the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Development and Evaluation of an Improved Off-Line Aerosol Mass Spectrometry Technique
Christina N. Vasilakopoulou
Kalliopi Florou
Christos Kaltsonoudis
Iasonas Stavroulas
Nikolaos Mihalopoulos
Spyros N. Pandis
Abstract. The off-line Aerosol Mass Spectrometry (AMS) technique is a useful tool for the source apportionment of organic aerosol (OA) in areas and periods during which an AMS is not available. However, the technique is based on the extraction of aerosol samples in water, while several atmospheric OA components are partially or fully insoluble in water. In this work an improved off-line technique was developed and evaluated in an effort to capture most of the partially soluble and insoluble organic aerosol material, reducing significantly the uncertainty of the corresponding source apportionment. A major advantage of the proposed approach is that no corrections are needed for the off-line analysis to account for the limited water solubility of some OA components. The improved off-line AMS analysis was tested in three campaigns: two during winter and one during summer. Collocated on-line AMS measurements were performed for the evaluation of the off-line method. Source apportionment analysis was performed separately for the on-line and the off-line measurements using Positive Matrix Factorization (PMF). The PMF results showed that the fractional contribution of each factor to the total OA differed between the on-line and the off-line PMF results by less than 15 %. The differences in the AMS spectra of the factors of the two approaches could be significant suggesting that the use of factor profiles from the literature in the off-line analysis may lead to complications. Part of the good agreement between the on-line and the off-line PMF results is due to the ability of the improved off-line AMS technique to capture a bigger part of the OA, including insoluble organic material. This was evident by the significant fraction of submicrometer suspended insoluble particles present in the water extract, and by the reduced insoluble material on the filters after the extraction process. More than half of the elemental carbon (EC) was on average missing from the filters after the water extraction. Significant EC concentrations were measured in the produced aerosol that was used as input to the AMS during the off-line analysis.
- Preprint
(1484 KB) -
Supplement
(1799 KB) - BibTeX
- EndNote
Christina N. Vasilakopoulou et al.
Status: open (until 01 Apr 2023)
-
RC1: 'Comment on amt-2023-16', Anonymous Referee #1, 10 Mar 2023
reply
This manuscript describes a method modification to off-line extraction of filter-based particle collections with re-suspension and analysis by high resolution AMS. Importantly, the method has been compared to on-line AMS (and ACSM) measurements from 3 different field studies. One limitation of the work presented here is that the off-line analysis is non-quantitative under the current method. However, the relative amount of contributing PMF components/factors shows reasonable agreement between the on-line analysis and off-line analysis.
Comments and recommended edits:
Line 31: not quite “automatically”, processing is needed. Please remove or revise.
Line 34: change “various sources” to “various sources or atmospheric transformations”
Line 34: already defined OA in line 29
Line 41: Regarding, “Its weight and size”, add “Its weight, size, and power consumption”
Line 74: include sample volume here (flow rate and sample time)
Section 2.1: does the quartz filter survive sonication? Or do they break apart? I’d imagine they break apart, in which case are there not issues with quartz fibers interfering with the atomizer stability if you don’t filter the water?
For blank measurements, has that water gone over a clean filter?
Line 93: change “variables related to CO2 “ to “variables related to CO2 and CO2-based corrections“
Line 96: used the name Thissio campaign here, but previously just referred to it as Athens campaign. Define this as Thissio earlier if that it the name you will use in the document.
Section 3: Is there any supporting information that could be provided here. Temperatures during each campaign? Maps of sample sites within the cities/regions? Plots of wind directions/speeds? Something to give the reader a better sense of each study.
Line 157: Any ideas on what is driving this difference between BBOA types?
Figure 2: this timeseries appears in low resolution on the file I accessed. Also, are the highest concentrations being cut off at the max 10 ug/m3 y-axis scale?
Also, would be helpful to see a stacked plot timeseries of the MO-OOA, LO-OOA, and HOA components here. And if showing them in this plot, the MO- and LO- categories should be described in the document when figure first introduced.
Figure 3: same points as fig. 2, except the resolution does look better for this figure in the file I have. Specify in figure description that these are results from PMF of on-line AMS. Also revise how it is stated in Figure 2. It is not on-line PMF. It is PMF of on-line AMS data.
