Source apportionment of highly time-resolved elements during a firework episode from a rural freeway site in Switzerland

Rai, Pragati; Furger, Markus; Slowik, Jay G.; Canonaco, Francesco; Fröhlich, Roman; Hüglin, Christoph; Minguillón, María Cruz; Petterson, Krag; Baltensperger, Urs; Prévôt, André S. H.

doi:https://doi.org/10.5194/acp-20-1657-2020

Articles | Volume 20, issue 3

https://doi.org/10.5194/acp-20-1657-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/acp-20-1657-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 20, issue 3

Research article

|

11 Feb 2020

Research article |

| 11 Feb 2020

Source apportionment of highly time-resolved elements during a firework episode from a rural freeway site in Switzerland

Pragati Rai, Markus Furger, Jay G. Slowik, Francesco Canonaco, Roman Fröhlich, Christoph Hüglin, María Cruz Minguillón, Krag Petterson, Urs Baltensperger, and André S. H. Prévôt

Download

Final revised paper (published on 11 Feb 2020)
Supplement to the final revised paper
Preprint (discussion started on 21 Jan 2019)
Supplement to the preprint

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'Review of MS ACP-2018-1229 Source apportionment of highly time resolved trace elements during a firework episode from a rural freeway site in Switzerland by Rai and co-authors.', Anonymous Referee #2, 12 Mar 2019
- AC1: 'Response to RC1', Pragati Rai, 05 Jul 2019
RC2: 'Source apportionment of highly time resolved trace elements during a firework episode from a rural freeway site in Switzerland', Anonymous Referee #1, 23 Apr 2019
- AC2: 'Response to RC2', Pragati Rai, 05 Jul 2019
RC3: 'Reviewer's comment to acp-2018-1229: Source apportionment of highly time resolved trace elements during a firework episode from a rural freeway site in Switzerland', Anonymous Referee #3, 24 Apr 2019
- AC3: 'Response to RC3', Pragati Rai, 05 Jul 2019

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

AR by Pragati Rai on behalf of the Authors (05 Jul 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (09 Jul 2019) by Willy Maenhaut

RR by Anonymous Referee #2 (23 Jul 2019)

Suggestions for revision or reasons for rejection

The manuscript has improved considerably and can be published if the remarks listed below are implemented.
In the methodological section it must be clarified the total apportioned variable is the sum of the 14 trace elements. This is not obvious as SA studies, including some of those cited by the authors (Liu et al, 2019; Ji et al., 2018), most commonly apportion the total PM on the basis of elements and other chemical components. Similarly, in section 4.2, the sentence “In this section the final results of the elemental PM10 source apportionment are presented and validated” must be reworded to clarify the total variable is only the sum of the 14 elements used for the analysis. The same applies to Figure 2 where it must be stated that the average concentration is the sum of the elements.
Combining measurements obtained with different size cuts (in the fine fraction) introduces an additional degree of uncertainty but also gives the chance to handle a wider range of information at once. The additional uncertainty due to using different sizes could be modelled with PMF/ME-2 However, I agree that in this case it would be challenging to combine elemental data in the fine + coarse fractions with those of other components in the fine fraction only.
There are other recent source apportionment studies using Xact 620/625 data than the six listed in Page 3 line 11 (you find some suggested papers in the references below).
The discussion about road salt is more convincing now. However, the composition of the deicing products may vary considerably between countries. It would be more appropriate to check with European rather than US data.
The characterization of the industrial source is still rather vague. The authors should better support their statement about the similarity with the profiles in the literature. Two of the cited emission sources, smelter and coal combustion process, are pretty different from each other. In Ostrava, Vossler et al. identified two point sources associated with coal combustion: a power plant (Coal Power) and a “dirty” industrial profile (Industry 1) with high content of typical biomass burning components (K and Rb).
The conclusion that traffic is the main source of PM10 elements in a traffic site is weak as outcome of a source apportionment study. I suggest putting more emphasis on the influence of other sources (fireworks, as suggested by the authors) or background levels on the elemental composition at a traffic site.
References
Sofowote U.M., Healy R.M., Su Y., Debosz J., Noble M., Munoz A., Jeong C.-H., Wang J.M., Hilker N.,. Evans G.J, Hopke P.K., 2018. Understanding the PM2.5 imbalance between a far and near-road location: Results of high temporal frequency source apportionment and parameterization of black carbon, Atmospheric Environment, Volume 173, 277-288.
Jeong, C.-H., Wang, J.M., Hilker, N., Debosz, J., Sofowote, U., Su, Y., Noble, M., Healy, R.M., Munoz, T., Dabek-Zlotorzynska, E., Celo, V., White, L., Audette, C., Herod, D., Evans, G.J., 2019. Temporal and spatial variability of traffic-related PM2.5 sources: Comparison of exhaust and non-exhaust emissions. Atmospheric Environment 198, 55–69. https://doi.org/10.1016/j.atmosenv.2018.10.038. (already cited in page 2 line 25)
Belis, C.A., Pikridas, M., Lucarelli, F., Petralia, E., Cavalli, F., Calzolai, G., Berico, M., Sciare, J., 2019. Source apportionment of fine PM by combining high time resolution organic and inorganic chemical composition datasets. Atmospheric Environment, 116816. https://doi.org/10.1016/j.atmosenv.2019.116816.

