Machine Learning Downscaling of SoilMERGE in the United States Southern Great Plains
Round 1
Reviewer 1 Report (Previous Reviewer 1)
Comments and Suggestions for AuthorsPlease see the attached file
Comments for author File: Comments.pdf
Author Response
Reviewer 1 and 2 – Figures
I am not sure why the reviewers are having issues with viewing the figures in this paper. In the word document of the latest version that I downloaded from your website they look fine. In the uploaded materials I have included a zipped file that contains all the figures at 300 dpi. If there is anything else that you need me to do, I will be happy to do it but since I cannot see the problem, I am not sure how to proceed.
Reviewer 1 – Abstract
As I indicated earlier the abstract is significantly over the recommended 200-word limit and there is not the room to add more details about the machine learning methods. The abstract already presents the key findings of this study.
Reviewer 1 – Introduction
- The main limitation of RZSM products besides spatial resolution is that these products are is extrapolated using surface data. This text was included in the original draft, and we took this language and reinserted into the current draft with the relevant references in a condensed fashion. There was concern in the original draft that the introduction was not focused, and we strived to include the requested details without making this section too expansive and unfocused.
- I added some additional details about SMERGE in the introduction.
Reviewer 1 – Problem Statement
Great specificity has been added to the problem statement and specific research questions are more clearly articulated.
Reviewer 1 – Methods
- We added the specific function used to split the data for training and validation.
- Additional pre-processing details have been added to methods section 3.1 as requested.
Reviewer 1 – Results
- I believe we have been quite through in describing and interpreting the sensitivity data in this study. I am not sure what additional interpretations could be added that would provide greater insights to help the reader evaluate this work.
- Previous reviewers indicated that the paper had too many tables. Being responsive to this concern we attempted to minimize the number of tables in the main narrative and converted tables to figures. The baseline performance metrics are present in tables in the appendix in the revised paper, so no information has been lost allowing the reader to fully evaluate the results of this work.
Reviewer 1 – Discussion
- Additional discussion of the limitation of in situ data have been added to the discussion section.
- Some additional discussion about the implications of this study have been added.
Reviewer 1 – Conclusions
We want to keep the conclusions concise and not repeat information already indicated at the end of the discussion section. Therefore, we have opted not to implement this recommendation.
Reviewer 2 Report (Previous Reviewer 3)
Comments and Suggestions for AuthorsThe authors have responded my comments, but Figures 2, 3, 4, 5, and 6 are missing. Please add these figures to the article.
Author Response
I am not sure why the reviewers are having issues with viewing the figures in this paper. In the word document of the latest version that I downloaded from your website they look fine. In the uploaded materials I have included a zipped file that contains all the figures at 300 dpi. If there is anything else that you need me to do, I will be happy to do it but since I cannot see the problem, I am not sure how to proceed.
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsPlease furnish all the figures
Please see the attached file
Comments for author File: Comments.pdf
Comments on the Quality of English LanguageThe quality of the English language could be improved
Author Response
please see the attachment
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThis article realizes the downscaling of SoilMERGE data based on three machine learning methods, and obtains the best downscaling method and the best spatial resolution, which provides reference for subsequent related work. However, there are several issues that need to be resolved.
(1)The introduction of the in situ data used in the paper is a bit confusing. It is recommended to modify the map of the original data usage and display the in situ data in more detail and clearly using tables and text.
(2)The purpose and significance of the research in this article are open to question. This study divided a relatively small research area into different time periods and several sub-regions based on data. Three different machine learning models were applied to each region to investigate the performance at various downscaling resolutions. The results showed that the performance varied for each period and region, as machine learning models can potentially overfit, and features that perform well within a very small time frame and spatial range may not necessarily be applicable to broader contexts. The article did not discuss the downsizing of different machine learning models on a global scale.
(3)The hyperparameter settings of the machine learning models could potentially impact the results. The article does not provide details about the tuning and optimization processes for each model. It's important to note that the performance of machine learning models can vary significantly between models that have not been optimized and those that have been properly tuned and optimized.
(4)The study discusses the effects of downsampling at different resolutions, which is meaningful. However, there is an issue to consider within this discussion. It's important to address the scale effect when comparing data at different downsampling resolutions with the same in-situ observational data. Whether the use of the same accuracy validation methods is appropriate should also be questioned.
Author Response
please see the attachment
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsRemote sensing- 2598281 This study used advanced machine learning techniques to downscale SMERGE to spatial resolutions between 100 to 3000 m. The method in the study is simple, and the analysis about result is insufficient. The following is the specific suggestions.
Major issues:
1. The analysis about result is insufficient. The tables such as Tables 10, 11 are too long, and are not clear. These tables including Tables 7, 8, 10 and 11 may be replaced by figures, because the figure may be more intuitive. The tables are only listed in the article, and are not able to provide meaningful information to the readers.
2. The figures in the study are not appropriate for the scientific article. Figures 4, 5, 6 lack legends. Figures 1 and 3 lack coordinate axis.
3. The meaning of the data in Table 2 is not clear. Is this data average values? 4. Figure 2 is useless, and may be replaced by text.
Minor issues:
1. Table 3, “Era/Network/Region” is in disagreement with “ARM_3_1”.
2. From Formula 3 to Formula 7, the situation when Delta r = 0 or Delta ubRMSE = 0 is not included.
Author Response
please see the attachment
Author Response File: Author Response.pdf
Reviewer 4 Report
Comments and Suggestions for AuthorsThe overall quality of the paper is good. However, there are several issues that need to be addressed:
1. The research area of this article was not displayed using maps. Suggest modification.
2. The meaning of this figure, Fig. 1, is not clear. Suggest further explanation. What are the regions corresponding to Era 2?
3. The flowchart , Fig 2, is too brief to provide a detailed illustration of the method used in this article. Especially how to use machine learning for downscaling.
4. The shortcomings for improvement in this study have not been addressed. Suggest explaining the shortcomings of this study.
Author Response
please see the attachment
Author Response File: Author Response.pdf