Figure 4: Legend does not indicate which is on-line and which is off-line.
Section 4.1.2.: it would be important to make these online/offline comparisons in terms of mass concentrations, not just percentages of total. It’s my understanding that a calibrant was not used for offline samples however.
Line 183: avoid statements such as “The only minor weakness”. The truth is it could go either way, maybe it is a weakness, maybe it is a strength, unclear without quantification. Just call it a “difference”
Line 189: Regarding bootstrapping comparison amongst factors: again, would be better to see this as mass concentrations
Figure S5: This could match better on a mass basis as opposed to a % of total, had a calibrant been used.
Line 213: I could imagine a third reason. With high time resolution online AMS, there are more time periods where individual components/factors dominate over the others, producing a more defined factor profile. The offline method always has a higher contribution from all factors due to the daily integration.
It seems there are some major contributions in the offline method from m/z’s < 20 relative to the online method. You may get better matches overall if you excluded m/z <20 from your comparison. And in doing so, consider why these may be higher in the offline method (residue from water?)
Line 217: Yes, here you accounted for the difference at m/z <20….why not go back and do the same for factors in previous paragraph (shown in fig S6)
Figure S18: not OOAI and OOAII according to previous figures, MO- and LO-
Line 296: fig. S19 not S20
Line 300: fig. S20 not S21….I think this is off going forward in figures, please fix
Line 303: how did authors get O:C from online ACSM? Did it have a ToF-MS?
Line 335: What do authors mean by “filtration phase”? It was explained that the water extracts were not filtered in their process.
Line 342: Were the blanks extracted from clean/unused filters, or just looking at the water? There could be suspended particles released from a clean filter.
Citation: https://doi.org/10.5194/amt-2023-16-RC1 -
AC1: 'Response to Referee #1', Spyros Pandis, 17 Mar 2023
reply
(1) This manuscript describes a method modification to off-line extraction of filter-based particle collections with re-suspension and analysis by high resolution AMS. Importantly, the method has been compared to on-line AMS (and ACSM) measurements from 3 different field studies. One limitation of the work presented here is that the off-line analysis is non-quantitative under the current method. However, the relative amount of contributing PMF components/factors shows reasonable agreement between the on-line analysis and off-line analysis.
We appreciate the helpful suggestions and comments from the reviewer. Our responses (in regular font) and the corresponding changes in the manuscript follow each comment of the reviewer (in italics)..
Our proposed off-line method can easily become quantitative, if the corresponding samples are also analyzed for OC/EC (e.g., by thermal optical analysis). This standard analysis by can be easily performed in parts of the collected filter samples. For example, the OC/EC analysis using Sunset analyzer requires just a small punch per filter (usually 1x1 cm or 1x1.5 cm), so sample availability is not an issue. The organic aerosol (OA) concentration can then be calculated by multiplying the measured OC with the OA/OC that is determined from the high-resolution AMS measurements. The product of the factor fractions determined by our proposed method with the OA concentration will then give the mass concentrations of each factor. We have now added a discussion about this important point raised by the reviewer, in the revised paper.
(2) Line 31: not quite “automatically”, processing is needed. Please remove or revise.
The word “automatically” has been removed.
(3) Line 34: change “various sources” to “various sources or atmospheric transformations”.
The phrase “various sources” has been replaced by “various sources or atmospheric transformations”.
(4) Line 34: already defined OA in line 29.
The “organic aerosol (OA)” has been replaced by “OA”.
(5) Line 41: Regarding, “Its weight and size”, add “Its weight, size, and power consumption”
We have rephrased the sentence following the suggestion of the reviewer.
(6) Line 74: include sample volume here (flow rate and sample time).
We have added the sampling time and average sampling volume for each campaign in the revised paper.
(7) Section 2.1: does the quartz filter survive sonication? Or do they break apart? I’d imagine they break apart, in which case are there not issues with quartz fibers interfering with the atomizer stability if you don’t filter the water?
