Detection of Air Pollution in Urban Areas Using Monitoring Images

Round 1

Reviewer 1 Report

 

This article explores the possibility of being able to discern the quality of air and air pollution levels through visual means. After analysis and verification, the following conclusions were reached. First, traditional image quality assessment methods, especially the objective measurement index of image haze characteristics and deep learning methods, should reflect air quality to a certain extent, with the same

effect as the subjective index, even better. Second, the proposed air pollution assessment

method, i.e. fastDBCP, has a faster speed than the previous methods and has a higher

degree of correlation with air quality indices, so it can be applied to a custom air

pollution monitoring system. Third, the new metric presented must be integrated into

an actual air pollution alert scheme to warn of worsening air pollution and should

be used in fog monitoring for auto driving, tornado monitoring for

the weather forecast, or forest fire forecast for disaster alerts. So current research still has a lot of room to improve. In addition, it is expected to apply our indicators in a video surveillance system, but at present, there are few outdoor video and image quality assessment databases for foggy weather, and no air quality indicator data is provided. The article is well-written and well-reasoned with new contributions, although preliminary, that is more than suitable for publication due to the work done. Therefore, the article is accepted for publication.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The authors of the article propose a new original method for monitoring of the degree of urban air pollution – automatic analysis of photo vision of the hazy degree from a far distance. The tests of the method was carried out by the authors on photos from the available database. The comparison of the obtained results with real air pollution indexes (instrumentally measured) revealed quite satisfactory accuracy of the method (more than 80%). Some of additional advantages of the proposed method in relation to traditional instrumental methods are shown: apparently it is cheaper and simpler, has a wider spatial coverage, but especially useful the presented novel method could be in an actual air pollution warning scheme particular to warn about the exceedance of air pollution threshold.

      Unfortunately, the authors have tested the method only on a sample from the existing database, and it is not yet clear how effective the method will be in real conditions - with different cloudiness, with other atmospheric phenomena (fogs, precipitation, inversions, winds, etc.). The article would be much more interesting if the authors, in addition to tests on the database, could provide the results of real observations at least within a month. Nevertheless, the article is of considerable interest to specialists in the field of atmospheric pollution monitoring and can be published in the journal Atmosphere.

Some minor comments:

Page 20 - the abbreviation "fastDBCP" at the first mention should probably be deciphered.

Fig 3 - correlation coefficients “MOS values vs. environmental indexes” must be negative (like AQI, PCC = -0.85 and so on).

Formula 1 - is optional, it is too well known and a reference to Pearson correlation coefficients (PCC) is usually sufficient.

Author Response

Please see the attachment.

Author Response File: Author Response.docx