, the conclusion has been updated.
- Additionally, English should be polished well.
Response
We are grateful for the suggestion.
We have now worked on both language and readability and have also involved native English speakers for language corrections. We really hope that the flow and language level have been substantially improved.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
The authors have answered my comments.