This is a good point. To address this issue we have used the Z-Sizer to measure the particles in water extracts of sonicated clean pre-baked filters and we compared them with those in which we just sonicated clean water. The results showed that no significant difference was observed in the sub-micrometer range. However, in the case of the clean filter, a small peak was observed for particles with sizes around 5 μm. This is probably due to fragments of the quartz filter. However, these larger particles do not make it into the AMS (they do not pass through its aerodynamic lens) and therefore do not affect our measurements. Their presence did not cause any problems in the operation of our atomizer for the hundreds of samples that we have analyzed so far. As a quality assurance measure, we always compare the OA mass spectra in the beginning and in the end of the off-line measurement and we have not seen any change. If something goes wrong with the atomization during a measurement, it would be probably detected this way. We also clean frequently the atomizer (every few samples) to minimize any potential contamination. This would also minimize problems due to the small fragments of the filter. A discussion of these issues has been added to the paper.
(8) For blank measurements, has that water gone over a clean filter?
We now clarify that the presented blank measurements refer to atomized clean water. We have also performed measurements for water that has gone over a clean filter using the Z-Sizer. These were the same as that of the clean water for the sub-micrometer particles that are the focus of this work. There were some larger (of the order of 5 μm) particles released from the filter, but these are not measured by the AMS so they do not affect our measurements. Please see also our response to Comment 7 above.
(9) Line 93: change “variables related to CO2” to “variables related to CO2 and CO2-based corrections”.
We have rephrased the sentence as suggested by the reviewer.
(10) Line 96: used the name Thissio campaign here, but previously just referred to it as Athens campaign. Define this as Thissio earlier if that it the name you will use in the document.
To avoid confusion with the two names we refer to the “Athens campaign” everywhere.
(11) Section 3: Is there any supporting information that could be provided here. Temperatures during each campaign? Maps of sample sites within the cities/regions? Plots of wind directions/speeds? Something to give the reader a better sense of each study.
We have added the requested supporting information, including the maps of the sampling sites and the average meteorological conditions during each campaign.
(12) Line 157: Any ideas on what is driving this difference between BBOA types?
There are a few possible explanations. One is that these factors correspond to fresh and more processed BBOA. The second is that these correspond to different wood types or combustion conditions. This issue clearly deserves additional investigation. We have added a brief discussion of these possible explanation in the revised paper.
(13) Figure 2: this times eries appears in low resolution on the file I accessed. Also, are the highest concentrations being cut off at the max 10 ug/m3 y-axis scale?
The image resolution has been improved and the scale of the y-axes has been corrected.
(14) Also, would be helpful to see a stacked plot time series of the MO-OOA, LO-OOA, and HOA components here. And if showing them in this plot, the MO- and LO- categories should be described in the document when figure first introduced.
A stacked plot time series of the different factors has been added to the supplementary material of the paper.
(15) Figure 3: same points as fig. 2, except the resolution does look better for this figure in the file I have. Specify in figure description that these are results from PMF of on-line AMS. Also revise how it is stated in Figure 2. It is not on-line PMF. It is PMF of on-line AMS data.
The description of Figure 3 has been corrected. We now clarify that these results refer the PMF analysis of on-line AMS measurements. The caption of Figure 2 has also been improved.
(16) Figure 4: Legend does not indicate which is on-line and which is off-line.
A legend has been added to indicate which is the on-line and which is the off-line. Also this information is now provided in the figure caption.
(17) Section 4.1.2.: it would be important to make these online/offline comparisons in terms of mass concentrations, not just percentages of total. It’s my understanding that a calibrant was not used for offline samples however.
Indeed, a calibrant was not used for the off-line samples in the current study. Off-line measurements of OC in the same filter samples can be used to convert the fractions calculated here to absolute factor mass concentration in future studies. This is discussed also in our response to comment 1 above. A paragraph has been added to the revised paper to explain this suggested approach for future applications.
(18) Line 183: avoid statements such as “The only minor weakness”. The truth is it could go either way, maybe it is a weakness, maybe it is a strength, unclear without quantification. Just call it a “difference”.
We have followed the suggestion of the reviewer and the phrase “the only minor weakness” has been replaced with “the difference”.
(19) Line 189: Regarding bootstrapping comparison amongst factors: again, would be better to see this as mass concentrations.