Referee Report: PDF

Hide

RR by Anonymous Referee #3 (29 Jul 2019)

Suggestions for revision or reasons for rejection

Reviewer’s comments.
The authors of the paper ACP-2018-1229 performed a huge job to revise the paper and they partially answered the concerns risen during the discussion phase. The reviewer appreciated their efforts, which gave a better explanation of data treatment helping the reader to follow the analysis. Nevertheless, in the reviewer’s opinion the procedure presented here is not fully convincing and it is still affected by a degree of subjectivity whose impact cannot be evaluated. Thus, in the reviewer’s opinion it does not meet the scientific standards for publication in Atmospheric Chemistry and Physics.
More in detail:
1) The authors define the fireworks hour FH as July 31st 21:00 to August 1st 7:00. They run PMF on such database to constrain fireworks profile. The period includes 10 hours – i.e. 10 time slots. The authors state that they performed PMF on such data. At the reviewer’s knowledge, 14 variables and 10 time slots used to resolve 5 factors by PMF should lead to scarcely reliable results due to lack of statistics (see https://source-apportionment.jrc.ec.europa.eu/Docu/european%20_guide_SA_RMs_revision_2019.pdf). Please consider that the results of the analysis are used to constrain fireworks profile in the final solution. Whether deriving constraints within a dataset is in principle possible (as proposed by Sofowote et al. 2015, http://dx.doi.org/10.1016/j.atmosenv.2015.02.045) in that case a very high number of samples was available to identify the constraints.
2) No discussion on the Q increase due to the constraints is carried out. Opposite, this should be used to support the reliability of the imposed constraints. Excessive increase of Q may indicate an excess of subjectivity in the solution (Paatero and Hopke, 2009 DOI: 10.1002/cem.1197)
3) Implemented constraints e.g. to FD and FH analyses arise from different solutions from the same dataset. It is noteworthy that 9-factor unconstrained PMF on NFD was used to constrain the sulphate factor, whereas sea-salt was constrained using the 7-factor solution from the same dataset. In the reviewer's opinion, only one solution from the same dataset should be used as reference for constraints, as the sea-salt solution could have been impacted by the increase of factors needed to derive the sulphate one. Furthermore, the authors performed also background concentration interpolation of the concentrations before and after fireworks to obtain background levels during the episode. The absence of a detailed analysis of the Q-value increases compared to FD and FH unconstrained solutions do not allow to detect the degree of subjectivity in all these assumptions.
4) The authors report scaled residuals for the sulphate source in the range [-0.45,0.85] for the unconstrained 9-factor NFD analysis: it is asymmetric and far from expected [-3,3]. This unique factor merit to be commented more in detail.
5) Performing ten pseudorandom run is quite limited to explore possible multiple minima. Especially for the chosen solutions, the reviewer would expect more runs (30-50). Moreover, it is now evident from the text that only in one case out of ten the authors were able to identify the unique factor sulphate mentioned above. Nevertheless, no comments on being the solution with the lowest Q or not. If not, what is the value compared with the lowest Q obtained in the ten pseudorandom runs?

