Improving Intra-Urban Prediction of Atmospheric Fine Particles Using a Hybrid Deep Learning Approach
Round 1
Reviewer 1 Report
In the Introduction section, the authors should consider using the chemical formula for O3, NOx, SO2, CO, and PM2.5. It would be clearer to the reader and should be introduced at the beginning of the text. The first paragraph of the introduction had no citations, which should be corrected.
CTMs should be Chemical Transport Models instead of Atmospheric transport models. Please make sure the terms are correct.
Materials and Methods should be in Section 2.
All of the chemical formulas, including PM2.5 need to be in subscripts. Please correct the mistakes throughout the text.
The caption of the Figures should be short and concise, not a lengthy paragraph
The subsection is misnumbered. Need to be corrected.
The equations are not listed correctly. Please follow the format required by the journal.
Too many equations have been listed, which is confusing to the readers. Please only show the important ones.
Section 3 should be Result and Discussion. Please follow the format required by the journal.
In Tables 1 and 2, why is the RMSE of ARIMA so high? Has the model been tuned thoroughly and achieved the best possible result? Please redo the calculation of the ARIMA model.
Section 4 should be the conclusion of the text. Please follow the format of the journal.
The references (only 16) are too few for the state-of-the-art literature review for this paper. Please increase the number of references for the paper.
Overall, the novelty and findings of this paper are low, and it would require major revision before the paper can be considered for publication.
Author Response
thank you for these valuable comment and suggestions, and have resumed your questions by point in the attached document.
Author Response File: 
Author Response.docx
Reviewer 2 Report
Overall the paper is well-written and the topic is interesting. The authors should address the following comments before it can be accepted.1) Have the issues related to multicollinearity has been addressed before developing a model?
2) Among Meteorological Factors and Pollutant Concentration, speically for Meteorological Factors(including temperature, relative humidity, wind direction, and wind speed), do the authors have employed feature reduction method? If no, why?
3) Inter-comparison Parameters of Model Performance matrices should be increased. Only 3 indices are not enough to judge the quality of the models Some good examples can be found in "Predicting the settlement of geosynthetic-reinforced soil foundations using evolutionary artificial intelligence technique"
4) Please improve the writing pattern/style. After evaluating the prediction performance of the proposed model in single station, we further validate the spatiotemporal prediction efficiency of the proposed model under equal geographical weight, which means the influence spatially and temporally of surrounding sites on the target site are considered equally. Why this sentence is bold?
5) Figure 7. 8. the distance and weight matrix of 6 counties. Correct and use the same pattern in the figures.
Also check the other figures
6) It was mentioned that data were normalised, please state the range.
7) Kindly present the overall comparison of the developed models on a single platform (e.g., taylor's diagram or violin plots). Some good examples are 1) https://doi.org/10.1080/15376494.2022.2114048 and (2)sciencedirect.com/science/article/pii/S1674775522001093.
Author Response
thank you for these valuable comment and suggestions, and have resumed your questions by point in the attached document.
Author Response File: Author Response.docx
Reviewer 3 Report
The research looks very interesting but frankly speaking, it is not very easy for me to follow the logic of the paper. More specific, I cannot figure out the major goal of this study. It seems that the authors used hourly meteorological and PM2.5 measurements at 6 stations to predict next hourly PM2.5 at these stations. The spatio-temporal information is used in the prediction by using a sophisticated manner. The authors also claimed the method can be used to predict PM2.5 at any other site. My big question is what are the inputs and what is the output. Most of the words in the paper are focused on the specific machine learning technique, but there are few words on this basic logic of this study. therefore, I suggest the authors to say some words on the logic of this study in the revision version.
Author Response
thank you for these valuable comment and suggestions, and have resumed your questions by point in the attached document.
Author Response File: Author Response.docx
Reviewer 4 Report
Comments for author File: 
Comments.pdf
Author Response
thank you for these valuable comment and suggestions, and have resumed your questions by point in the attached document.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The authors have corrected the mistakes that were mentioned in the previous review report. Therefore, I have no objection to this manuscript being published.
Reviewer 3 Report
the authors addressed my concerns very well, so I suggest to accept this MS