We have added the corresponding information (absolute concentrations) next to the percentages. The absolute concentrations were calculated multiplying the results of the bootstrap analysis with the absolute concentrations of the on-line AMS measurements.
(20) Figure S5: This could match better on a mass basis as opposed to a % of total, had a calibrant been used.
We have added in the same figure graphs with the absolute concentrations using as a basis the absolute concentrations of the on-line AMS measurements.
(21) Line 213: I could imagine a third reason. With high time resolution online AMS, there are more time periods where individual components/factors dominate over the others, producing a more defined factor profile. The offline method always has a higher contribution from all factors due to the daily integration.
We agree with the reviewer, but this is part of our first explanation and it was analyzed in more detail Vasilakopoulou et al. (2022). We have added a couple of sentences here to clarify the point made by the reviewer.
(22) It seems there are some major contributions in the offline method from m/z’s < 20 relative to the online method. You may get better matches overall if you excluded m/z <20 from your comparison. And in doing so, consider why these may be higher in the offline method (residue from water?).
The reviewer is correct; the water appears to contribute to these differences in the low m/z values. For example, the signal at m/z 18 is higher in all off-line spectra compared to the on-line. We have added a brief discussion of this issue in the revised paper.
(23) Line 217: Yes, here you accounted for the difference at m/z <20…. why not go back and do the same for factors in previous paragraph (shown in fig S6).
We have followed the suggestion of the reviewer and we have compared the PMF factors between the on-line and the off-line PMF results without accounting for the m/z <20. These comparisons are now discussed in the revised paper.
(24) Figure S18: not OOAI and OOAII according to previous figures, MO- and LO-.
The legend in the figure has been corrected, and the OOA I and OOA II have been replaced by MO-OOA and LO-OOA.
(25) Line 296: fig. S19 not S20
The figure number has been corrected.
(26) Line 300: fig. S20 not S21….I think this is off going forward in figures, please fix.
All the figure numbers have been corrected.
(27) Line 303: how did authors get O:C from online ACSM? Did it have a ToF-MS?
In order to estimate the O:C from the on-line ACSM measurements we used the approach of Canagaratna et al. (2015) for the unit-mass resolution data. The O:C estimates are clearly more uncertain in this case. This clarification has been added to the paper.
(28) Line 335: What do authors mean by “filtration phase”? It was explained that the water extracts were not filtered in their process.
We now clarify in the paper that filtration (1 μm pore size) was used only for these Z-Sizer measurements not for the off-line AMS analysis.
(29) Line 342: Were the blanks extracted from clean/unused filters, or just looking at the water? There could be suspended particles released from a clean filter.
In this case only the water was used as blank. We have also performed measurements for water that has gone over a clean filter using the Z-Sizer. These were the same as that of the clean water for the sub-micrometer particles that are the focus of this work. There were some larger (of the order of 5 μm) particles released from the filter, but these are not measured by the AMS so they do not affect our measurements. Please see also our response to Comment 7 above. This information has been added to the paper.
Citation: https://doi.org/10.5194/amt-2023-16-AC1
-
AC1: 'Response to Referee #1', Spyros Pandis, 17 Mar 2023
reply
-
RC2: 'Comment on amt-2023-16', Anonymous Referee #2, 19 Mar 2023
reply
This manuscript offers an improved protocol to measure aerosol source apportionment that can be applied to many more areas where online AMS measurement is challenging. The authors provide comprehensive comparison between this improved offline method with corresponding online method very clearly.
However, more emphasis should be focused on the discussion of the difference, instead of only showing results: in Section 4, multiple paragraphs only report different values and do not report why such variations exist. Meanwhile, it would be helpful if cross-comparison among different campaigns is expanded upon the brief discussion in Section 6. For example, why does summer measurement agree better in offline vs online than winter measurement? Why secondary OOA factors agrees better than HOA and COA?
On the other hand, there is an absence of evidence on how this improved method out-performs the original method proposed in Daellenbach et al. 2016. It would be helpful if the authors can provide additional evidence and explicit explanation on the results of the original method and improved one.
Line 65: Please provide reasoning behind “significant differences observed in factor spectra” and why the differences matter in the context of offline AMS PMF.
