egusphere-2023-1039/Journal relation: ACP     

Submitted on 17 May 2023

Title: Source Apportionment of PM_2.5 in Montréal, Canada and Health Risk Assessment for Potentially Toxic Elements

Authors: Nansi Fakhri, Robin Stevens, Arnold Downey, Konstantina Oikonomou, Jean Sciare, Charbel Afif, and Patrick L. Hayes

General:

The manuscript presents a source apportionment study in the city of Montreal, Canada and looks at associated health risks. Daily filter samples were used during a 3-month period and analysed for a comprehensive chemical composition, including a number of organic molecular markers to better identify sources. Further the study utilises a chemical transport model to identify source regions and evaluates the health risks of measured components.

In my opinion the manuscript represents a good contribution to existing literature and the topic is relevant. The scientific quality is sound, and the analysis has been performed and presented with care. The structure of the manuscript, the results and presentation are clear. Thus, I believe, the manuscript is worth publication in ACP/EGUsphere, however, I do have some comments below:

Major Comments:

1. The study took place during the Covid Pandemic in 2020 but there is no mention of what impact this may have had on the outcome of the study. Even though, from what I can find, Montreal was not in a lockdown during that period, activities will have altered and thus might have influenced local and transboundary pollution. I think this point needs to be addressed.
2. Summary needs to be clearer in what species have been used for source apportionment. The sentence starting “this source apportionment study, which examined…” (line 15) sounds like the large suit of organic markers are the chemicals used for source apportionment. This section needs reworking to be clearer.
3. Section 2.3 Enrichment factors and respective results: The enrichment factors were calculated with Al as reference element, however, later in the text (line 442, p19) there is mention of Aluminium production. Will this impact the Al concentrations within Montreal and thus is Al an appropriate reference element?
4. Section 2.5 Source apportionment: Given this is a source apportionment study as per title I think there needs to be some more information on the source apportionment: a) Uncertainties are important in PMF analysis and therefore it would be useful to know how uncertainties were calculated. I could not find these calculations in the method section or method reference Fakhri et al. (2023). Can this please be detailed, possible in the supplementary material. b) It would be good to list the species used in PMF and the percentage below LOD, maybe this can be indicated in table 3 or given in the appendix. c) Figure 5 on page 21 should include the concentration of species and % of species in each factor as both information is useful for identifying the factors. Possibly this should also include confidence intervals that should be available through bootstrap. d) in the supplement it would be useful to also display Q/Qexp. e) It would be good to have an idea of the residuals as well.

Minor Comments:

P3L71: “…elements in the PM2.5 are investigated…” “the” should be deleted

P12Figure 2 and respective section 3.1: It is not entirely clear if the concentrations given are for the period of 13/Aug-11/Nov for all years and sites or just for the year 2020; if just for the year 2020, it might be useful to only use the same period in previous years also or indicate clearer if this is not the case.

P13Section3.2L340onwards: I think it would be useful to include more information on this in the supplementary, like a figure or what EC/OC min is used and how it was derived, and also a reference of the method used.

P17L397-398: Sentence: “No correlation was found between Cu…” - is this finding confirmed by the source apportionment or is it that Cu has a more dominant source but still has a brake wear component?

P17L406-407: Sentence: “Lastly, no corelation was found between…” – this needs a reference for Zn, Pb and Sb, Cl as incinerator traces.

P18section3.5 This references the mass closure results. I think the mass closure should be mentioned in the text or even the methodology.

P20L463 onwards: the traffic exhaust factor still has some Al in it and Fe, thus Might there still be some mixing with road dust/crustal dust? Especially as the road dust has less aluminium than the traffic exhaust – see also comment about PMF in general as the factor profiles in the figure would benefit from more information.

P20L481 onwards: Similar to the previous comment, I wonder who there is some Cu, Sb, Fe in the biogenic SOA – is there still some mixing? I guess from the supplementary material it sounds like a higher solution split the factors too far, so maybe just a comment or a reference that may have experienced the same issues would be useful.

Review of: Source Apportionment of PM2.5 in Montreal, Canada and Health Risk Assessment for Potentially Toxic Elements.

This work dealt with an analysis of PM2.5 collected in a sampling campaign that lasted roughly 3 months (actually 80 days) in Montreal, a populous city in Quebec, Canada. The analysis involved factor analytical source apportionment with positive matrix factorization, use of enrichment factors, a chemical transport modelling exercise with GEOS-CHEM and a health risk assessment of components of the sampled PM.

The authors make some statements either implicitly or explicitly that can be considered as the main results/conclusions of the work:

1.    Their chemical analyses are an exhaustive characterization of PM2.5. By contrast, the analyses done by Environment Canada within the National Air Pollution Surveillance (NAPS) framework is inadequate in fully characterizing the organic species in PM2.5.

2.    The inclusion of their chosen tracers helps them identify and distinguish certain factors in their PMF analyses. Implying that these factor identifications would not have been possible/successful otherwise for the 11 factors found.

3.    Certain factors with low mass are likely more critical from their health risk analyses perspective. Thus, implying that reductions in PM2.5 mass concentrations do not necessarily translate to healthier air quality.

4.    Their GEOS-CHEM analyses results for SOA, ammonium sulphate and ‘Dust in PM2.5’ are said to show that SOA and ammonium sulphate have substantial origins in the US. ‘Other’ sources dominate the ‘Dust in PM2.5’.

This review will focus on the PMF analyses in detail, while other experts may be able to evaluate the health-risk assessment portions. Taking the source apportionment aspects all together, this study cannot be recommended for publication. There are substantial flaws in methodology and interpretation.

To start, the chemical analyses is not a complete characterization of particulate organic matter. Thus, to suggest that this study in some way improves on the NAPS method for organic PM is a stretch. There are entire compound classes of organic compounds that are missing in the proposed approach ranging the entire gamut of non-polar to polar compounds. Also, the practicality of perpetually running a chemical laboratory for exhaustive characterization of all organic compound classes for air monitoring locations across an entire country is glossed over by the authors likely due to the fact that this study is an intensive 90-day sampling campaign, where it may be possible to analyze some more compounds than the standard NAPS protocol. There is always a trade-off between the frequency of analyses and how many components can be reliably analyzed. For long term monitoring, determining all organic particulate matter components is unrealistic for analytical laboratories, even if it is feasible for short-term campaigns such as this study of 80 near-consecutive days. 

Moreover, the determination of organic carbon is sufficient to account for about or more than half the mass of particulate organic matter.  The prescribed remedy proposed by the authors wherein some organic compounds are individually determined is also flawed from a mass balance perspective since it leads to double-counting of organic carbon mass. Their remedy cannot be considered an exhaustive analysis of particulate organic matter but is neither insignificant enough to be harmless to an overestimation of organic mass.

Confidence in the PMF analysis itself is very low. The slope in Figure S2 shows that at any given time, their PMF analysis accounts for only 18% of observed PM2.5 mass. The authors have focused solely on the R2 metric but failed to realize that the sum of factors must account for 80 – 100 % of the measured mass (as seen from the slope) for the apportionment to be considered relevant. Based on this fact alone, this work should not be published.

To discuss the extra factor identities found in this work, it is always the case that the more disparate variables added in the input matrix, the more factors will be resolved. The challenge that arises though is establishing the linkage between ‘factor’ and ‘source’. There is nothing in this work that goes the extra step to establish the actual sources of these novel factors that the authors claimed would not have been found without their analytical method. No attempts were made at showing temporal trends or spatial apportionments. How can it be conclusively shown that some of these new factors do not represent factor splits? Plant wax, biogenic SOA may in fact be an overextraction of the same factor that has now been split into two separate factors.

While the GEOS-CHEM analysis is to be lauded, its value accrues when it is used in a framework that exhaustively analyzes the receptor modelling data first before a comparison can be done. PMF factor contribution results themselves are usually subjected to spatial analyses in the form of polar meteorological plots as well as air mass back trajectory analyses for both local and regional apportionments. If these were done then compared with GEOS-CHEM, more support could have been said to derive from the latter. Thus, it is hard to believe that Quebec, a province with no coal-fired power generation, is a source of more particulate secondary sulphate than the US or the rest of Canada, as seen just by relying on the GEOS-CHEM results alone. The authors are enjoined to study the use of conditional probability plots for both local (CPF) and regional (PSCF) spatial apportionments at the very minimum. For a more thorough analyses on local and regional scales, CBPF and CWT are respectively recommended.

Finally, the use of enrichment factors does not belong in contemporay source apportionment studies. Enrichment factors are flawed for incontrovertible scientific reasons, e.g., see Reimann and De Caritat. Environ. Sci. Technol. 2000, 34, 5084-5091.

In conclusion, the work in its current state cannot be recommended for publication. The PMF analyses that much of it relies on is flawed and will require a complete redesign. The authors will however have to address these issues before further consideration for publication can be contemplated.