Round 1

Reviewer 1 Report

The title of the manuscript concerning the monitoring of air pollutant emission in factories is appropriate and reflects the contents of the paper. No changes are necessary.

The abstract provides a concise but complete summary about the main objective, general picture of the methodological approach, general picture of the methodological approach, results and conclusions of the paper.

Keywords are well suited with the paper text.

The introduction provides sufficient background information for readers and adequately explains the framework and problems of the air pollution which is not a major issue only China but all over the world. The general topic, issue, and area of concern are given to illustrate the context.

In the literature review paragraph there are new and well pointed out information concerning research results in environment pollution and some solutions found for emissions mitigation.

The methods used are appropriate to the aims of the study. Sufficient information is provided for a capable researcher to reproduce the experiments described. There are no additional experiments that would greatly improve the quality of this paper.

The results are clearly explained and presented in an appropriate format. All the figures and tables are mostly easy to interpret and show essential data that could not be easily summarized in the text; however there is unnecessary duplication of data in the text.

The work is original by using Single-location recovery and Multiple-location recovery algorithm and contains new results that significantly advance the research field, and show incremental advance over prior research results.

The published literature is presented as background information and is connected to the specific findings of this study.

The conclusions are logically supported by the obtained results.

Author Response

Dear Reviewer,

Thanks for your valuable comments. We enlist all your comments in the attachment and highlight our response and actions in red. The revised part in the manuscript is marked in blue. Please see the attachment. Your patience is deeply appreciated.

Authors

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper presents three approaches for filling in missing or manipulated data on air pollution with fine dust particles (PM2.5). These three approaches are quite innovative solutions not only for dealing with missing or manipulated data. Results show a very good description of the predicted compared to the actually measured. The application of such methods would help to reduce air pollution in industrial regions where it is not possible to fully investigate the pollution or there are technical interruptions of the monitoring stations.

Author Response

Dear Reviewer,

Thanks for your valuable comments. We enlist all your comments in the attachment and highlight our response and actions in red. The revised part in the manuscript is marked in blue. Please see the attachment. Your patience is deeply appreciated.

Authors

Author Response File: Author Response.pdf

Reviewer 3 Report

Review of the paper „Multimodal Data Based Regression to Monitor Air Pollutant Emission in Factories”

The article concerns the important problem of air pollution, resulting from the construction of low-tech factories close to places where people live and work. The authors honestly admit that factory owners can distort air quality measurements in order not to carry out costly retrofit installations and productivity. It is estimated that about 10% of factories faked the monitoring data in 2015. The authors designed three ways to detect particulate matter exceedances based on other factory-reported conditions and independent air quality measurements taken around the factories.

I offer a few suggestions to help you improve the manuscript:

- abstract: “Pollutant emission data, such as air pollution…” – I propose to write “air pollution originating from anthropogenic emission”,

- chapter “2. Literature review” after “1. Introduction”: I propose to integrate the review with the “Introduction”, which currently includes only the reason for the manuscript preparation and selected indicators to complement the fake air measurements from factories. I suggest that the present chapter 2 should start “Introduction” and recommend to work on it so that the necessity for this manuscript results from the literature. In this chapter, I also propose to define the intended auditory of the manuscript,

- I am confused with “3. Motivations and methods”, it makes the paper more popular than specialized. I suggest using traditional chapters – everything related to “motivations” should be placed in “Introduction”, while “methods” should be organized as a separate chapter, in which you can present the methodology of calculating real air pollution from factories proposed for SVR, GPR and combined (in separated sections); preliminary results can be simply written in the chapter “Results”,

- lines 144–145: “Another possibility of recovering local air quality readings is to predict them from surrounding air quality stations” – what is the guarantee that these stations give reliable air quality measurements? Maybe it is worth discussing which offices control these stations and measurements, and is there a possibility of erroneous or false measurements?

- line 156: Gaussian Process Regression (GPR?),

- „4. Results and Discussion” – you don’t have to write what is in the chapter at its beginning. In general, the entire article has too much written about what will be in the next paragraphs, it is worth reading the whole manuscript and creating a more specific text,

- „5. Conclusion”: “The study introduces a novel approach to recover the missing or mislabeled air quality measurements from factory USING? indirect factors or surrounding air quality readings.” – this is not entirely true as the article is not intended to detect or complement missing measurements from factories, but the aim is to find out the truth about the air quality in the vicinity of factories that collect complete, although fake measurement,

- implementation of analysis in practice is missing; the proposed analytical tool is quite difficult – do authors think that it can be used by environmental officials, or is it supposed to serve scientists interested in air pollution from surrounding factories?

Author Response

Dear Reviewer,

Thanks for your valuable comments. We enlist all your comments in the attachment and highlight our response and actions in red. The revised part in the manuscript is marked in blue. Please see the attachment. Your patience is deeply appreciated.

Authors

Author Response File: Author Response.pdf

Reviewer 4 Report

Dear Authors,

Here below my comment about your paper.