Line 79-80: “The AMS measures particles smaller than 1 μm. If there are larger particles (e.g. fragments of the quartz filter) in the produced aerosol these do not make it through the AMS aerodynamic lenses” change to “The AMS measures particles smaller than 1 μm because larger larger particles (e.g. fragments of the quartz filter) in atomization are unable to pass through the AMS aerodynamic lenses”. Formality of language.
Line 92-93: Please provide additional information on how you define signals to be weak or bad.
Line 128: Please be consistent with terminology. PM1 mass concentration in Fig 1 is described as OA concentrations. Either change Fig 1(a) y axis to PM1 mass concentration or change Line 128 to OA concentrations.
Line 156 -Line 160: Why separate BBOA into 2 factors and what does each factor represent in the context of BBOA?
Line 183- Line 188: When not providing the meaning of BBOA I and BBOA II, it’s hard to understand why a difference of ~15% matters.
Line 197-200: Please provide additional information on why the high uncertainty comes from temporal resolution: in Vasilakopoulou et al 2022, it was explained that high uncertainty is a result of challenging separation in low temporal resolution. If this paper holds the same statement, please add it to the explanation.
In Vasilakopoulou et al 2022, it was shown in high concentration days, each factor has good agreement between 24 hour filter and 30 min filter temporal resolutions, meaning we are not observing a large difference in offline analysis of various temporal resolution. However, this paper shows 24 hour filter still has a large discrepancy when comparing with online 3 min measurement. Does it imply offline measurement in general unable to provide information as accurate as online measurement?
Line 208: “Make it to the AMS” should be changed to “detectable by AMS” to be more formal in language use.
Line 213-216: The difference in mass spec is an interesting observation worth exploring more in detail. For example, Figure S6 suggests offline method tends to underestimate m/z higher than 40 whereas overestimate m/z lower than 40. Is there any scientific hypothesis/publication that implies lower m/z offers higher extraction efficiency, etc. to further expand this paragraph of explanation?
Line 220: Following up on the previous comment, why should we exclude m/z values below 20, other than it shows a large variation between offline and online measurement method?
Line 230: Please provide additional information on why we should compare organics to sulfate ratio and what does it imply.
Line 270-278: It is good to observe better agreement in this summer campaign as compared to the winter campaign. But can you please add additional scientific discussion on why better agreement is obtained in HOA markers in summer measurement?
Line 304-305: Can you please provide O:C information of MO-OOA and LO-OOA factors in the offline analysis? In Vasilakopoulou et al. 2022, LO-OOA always have a lower O:C compared to that of MO-OOA. Such observation in this paper seems odd. If MO-OOA and LO-OOA starts to mix in offline analysis, what is the rationale behind separating the secondary factors into two factors in the first place?
Line 306-311: I would hope to see more in-depth discussion on why distribution differs between factors across different campaigns, or if such differences in within uncertainty and acceptable in authors’ opinions.
Line 310: “LO-OA” should be “LO-OOA”
Line 326-327: The insolubility of HOA and COA is an impressive point to bring up but at the same time, HOA and COA has relatively larger variation in the 3 campaign measurements, compared to other factors. Could this insoluble characteristic explain such large variation?
Section 5.2: It is unclear how this section helps to explain the overhead hypothesis of section 5. I think the authors are trying to imply that some insoluble materials are sent and detected to AMS in measurement step, but it was not emphasized how the results support the hypothesis in this section.
Line 353-354: How do we know most of the insoluble materials are BC? Or why is BC representative of the insoluble materials?
Figure 4: (1) Please provide legends for which bar is online and offline results. (2) This figure will offer more useful information if uncertainty from figure S4 can be incorporated here.
Figure 6: I suggest adjusting this figure to the style of Figure 9 where state which plot is CxHy or CxHyO directly on the figure.
Figure 8: Can we add uncertainty information onto this figure?
Figure 9: Inconsistency in figure styles compared to Figure 6, can add the ion fragment formulae to x-axis as Figure 6.
Fig S5: This figure can better aid your explanation in Line 197-200 if it can be switched to a diurnal time series plot for each factor comparing online vs offline.
Figure S8: This figure will be more helpful if you add the total OA concentrations of the 2 days onto (a) and (b) instead of only showing the normalized values.
