Secondary organic aerosol formation from biomass burning emissions

Biomass burning is an important source of aerosol and trace gases to the atmosphere, but how these emissions change chemically during their lifetimes is not fully understood. As part of the Fire Influence on Regional and Global Environments Experiment (FIREX 2016), we investigated the effect of photochemical aging on biomass burning organic 15 aerosol (BBOA), with a focus on fuels from the western United States. Emissions were sampled into a small (150 L) environmental chamber and photochemically aged via the addition of ozone and irradiation by 254 nm light. While some fraction of species undergoes photolysis, the vast majority of aging occurs via reaction with OH radicals, with total OH exposures corresponding to the equivalent of up to 10 days of atmospheric oxidation. For all fuels burned, large and rapid changes are seen in the ensemble chemical composition of BBOA, as measured by an aerosol mass spectrometer (AMS). 20 Secondary organic aerosol (SOA) formation is seen for all aging experiments and continues to grow with increasing OH exposure, but the magnitude of the SOA formation is highly variable between experiments. This variability can be explained well by a combination of experiment-to-experiment differences in OH exposure and the total concentration of non-methane organic gases (NMOGs) in the chamber before oxidation, measured by PTR-ToF-MS (r2 values from 0.64 to 0.83). From this relationship, we calculate the fraction of carbon from biomass burning NMOGs that is converted to SOA as a function of 25 equivalent atmospheric aging time, with carbon yields ranging from 24 ± 4 % after 6 hours to 56 ± 9 % after 4 days.

I wonder because the SP2 can saturate at high number concentrations, and the SP-AMS CE for BC requires some additional considerations (Ahern et al., 2016;Onasch et al., 2012;Willis et al., 2014.)Author response: SP2 and SP-AMS black carbon mass loadings generally agree within a factor of two, with SP2 mass loadings consistently higher than the measurements from the SP-AMS and a reasonably linear relationship (r2 = 0.9).BC measurements from the SP2 are those reported in the manuscript.The SP2 was operated with timevarying flow rates to account for the large dynamic range in the BC concentrations over the course of an experiment.In this manner, coincidence and under-counting (saturation) issues were avoided.The SP2 was calibrated with size-selected fullerene soot, and mass concentrations were corrected for "missing mass" outside of the SP2 detection window via multi-modal log-normal fitting.However we note that SP-AMS BC measurements were used only to determine which experiments to filter out of the analysis (due to enhanced wall loss), and so any CE-related errors will not affect the results presented.Black carbon measurements and other details about primary emissions will be discussed further in a future publication (Cappa et al., in preparation).
Referee #1: P5.26-30I find your parameterization of CE very interesting and possibly broadly applicable.But why was it necessary at all?If you can calculate MFR, why not use the SEMS-measured size distribution to correct for CE directly?Given the large amount of variability in your FigS3, it is not obvious that using the correlation is an improvement in accuracy or precision over a size distribution correction.Additionally, I don't agree with the statement that there was no good internal standard available.While I think that your CE parameterization could be very useful, it warrants verification by looking at other measurements more closely.For example, you state that there was an SO2 monitor, which would allow for a sulfur mass balance.Black carbon, measured by SP-AMS or SP2, was present at useful concentrations for some of the experiments.
Author response: While an SO2 monitor was present, we observed that the measurements were unreliable due to, most likely, strong interferences from large concentra-C3 tions of PAH's and other molecules, despite the internal scrubber in the SO2 monitor.Further, it is not entirely clear to us how sulfur balance (gas + particle) would provide clear insights into the particle collection efficiency; we did not have an independent particulate sulfate measurement, but a gas-phase sulfur measurement.Given that the SO2 measurements were compromised by interferences, we have now removed them from the list of measurements that were made.While we can use the SEMSdetermined MFR, it is not clear how this would lead to a "direct" correction for the CE.It would provide an alternative approach, but with a complication that the MFR from the SEMS would include contributions from non-refractory material and thus does not directly address the issue of how the organic component of the particles responded to temperature changes.The CE changes observed derive, in part, from changes in the organic MFR that result from oxidation leading to less volatile OA.This would only partially be captured by the SEMS because of the contribution of non-refractory components (e.g.BC).BC cannot be used as an internal standard since the BC and organic mixing state varied dramatically between experiments, with some having the majority of the organic and BC being internally mixed (at low [OA]/[BC] ratios) and some having most of the organic material externally mixed from BC (at high [OA]/[BC] ratios) (McClure et al., in prep.).Given these overall issues, we believe that the organic MFR links more closely to the physical changes that occur.The text in the manuscript was amended (page 5, line 31): "However, we were unable to find a suitable tracer in these experiments: sulfate changes as a result of oxidation of emitted SO2, black carbon (when present in high concentrations) exhibited wall losses different from OA (as described below) and appeared not to be homogeneously mixed with the OA (McClure et al., in prep), and POA tracers (such as the C7H11+ ion, recently used by Ahern et al. (2019)) are likely to be lost via heterogeneous oxidation at the high OH exposures examined here.Thus, corrections for CE, dilution, and particle wall loss were carried out individually, as described below." Referee #1: P5.31 Would you please confirm that the experiments used to calculate your CE were devoid of nucleation, particles grown outside the SEMS or AMS transmission/measurement ranges, and weren't unduly influenced by rBC after thermodenuding?Also, out of curiosity, how frequently was the thermodenuder valve switched, and therefore a new CE able to be calculated?Author response: Yes, only data that did not show significant nucleation and had low rBC loadings were used to calculate the CE parameterizations.In addition, only SEMS and PToF size distributions that could be fit to a lognormal function were used.The thermodenuder valve was switched from thermodenuder to bypass every two minutes.Some additional text clarifying these details is included in the main text (page 6, lines 4-6): "CE and particle density were calculated by comparing AMS particle time-of-flight (PToF) and SEMS size distributions (Bahreini et al., 2005) for a subset of data points with PToF and SEMS distributions that could be fit to lognormal functions, did not show significant particle nucleation, and had low rBC concentration (see below)." Referee #1: P6.2-3 Please provide a citation or clarify regarding the relationship between volatility and phase.It might be easier to provide citations that claim that SOA has been observed to be an amorphous solid with low volatility, and therefore is likely to bounce.
Author response: This is an excellent idea; we have included citations for Matthew et al. (2008) describing collection efficiencies as a function of particle phase in the AMS, as well as Virtanen et al. (2010) showing that SOA can be an amorphous solid.
Referee #1: P8.5 What are the possible implications for the chamber OA concentration having decreased by two orders of magnitude from the beginning to the end of the experiment (FigS4)?It stands to reason that fewer of the semi-volatile SOA products will condense at low OA concentrations late in the experiment, but that any that do condense will have a larger impact on OSc.

C5
Author response: From Fig. S4, the organic concentration decreases by only one order of magnitude (200 ï Ą g/m3 to 20 ï Ą g/m3).Although semi-volatile gases are less likely to condense at low OA concentrations observed at the end of experiments due to significant dilution over the course of each experiment, most reactive gases are likely to have reacted relatively early in the experiment when dilution has less of an impact (see Fig. 6, most SOA growth occurs within ï Ą¿2 days).Indeed, those gases (and potentially secondary products) that do react at longer OH exposures and condense are likely to have a large impact on the calculated average OSC and elemental ratios of the OA.Thus, the calculated OSC and elemental ratios may be more representative of the oxidized long-lived gases (which condense) rather than the BBOA + SOA as a whole.The following additional text was added to the manuscript to clarify this point (page 8, lines 7-8): "Over longer timescales, when dilution is more significant and OA concentrations are lower, calculated OSC are likely to reflect the oxidation of longer-lived gases." Referee #1: P10.18 How does this approach compare with the measured PTR-MS NMOG concentrations?Author response: We are not able to directly measure the amount of NMOG reacted, due to the dilution loss of NMOG (gases removed from the chamber before they can be reacted with OH), formation of secondary gas-phase products, and potential offgassing of non-SOA forming low molecular weight NMOG from the chamber walls, leading to calculated SOA yields greater than unity.Thus, only initial NMOG measurements are used for the analysis and reacted NMOG are calculated as described in the text.This explanation is now included in the main text (page, lines) Referee #1: P11.19Given the importance of dilution and volatility on the results pre-Williams, L. R., Lambe, A. T., Worsnop, D. R., and Abbatt, J. P. D.: Collection efficiency of the soot-particle aerosol mass spectrometer (SP-AMS) for internally mixed particulate black carbon, Atmos.Meas. Tech., 7, 4507-4516, https://doi.org/10.5194/amt-7-4507-2014, 2014.Referee #2 Scientific Referee #2: On p. 4 potential loses in the sampling line are discussed, and a figure is provided in the supplement illustrating the difference between gaseous emissions sampled in the stack directly, to those in the community inlet, binned by saturation concentration.The binned comparison does not suggest a systematic loss in the community inlet either across bins or as a function of volatility.However, there is a significant difference in one of the bins (C* 107-108).Is this difference well understood (e.g., likely due to a specific class of compounds)?And, how might this difference affect the results and analysis?
Author response: The compounds in this bin do not correspond to a specific class of compounds but to a variety of compounds, including: acetic acid, furfural, furanone, monoterpenes, and methyl glyoxal/acrylic acid (in order of decreasing average abundance across all experiments).The presence of well-known SOA precursors (e.g., monoterpenes) in this volatility bin suggest that it is possible that some SOA precursors are in lower abundance in the mini-chamber relative to the stack.Another possibility is that acetic acid, the most abundant compound in this volatility bin, is preferentially lost during transport to the chamber due to accumulated water on the surfaces of the community inlet.Although preferential loss of some compounds in this volatility bin may affect SOA yields, the observation that total NMOGs correlates well with SOA suggests that some preferential loss of a small subset of compounds would not greatly affect SOA formation.
Referee #2: On p. 7, line 4 the authors state that the dilution factor prior to oxidation C9 influenced observed initial aerosol mass and reference Table 1 in the supplement.It is not clear how the dilution factor is represented in the table.Is it a function of sampling time?This needs a bit more clarification/explanation.
Author response: This text simply refers to the fact that after sampling, experiments are diluted by varying amounts depending on the length of time between sampling and initiation of oxidation (254 nm UV lights).The sentence has been reworded for clarity (page 7, line 8-9): "The total, initial aerosol mass in the chamber varied widely from experiment to experiment (SI Table 1), averaging 130 ï Ćś 103 ï Ą g m-3 (mean ï Ćś 1ï Ąş), depending on the amount of fuel burned, fuel type, sampling time, and dilution prior to oxidation."Referee #2: The relationship between reported enhancement ratios in this work with previously published ratios is discussed on p. 7, first paragraph.The authors suggest that once aging and collection efficiency are taken into account, the results are broadly consistent with other results, and not overestimates.However, based on Fig. S6 panel b, where OA enhancement ratios are plotted as a function of aging time, the reported enhancements are still a factor of 2+ higher than Ortega et al.Are the Lim et al. enhancement ratios in the right panel actually for CE =1?They seem significantly higher than what is presented in the left panel (and it is assumed that both are for CE = 1).In addition, the average appears to be â Ĺij3, which is the CE corrected-value reported in the main text.
It's not obvious to me why you compare [calculated VOC reacted]/[measured SOA formed] instead of [measured VOC reacted]/[measured SOA formed]?Or to go backwards, can you use your [measured NMOG reacted]* this calculated SOA yield to predict SOA formation?