Further comment
Mass apportionment: the reviewer is still convinced that an attempt to insert PM10 TEOM data to provide more information on PM10 source apportionment should be performed and commented.

Hide

RR by Anonymous Referee #4 (27 Aug 2019)

Suggestions for revision or reasons for rejection

Regarding the manuscript entitled “Source apportionment of highly time resolved trace elements during a firework episode from a rural freeway site in Switzerland”. The study is very interesting as the authors are using an approach that it is commonly used for the organic fraction of PM to a “portion” of the inorganic fraction of PM. The methodology that is used is thoroughly explained, the study is brilliantly written and the results are interesting. As it happens in every case that a new approach is tried, there are some fundamental assumptions that the authors need to make in order to use it. In my opinion there are some important questions to be answered before the study is published.
I understand that the authors aim to perform source apportionment analysis not on the full PM10 mass, but on the identified/used elements only. My first minor comment regarding this, is that it should be made very clear from the beginning of the manuscript and in the Abstract. In line 14 of the Abstract, the term “elemental PM10” is used to probably describe that approach. In my opinion this term is unclear and wrong. In line 17 the term “PM10 elemental mass” is used, which is much more accurate but still unclear. It must be stated clearly that the work is representative of PM10 mass that corresponds to the used variables (around 20% of the total).
My main concern and the question that rises, is even when the target is clear, is it possible to perform source apportionment analysis without major PM components? In principle you can perform SA in any type of dataset and the results will be representative of that mass. As the authors state correctly, when you do SA in organics, the results are representative for the organic mass only. The problem is that when the organics are used, all the possible (or at least most of them) organics compounds are taken into account. Here only a small fraction of the inorganics is used. My concern in this case, is not (only) the mass that it is missing when the ions and EC are not taken into account, but the tracers that are not present. For example, you can run PMF for the organics without m/z 43, 44, 55, 57 and 60 (or some of them), you still have enough mass to apportion, but you are missing key tracers that will help resolve the factors. It is the same as trying to identify traffic without EC. The authors should thoroughly discuss those limitations in the manuscript and explain why by the way the methodology is implemented the limitations are resolved. For example, in the conclusions the authors state that “We show that the rotational control available in ME-2 provides a means for treating extreme events such as fireworks within a PMF analysis”. Even though that is true, the authors need to explain that the fact they have to resort in so complicated methodology to get a good solution, is not only due to the nature of the environmental situation at hand, but because of the lack of data. I strongly believe that if the ions and EC/OC were available the situation would be much simpler. It is extremely important to give the right message to the readers, that they need to use full and complete datasets in order to perform good source apportionment analysis. The authors may have the experience and expertise to work around it, but this fact must not be used as an example.
My other fundamental concern is the use of FH PMF resolved profile for constraining the firework profile. My problem is the number of “samples” that is used to run PMF is this case (for FH). How many samples are used? Are there enough to even run PMF let alone have good statistics? It has been proved that the suggested methodology works brilliantly for a usual organic dataset where thousand of measurements are available, but I am not certain that it works just as well with 500 samples (when you split the dataset and therefore reduce the variability) and I am certain that it does not work with less than 50. One can argue that this is a PMF/CMB hybrid and in CMB you can run the model successfully with one sample in principal. The problem is that in CMB external information is used to constrain the factors (regarding the source profiles), while in this case the information is extracted by the dataset itself, which in a way recycles the uncertainty.
The naming of the of the “Road dust” and “Traffic related” factors is wrong. The authors describe the Traffic related factor as “Coarse particles from brake/disc wear…” which is the exact definition of road dust! On the other hand, they describe “Road dust” as “The sampling site is located close to the freeway and must be influenced by wear and tear of the asphalt/concrete roads because of heavy traffic.”. I strongly suggest that the “Traffic related” factor is renamed to Road dust and the factor named “road dust” should be named “road abrasion”. The are two types of traffic related sources: road dust and exhaust emissions. Since the factor that the authors name "road dust" cannot be exhaust emissions (there is no EC and authors describe it differently), it must be road dust. The naming makes much more sense that way since in this case Road dust contains actual tracers of road dust such as Zn and Cu, and “road abrasion” can justify the presence of only Ca in the factor.
“The time series of secondary sulfate exhibits peaks during the fireworks event”. Does this not mean that the factors are somewhat mixed? Secondary sulfate should not be related with fireworks. Please comment about that in the manuscript.