Figure S13, S14, S20: Figures would be clearer if each factor is color coded as in Figure S6
Figure S18: Should OOA I and OOA II correspond to LO-OOA and MO-OOA as Figure S17? Please be consistent with terminology.
Citation: https://doi.org/10.5194/amt-2023-16-RC2 -
RC3: 'Comment on amt-2023-16', Anonymous Referee #3, 19 Mar 2023
reply
Review of “Development and Evaluation of an Improved Off-line Aerosol Mass Spectrometry Technique” by Vasilakopoulou et al.
General Comments: In this manuscript, the authors present results from the comparison of three campaigns between offline and online AMS measurements. They find good agreement with the overall PMF characterizations between both methods and suggest that this improvement is due to an increase in the extraction and nebulization of the OA material to include more insoluble material. The paper is well written, and the figures are clear. Overall, the work is a good demonstration of an improved technique. However, there are numerous points where more explanation or clarity is needed. I recommend accepting this work after the following concerns are addressed.
Specific comments.
- Please provide more information on the blanking methods that were used throughout the experiments.
(a) On line 81, the blanking looks to be ultra-pure H2O. Was this used to extract blank filters, or is it just a solvent blank? Were any experiments run to test the background with an internal standard, like a labeled salt, included? If the total concentration of the material in the solution is low, the dried particles will be too small to enter the AMS. However, there may still be organic material that is observable when a salt is included to increase the concentration.
(b) What are the blank spectra for the different campaigns? Were these collected off blank filters? How were the blank filters collected and handled? How many blank filters were collected and run and how much variation was observed?
(c) In the paragraph under 2.3 Off-line source apportionment the subtraction of blanks is mentioned. How was blank subtraction carried out for the offline analysis? Was this done prior to PMF analysis?
(d) In section 5.1, is the blank that is used for comparison an extraction of a blank filter?
- In section 4.1.1 two BBOA factors were observed, but the mass spectra and time series are very similar. Why was a solution with two BBOA selected?
- In all the diurnal patterns in the supplemental, what was the variation throughout the day? I often see this as a shaded region. Please include something like this or another way to communicate the variability.
- The numbering for the figures is confusing because Figures 2 and 3 are not discussed until much later in the manuscript. I recommend renumbering the figures so that they are listed in the order they are mentioned in the text.
- In Figure 4 there is no label for offline vs. online, please add that. Also, bootstrapping analysis is carried out for all the offline work. Can you include these results as error bars on the figures?
- On line 200 it is mentioned that the uncertainty on a daily basis can be really high and that this is due to the temporal resolution of the offline measurements with a citation. Please provide more text here to describe what is meant by this. Also, this suggests that the uncertainty on a daily level might decrease if the same time resolution is used for the online measurements. Essentially downgrading the time resolution. Was something like this done? That would be a more direct comparison for this portion of the analysis.
- On line 215 a few different reasons for the differences are listed. What was the particle size range that was collected for each filter? Were cyclones included and if so, what was the cut-point? What was the photochemistry like during the collection? Might there be some additional aging of the material on the filter during collection? What about volatilization of semi-volatile organic compounds off the filter? These are a few other reasons I can think of that might contribute to the differences observed between the offline and online measurements and I would recommend including at least some of them in the discussion.
- On lines 221-223 there is a discussion of the offline method capturing 64% of the CxHy+ and 82% of the CxHyO+. I do not understand what is being communicated here (and in the later portions of this text where this same analysis is carried out). What does it mean that these percentages are being captured?
- In section 4.2.1 the difference between the MO-OOA and LO-OOA factors is noted to be small and the spectra look very similar. For the offline analysis however, they are larger. How does this compare to the differences that are observed in other campaigns between these secondary factors?
- There are some errors in the numbering of the supplemental figures in section 4.3.2, please correct that.
Citation: https://doi.org/10.5194/amt-2023-16-RC3
Christina N. Vasilakopoulou et al.
Christina N. Vasilakopoulou et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 310 | 97 | 14 | 421 | 38 | 3 | 6 |
- HTML: 310
- PDF: 97
- XML: 14
- Total: 421
- Supplement: 38
- BibTeX: 3
- EndNote: 6
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1