The authors have made notable revisions to their manuscript, and I find it most certainly improved. I do, however, have remaining questions that derive from the revisions that I think should be addressed prior to publication.
The caption for Fig. 9 seems incorrect. I think that the blue and green are switched, as a ZSR mixing model would lead to a lower calculated GF at a given OVF when the organics are assumed to have a GF = 1 compared to a GF = 1.6.
When the authors now state “An increase in the compressed film model HGF relative to the ZSR modelled HGF is therefore due to a reduction in surface tension, not to changes in the water activity,” I’m not sure this is correct. The activity is changing because material is partitioning to the surface and therefore no longer contributing to solubility. So, the reduction in surface tension here offsets the change in water activity, I think.
P1/L9: For clarity, would be good to say “The composition of the submicron organic fraction” rather than just “the organic fraction” as the previous sentence breaks things into size ranges.
P3/L2: I suggest adding “dependent on the method used and size range investigated.”
P3/L5: I suggest changing to “based on water uptake methods, where volume mixing rules are assumed,
P3/L12: I suggest it be clarified that the volatilizable component is not the only type of organic present.
P4/L20: The authors updated this sentence, but it is now unclear to me how this is consistent with ZSR. Consistent in what manner? Above, the authors noted that the organic fraction in small SSA can be much greater than 4-17%. Also, in the list of references given few have explicitly shown that ZSR works by making independent measurements of composition and hygroscopicity, and then comparing. I suggest this be clarified.
Regarding the organic volume fraction derived by volatility, the authors note in their response that if material charred it would be assigned to the low volatility organic fraction. However, this assumes that the charred material would evaporate at all. This is not guaranteed. In OC/EC analysis, charred organics typically do not evaporate until very high temperatures. These organics would be mis-attributed to salt. This aspect could be clarified further.
P11/L5: It would be good if the authors clarified here that this is now for PM1, as the discussion to this point about volatility is for small particles.
P14/L4: Certainly the size distributions obtained are a result of the glass filters used. It is established that use of different generation methods leads to different size distributions. Perhaps the authors’ point here is that the particulars of these distributions are not a result of using four filters, but of using filters in general. Also, this seems to conflict with the authors statement on P15/L9 that the use of four filters is key to getting a broader distribution compared to studies that use different filter configurations. I suggest that this could be clarified.
P15/L20: I’ll encourage the authors to modify this to say “(generally greater than 0.4 and as high as 0.85)”
Fig. 6: There are a few errors in the revised Fig. 6. The slope of the blue lines is definitely not 1.3. It is much larger. I think this was mislabeled. Also, I highly doubt the R2 for the blue points is 0.68. I think it is closer to 0.35. I think this higher R2 reported is only obtained when the volatility-derived OVF values near zero when the filter measurements are > 0 are excluded. Similarly, I do not think the R2 value for the red points is 0.27 based on the measurements shown; it is likely lower. I’ll note that these near-zero points were not included in the previous version of Fig. 6. It appears that these near-zero points had non-zero values in the previous version. Have these been revised? Or is this a typo? It would be useful if it were clarified where these additional measurements came from. Related, all R2 values discussed on P17 and P18 should be revised accordingly if these near-zero points are correct.
P17/L8: I would argue that if the volatilities differ by hundreds of degrees (as is true for the semi-volatile and low-volatilty components) their composition is necessarily quite different. Perhaps the authors could clarify here that they seem to mean the fractional contributions of the different organic components were similar, which is not the same as saying that the composition is similar.
P19/L11: It would still be good to indicate this is for PM1.
P19/L31: The authors now note explicitly that the estimated OVF from the UFO-TDMA measurements do not correlate with either the semi-volatile or low-volatility OVF values, which is an interesting addition. Yet, they note that the UFO-derived OVF is likely associated with the semi-volatile components, and further that it must be a subset of the semi-volatile components. If this is the case, then I would think that the UFO-derived OVF values should be universally lower than the semi-volatile OVF values. However, the UFO-derived OVF values are larger (35%) than the semi-volatile OVF values (max ~15%). Can this be explained.
Fig. 9: The x-axis values seem to have changed from the previous version. It is unclear (i) why they have changed and (ii) why the x-axes now differ between panels (a) and (b). This needs to be reconciled.
P21/L23: It is still not clear that this enhancement in droplet diameter is important at much higher RH than considered here. I suggest the authors make this aspect clear.
P22/L9: It is unclear whether this organic fractions are for PM1 or for the 50 nm particles.
P23/L10: While I generally try to avoid bringing up new things during re-review, have the authors considered the sensitivity of their OCEANFILMS results to the assumed bubble thickness? They note in the methods that the assumed bubble thickness affects the resulting organic fraction. Can better agreement be obtained using a different, yet still reasonable, value for the bubble thickness? That said, the model/observation agreement is better at high OMF and worse at lower values, so changing the bubble thickness is likely to affect the intercept more than the slope, unless the authors were to allow the bubble thickness to vary with the water composition in some manner.
Compressed film model: I appreciate the updates that the authors have made. However, I still have concerns and questions over this discussion and implementation.
(i) For the ZSR comparison, the authors used the OVF values derived from the volatility measurements, specifically it seems that they used the total OVF. (Although it is not clear whether these are actually OMF or OVF…see question above regarding Fig. 9 discrepancy.) However, in the CF discussion, the authors use the OMF from the PM1 measurements, which they note on P26/L10. But why? This does not lead to a fair comparison, and makes it difficult to compare with Fig. S8, which shows the surface tension vs. the volatility-derived OVF. Also, the OVF is, presumably, used as the input to the model, and thus seems to be a better parameter to compare with.