Hide

ED: Reconsider after major revisions (05 Sep 2019) by Willy Maenhaut

AR by Pragati Rai on behalf of the Authors (22 Oct 2019) Author's response Manuscript

ED: Publish subject to minor revisions (review by editor) (07 Nov 2019) by Willy Maenhaut

The authors have reasonably replied to the critical comments of the three Anonymous Referees and they have modified their manuscript accordingly. However, the comments given below should be addressed and many alterations are needed for the Main text and Supplement before the manuscript can be published in ACP.

For the Main text:
Page 1, line 10: Replace "were performed" by "was performed".
Page 1, line 15: Replace "brackets" by "parentheses" or alternatively by "round brackets".
Page 2, line 20: Replace "Yu, 2013" by "Yu et al., 2013".
Page 2, line 29: "a very limited" is exaggerated; I suggest replacing it by "a limited".
Page 4, line 15: Replace "balck carbon" by "black carbon".
Page 4, line 22: Replace "1μm" by "1 μm".
Page 5, line 11: Place the "a" of "a priori" also in italic.
Page 6, line 16, page 10, line 12, and page 13, line 7: Replace "e.g." by "e.g.,".
Page 7, lines 7-9: The authors write that "solutions with three to 10 factors" were examined, and then they state that "The unconstrained PMF solution resulted in mixed factors ... even for higher numbers of factors (Fig. S1)" but Fig. S1 shows a solution with only 9 factors. This is quite confusing. Rephrasing is needed here.
Page 7, line 10: Replace "is very" by "ratio is very".
Page 7, line 16: Abbreviations and acronyms (here "LT") should be defined (written full-out when first used; therefore replace "LT" here by "local time (LT)".
Page 7, line 28: Replace "Fig. S2" by "Fig. S3".
Page 7, line 32: The sentence "Although these r2 are quite similar, the solution characteristics are notably different" is unclear to me; "similar" to what and "different" from what? Rephrasing is needed.
Page 7, line 30: I presume that "r2 = 0.91" refers to the correlation of the sulfate factor with ACSM sulfate in the nine-factor solution shown in Fig. S1. If so, it is strange to call the nine-factor solution "the r2 = 0.91 solution"; it should in this case be replaced by "the nine-factor solution shown in Fig. S1".
Page 8, line 2, and page 10, line 8: Replace "i.e." by "i.e.,".
Page 8, line 5: Replace "composition yields" by "composition yield".
Page 8, line 27: Replace "fireworks factor" by "a fireworks factor".
Page 8, line 34: Replace "but same as" by "but the same as the".
Page 8, line 34, and page 9, lines 1-2: The sentence "The increase in Q is significant over the constrained five-factor FH PMF and unconstrained five-factor because its UEV for each variable is between 2 % to 16 % (average ~10 %), while is it between 1.5 % to 15 % (average ~5 %) (Fig. S5)" is unclear to me. "increase in Q" for which PMF solution?; in "and unconstrained five-factor" which PMF solution is meant here? and for which PMF solutions do the percentages apply?
Page 9, line 23: Replace "variables were" by "variables showed".
Page 10, lines 27-29, and page 32, caption of Figure 3a: The shaded areas that represent the uncertainties for the fireworks period are only visible after magnifying the Figure by 800%. Modification of the Figure and alteration in the text are needed.
Page 11, lines 22-23, and caption of Fig. 3b: It is not mentioned in the caption of Fig. 3b that the fireworks factors are excluded in the average concentration of 3134.43 ng m-3. If they are not, "(excluding the fireworks factors)" should be removed in line 23. Furthermore, there are too many significant figures (significant digits) given for that average concentration; 3 significant figures suffice.
Page 11, line 29: Replace "Fig.3a" by "Fig. 3a".
Page 12, line 9: Replace "atomic ratio" by "molecular/atomic ratio".
Page 12, line 17: Replace "in (Pongpiachan et al., 2018)" by "in Pongpiachan et al. (2018)".
Page 12, line 20: Abbreviations and acronyms (here "CEST") should be defined (written full-out when first used; since "CEST" is not used further in the manuscript, it should be replaced here by "Central European Standard Time".
Page 14, line 10: Replace "where they are" by "where the two elements are".
Page 14, lines 12-13: Replace "where the non-exhaust non-exhaust traffic-related" by "where the non-exhaust traffic-related".
Page 15, line 26: Replace "in secondary" by "in the secondary".
Page 15, line 31: Replace "during fireworks" by "during the fireworks".
Page 15, line 32: It is unclear to me what the authors mean by "because secondary sulfate formed maximum peak 1 h later"; should this text perhaps be replaced by "because the secondary sulfate formed maximum peaks 1 h later"?
Page 16, line 1: Replace "makes secondary" by "makes the secondary".
Page 16, line 16: Replace "Switzerland;" by "Switzerland,".
Page 16, line 24: Replace "of data" by "of the data".
Page 17, line 8: Replace "performed measurement. MF and RF analysed data" by "performed the measurement. MF and RF analysed the data".
Page 18, line 15: Replace "Environ." by "Environ.: X".
Page 20, line 18: Replace "Geophys. Res.," by "Geophys. Res.: Atmos., 124,".
Page 21, line 24: Replace "Meas.Tech." by "Meas. Tech.".
Page 25, line 9: Replace "of Reseach" by "of Research".
Page 26, line 24: Replace "Environ." by "Environ.: X".
Page 28, line 21: "Wang et al., 2018" should start on a new line.
Page 28, lines 27-28: There is not referred to this reference within the text.
Page 31, line 2: Replace "of PMF" by "of the finally retained 8-factor PMF".

