the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Source apportionment resolved by time-of-day for improved deconvolution of primary source contributions to air pollution
- 1Department of Mechanical Engineering, University of British Columbia, Vancouver, Canada
- 2Department of Civil Engineering, Indian Institute of Technology Delhi, New Delhi, India
- 3Department of Civil and Environmental Engineering, UC Berkeley, California, USA
- 4School of Public Health, UC Berkeley, California, USA
- 5McKetta Department of Chemical Engineering, The University of Texas at Austin, Texas, USA
- 1Department of Mechanical Engineering, University of British Columbia, Vancouver, Canada
- 2Department of Civil Engineering, Indian Institute of Technology Delhi, New Delhi, India
- 3Department of Civil and Environmental Engineering, UC Berkeley, California, USA
- 4School of Public Health, UC Berkeley, California, USA
- 5McKetta Department of Chemical Engineering, The University of Texas at Austin, Texas, USA
Abstract. Present methodologies for source apportionment assume fixed source profiles. Since meteorology and human activity patterns change seasonally and diurnally, application of source apportionment techniques to shorter rather than longer time periods generates more representative mass spectra . Here, we present a new method to conduct source apportionment resolved by time of day using the underlying approach of positive matrix factorization (PMF). We call this approach “time-of-day PMF” and statistically demonstrate the improvements in this approach over traditional PMF. We report on source apportionment conducted on four example time periods in two seasons (winter and monsoon 2017), using organic aerosol measurements from an Aerosol Chemical Speciation Monitor (ACSM). We deploy the EPA PMF tool with the underlying Multilinear Engine (ME-2) as the PMF solver. Compared to the traditional seasonal PMF approach, we extract a larger number of factors as well as PMF factors that represent the expected sources of primary organic aerosol using time-of-day PMF. By capturing diurnal time series patterns of sources at a low computational cost, time-of-day PMF can utilize large datasets collected using long-term monitoring and improve the characterization of sources of organic aerosol compared to traditional PMF approaches that do not resolve by time of day.
- Preprint
(791 KB) -
Supplement
(7451 KB) - BibTeX
- EndNote
Sahil Bhandari et al.
Status: final response (author comments only)
-
RC1: 'Comment on amt-2022-76', Anonymous Referee #2, 19 Apr 2022
The manuscript by Sahil Bhandari et al. presented a new method to conduct source apportionment, which can utilize large datasets collected using long-term monitoring compared to traditional positive matrix factorization approaches that do not resolve the diurnal pattern of factor profiles. In addition, the results showed that the new method resolved a greater diversity of factors compared to the traditional seasonal PMF approach in winter and monsoon seasons. In general, this manuscript is well written, but the following aspects should be fully addressed before it can be considered for publication.
- The authors split the data into six 4-hour time windows, and found the differences of MS and TS of OA factors between new method and traditional positive matrix factorization approaches. My major concern is that are these differences (or the characteristics of MS/TS) affected by time division? For example, what are the differences between the results in 11:00-13:00 LT, 13:00-15:00 LT and 11:00–15:00 LT? The authors need to address such uncertainties in the revised manuscript.
- More information needs to be listed to support source apportionment results in HOA and COA, as the current mass spectra appear to be confusing. What about their correlations with tracer species? In fact, the authors showed the correlations in Fig. S22-23 and S7-S8, but more discussion should be included in the main text. In addition, how about the results of 4/5/6-factor solutions?
- What is the justification for distinguishing between local OOA and regional OOA? Figs. S27-S30 did not support your conclusion in lines 412-414 in my sense.
- What are the correlations of same type of OA factors between the daytime and nighttime? It would be nice to have some comparison of MS of the same type of OA factors between daytime and nighttime. In my sense, the differences in MS between day and night in OA factors are the highlights of this paper. However, the potential differences between day and night and the reasons have not been discussed in depth.
- Zoom the legend in axis in Fig. S22-23 and S7-S8, so that the readers can see them clearly.
- Repeated descriptions: lines 216-217 and lines 159-162.
-
RC2: 'Comment on amt-2022-76', Anonymous Referee #3, 23 Jun 2022
General Comments:
The manuscript by Bhandari et al. presents an innovative application of the PMF for long term and highly time resolved datasets. The fact that various sources influence a site at specific hours throughout the day, running PMF at different time of the day appears to be a logical approach, as it also allows for more MS variability. The application of this time-of-day PMF on a long term ACSM dataset (Delhi, winter and monsoon 2017) improved the source apportionment of OA, by further separating source-specific POA compared to results obtained with standard seasonal PMF. This paper is clearly written and relatively well structured. Some minor comments need to be addressed before being accepted.
Minor Comments:
- “Results from PMF analysis for all times of the day are presented in a companion paper (Bhandari et al., 2022).” I find that at least a brief overview of the different factors observed for all time-of-day results should be described in the supplement. Indeed, the change in POA factors from non sequential time-of-day, here 11am-15pm and 23pm-3am, is possible assuming dilution, atmospheric processing, or drastic change in air masses influencing the site. While, for time-of-day following one another (e.g. 11am-3pm and 3pm-7pm), I wonder if POA factors and their concentrations show a decrease before disappearing at a later timing (e.g after 7pm). Reconstructing the diurnal profiles of all POA and SOA of all time-of-day results compared to seasonal one could support the advantage of the new approach.
- Table S2: how do you explain the differences in term of mass concentrations for OOA during W172303 even though similar factors are identified with both PMF type?
- Page 7 line 215: You mentioned that focusing on the 11am-15pm time of day “we expect to see more oxidized aerosols”. Two SOA were identified regardless of the type of PMF applied. Is the ACSM mass resolution limiting further separation or could it be that some of the seasonal SOA are identified as oxidized POA in the time-of-day PMF (e,g oxidized BBOA)?
- Page 7 line 222: “Future work should investigate the optimal length of the time window to sufficiently represent the diurnal variations in mass spectral profiles while managing computational burden”. I also think that more explanations should be provided regarding your choice of using a 4 hours window and to focus on 11-15 and 23-03.
- Adding the F44 vs F43 diagram could help segregating the different type of OA.
- Figure 3 and later: keep consistent writing of the unit “µg m-3” in text/captions/figures (main text and SI).
- Lines 389-390: change “at” to “in the afternoon”.
- I think that the different MS identified for the time-of-day PMF would add more value to the discussion and would be more useful in section 3.2.1 and 3.2.2 instead of having them in the SI.
Supplement:
- Page 3 line 32 and 35: Please, correct the reference and add punctuation “Fig. S2a–S” and “Fig. S3a–S”
- Fig S14 and Fig S15: I suggest to put in parallel day (a,c,e) and night (b,d,f) time-of-day results for easier comparison.
Sahil Bhandari et al.
Sahil Bhandari et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 288 | 165 | 9 | 462 | 24 | 6 | 5 |
- HTML: 288
- PDF: 165
- XML: 9
- Total: 462
- Supplement: 24
- BibTeX: 6
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1