Next Article in Journal
Atmospheric Observation of Fluorinated Greenhouse Gases around Four Chemical Plants in China
Previous Article in Journal
Regional Tropospheric Correction Model from GNSS–Saastamoinen–GPT2w Data for Zhejiang Province
 
 
Article
Peer-Review Record

Application of a Deep Learning Fusion Model in Fine Particulate Matter Concentration Prediction

Atmosphere 2023, 14(5), 816; https://doi.org/10.3390/atmos14050816
by Xizhe Li 1, Nianyu Zou 1,* and Zhisheng Wang 2
Reviewer 1:
Reviewer 2: Anonymous
Reviewer 3: Anonymous
Atmosphere 2023, 14(5), 816; https://doi.org/10.3390/atmos14050816
Submission received: 16 March 2023 / Revised: 27 April 2023 / Accepted: 28 April 2023 / Published: 30 April 2023
(This article belongs to the Section Atmospheric Techniques, Instruments, and Modeling)

Round 1

Reviewer 1 Report

This paper presents a CNN-SSA-DBiLSTM-attention deep learning fusion model for the prediction of PM2.5 concentration. This research has important implications for human health and environmental protection. The proposed method could be inspire researchers to develop even more effective methods for atmospheric environment pollution modeling. As to the paper structure, I think it is clear and easy to follow. Apart from the strengths, I also have some suggestions about the paper:

 Q1. The technical details should be provided more clearly. How to perform the Sparrow Search Algorithm in the proposed model?

 Q2. The details of experiment are not sufficient. It is suggested that supplementing the settings of parameters such as learning rate and optimizer.

 Q3. According to the experimental results, the proposed model are better than those of the other models, but the short-term prediction effect is not good, what is the reason? And how do you plan to solve this problem in future work?

 Q4. What do the horizontal and vertical coordinates of Figure 10 represent?

 Q5. Correlation analysis is not reflected in the abstract. What are the main factors that input the model?

 Q6. Figure 7 should describe the image.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

PM10 and PM2.5 need to be in subscript. Please revise.

 

Why is the observation period only 2 years? This period is too short to draw any conclusions. Should consider 5 to 10 years of data when building the models.

 

R2 is the coefficient of determination.

 

PM2.5 has an aerodynamic diameter of <2.5 um, not in ug. Please be careful in the use of units.

 

What do you mean by one monitoring station for each meteorological factor? This sentence does not make sense. Please revise.

 

Why the units for the air pollutants are not uniform? Consider using ug/m3 for all pollutants.

 

The equations are not listed correctly. Need to be listed in the following format (eq.1), (eq.2)…

 

The text size is not uniform throughout the text. Please revise.

 

Section 3. Data Preprocessing should not be in a new section. Please follow the journal format.

 

There must be more text and explanation to justify Figure 9, or the authors should consider removing it.

 

Caption in Table 3 PM2.5 prediction (24/72) results is not clear. What do you mean by 24/72?


The same problem is in the Table 4 caption. What do you mean by WK/M? The author must be clear about what that means.

 

Why does Table 5 only show the results for Spring, Summer, and Autumn? What about Winter? This does not really make sense to me.

 

The conclusion should not be in a point form. Need to be integrated into a paragraph.

 

Overall, the novelty of this paper is average, and there are many mistakes in the text writing. Also, the observation period is too short, so the data used to build this model is not enough. The author must use more years of data to build the models. At this stage, I do not recommend that the manuscript be accepted unless a major revision is made.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Manuscript Number; Atmosphere-2304824

Title; Application of Deep Learning Fusion Model in Fine Particulate Matter Concentration Prediction

Although the topic is of interest to the Scientific community, before consideration for publication, this paper should be improved. Authors should reconsider the main objective of the paper according to the content. They should try to synthesize and emphasize the main findings of the study and avoid long sentences. Furthermore, authors should avoid drawing risky conclusions.

Evaluation; Major Revision.

1.    Abstract; The determination coefficient (R2) is stable at about 0.935, which is 2% to 6% higher than other models.

Many numeric data are given with too many significant figures; 2 significant figures suffice, and 3 suffice in case the first significant figure is "1". 

2.    Keywords; Must to revised; spelling and avoiding general and plural terms and multiple concepts (avoid, for example, 'and', 'of').

Unsuitable >>> deep learning fusion model (too long).

3.    Line 35-36; PM2.5 is a kind of fine particle with an aerodynamic diameter of <2.5 µg

Should be PM2.5 is a kind of fine particle with an aerodynamic diameter of 2.5 µm

4.    PM2.5 should be PM2.5  (subscribed). Please carefully use the word PM2.5, (subscript) in all of the main text.

5.    You must provide all the figures and tables in high resolution. Make all the labels and legends more legible. e.g. Figure 10, 11: The x and y exit information must be put.

 

6.    Conclusion; Many paragraphs are too short.  Please revise and combine them into only one paragraph in the conclusion. The conclusions could be further developed, there is a lot of interesting data in the article.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Although the authors did not give their best efforts to improve their manuscript, I do not oppose the paper being published by the editorial office.

Reviewer 3 Report

This revised version is suitable for publication.

 

Back to TopTop