For the Supplement:
Page 4, line 4: Replace "Paatero et al." by "Paatero".
Page 4, line 5: Replace "S1, for variable j)" by "S1) for variable j".
Page 5, line 3: Replace "11 data" by "the 11 data".
Page 5, line 4: Replace "number of" by "the number of".
Page 5, line 12: Replace "constrain fireworks" by "constrain the fireworks".
Page 12, line 3: Replace "underlain in" by "indicated in".

Hide

AR by Pragati Rai on behalf of the Authors (13 Nov 2019) Author's response Manuscript

ED: Publish subject to minor revisions (review by editor) (29 Nov 2019) by Willy Maenhaut

AR by Pragati Rai on behalf of the Authors (30 Nov 2019) Author's response Manuscript

ED: Publish subject to minor revisions (review by editor) (09 Dec 2019) by Willy Maenhaut

AR by Pragati Rai on behalf of the Authors (09 Dec 2019) Author's response Manuscript

ED: Publish as is (19 Dec 2019) by Willy Maenhaut

AR by Pragati Rai on behalf of the Authors (19 Dec 2019)

Download

Article (5449 KB)
Full-text XML

Short summary

A source apportionment study of hourly resolved elements in PM₁₀ measured at a traffic-influenced site in Härkingen, Switzerland, using positive matrix factorization (PMF) and multilinear engine-2 (ME-2) offered resolution of robust and unambiguous factor profiles and contributions. We show that the rotational control available in ME-2 provides a means for treating extreme events such as fireworks within a PMF analysis.