Overall, it is an important study. The statistical issues as well as the following points should be resolved. ===================================

Overall comments

The study examined spatial interpolation using support vector regression model, and Multiple-location recovery. The paper tries to estimate the true emissions of pollutants and evaluate the exhausted gas illegally emitted by the factories in China. The problem presented and proposed solution in the paper is very interesting and actual. The paper itself is well written, although somewhat descriptive without a specific contribution from methodological point of view. The authors have conducted a thorough literature review, undertaken a rigorous piece of data collection. The presented materials and method description is clear, but the results need improvements. In fact, the authors should provide more details about the statistical test comparison between different apartments, regions as well as from time variability point of view. I think that the motivation use of some methods need to be made clearer. In a word I don't think the authors have made a strong enough case for why their methods are better than others. This paper combines a multidisciplinary scientific field is insightful, well written and interesting. 

Specific Comments

Major comments

1. I have several concerns, but the overarching one is that what is being proposed is not new. The spatial interpolation problem has been well documented in the literature. I have personally read dozens of paper about monitoring campaign results (as well as surveys, ...), and there are several reviews and cross-sectional studies. The readers need to answer to why is the proposed methods are better than others? Why isn't the problem already solved using a simple algorithm ?
2. The abstract (and/or Introduction) should include: contributions and advantages. So more details about similar studies should be provided.
3. In Figure 7, and 9, it's very difficult to understand what air quality measurements are. Is it PM2.5, .... ?
4. In Figure 8, it is not enough what the figure is, it will be necessary to detail which pollution you are talking about, on which scales, and how is it important. In what way does the gradient inform us about the pollution levels. You must go beyond a simple description, conclusions about the quality of the interpolation must be interpreted with quantitative aspects.
5. Neither SVR nor GPR is a new method for spatial interpolation. In the Introduction section, the authors identified themselves as the precursors of these methods (lines 41 –58).
6. It is really unclear to me how authors justified the choice of examining daily average patterns (aim of the study) and say that there is no seasonal changes in thermal conditions : «... almost absent in most parts of the tropical Southeast Asia»
7. As far as I can tell, no comparisons were made with other methods, and definitely no statistical comparisons. I suggest examining the predictive modeling (Ordinary Kriging, Distance weighted average, inverse distance methods, or K-nearest neighbor, ... ) as baseline model. The comparison will make this study more comprehensive on the existing literature and different methods.
8. There is a lack of specification on which pollutants the interpolation is used and how the models have been evaluated (in terms of ML techniques). MFAs were discussed without presenting the unit of measurement to apprehend the quality of the fit. These aspects are extremely important.
9. A strengthened background and discussion section would be needed for acceptance. I found the conclusions are overstated.
10. Finally, can you offer any conclusions (or a clear recommendation) based on your study to «Ministry of Environmental Protection» faced with such aspects. Can other identified risk factors be used to aid the engineer in design decision?

Minor Comments

1. I suggest moving Figure 6 after Figure 7.
2. Line 156 : Gaussian Process Regression (GP) ⇾ GPR ?

 

Comments for author File: Comments.pdf

Author Response

Dear Reviewer,

Thanks for your valuable comments. We enlist all your comments in the attachment and highlight our response and actions in red. The revised part in the manuscript is marked in blue. Please see the attachment. Your patience is deeply appreciated.

Authors

Author Response File: Author Response.pdf

Round 2

Reviewer 4 Report

Dear Authors,

I have completed the review of your second draft and can see that it is much improved from the first version.
The manuscript is easy to read and interesting. I really appreciate your answers and the addition of several sections, particularly discussion section. 
I have a few comments for you—I apologize to not get these to you earlier. Furthermore, I hope these comments are still helpful.

There are some research have been done in China as well as in E.U regarding the energy-related emissions of particulate matter. 
Thus, I think that your paper should cite these works, take for example : 

1. *-- Meng Gao, Gufran Beig, Shaojie Song, Hongliang Zhang, Jianlin Hu, Qi Ying, Fengchao Liang, Yang Liu, Haikun Wang, Xiao Lu, Tong Zhu, Gregory R. Carmichael, Chris P. Nielsen, Michael B. McElroy,
   "The impact of power generation emissions on ambient PM2.5 pollution and human health in China and India,
   Environment International,
   Volume 121, Part 1,2018,
   Pages 250-259,
   ISSN 0160-4120,
   https://doi.org/10.1016/j.envint.2018.09.015.
2. *-- Siyuan Yang, Bin Chen, Muhammad Wakeel, Tasawar Hayat, Ahmed Alsaedi, Bashir Ahmad,
   PM2.5 footprint of household energy consumption,
   Applied Energy,
   Volume 227,
   2018,
   Pages 375-383,
   ISSN 0306-2619,
   https://doi.org/10.1016/j.apenergy.2017.11.048.

I have one last question !
Consumer price subsidies for energy have indirect effects on pollution, which might be either positive or negative, depending on a number of factors, including the energy sources and the uses they target. My question how industries about adapt their consumption according to energy price and thus affect PM2.5 concentrations ? 

 

I believe you have addressed my concerns. In general, congratulations on improving your paper considerably.

Best regards

Author Response

Dear Reviewer,
 
Thanks for your valuable comments. We enlist all your comments in the attachment and highlight our response and actions in red. The revised part in the manuscript is marked in red. Please see the attachment. Your patience is deeply appreciated.
 
Authors

Author Response File: Author Response.pdf