A study on the performance of low-cost sensors for source apportionment at an urban background site
- 1Division of Environmental Health and Risk Management, School of Geography, Earth and Environmental Sciences University of Birmingham, Edgbaston, Birmingham B15 2TT, United Kingdom
- 2Department of Engineering, University of Cambridge, Trumpington Street, Cambridge, CB2 1PZ, United Kingdom
- 3Wolfson Atmospheric Chemistry Laboratories, Department of Chemistry, University of York, Heslington, York YO10 5DD, United Kingdom
- 1Division of Environmental Health and Risk Management, School of Geography, Earth and Environmental Sciences University of Birmingham, Edgbaston, Birmingham B15 2TT, United Kingdom
- 2Department of Engineering, University of Cambridge, Trumpington Street, Cambridge, CB2 1PZ, United Kingdom
- 3Wolfson Atmospheric Chemistry Laboratories, Department of Chemistry, University of York, Heslington, York YO10 5DD, United Kingdom
Abstract. While the measurements of atmospheric pollutants are useful in understanding the level of the air quality at a given area, receptor models are equally important in assessing the sources of these pollutants and the extent of their effect, helping in policy making to deal with air pollution problems. Such analyses were limited and were attempted until recently only with the use of expensive regulatory-grade instruments. In the present study we applied a two-step Positive Matrix Factorisation (PMF) receptor analysis at a background site using data acquired by low-cost sensors (LCS). Using PMF, the identification of the sources that affect the air quality at the background site in Birmingham provided results that were consistent with a previous study at the site, even though in different measuring periods, but also clearly separated the anticipated sources of particulate matter (PM) and pollution. Additionally, the method supplied a metric for the contribution of different sources to the overall air quality at the site, thus providing pollution source apportionment. The use of data from regulatory-grade (RG) instruments further confirmed the reliability of the results, as well as further clarifying the particulate matter composition and origin. Comparing the results from a previous analysis, in which a k-means clustering algorithm was used, a good consistency between the results was found, and the potential and limitations of each method when used with low-cost sensor data are highlighted. The analysis presented in this study paves the way for more extensive use of LCS for atmospheric applications and receptor modelling. Here, we present the infrastructure for understanding the factors that affect the air quality at a significantly lower cost that previously possible, thus opening up multiple new opportunities for regulatory and indicative monitoring for both scientific and industrial applications.
- Preprint
(5640 KB) -
Supplement
(318 KB) - BibTeX
- EndNote
Dimitrios Bousiotis et al.
Status: final response (author comments only)
-
RC1: 'Comment on amt-2022-84', Anonymous Referee #1, 28 Apr 2022
General comments:
This manuscript showcases an important method that can be applied to measurements from low-cost sensors for source apportionment. I recommend the following major revisions:
Specific comments
1) Some of the sentences in the Introduction are very long and should be shortened to make the manuscript clear
2) In the methods, it might be useful to have a table detailing the different instruments, their method of operation, pollutants measured as well as if they were low-cost/ reference, and location. This was not clear for some of the instruments mentioned, for example, the Box of Clustered Sensors. This section was a little hard to follow with the number of instruments mentioned but not described in detail. It also wasn't clear why indicators such as LDSA were mentioned in this section and what that had to do with source apportionment. I think including a few more details about the method and the pollutants used in the Introduction would be helpful to readers.
3) The last paragraph in section 2.1 was not about the instruments at all. I suggest moving this paragraph to the next sub-section.
3) When explaining the PMF method I suggest that the authors actually include equations to describe the two-step PMF process used in this analysis. The authors do not explain the limitations of using a combination of PNSD and particle composition, and the need to use the two-step PMF method. I think this is a critical point and needs to be elaborated on. How did this method differ from that used by Hagan et al.- the study the authors cited in the Introduction?
4) More details of the PMF method were included in the Results instead of the Methods section (eg section 3.2). This again makes it hard for the reader to follow with the authors did.
5) It appeared that without data from reference monitors, the four factors identified from the OPC data alone were hard to interpret. If so- why bother conducting a source apportionment analysis with low-cost sensors?
6) Given that the OPCs do not measure particles < 0.3 micrometers, how useful is this technique in areas dominated by vehicle emissions?
-
AC1: 'Reply on RC1', Francis Pope, 12 May 2022
The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2022-84/amt-2022-84-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Francis Pope, 12 May 2022
-
RC2: 'Comment on amt-2022-84', Anonymous Referee #2, 30 Apr 2022
The authors present a new methodology for using LCS for source apportionment. This is an important topic as being able to extract source information from LCS AQ data would immensely improve the utility and power of LCS. Overall I think the paper is adequate for publication subject to minor revisions.
Specific comments:
1. I think the paper overall, but especially the abstract, could be a little more quantitative in its description. The abstract contains several instances of describing results qualitatively (e.g., "provide results that were consistent with a previous study" line 28; "good consistency between results", line 35, etc). It would be better to provide the numbers/statistics that show this rather than just telling the reader that the results were consitent.
2. There is no discussion or citation of the performance of the Alphasense OPC-N3, which is critical in interpreting the source apportionment results. Have the authors compared the PNSDs from the Alphasense to any reference field monitors or lab instruments? The performance of these optical particle counters through publicly available resources such as AQ-SPEC is fairly mixed.
3. Line 203 mentions separate NO/NO2 LCS data. Is this from the "Box of Clustered Sensors"? It's a little unclear what devices are being used here. I have a similar concern with the quality of the data here as well, as several studies have shown that the NO2 from alphasense gas sensors are not very reliable
4. The data showing the source apportionment from the LCS alone (particles and gases) seems to be of weaker utility than when the ACSM is brought in. In particular LC4 does not really have any source condition associated with it, as the authors mention. I find the statement on line 461-462, saying that hyperlocal source apportionment is now possible with only LCS, to be exaggerating a little bit. I'd recommend softening that or at least adding in the caveats that some sources can't be well characterized. The way it is written now somewhat oversells the results, I think.
-
AC2: 'Reply on RC2', Francis Pope, 12 May 2022
The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2022-84/amt-2022-84-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Francis Pope, 12 May 2022
Dimitrios Bousiotis et al.
Dimitrios Bousiotis et al.
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
306 | 99 | 14 | 419 | 17 | 4 | 4 |
- HTML: 306
- PDF: 99
- XML: 14
- Total: 419
- Supplement: 17
- BibTeX: 4
- EndNote: 4
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1