The hygroscopicity parameter (κ) of ambient organic aerosol at a field site subject to biogenic and anthropogenic influences: relationship to degree of aerosol oxidation

Cloud condensation nuclei (CCN) concentrations were measured at Egbert, a rural site in Ontario, Canada during the spring of 2007. The CCN concentrations were compared to values predicted from the aerosol chemical composition and size distribution using κ-Köhler theory, with the specific goal of this work being to determine the hygroscopic parameter ( κ) of the oxygenated organic component of the aerosol, assuming that oxygenation drives the hygroscopicity for the entire organic fraction of the aerosol. The hygroscopicity of the oxygenated fraction of the organic component, as determined by an Aerodyne aerosol mass spectrometer (AMS), was characterised by two methods. First, positive matrix factorization (PMF) was used to separate oxygenated and unoxygenated organic aerosol factors. By assuming that the unoxygenated factor is completely non-hygroscopic and by varyingκ of the oxygenated factor so that the predicted and measured CCN concentrations are internally consistent and in good agreement, κ of the oxygenated organic factor was found to be 0.22 ±0.04 for the suite of measurements made during this five-week campaign. In a second, equivalent approach, we continue to assume that the unoxygenated component of the aerosol, with a mole ratio of atomic oxygen to atomic carbon (O/C) ≈ 0, is completely non-hygroscopic, and we postulate a simple linear relationship between κorg and O/C. Under these assumptions, the κ of the entire organic component for bulk aerosols measured by the AMS can be parameterised as κorg=(0.29±0.05)·(O/C), for the range of Correspondence to: R. Y.-W. Chang (rchang@chem.utoronto.ca) O/C observed in this study (0.3 to 0.6). These results are averaged over our five-week study at one location using only the AMS for composition analysis. Empirically, our measurements are consistent with κorg generally increasing with increasing particle oxygenation, but high uncertainties preclude us from testing this hypothesis. Lastly, we examine select periods of different aerosol composition, corresponding to different air mass histories, to determine the generality of the campaign-wide findings described above.


etc.
We thank the reviewer for his helpful comments.We have extensively re-written the paper to make the analysis techniques, and in particular the associated assumptions, clearer.
In particular, we fully agree that we have not demonstrated (in the original manuscript) that κ org correlates with the degree of oxygenation of the aerosol, whether with the O/C or with the oxygenated factor.Although we thought the assumptions that we made in the analysis were clear in the original manuscript, we have now made these statements explicitly in the Abstract and a number of other parts of the paper to avoid confusion.
To briefly summarize, we make an assumption that the hygroscopicity of the organic component varies with its degree of oxygenation.Specifically, we assume that the oxygenated factors contribute to the hygroscopicity of the aerosol, and that the unoxygenated factor does not, whether due to its solubility, as suggested by Petters et al. (2009), or its intrinsic hygroscopicity.Under this assumption, we then determine how our field data inform us about the hygroscopicity of the oxygenated fraction.
Although attempting to fit for the κ of both the oxygenated and unoxygenated components would have been the ideal approach, the uncertainties in our measurements were too great to evaluate the difference between such sets of κs.As such, we constrained our model by assuming that the unoxygenated (HOA) component of the organic aerosol contributed minimally to the aerosol hygroscopicity.This is a reasonable assumption because HOA is thought to be hydrocarbon-like (Zhang et al., 2005) and aliphatic in nature, and in general, these types of compounds have not been found to be CCN-active in laboratory experiments (e.g.Raymond and Pandis, 2002;Pradeep Kumar et al., 2003).In addition, ambient measurements of particles at locations in which HOA is high, especially at the smaller sizes, have shown that the organic component does not appear to contribute to the CCN-activity (Cubison et al., 2008;Quinn et al., C12058 2008).Based on these published results, we felt that the assumption that the unoxygenated component was non-hygroscopic was reasonable.
In contrast to the HOA factor, the OOA factor, which makes up most of the oxygenated component of the organic aerosol, is thought to be characteristic of highly processed organic aerosols (Zhang et al., 2005) formed from secondary processes (Zhang et al., 2007).Again, laboratory experiments of compounds that are formed from secondary processes such as glutamic, glutaric, pinonic and norpinic acids have found that these aerosols are moderately CCN-active (Raymond and Pandis, 2002) and smog chamber studies of gaseous organic precursors with ozone or OH have also found that the organic aerosols formed are moderately CCN-active (e.g.Duplissy et al., 2008).These findings support the idea that the oxygenated component of the organic aerosol is CCN-active and that it is not unreasonable to assume that it contributes to the overall aerosol hygroscopicity.As such, we believe that assuming that the unoxygenated component is non-hygroscopic and that the oxygenated component is hygroscopic is consistent with what is known from the literature.Under this assumption, the goal of the paper is to extract only one piece of information, i.e. the hygroscopicity of the oxygenated component of the organic aerosol.This discussion is elaborated upon greatly in the revised manuscript (Sect.3.3).The point the Reviewer is making here is similar to his first, i.e. in the original manuscript he is stating that we do not show that a correlation exists between κ org and O/C.We agree.As stated in the original version of the paper, the functional form of Eq. 5 and 6 is assumed.
To address this point, we now include a graph (new Fig. 4) of inferred κ org vs O/C in Sect.4.1.2in the revised manuscript, where we have averaged κ org values for different O/C ratios.The scatter in this plot was the reason that we originally did not include it in the manuscript, i.e. we are hesitant to derive a functional relationship from these data preferring, instead, to propose Eq. 5 based on chemical intuition (i.e. that the κ org value would converge to zero for O/C of zero and to higher values for higher O/C). Figure 4 illustrates that the uncertainties are quite large for the average inferred κ org .Nevertheless, disregarding the uncertainties for the time being, there is a positive correlation between κ org and O/C over our relatively narrow range of measurements.As noted in the response to Reviewer 3, the data points in Fig. 4 are quite similar to those of Jimenez et al. (2009) for the κ org values derived from both lab and field data for subsaturated conditions.We also include in the paper the functional form of the correlation that is presented in Fig. 4.However, given the uncertainties involved, we do not consider our data to be a strong test of the inherent assumptions in our analysis, i.e. that the oxygenated component of the organic aerosol drives the hygroscopicity in a linear manner.Furthermore, in Sect.4.4 in the revised text, we point out that if a constant κ org of 0.15 is used in our analysis, the organic-rich particles tend to be slightly overpredicted compared to the inorganic-rich particles.
Section 4.1.2,Table 1 and Fig Although only representing 11% of the full study, these time periods to which the reviewer is referring do exhibit different CCN behaviour than for the rest of the periods examined.So the reviewer has a fair comment.Although present in the original manuscript, we now stress in the Conclusions that not all time periods adhere to the average results for the entire study.