(ii) Are the surface tensions reported in Fig. S8 the value at 90% RH? It is not clear.
(iii) The authors state that “The reduction in surface tensions at low OVFs for the lipids plus polysaccharides and all organics partitioning cases results in over prediction of the HGF” and “the error in the modelled HGF is reduced at low OVFs when only the lipids are partitioned to the surface.” It is not clear from Fig. 13 how this is the case. First, the use of absolute values of the measurement-model difference makes it not possible to tell what is an over vs. underestimate. Second, it is unclear what is being compared to what when it is stated that error is reduced. Looking at Fig. 13, the only difference between the panels is for the “functional groups” calculations for the lipid-only partitioning. The “OCEANFILMS” calculations for lipid-only are the same as for the total organic partitioning. Thus, how is error reduced comprehensively?
(iv) The authors have selected to use a more reasonable organic molecular volume now. However, they assume this is appropriate for lipids. Given a typical lipid density of 1 g/cm3, a molecular volume of 4x10^-5 implies a molecular weight of 40 g/mol. This is inconsistent with the properties of lipids. Also, is there a reason why they have chosen to only consider the upper-limit organic HGF case? The data presented do not seem to support a large proportion of organosulfate species, which are the compounds that were observed to have such large HGF values. The lipopolysaccharides, which dominate the mass here (according to the FTIR measurements) have HGF values closer to 1.25 at 90%, per the cited Estillore et al. (2017) paper.
(v) Differences are seen between the functional group and OCEANFILMS model partitioning for lipid-only partitioning at low OVF. This is not discussed explicitly; instead, the authors focus on one of these results (from the functional group analysis) as this is the one result that differs from the others. A more complete discussion should be provided.
(vi) It is not clarified in Fig. 13 that the results shown are likely for the functional group analysis, and not OCEANFILMS.
(vii) In Fig. 12 the authors show the absolute value of the measured minus modeled. At minimum, the y-axis label should be updated to indicate the absolute value has been used. However, by showing this as an absolute value, this masks the fact that at low OVF the modeled overpredicts the HGF and at higher OVF the model underpredicts. I suggest that this figure would be more meaningful if the true difference were shown, rather than the absolute value.
(viii) I am not certain that Fig. 14, as presented, “highlights the contribution of surface tension to the observed SSA HGF.” The authors show results from the functional group analysis when only lipids partition. But, this is the case where surface tension has the smallest impact, per Fig. S8, since many points are unaffected. Only the points at the largest OVF are impacted by surface tension reduction. This is the reason that the ZSR and CF model results are so similar at low OVF. In particular, since the lipids are only ~10% of the total organics, this leaves 90% to contribute to the hygroscopicity with an apparent HGF of 1.6 (per the authors’ constraints) and little influence on the surface tension. When the OVF is < 0.2, the fractional amount of organics that does not contribute to hygroscopicity is 0.2*0.1 = 0.02. At larger OVF, the majority of the organics still contribute to the hygroscopicity (since lipids are ~10%) but now the surface tension is reduced and there is slightly greater growth. The authors do note in the text that at high organic fraction the HGF is increased by 0.05 owing to the decrease in surface tension. I’ll suggest that, if the authors’ aim is to “highlight” the influence of surface tension, they add a panel that shows the difference between the ZSR and CF model results as a function of surface tension.
(ix) I find the sentence on P26/L19 starting “The role of surface tension…” is a run-on. I think a period is missing after the first clause.
(x) Now that the authors CF calculations are clearer, I will suggest that the remainder of the paragraph starting on P26/L21 with “For example…” be substantially altered. The key point it seems the authors are trying to make here is that OVF values inferred from HGF measurements might lead to underestimates in the actual OVF. I think that the authors could better make this point by simply stating this, rather than trying to make a potentially apples-to-oranges comparison between different studies that took place at different times (which they seem to acknowledge in the last sentence). As currently written, the main point, I find, gets a little lost. I’ll further encourage the authors to be semi-quantitative here. They should be able to say, at least approximately, by how much the OVF might be underestimated. The magnitude of the differences in the HGF values shown in Fig. 14 b/t the ZSR and CF calculations suggest the potential mis-application of the ZSR could lead to a differences in the OVF of nominally 5% when the OVF is large (e.g. 0.75 instead of 0.8). I do not think that the authors’ arguments or calculations would support differences of up to 40% in the OVF, as implied by the cited studies and numbers.
(xi) The authors note that the “discussion of the potential for surface tension to impact CCN concentrations has been removed. This section has been updated to discuss the potential difference between the CCN computed using ZSR modelled and measured HGFs.” However, I still have concerns regarding this discussion. The authors implicitly assume that whatever phenomenon that is leading to the measured HGF here being relatively independent of the OVF persists to the point of activation. This is not guaranteed, and the authors seem to imply from the discussion previous to the CCN discussion that surface tension is responsible for at least some of the influence. I strongly encourage the authors to restrict their discussion to what they have measured (HGF values) or clearly state the assumption regarding translation of the HGF values measured at 90% to the critical supersaturation.
(xii) I suggest the authors report the lower-limit surface tension and the C0 value used in the main text. The lower-limit surface tension is important to the results.
Minor:
P1L4: Should be “and therefore”
P7/L18: Typically, Figures are numbered in the order they are presented. Thus, this should probably be Fig. 3.
P5/L6: The Aitken mode composition was also estimated from the UFO-TDMA. This might be noted.
P12/L35: There is extraneous text where I believe a reference should be.
P19/L25: This is an incomplete sentence. |
- Printer-friendly version
- Supplement
