Improved Monthly and Seasonal Multi-Model Ensemble Precipitation Forecasts in Southwest Asia Using Machine Learning Algorithms

Round 1

Reviewer 1 Report

The paper contains useful information, and the efficacy of the methods utilized is clearly stated. It is suitable for publishing.

Author Response

We thank the reviewer for the time spent on reading the paper, and for the positive comments.

Reviewer 2 Report

The topic of the paper is within the scope of the journal. However, there are some flaws in the manuscript as it stands at the moment. Corrections should be made as needed to ensure that the publication recommendation is warranted.

• Line 103 speling [previosu]

• In a clear and concise manner, address the research gap in relation to the objectives of the study. The paper's objectives are not articulated very well, and it needs a few more assertions that are more detailed. Rewrite the objectives so that they are more understandable.

• Total Coverage Area of the study region?

• What is the novelty of this work? 

• Dictate the numerical value of the statistical indices in the text also, directly indicating figure causes a break in the study.

• What is the usefulness/applicability of the study? Must include the limitations also

• Instead of Correlation use the Correlation coefficient

• Modellers should exercise extreme caution when it comes to allowing their understanding of NSE values to impact their interpretation of KGE values. What is the rationale to use NSE and KGE?

• In this paper, a multi-model ensemble strategy employing well-known ANN and RF algorithms was presented for monthly forecasting of precipitation for southwest Asia. The efficacy of such methods relies so substantially on tweaking the parameter optimization, Discuss these issues.

• The gentle discussion part is weak and must be addressed. The discussion section of a publication ought to contain a synopsis of the paper's primary findings, an explanation of how those findings fit into the greater body of scientific research, and a discussion of any issues with the study or differences between its findings and those of other related published works.

• In Figure 13. the a, b,c and d  are not complete, adjust the size

• In the Abbreviations section, the ANN is misspelt

• Once the aforementioned revisions have been made, I believe this study is ready for publication. I am optimistic that the authors will find my comments useful in reworking the manuscript. Many thanks!

Author Response

We appreciate your precious time in reviewing our paper and providing valuable comments. It was your valuable and insightful comments that led to possible improvements in the current version. The authors have carefully considered the comments and tried best to address each one. We hope the manuscript after careful revisions meets your high standards. Below we provide our point-by-point responses. All the important modifications in the manuscript have been highlighted in red.

• Comment 1. Line 103 spelling [previosu]
• Thank you for the comment. The spelling is corrected.

• Comment 2. In a clear and concise manner, address the research gap in relation to the objectives of the study. The paper's objectives are not articulated very well, and it needs a few more assertions that are more detailed. Rewrite the objectives so that they are more understandable.

• Details about the objectives are mentioned in red text that we added (lines 93-103 of the revised manuscript).

• Comment 3. Total Coverage Area of the study region?
• Thank you for this comment. The total coverage area is depicted in Figure 1 (southwest Asia).

• Comment 4. What is the novelty of this work?

• The machine learning techniques are widely used in literature for prediction, in this paper we used two machine learning techniques for monthly forecast of precipitation for southwest Asia which is a large region. Also, the proposed approach is a multi-model ensemble one. In the red text in the introduction section, we had already highlighted these improvements on previous work.

• Comment 5. Dictate the numerical value of the statistical indices in the text also, directly indicating figure causes a break in the study.
• Thank for the comment. The numerical value of the statistical indices is now indicated in the text. For each Figure, details of the improvements are illustrated by mentioning the numerical values of statistical indices (lines 200, 213, 222 and 230).
•  
• Comment 6. What is the usefulness/applicability of the study? Must include the limitations also
• We added some possible applications and limitations of the proposed algorithms in Conclusion part.

• Comment 7. Instead of Correlation use the Correlation coefficient
• We replaced “correlation” with “correlation coefficient” throughout the paper.

• Comment 8. Modellers should exercise extreme caution when it comes to allowing their understanding of NSE values to impact their interpretation of KGE values. What is the rationale to use NSE and KGE?
• In this article, we tried to calculate various indicators to evaluate the proposed method. This helps us in multi-dimensional evaluation from different points of view.

• Comment 9. In this paper, a multi-model ensemble strategy employing well-known ANN and RF algorithms was presented for monthly forecasting of precipitation for southwest Asia. The efficacy of such methods relies so substantially on tweaking the parameter optimization, Discuss these issues.
• The adjustable parameters of RF (e.g., number of trees) and ANN (e.g., number of layers and neurons) were selected based on an experimental test. Indeed, several values of parameters were tested to select the optimal values. The main criteria were the values of the statistical indicators. For example, increasing the number of trees in RF or increasing the number of layers and the number of neurons in ANN, cause overfitting. On the other hand, low number of trees in RF and low number of neurons in each layer of ANN cause underfitting. This is indicated in the revised manuscript at line 162 in the red additional text.

• Comment 10. The gentle discussion part is weak and must be addressed. The discussion section of a publication ought to contain a synopsis of the paper's primary findings, an explanation of how those findings fit into the greater body of scientific research, and a discussion of any issues with the study or differences between its findings and those of other related published works.
• The authors tried to add more discussion at conclusion part. For example, at line 254 we added some discussions for Figure 13.

• Comment 11. In Figure 13. the a, b,c and d  are not complete, adjust the size
• This is corrected.

• Comment 12. In the Abbreviations section, the ANN is misspelt
• This is corrected.

Reviewer 3 Report

Thank you for providing me the opportunity to review the paper by Pakdaman et al. on precipitation forecasting in Asia.  I enjoyed the short piece and I found the paper to be well organized and well written.  The key references are cited, the abstract is informative, the length is fine, the methods are appropriate, the graphics are publishable (more than publishable – they are very nice), and the results are interesting. I found the entire effort to be of the highest quality.

Author Response

We thank the reviewer for the time spent on reading the paper, and for the very positive and encouraging comments.

Round 2

Reviewer 2 Report

The corrections have been incorporated and the manuscript may be published.

Author Response

We thank the reviewer for the time spent on reading the new version, for the positive conclusion.