Spatiotemporal Air Pollution Forecasting in Houston-TX: A Case Study for Ozone Using Deep Graph Neural Networks

Round 1

Reviewer 1 Report

Dear Authors,

Please find attached my review as a pdf file.

Reviewer

Comments for author File: Comments.pdf

Author Response

Dear Reviewer,

Please see the attachment for point-by-point response to your suggestions.

Best regards,

The Authors.

Author Response File: Author Response.docx

Reviewer 2 Report

Authors predict O3 concentration over a city based on different approcches and data inputs by using neural networks and GraphSAGE paradigm. The paper is too long with 17 figures. Authors should show only max 5 figures. The text in redundant and repeating again and again the same. Authors should write crystal clear what they mean and be as short as possible. English should be written in advanced form too. Paper has no discussion. It is the first time I see "Discussion" without a single reference. Although I think paper should be rather rejected, there is some chance, that authors will rewrite the paper significantly.

Here are some detailed comments:

line 34: O3 is not emitted from fossil fuels

line 93: using deep forests - sounds strange

Introduction: much better prediction is achieved, when very detailed spatial analysis of the plant cover is performed (especially in case of urban environment this seems to be important). See paper 10.1016/j.ufug.2022.127757 which you might include in the text.

line 142: Use ozone or O3, but be consistent.

line 174: indicate how did you measure the data. What is the time resolution of the data. At which height and by which analyser are they measured? Or where did you get the data?

lines181-191: Why do you repeat this again here?

fig 2: Describe more the Fig. What are the differences between a-d? Why you do not show only one map and not 4 maps which are the same? Make text bigger, it is not possible to read.

lines 198-211: write in more condensed way, it is not a thesis, but scientific paper.

fig 3: i do not see any point in showing this figure. Especially when you write it is the same. just write in the text that wind has predominant East and South direction.

fig 4: what should that be? Why you do not describe anything in the figure?

fig 5: describe the figure, do not write only GraphSAGE structure.

Eq 2, 3, 5. This is paper book knowledge. Do not write this Eqs here.

fig 9, 12, 15: Do not show this graphs here. Similar information even fig 8 has and others onward. Why to double the information?

lines 486: Why to repeat something again? Write more straightforward!

Discussion: You wrote discussion without a single reference? Then you did not discuss anything. Write proper discussion.

Author Response

Dear Reviewer,

Please see the attachment for a point-by-point response to your suggestions.

Best regards,

The Authors.

Author Response File: Author Response.docx

Reviewer 3 Report

The manuscript needs to be revised to guarantee publication. 

Your SHAP analysis is a novelty for your domain. So, I kindly invite you to consider the following comments:

- You need to highlight the use of local XAI in your manuscript. Readers unfamiliar with SHAP analysis need to understand why the local explainability of your models is important here. I kindly invite you to introduce this in the beginning, move the description of SHAP analysis to materials and methods, and elaborate on previous local XAI implementations on regression tasks. Recommended readings: 

 

• https://doi.org/10.1016/j.habitatint.2022.102660

• https://doi.org/10.1038/s41598-022-11429-9

• https://doi.org/10.1007/s11356-021-17135-9
• - There are some problems related to SHAP dot plot. First of all, the colormap is extremely small. Then, it seems that you added the dependent variable to the dot plot. Please remove that, and keep the independent variables only. After that, you'll see more distinguishable sample dots. 
• - I highly recommend adding a colorized bar plot for your mean SHAP values (see the recommended readings) to understand the direction of the contribution of each independent variable (positive or negative). 
• - I see that you only discussed and compared the performance metrics of other studies. Since you made an explanation of your models, you need to discuss and compare the importance of your independent variables. Does your list of important variables confirm or contradict other studies? 
• - Please mention and cite the library/package for your SHAP analysis. 

Author Response

Dear Reviewer,

Please see the attachment for a point-by-point response to your suggestions.

Best regards,

The Authors.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

thank you for the changes

Reviewer 3 Report

Dear Editor and Authors,
Thank you for inviting me to evaluate the revised manuscript. In my opinion, the authors have made a substantial overhaul and have attempted to address all of my concerns. I think the paper is now ready to go to print. 

Improvements include:

-Providing a better explanation for SHAP technique.
-Incorporating recent literature more effectively.
-Editing the figures in order to make them more comprehensible.
-Comparing and discussing the importance of independent variables with recent literature.