C12062
We agree that the LV-OOA and SV-OOA terminology is more meaningful in terms of understanding the properties of the organic component.However, tracer correlations suggest that the time series in this study are primarily driven by factors other than volatility, despite the fact that the OOA-1 and OOA-2 factor mass spectra are similar to those of LV-OOA and SV-OOA, respectively (Slowik et al., 2010).As such, we continue to use the more general OOA-1/OOA-2 terminology in the manuscript, but we include the newer terms after each entry, e.g.OOA-1 (LV-OOA).This also allows this manuscript to be consistent with the previously-published papers from this study.This is now mentioned in the paper (Sect.2.4).
p. 25332, lines 14-19: Is the simplification to approximate ammonium nitrate by ammonium sulphate needed.The calculations would hardly become more complex with distinguishing them.Furthermore, if this simplification is maintained, then the associated uncertainty in the derived kappa of the organic aerosol fraction has to be added to the (nice) sensitivity analysis presented in Section 4.3).Furthermore, please make sure that the kappa value of ammonium nitrate at the point of activation at SS=0.42% is indeed 0.67 as stated.
The κ for ammonium nitrate is in fact 0.72 at 0.42% supersaturation (thermodynamic data from Clegg et al. (1998); Wexler and Clegg (2002)).This was an oversight on our part.It is true that calculating ammonium nitrate separately is not complicated, however, we felt that it was easier to simplify the inorganic component.The uncertainty associated with this assumption is not great and has been added to the Uncertainties section.This statement refers to the reported increase of κ from 0.09 to 0.12 in paragraph 6 and the open symbols in the middle panel of Fig. 2 of the Duplissy et al. (2008) paper.While it is true that Petters et al. (2009) discuss the difference between kappa at super-and subsaturated conditions, their companion paper, Wex et al. (2009) did measure a lower κ under supersaturated conditions for their SOA generated in the presence of butanol (an OH scavenger) compared to experiments without butanol, which were inferred in their paper to be more oxidized (refer to Fig. 8 of their paper).
This paragraph has been moved to the new section in the revised manuscript that discusses the relationship between degree of oxygenation and hygroscopicity (Sect.

3.3) and now reads:
"This is supported by laboratory studies that have found that oleic acid aerosols that have undergone ozonolysis have become somewhat more CCN-activity (Broekhuizen et al., 2004;Shilling et al., 2007) and chamber studies that show that ageing organic aerosols can also increase their CCN-activity to varying amounts (Duplissy et al., 2008;Petters et al., 2006;Wex et al., 2009).We do note that there is also some evidence C12064 that the hygroscopicity of secondary organic aerosols under supersaturated conditions in chamber studies do not vary with ageing (Prenni et al., 2007;Jurányi et al., 2009)."This sentence has been changed to: "Under these assumptions, there is the expectation that the aerosol's hygroscopicity will increase with the degree of oxygenation, although this relationship would intrinsically arise from factors such as molecular weight or solubility, that would affect hygroscopicity and be related to the oxygenation of the organic components." and it is further discussed in the new Sect.The organic component has been made more consistent throughout the paper and is now treated as two components, oxygenated and unoxygenated, for all the calculations including the density.The oxygenated component, in particular the OOA component, is thought to be similar to secondary organic aerosol (SOA).As such, a density of 1500 kg m −3 (measured for monoterpene SOA by Kostenidou et al. (2007)) was used for the oxygenated component.On the other hand, the unoxygenated component, i.e.HOA, is thought to be composed of, among other components, lubricating oil, which has a density of 900 kg m −3 (e.g.Cylinder Lube 1000 from NOCO Energy Corp.).As such, this was the density used for the unoxygenated component.In regards to the solubility of the oxygenated component, increasing solubility by ten times does not affect our final results, suggesting that at activation, it has completely dissolved.p. 25336, lines 4-7: An average ratio (R_CCN) of 1.20 between predicted and measured CCN concentration seems inconsistent with a fitted slope of 1.02 between the same two quantities.First, is the distribution of individual R_CCN values rather normal or lognormal?In the latter case it is better to report the geometric mean instead of the arithmetic mean of the R_CCN values.Second, Fig. 2 shows that the R_CCN are biases towards values >1 mainly at medium and low CCN concentrations, while they are near unity at the highest CCN concentrations.Fitting a line with axis intercept at the origin is most sensitive to the highest CCN concentrations unless a weighting is introduced.
The distribution was indeed lognormal, and the geometric mean was lower than the arithmetic mean (1.18 and 1.25 for the organic-and inorganic-rich particles, respectively).However, it is still not enough to account for the difference.In an attempt to reduce the weighting of the higher data points, the logarithm of the data points were taken and then linearly fit, but the slope only increased slightly to 1.03.If the data points are weighted by their estimated error (20% based on the uncertainties of the measured CCN concentrations) and the intercept set to zero, then the slope can C12066 be increased to 1.10 (1.05 if the intercept is not forced through zero), which halves the discrepancy with the value obtained from R CCN , but still does not explain all of it.We agree that these values are inconsistent but we cannot reconcile them.Any suggestions that you may have would be welcome.This discussion has been included in Sect.4.1.1 of the revised manuscript."Results from laboratory studies are inconclusive about the relationship between organic oxygenation and hygroscopicity under supersaturated conditions.Unoxygenated organic aerosols that are non-hygroscopic such as oleic acid, stearic acid and bis-2-ethylhexyl sebacate, can be heterogeneously oxidised such that they become moderately hygroscopic (Petters et al., 2006;Broekhuizen et al., 2004;Shilling et al., 2007;George et al., 2009).A smog chamber study by Wex et al. (2009) also showed that the hygroscopicity of SOA increased in the absence of an OH scavenger, which the authors suggested was due to increased oxidation in the organic aerosol.Finally, Duplissy et al. (2008) also showed that the hygroscopicity increased slightly throughout the photoxidation of α-pinene.However, there have also been several studies of SOA in smog chambers that have not observed any dependence of hygroscopicity on degree of oxygenation, whether by varying precursor concentrations that should result in varying oxygenation in the final aerosol (Jurányi et al., 2009), or by monitoring the hygroscopicity throughout a reaction (Prenni et al., 2007).As such, the dependence of hygroscopicity on the degree of oxygenation from laboratory studies is unclear.However, the applicability of these lab experiments to ambient conditions has not been fully confirmed, given that chamber-generated SOA tend to be less oxygenated than the OOA-1 (LV-OOA) component found in highly processed ambient aerosols (Ng et al., 2009).Although the focus of this work is on supersaturated conditions, and it is unclear whether such κ parameters are fully the same as those derived from subsaturated conditions, we do note that κ for organic aerosols measured under subsaturated conditions have recently been reported to scale with the O/C of the aerosol (Jimenez et al., 2009)."p. 25343, lines 1-4: As detailed above I do not at all agree with this summary statement.

C12068
This statement has been deleted.p. 25343, conclusions, lines 11-12: "most of the specific time periods for the study".Call 2 out of 3 "most" is audacious (see also corresponding major comment).
The wording has been changed in the Conclusions.

