Machine-Learning-Assisted Characterization of Regional Heat Islands with a Spatial Extent Larger than the Urban Size

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The authors proposed a machine-learning-assisted solution to quantify SHUIs. The research methodologies are reasonable, and the findings are interesting. However, there are still some aspects that should be improved to make the paper publishable. I focus here only on some points, which are hopefully easy for the authors to take into account in the revision.

 

1.       Part Introduction - the innovation of the work should be further highlighted, and the factors influencing SHUIs / LST can be introduced in this part. There are some references on this topic, I suggest you supplied it in this part, as follows. 1) doi: 10.1016/j.rse.2023.113650, 2) doi: 10.1016/j.scs.2023.104933, 3) doi: 10.1016/j.rser.2022.112350.

2.       Sec 2.1 - many data were used in this study. It is good to add more details to the section, e.g., the reasons for choosing the data (FVC, human settlement, topographic variable…)

3.       Line 153 - it is good to modify the expression, ‘more than one hundred, one hundred 153 to fifty, fifty to five, and below five km2.’ Numbers are more appropriate.

4.       How to verify the accuracy of the results? Or comparison with previous studies (advantage and disadvantage)? It is worth being further discussed.

Overall, it is a good work.

Comments on the Quality of English Language

Minor editing of English language required.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The study characterized SUHI spatial patterns and scales, estimated FVC and LST backgrounds, quantified SUHI intensity, and analyzed seasonal variations in the YRDUA region. Challenges in identifying a suitable intensity threshold for optimal SUHI characterization were highlighted, emphasizing the importance of considering spatial extents compatible with LST annual cycles. The RF models demonstrated reliability in simulating non-SUHI zone LSTs.

 

The language used in the article is generally technical and appropriate for a scientific publication. It demonstrates a solid understanding of the subject matter.

Sentences are well-structured, and the terminology is consistent with the field of remote sensing and climatology. The use of acronyms is prevalent, which is common in scientific writing. However, it's essential to ensure that acronyms are defined upon first use to enhance reader comprehension. The article follows a standard scientific paper structure with sections such as Introduction, Methods, Results, Discussion, and Conclusions. Each section seems logically organized, providing a clear flow of information. The transition between sections is generally smooth.

Headings and subheadings are appropriately used to guide the reader through different parts of the article. The graphics are clear, labeled appropriately, and contribute effectively to the understanding of the presented data.

 

Before finalizing the publication, we kindly request you to consider and incorporate the following suggestions for improvement:

Abstract:

 

1. Emphasize the innovative aspects of the proposed machine-learning-assisted solution early in the abstract. Clearly state how this approach advances existing methodologies.

2. Explicitly state the existing gap in research that your study aims to address. This could be related to the limitations of current models or challenges in characterizing SUHIs in urban agglomerations.

3. Consider adding a brief mention of potential limitations of the proposed machine-learning approach. This adds transparency to your methodology and sets realistic expectations for readers.

4. Explicitly connect the findings and methodology to practical implications for urban planning and environmental management. Explain how the results can inform decision-makers and contribute to sustainable urban development.

 

 Methods

1. Consider briefly addressing the quality assurance procedures applied to the datasets. Mention any preprocessing steps or data validation techniques to assure readers of the reliability of the information used in the study.

2. Provide a brief rationale for the chosen temporal and spatial resolutions of the datasets. Explain how these resolutions align with the study's objectives and the urban characteristics of the Yangtze River Delta urban agglomeration (for example why you don’t use Imperviousness Density or building heights datasets).

3. If applicable, briefly mention any critical hyper-parameters used in the RF models (e.g., the number of trees, tree depth) and explain how these were determined. This adds transparency to the modeling process.

4. Offer a brief overview of the training process for the machine learning models. Highlight any specific considerations or challenges encountered during the model training phase, such as handling missing data or optimizing model parameters.

 

Results

1. Begin by providing a brief overview of the key findings before delving into specific details. Clearly outline the subsections within the Results section to guide the reader through the different analyses.

 

Discussion

1. Relate the identified SUHI zones to practical applications in urban planning and decision-making, discussing how different zones can inform mitigation strategies.

2. Discuss the applicability and generalizability of the RF models to other regions or similar studies, considering the robustness of the selected features.

3. Connect your findings to existing literature on SUHI quantification, highlighting any novel contributions your study brings to the field. Discuss how your approach addresses limitations or gaps identified in previous research.

 

Conclusion

 

 

1. Propose potential future research directions, addressing any limitations identified in the current study and suggesting areas for further exploration.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

In my opinion, the article does not have very big gaps, however, for better reception, reliability and readability, corrections/additions should be made:

 - in the theoretical part, the authors very sparsely cited the study of heat island using machine learning. The literature review in this area should be significantly expanded, as many papers on this topic have been published.

 - In the methodology, the authors did not specify (or I did not find it in the text) what library/program was used for  machine learning random forest method. 

 - In Figure 7, the green dots are very unreadable, it would be necessary to either use a different color or for other layers to set the colors brighter, or otherwise improve the readability of the maps.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf