General Circulation Model Downscaling Using Interpolation—Machine Learning Model Combination—Case Study: Thailand
Round 1
Reviewer 1 Report
Please see the attachment.
Comments for author File: 
Comments.pdf
Please revisit the entire manuscript for minor errors in the English language.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
The English is easily understandable for me. However, there are some places where grammar should be corrected.
The purpose of the research is not clearly stated in the introduction section.
What are the advantages and disadvantages of your study?
The list of references could be enriched with more recent papers on the subject.
Please increase the resolution of Figure 1.
Please provide more information about the dataset.
Why is the ANN method used for downscaling? Why not other methods?
Avoid long sentences throughout the manuscript.
In the results and discussion section, please add some findings from the previous studies and try to discuss them here. This will enrich this section.
What would be the contribution to the scientific community?
In the conclusions section, providing limitation(s) of the study and future suggestion(s) would be useful for the journal readers.
The English is easily understandable for me. However, there are some places where grammar should be corrected.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
In general, the methodology of this manuscript is reasonable, but the quality and presentation of data used and results must be improved.
General Comment
Overall, the manuscript requires serious edits.
Point 1: Figure 3, Under Data Processing – it should read Train-test split. Quality of the figure should be improved.
Point 2: Correlation of variables could be better presented by a heatmap.
Point 3: Table 4, are the observed values monthly?
Point 4: Table 5, there should be a horizontal rule to demarcate rows for ANN and IDW.
Point 5: More model evaluation metrics like coefficient of determination should be added to further explain the performance of models.
Point 6: I propose that authors should include training loss to show how well the model is fitting the training data and validation loss to show how well the model fits new data.
Point 7: Sections 4.1 and 4.2 should be placed under materials and methods. Section 4 should only present results. I propose that authors should consider restructuring the manuscript for readers to easily follow and understand the paper. In the current state it is difficult to follow the paper.
Point 8: The authors should explain the choice of GCM used (IPSL-CM6A-LR) and why they did not compare some GCMs for the robustness of their results
Point 9: In order to further highlight the innovation of this article, it is better to compare the results of this study with other studies.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
1. The article writes a lot about the theoretical foundations of the operation of neural networks (ANN) and separately about downscaling using IDW, but there is no detailed example of the synchronous operation of both methods - a step-by-step description of the operation is necessary.
2. Already in Figure 1 illustrating downscaling, a series of light/dark points appear on the map in an almost regular manner, where do they come from?
Are they the result of the topography/use of the land, or is it some method error?
Similar points appear in later drawings. This should be clarified.
3. The article is intended to focus on downscaling in the future climate and should remain so, so there is no need to use two different climate scenarios in different time horizons - one scenario and one time horizon is enough.
4. The article lacks explanations as to how the data for future climatic conditions were obtained, whether they were generated, or whether they were data from global forecasts - a detailed description is necessary.
5. The term historical scenario is used in the text as well as in the figure captions, it should be corrected to historical data.
6. In the discussion, the authors write about the decrease/increase in precipitation/temperatures in different parts of the country - the decrease/increase may refer to changes over time and not to the area.
7. Both the discussion and conclusions should be more about downscaling than about climate change.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The manuscript is enhanced in quality and ready for publication.
Author Response
Thank you for your detailed review and publication acceptance.
Reviewer 2 Report
The quality of the paper has increased to a great extent; therefore, I think the article can be accepted in this format.
Author Response
Thank you for your detailed review and publication acceptance.
Reviewer 3 Report
The authors have made improvements to the article. But a serious methodological problem remains to be solved.
I am not at all convinced by your answer on the performance of the models: you are using very short time series. As you do not want/think to add indicators, the other option is to take time series of at least 30 years. Indeed, monthly meteorological data over 6 years, has no physical meaning for a given country.
As it stands, this article cannot be accepted for me without this substantial methodological addition.
Minor edits
Point 1: Figure 1 legend to modify: add “left” (for non-downscaled data) and “right’ (for downscaled data)
Point 2: To be coherent with previous figures, for figures 11 to 13 replace “Celsius” by “°C”
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 4 Report
The authors have sufficiently corrected the article and clarified my comments.
Author Response
Thank you for your detailed review and publication acceptance.