Technical corrections:
p. 25332, line 2: What do you mean with "initially"?Isn't it simpler to directly state that you assume surface tension of pure water in all closure calculations.
We agree.The "initially" was extraneous and has been removed.
it does not give evidence of a correlation between O/C and kappa_org, similar to the previous comment.Starting of with the presumption that kappa_org correlates with O/C (i.e.assuming kappa_org = a * (O/C)) will always result in a positive value for a, if the campaign mean kappa_org is larger than zero.With the "opposite" assumption that kappa_org = b*(C/O), one would end up with something like b ∼ 1/a I guess, which is also a positive value.The key thing to be shown with the data set of this study is not that the assumption kappa_org = a * (O/C) performs better than the assumption kappa_org=0.Instead it has to be shown that kappa_org = a * (O/C) gives significantly better CCN predictions for the individual data points compared to alternative assumptions such as kappa_org=constant (optimized of course) or kappa_org = b * (C/O).
. 5: Deriving kappa_ox and kappa_unox separately C12060 for different time periods is a step towards testing the generality of the "postulated" relationship kappa_org = eps_ox*0.20.However, my interpretation of the results is opposite from the author's interpretation.The fact that the retrieved kappa_ox is massively different for one out of just three periods rather indicates that kappa_org is mainly controlled by other factors than the just ratio of HOA to OOA, and that the above relationship cannot be recommended for general use, as it is put in abstract and conclusions.
p. 25339, line 24 -p.25340, line 6: This whole paragraph including Table2is pure speculation and has to be removed.The kappa values of the different PMF factors were not independently derived, they are just calculated under the assumption that the relationship inferred between k_org and O/C is generally applicable.Above I argued that validity of this relationship has not been proven and the authors themselves state that "The oxygenated factors were grouped together because our model was not sensitive enough to elucidate separate hygroscopicities for each factor."on p. 25333 of their manuscript.We have removed this paragraph and the associated parts of the Table from the paper.Minor comments p.25327, lines 2-5: In the mean time Jimenez et al. (2009) have published a paper C12061 in which the O/C of the organic aerosol component of laboratory generated and ambient aerosols is linked to the hygroscopicity at subsaturated relative humidity (RH).Another study by Raatikainen et al. (2009) investigates the link between different AMS-PMF-derived OA subfractions and hygroscopicity at subsaturated RH, volatility, etc. References to these papers have been made in the revised manuscript in the Introduction.p. 25329, lines 3-7: Please cite the source of thermodynamic data used for calculating the CCN properties of calibration substance ammonium sulfate.Did you assume a kappa value of 0.61 as mentioned later in the manuscript for the closure calculations?The kappa value of pure ammonium sulfate at the point of CCN activation depends on the supersaturation (SS).It changes from 0.65 at SS=0.1% to 0.55 at SS=1.0% and it is 0.59 at SS=0.42.Thank you for bringing this to our attention.A value of 0.59, as calculated from activity coefficients fromClegg et al. (1996) at 0.42% supersaturation, is now used and carried through in the analysis.This is now explained in the Sect.3.1 of the revised manuscript.[Note: The original calculation used 0.61 both to calibrate the effective supersaturation of the chamber and for the data analysis, so the results do not significantly change by using the (correct) value of 0.59.]p. 25330, lines 12-20: I recommend using the terminology of Jimenez et al. (2009) for the PMF factors."low volatility OOA" and "semi-volatile OOA" are more meaningful terms than OOA-1 and OOA-2.

p. 25333 ,
line 11: Duplissy et al. (2008) have shown that O/C has a strong influence on hygroscopicity of alpha-pinene SOA at RH=95%.The effect of varying O/C, caused by varying precursor concentration, on CCN activation of alpha-pinene SOA was not shown in Duplissy's study, instead it was just inferred from the behavior at RH=95% and the presumption that CCN activity will change similarly (CCN data very only available for high O/C from low precursor concentration experiments).However, a follow up study by Juranyi et al. (2009) showed that varying O/C of alpha-pinene SOA has virtually no influence on CCN activation, while the strong effect on hygroscopicity at lower RH was confirmed.Petters et al. (2009b) provides theoretical arguments that for this difference between "low" RH and very dilute solutions at the point of CCN activation.See also references therein for further experimental studies.

p. 25333 ,
lines 14-15: According to the arguments brought up in Petters et al. (2009b) one does not necessarily expect a strong dependence of CCN activity (CCN-derived kappa_org) on O/C from a theoretical point of view.Molecular weight gains on importance compared to O/C when it comes to very dilute solutions, in contrast to the influence of O/C on hygroscopicity at lower water activity.
p. 25338, lines 3-16: This is largely a repetition of text already given on page 25335.Please try to reduce redundancy.Section 3 has been restructured to reduce redundancy and to make the text more clear.p. 25341, line 16-20: The studies by Juranyi et al. (2009) and Petters et al. (2009b) indicate that O/C has little influence on CCN activity of organic compounds.p. 25341, line 27ff: Ensemble mean molecular weight is also a key factor for CCN activity of organics, not just O/C.These issues are discussed in the newly added Sect.3.3 and sections of this paragraph have been modified as such: