In situ measurements and modeling of reactive trace gases in a small biomass burning plume

An instrumented NASA P-3B aircraft was used for airborne sampling of trace gases in a plume that had emanated from a small forest understory fire in Georgia, USA. The plume was sampled at its origin to derive emission factors and followed ∼ 13.6 km downwind to observe chemical changes during the first hour of atmospheric aging. The P-3B payload included a proton-transfer-reaction timeof-flight mass spectrometer (PTR-ToF-MS), which measured non-methane organic gases (NMOGs) at unprecedented spatiotemporal resolution (10 m spatial / 0.1 s temporal). Quantitative emission data are reported for CO2, CO, NO, NO2, HONO, NH3, and 16 NMOGs (formaldehyde, methanol, acetonitrile, propene, acetaldehyde, formic acid, acetone plus its isomer propanal, acetic acid plus its isomer glycolaldehyde, furan, isoprene plus isomeric pentadienes and cyclopentene, methyl vinyl ketone plus its isomers crotonaldehyde and methacrolein, methylglyoxal, hydroxy acetone plus its isomers methyl acetate and propionic acid, benzene, 2,3-butanedione, and 2-furfural) with molar emission ratios relative to CO larger than 1 ppbV ppmV. Formaldehyde, acetaldehyde, 2-furfural, and methanol dominated NMOG emissions. No NMOGs with more than 10 carbon atoms were observed at mixing ratios larger than 50 pptV ppmV CO. Downwind plume chemistry was investigated using the observations and a 0-D photochemical box model simulation. The model was run on a nearly explicit chemical mechanism (MCM v3.3) and initialized with measured emission data. Ozone formation during the first hour of atmospheric aging was well captured by the model, with carbonyls (formaldehyde, acetaldehyde, 2,3-butanedione, methylglyoxal, 2-furfural) in addition to CO and CH4 being the main drivers of peroxy radical chemistry. The model also accurately reproduced the sequestration of NOx into peroxyacetyl nitrate (PAN) and the OH-initiated degradation of furan and 2-furfural at an average OH concentration of 7.45± 1.07× 10 cm in the plume. Formaldehyde, acetone/propanal, acetic acid/glycolaldehyde, and maleic acid/maleic anhydride (tentatively identified) were found to be the main NMOGs to increase during 1 h of atmospheric plume processing, with the model being unable to capture the observed increase. A mass balance analysis suggests that about 50 % of the aerosol mass formed in the downwind plume is organic in nature. Published by Copernicus Publications on behalf of the European Geosciences Union. 3814 M. Müller et al.: Reactive trace gases in a biomass burning plume


Introduction
Understanding and predicting the impacts of biomass burning emissions on air quality is a challenging but important task.Fire emissions include a plethora of inorganic and organic species, both in the gas and the particulate phase, and many of them undergo rapid chemical transformations and phase changes after their release to the atmosphere (e.g., Simoneit, 2002).These processes are the focus of intense research efforts, both in the laboratory and in the field.Over the last decade, many airborne field studies have been undertaken to characterize emissions and evolution of gases and particles in the aging plume (e.g., Akagi et al., 2012Akagi et al., , 2013;;Yokelson et al., 2009).In general, these studies have targeted emissions from medium and large-scale fires.Small fires (< 500 m diameter of burned area) have been undersampled although they may contribute 35 % or more to global biomass burning carbon emissions (Randerson et al., 2012).Emissions from small fires are often not included in emission inventories, and local and regional air quality assessments seldom include emissions from small fires.In addition, the chemical complexity of emissions poses a major challenge to modeling efforts.Lumped mechanisms are thus typically used in chemical models to predict the evolution of trace gases in biomass burning plumes.Lumping of species may, however, result in an oversimplification of the involved chemistry, which will ultimately yield erroneous model predictions.
In this work, we present the results from an airborne study in which inorganic and organic trace gases emanating from a small forest understory fire were measured with state-of-theart analytical tools.A proton-transfer-reaction time-of-flight mass spectrometry (PTR-ToF-MS) instrument delivered nonmethane organic gas (NMOG) data at unprecedented spatiotemporal resolution.We sampled the plume at its origin to derive emission factors and followed it downwind to observe chemical changes during the first hour of atmospheric aging.We also found that a 0-D photochemical box model, run on a nearly explicit chemical mechanism and properly initialized with the measured emission data, adequately described key chemical processes (ozone and radical formation, NO x sequestration) in the aging plume.

Sampling strategy and conditions
A small biomass burning plume was intercepted by the NASA P-3B research aircraft in Laurens County near Dublin, GA, USA, on 29 September 2013, during a flight from Houston, TX, to Wallops Island, VA.The plume emanated from a managed forest understory fire located at 32 • 23 42 N, 82 • 51 7.2 W which had been applied after logging and forest clearance activities.Historic Google Earth imagery shows that the area to the SW of the fire location had undergone in- tense forest clearing between 2011 and 2014.After the flight, the burned area was inspected by a local official who identified residual tree logs (pine, oak) and weeds as fire fuels.Figure 1a and b are two frames from the P-3B front camera showing the fire and the emanating plume at 17:33:32 UTC (UTC = local time + 4 h) and 17:42:51 UTC, respectively.
Figure 2 depicts the P-3B flight pattern color-coded in radar altitude, with blue lowest and red highest.The flight direction is indicated by black arrows.Winds steadily blew from the NE at an average speed of 3.5 m s −1 (Figure 2, wind rose inset in the upper left corner).The average temperature during the sampling period (17:30-17:55 UTC) as measured by the P-3B met sensors was 26.5 ± 5.3 • C, and the average relative humidity was 60.4 ± 2.3 %.The average vertical temperature gradient was −1.34 • C per 100 m, causing the plume to slowly rise downwind of the source.The turbulence condition of the boundary layer was neutral to slightly unstable.
The fire was sighted and approached from the SW.Following a 180 • turn, the aircraft overflew the fire for the first time at 125 m altitude (Fig. 1a) at 17:33:35 UTC (source emission profile 1).The plume was then followed downwind in a southwesterly direction for approximately 2 min, slowly climbing in altitude to reach a radar altitude of 190 m at a 13.6 km downwind location (longitudinal plume transect 1).The underlying terrain was forested and agricultural land.At an average wind speed of 3.5 m s −1 , the plume travel time for a 13.6 km distance is approximately 1 h.Following a horizontal loop maneuver, the ∼ 8 km broad plume was sampled transversely at 160 m radar altitude at the 13.6 km downwind location (transverse downwind plume transect 1).Subsequently, the P-3B returned to the fire, intercepting the freshly emitted plume at 17:42:57 UTC (Fig. 1b) and at 17:45:38 UTC, at 110 and 80 m altitude, respectively (source emission profiles 2 and 3).The downwind pattern was repeated with longitudinal plume transect 2 reaching 220 m altitude at the 13.6 km downwind location.The second transverse downwind plume transect was at 160 m altitude at the 13.6 km downwind location.The fourth and final fire overflight was at 75 m altitude at 17:54:25 UTC (source emission profile 4).By implementing this sampling strategy, we obtained (i) four source emission profiles within 21 min, (ii) two longitudinal plume transects (source to 1 h downwind), and (iii) four plume characterizations at 1 h downwind distance from the source (two longitudinal "spot" samples and two "integrated" cross-plume samples).The results (see Sect. 3) indicate nearly stable source conditions during the sampling period.This implies that the observed downwind differences in chemical composition were mostly due to dilution and photochemistry.

Analytical instrumentation
The NASA P-3B was returning from a DISCOVER-AQ (Deriving Information on Surface conditions from Column and Vertically Resolved Observations Relevant to Air Quality) deployment (http://discover-aq.larc.nasa.gov/) in Houston, which had it equipped with a payload for in situ atmospheric chemistry measurements.The data used in this study were obtained using the analytical instruments listed in Table 1.
This work focuses on NMOGs as measured by the PTR-ToF-MS instrument described in detail by Müller et al. (2014).The data presented herein were acquired at a frequency of 10 Hz, which makes the PTR-ToF-MS instrument ideally suited for airborne NMOG measurements at high spatiotemporal resolution.However, only the elemental composition of organic analytes can be determined, not their structure.In other words, the PTR-ToF-MS instrument does not resolve isomeric NMOGs (e.g., acetic acid and glycolaldehyde).The PTR-TOF Data Analyzer Toolbox (https://sites.google.com/site/ptrtof/)was used for data analysis (Müller et al., 2013).Accurate m/z information; element restriction to C, H, N, and O atoms; and isotopic pattern analyses were used to determine the elemental composition (C w H x N y O z ) of detected analyte ions.It has been shown in previous work that accurate m/z information can be obtained even at a moderate mass resolution m/ m in the range of 1000 to 1500 (Müller et al., 2011(Müller et al., , 2014)).The assignment of observed m/z signals to specific chemical compounds was based on the literature (see Sect. 3.1.2).
Methanol, acetonitrile, acetaldehyde, acetone, isoprene, methyl ethyl ketone, benzene, toluene, m-xylene, 1,3,5trimethylbenzene, and monoterpenes (α-pinene) were calibrated externally using a dynamically diluted certified standard.The measurement accuracy is ±5 % for pure hydrocarbons and ±10 % for oxygenates.Formic acid and acetic acid were calibrated (± 10 %) in a post-campaign study using a liquid standard nebulization device (LCU, Ionicon Analytik, Austria).The protonated formaldehyde ion signal was crosscalibrated to formaldehyde data collected by a difference frequency generation absorption spectroscopy (DFGAS; Weibring et al., 2007) instrument during the same flight and at the same humidity conditions.Although less accurate (±10 %), PTR-ToF-MS formaldehyde data were used instead of DF-GAS observations because of a higher data density in the plume.Instrumental response factors to furan, methylglyoxal, and 2-furfural were calculated from ion-molecule collision theory (Cappellin et al., 2012).The estimated measurement accuracy for these species is ±25 %.Peroxyacetyl nitrate (PAN) was quantified (± 40 %) using a calibration factor obtained in a previous study (unpublished data).All other organic signals were corrected for instrumental mass discrimination effects and converted to volume mixing ratios by using the acetone sensitivity as a proxy.Mixing ratios in acetone equivalents are estimated to be accurate to within ±40 %.This is also the maximum error we must assume for the total NMOG mass calculated by summing all individual signals calibrated as specified above.
The PTR-ToF-MS instrument also detects a few inorganic gases, nitrous acid (HONO) and ammonia (NH 3 ) being two prominent examples.Given the importance of HONO for fire plume photochemistry, we made an attempt to quantify HONO emissions.HONO dehydrates upon protonation, forming NO + ions, which are observed at m/z 29.997.The excess NO + signal in the plume was assigned to HONO.The contribution from organic nitrites was assumed to be minor.A positive measurement artifact from NO 2 -to-HONO conversion (1 % of NO 2 ) on instrumental surfaces was subtracted.The instrumental response to HONO and HONO inlet artifacts has been characterized in previous studies (Metzger et al., 2008;Wisthaler et al., 2003).Given that different inlet and drift tube configurations were used in those studies, the 1 % NO 2 -to-HONO conversion efficiency is to be considered an upper-limit estimate.Still, the NO 2 artifact only accounts for 10.4 % of the excess NO + signal measured at the source.The estimated accuracy of the reported HONO data is ±30 %.NH 3 measurements suffered from a high intrinsic background signal generated in the ion source of the instrument.This deteriorated the detection limit to 12 ppbV for 1 Hz measurements.

Data processing
Volume mixing ratios (VMRs) were obtained as described in Sect.2.2.When referring to the VMR of a species X, the italic style, X, is used throughout this work.
Given that the P-3B spent about 2 s in the plume during fire overflights and that CO was only measured at 1 Hz, it was not possible to perform linear regression analyses, X vs. CO, on data from individual plume intercepts.For each plume intercept, we calculated the excess mixing ratio of X in the fire plume, X, as the average mixing ratio of X inside the plume, X plume , minus the average mixing ratio of X outside the plume, X background : (1) X background was calculated from the data obtained immediately before plume interception.Background mixing ratios of all species discussed herein were stable in the investigated domain.This analysis was performed for each of the fire overflights, resulting in four data points, X vs. CO, for source emission characterization.A linear least-square regression analysis was then applied to these four data points, with the slope of the regression line describing the molar emission ratio (ER) of the species X relative to CO, ER X/CO , in ppbV ppmV −1 .The precision of the CO data is better than ±1 ppbV, which justifies the use of a univariate regression method.The standard error of the slope reflects both the natural variability in the plume and the measurement imprecision.A delayed instrument response was observed for formic acid and acetic acid.In-plume concentrations of these acids were derived as discussed in the Supplement.The dilution-corrected molar excess mixing ratio of a species X, dil X (in ppbV), at a downwind location was calculated from the excess mixing ratio of CO observed at the fire source, CO source , and the locally observed X and CO using the following equation: By introducing this parameter, we are able to study loss or formation processes in the plume without confounding contributions from dilution.On a 1 h timescale, no photochemical loss of CO occurs, and the contribution from photochemically formed CO to the large CO levels already present in the plume is negligible.Reported dil X are average values from two longitudinal plume transects for which data were binned at 1 km spatial resolution.
The emission factor of a species X, EF X , in grams per kilograms (g kg −1 ) was calculated according to Yokelson et al. (1999): with F C being the mass fraction of carbon of the fuel; MM X and MM C the molecular masses of the species X and of carbon, respectively; and C X /C T the fraction of moles emitted as species X relative to the total number of moles carbon emitted.F C was not measured during this study, but 0.50 is a typical value for biomass (Burling et al., 2010).The accuracy of C T is limited by unmeasured carbon.This fraction is assumed to be less than 2 %.EFs were calculated as averages from the four fire overflights.
The oxygen-to-carbon (O : C) ratio of all detected NMOGs was calculated as follows: with n O,i and n C,i being the number of oxygen atoms and carbon atoms in the species X i , respectively.The modified combustion efficiency (MCE) was calculated as follows (Ferek et al., 1998): Aerosol mass was calculated from the 60-1000 nm integrated optical aerosol volume as measured by the Ultra-High-Sensitivity Aerosol Spectrometer (UHSAS) instrument assuming an average biomass burning secondary organic aerosol density of 1.3 g cm −3 (Aiken et al., 2008).

Chemical box model calculations
We used the University of Washington Chemical Box Model (UWCM) (Wolfe and Thornton, 2011) run on Master Chemical Mechanism (MCM) v3.3 chemistry (Jenkin et al., 1997(Jenkin et al., , 2003(Jenkin et al., , 2015;;Saunders et al., 2003) to simulate the downwind processing of trace gases in the biomass burning where k dil represents the dilution rate coefficient obtained from the decrease of CO vs. plume travel time.k dil was calculated in 285 s time bins (equivalent to 1 km distance bins).More information on the UWCM and the underlying theory can be found in Dillon et al. (2002), Wolfe and Thornton (2011), and Wolfe et al. (2012).MCM v3.3 chemistry does not include the degradation of furan and 2-furfural, two highly reactive compounds with significant primary emissions from fires.We included these species in our chemical mechanism using the photolysis rates reported by Colmenar et al. (2015) and the OH reaction rates reported by Bierbach et al. (1992).We assumed that butenedial is the only primary reaction product of the reaction of furan with OH radicals (Aschmann et al., 2014).The atmospheric oxidation products of 2-furfural are unknown.

Inorganic gases
Table 2 summarizes ER X/CO and EF X values of major inorganic gases as obtained from four source emission profiles.An MCE of 0.90 ± 0.02 was derived from the measured CO and CO 2 data, indicating stable burning conditions and roughly equal amounts of biomass consumption by flaming and smoldering combustion.ERs and EFs of NO and NO 2 are within typical ranges reported in the literature (Akagi et al., 2011).The observed ER HONO/CO of 2.0 ± 0.7 ppbV ppmV −1 is also in  The slopes of the least-square regressions (dotted lines) correspond to the initial molar emission ratios (ER X/CO , in ppbV ppbV −1 ).good agreement with previously reported values (e.g., Veres et al., 2010) increasing our confidence in the tentative identification and quantification of HONO emissions by PTR-ToF-MS.Excess mixing ratios of NH 3 in the plume were below the detection limit, so only an upper limit for ER NH 3 /CO and EF NH 3 is reported.

Organic gases
Methane (CH 4 ) was the main organic gas emitted from the fire.ER CH 4 /CO and EF CH 4 are 108.4 ± 13.4 ppbV ppmV −1 and 6.25 ± 2.86 g kg −1 , respectively.This work, however, focuses on NMOG emissions.Figure 3a shows the 10 Hz time series of acetonitrile (CH 3 CN), furan (C 4 H 4 O), the sum of monoterpene isomers (C 10 H 16 ), and isoprene (C 5 H 8 ) as measured during the overflight at 17:54:25 UTC (source emission profile 4). Figure 3b shows the time series of benzene (C 6 H 6 ), toluene (C 7 H 8 ), C 8 -alkylbenzene isomers (C 8 H 10 ), and C 9 -alkylbenzene isomers (C 9 H 12 ) for the same time period.The data demonstrate that the airborne PTR-ToF-MS instrument generates high-precision NMOG data even for very localized emission sources.The two small plumes discernible in Fig. 1a and b are well resolved in the PTR-ToF-MS data shown in Fig. 3.All signals instantly drop to background levels outside the plume, confirming the excellent time response of the airborne PTR-ToF-MS instrument for analytes that do not adhere to instrumental surfaces.
It is currently not possible to fully exploit these highly time-resolved NMOG data to determine ER X/CO because CO is only measured at 1 s time resolution.ER X/CO values were thus obtained from average values for each source emission profile as described in Sect.2.3.
Figure 4 shows X vs. CO as obtained for 2-furfural, benzene, furan, and monoterpenes during each of the four fire overflights.The compounds were selected as representatives of different chemical classes (including furans, aromatics, aldehydes, terpenes) that can have different production mechanisms in the fire, e.g., furan being formed by pyrolysis and monoterpenes just being evaporated (Yokelson et al., 1996).A strong linear relationship was found not only for the species shown here but for all detected NMOGs, indicating that source emissions were nearly stable during the 21 min sampling period.This important finding will later allow us to draw conclusions from analyte ratios measured downwind.
In total, 57 m/z signals (NO + , NO + 2 , and 55 C-containing ions) in the PTR-ToF-MS spectrum showed an enhancement in the source emission profiles.Table 3 lists ER X/CO and EF X of the 18 ion signals that contain carbon atoms and that were observed with an ER X/CO > 1 ppbV ppmV −1 .These signals contribute 93 % of the total NMOG emissions as detected by PTR-ToF-MS.Emissions are dominated by formaldehyde, methanol, acetaldehyde, and 2-furfural (EF > 1 g kg −1 ).The complete list of all detected ion signals is given in Table S1 in the Supplement.It is beyond the scope and possibilities of this work to make an independent assignment of m/z signals to specific neutral precursors.The P-3B payload did not include any NMOG analyzer with higher analytical selectivity than the PTR-ToF-MS instrument.Our assignment of m/z signals to specific chemicals in Table 3 thus exclusively relies on two recent studies and the references used therein.Yokelson et al. (2013) used results from multiple analytical techniques for assigning m/z peaks.Stockwell et al. (2015) used a high mass resolution PTR-ToF-MS instrument for elemental composition determination and open-path FTIR data together with literature reports for mass spectral interpretation.In the case of multiple neutral precursors for a specific m/z signal, we considered only species with a relative contribution > 10 % to the total signal.Two ion signals (m/z 85.027 and m/z 111.041) were not reported previously.The assignment made is tentative, and the compounds (in italic in Table 3) were not included in the modeling study.The reader is cautioned that this is still an evolving field of research and some signals may be misassigned or suffer from as yet unknown interferences.
Total observed carbon emitted as NMOGs (55 ion signals) was 10 472 ppbC.The O : C ratio at the fire source was 0.41. Figure 5 shows the relative contribution of C 1 to C 10 compounds to total NMOG emissions on a carbon atom basis.
The dominant contribution to NMOG carbon emissions came from the C 5 -compound 2-furfural.Significant carbon emissions (ER X/CO > 50 pptV ppmV −1 ) were detected only up to C 10 (monoterpenes).

Plume evolution
The NASA P-3B sampled the downwind plume for approximately 2 min of flight time.At an average wind speed of 3.5 m s −1 , this corresponds to approximately 1 h of atmospheric plume processing.Volume mixing ratios of inert trac- ers (CO 2 , CO, acetonitrile, and benzene) consistently decreased by a factor of ∼ 13.5 during the two longitudinal plume transects.We used this decrease to derive dilutioncorrected molar excess mixing ratio of reactive trace gas species X, dil X (see Sect. 2.3).dil X were used to investigate downwind plume chemistry by observations and by a 0-D photochemical box model simulation initialized with measured emission data.

Ozone formation and sequestration of nitrogen oxides
Figure 6 shows dilution-corrected molar excess mixing ratios of O 3 NO, NO 2 , and NO z (= NO y -NO-NO 2 ) during 1 h of atmospheric plume processing.Point symbols refer to the measured data; solid lines represent the output of the UWCM based on MCM v3.3 chemistry.
Ozone is efficiently formed in the plume in the presence of NO x and NMOGs.Close to the source (t < 600 s), ambient O 3 reacts with abundantly emitted NO, resulting in negative O 3 excess mixing ratios (not displayed on the logarithmic ordinate of Fig. 6).After ∼ 10 min of plume processing net ozone formation starts, resulting in a dilution-corrected increase of O 3 on the order of 50-60 ppbV during the first hour the plume resides in the atmosphere.The UWCM (MCM v3.3The model also accurately captures the net formation of NO z (= NO y −NO−NO 2 ).Modeled NO z sums all species in the MCM v3.3 degradation scheme that include nitro or nitroso groups.The main contributors to NO z being formed are PAN and nitric acid (HNO 3 ).The model simulates dil PAN = 3 ppbV and dil HNO 3 = 2.4 ppbV after 1 h of plume evolution, which accounts for ∼ 90 % of all NO z formed.Under the operating conditions used in this study, PAN is predominantly detected at m/z 45.992 (NO + 2 ) by the PTR-ToF-MS instrument (Hansel and Wisthaler, 2000).Using a PAN calibration factor obtained in a previous study, we obtain an excellent agreement between measured and modeled PAN concentrations (Fig. 7).

Evolution of NMOGs
Fire emissions include many NMOGs that quickly react with OH radicals.OH radicals are abundantly formed in biomass burning plumes, causing highly reactive NMOGs to disappear even on the 1 h timescale investigated in this study (Akagi et al., 2012(Akagi et al., , 2013;;Hobbs et al., 2003).Figure 8 shows dilution-corrected mixing ratios of furan and 2furfural during 1 h of plume evolution.Point symbols refer to the dilution-corrected experimental data; solid lines represent the output of the UWCM.Measured and modeled data are in excellent agreement, confirming that we observed the OHinitiated degradation of furan and 2-furfural.The influence of interfering isomers (or fragment ions), if any, is small.The box model output indicates nearly stable OH radical concentrations of 7.45 ± 1.07 × 10 6 cm −3 along the 13 km downwind transect.Other studies (e.g., Yokelson et al., 2009) have reported similarly high average OH levels in biomass burning plumes.
Figure 9 shows dilution-corrected mixing ratios of four important oxygenated NMOGs: formaldehyde, acetaldehyde, methanol, and acetone/propanal.Point symbols again refer to the dilution-corrected experimental data; solid lines represent the output of the UWCM.Formaldehyde and acetone/propanal show a distinct increase after half an hour of plume processing, which is not captured by the model simulation based on MCM v3.3 degradation chemistry of the 16 most abundant NMOGs (as detected by PTR-ToF-MS).Interestingly, the experimental data indicate a significant loss of methanol during the initial 15 min of plume processing.This sink is also not included in MCM v3.3 chemistry, and heterogeneous loss processes should be investigated.The observed initial drop could, however, also be caused by an unknown highly reactive compound that interferes with the detection of methanol.In addition to the carbonyls discussed above, acetic acid/glycolaldehyde and the C 4 H 3 O + 3 signal, which is tentatively assigned to maleic acid/maleic anhydride, exhibited dilution-corrected increases of ∼ 1.5 ppbV and ∼ 1 ppbV, respectively.The model was unable to capture the observed increase.This does not come as a surprise since these species are typical higher-order degradation products that are not included in MCM v3.3 degradation schemes.
Figure 10 compares the relative contributions of C 1 to C 12 compounds to total NMOG carbon measured at the fire source and at the 1 h downwind location.C 1 , C 2 , and C 4 compounds exhibited the largest relative increase.The observed O : C ratio at the 1 h downwind location source was 0.56, compared to 0.41 observed at the source.This is consistent with the conceptual picture of a photochemical breakdown of NMOGs into smaller, more oxidized species.

Gas-to-particle conversion
A dilution-corrected mass balance analysis reveals that 40.8 µg cm −3 of the mass initially emitted as NMOGs was lost during 1 h of atmospheric processing.This equals 24 % of the carbon initially emitted as NMOGs.At the same time, the dilution-corrected total particle mass concentration as derived from UHSAS measurements increased by ∼ 78 µg cm −3 .These mass concentration calculations are only approximate (for details see Sect.2.2), but this analysis suggests that about 50 % of the aerosol mass formed in the downwind plume is organic in nature.This agrees with findings from previous studies that observed significant organic and inorganic aerosol formation in aging biomass burning plumes (Cubison et al., 2011;Yokelson et al., 2009).
Given that photooxidation of 2-furfural has the highest mass turnover, secondary organic aerosol formation from the 2furfural + OH reaction should be investigated in laboratory experiments.

Summary and conclusion
A plume emanating from a small forest understory fire was investigated in an airborne study.High-spatiotemporalresolution data were obtained for inorganic and organic trace gases, the latter being sampled for the first time at 10 Hz using a PTR-ToF-MS instrument.We generated quantitative emission data for CO 2 , CO, NO, NO 2 , HONO, NH 3 , and 16 NMOGs with ER X/CO > 1.0 ppbV ppmV −1 .NMOG emissions were dominated by formaldehyde, acetaldehyde, 2-furfural, and methanol.No NMOGs with more than 10 carbon atoms were observed at mixing ratios larger than 50 pptV ppmV −1 CO emitted.Downwind plume chemistry was investigated both by observations and by a model simulation using nearly explicit MCM v3.3 chemistry.The observed dilution-corrected O 3 increase on the order of 50-60 ppbV was well captured by the model, which indicated carbonyls (formaldehyde, acetaldehyde, 2,3-butanedione, methylglyoxal, 2-furfural) in addition to CO and CH 4 as the main drivers of peroxy radical chemistry.The model also accurately reproduced the sequestration of NO x into PAN and the degradation of furan and 2-furfural at average OH plume concentrations of 7.45 ± 1.07 × 10 6 cm −3 .Formaldehyde, acetone/propanal, acetic acid/glycolaldehyde, and maleic acid/maleic anhydride (tentative identification) were found to increase during 1 h of atmospheric plume processing, with the model being unable to capture the increase.A dilutioncorrected mass balance analysis suggests that about 50 % of the aerosol mass formed in the downwind plume is secondary organic in nature.
We conclude that the PTR-ToF-MS instrument is a powerful analytical tool for airborne plume studies.The generated data are highly valuable in characterizing point source emissions and near-field chemical transformations.Key chemical processes (ozone and radical formation, NO x sequestration) in an aging biomass burning plume were accurately simulated using a 0-D photochemical box model run with up-todate and nearly explicit MCM v3.3 chemistry.

Figure 2 .
Figure 2. Flight pattern of the NASA P-3B to obtain four point source emission profiles, two longitudinal plume transects (source to 1 h downwind), and two transverse downwind plume transects (1 h downwind from source).The inset shows wind rose data obtained during the two longitudinal plume transects when wind measurements are most accurate.

Figure 4 .
Figure 4. Average excess VMRs of 2-furfural, benzene, furan, and monoterpenes vs. average excess VMRs of CO.Each data point represents data from one fire overflight (source emission profile).The slopes of the least-square regressions (dotted lines) correspond to the initial molar emission ratios (ER X/CO , in ppbV ppbV −1 ).

Figure 5 .
Figure 5. Relative contributions of C 1 -C 10 compounds to total NMOG carbon emissions.C 1 to C 5 compounds each have relative contributions > 10 % and in sum contribute ∼ 80 % of the total NMOG carbon emissions.

Figure 6 .
Figure 6.Dilution-corrected molar excess mixing ratios of O 3 , NO, NO 2 , and NO z (= NO y -NO-NO 2 ) during 1 h of plume evolution (in 1 km bins).Point symbols refer to the measured data; solid lines represent the output of the UWCM based on MCM v3.3 chemistry.

Figure 7 .
Figure 7. Dilution-corrected molar excess mixing ratios of PAN during 1 h of plume evolution (in 1 km bins).Point symbols refer to the measured data; the solid line represents the output of the UWCM based on MCM v3.3 chemistry.

Figure 8 .
Figure 8. Dilution-corrected molar excess mixing ratios of furan and 2-furfural during 1 h of plume evolution.Point symbols refer to the measured data (1 km bins); solid lines represent the output of the UWCM fed with MCM v3.3 chemistry.

Figure 9 .
Figure 9. Dilution-corrected molar excess mixing ratios of formaldehyde (a), acetaldehyde (b), methanol (c), and acetone/propanal (d) during 1 h of plume evolution.Point symbols refer to the measured data (1 km bins); solid lines represent the output of the UWCM fed with MCM v3.3 chemistry.

Figure 10 .
Figure 10.Relative contributions of C 1 to C 12 compounds to total NMOG carbon measured at the fire source and at the 1 h downwind location.

Table 1 .
Excerpt of the P-3B analytical chemistry payload.
a Measurement frequency was 1 Hz for instruments except PTR-ToF-MS (10 Hz).

Table 2 .
Molar emission ratios (ERs) relative to CO and emission factors (EFs) of the major inorganic gases as obtained from four fire overflights.Compound ER X/CO (ppbV ppmV −1 ) EF X (g kg −1 ) The model was initialized using measured source concentrations of NO, NO 2 , HONO, O 3 , CO, CH 4 , and of the 16 most abundant NMOGs detected by PTR-ToF-MS (ER X/CO > 1.0 ppbV ppmV −1 ; compounds identified in previous studies as detailed in Sect.3.1.2).The model was run using the measured meteorological parameters (pressure, temperature, relative humidity, solar zenith angle) and the observed NO 2 photolysis rate.The model calculates dilution from the simple equation

Table 3 .
Measured accurate m/z, elemental composition C w H x N y O + z of the detected ion, neutral precursor assignment based on literature information (significant interferants in parentheses, tentative assignments in italic), emission factor (EF) and standard deviation (SD), and emission ratio (ER) and standard deviation for all detected NMOGs with ER X/CO >1 ppbV ppmV −1 .
chemistry; initialized with measured emissions of NO, NO 2 , HONO, O 3 , CO, CH 4 , and 16 NMOGs) simulates the evolution of O 3 , NO, and NO 2 well.An even better agreement in the ozone evolution is obtained if the model is constrained to measured formaldehyde values which slightly exceed the modeled values at t > 1500 s (see Sect. 3.2.2).O 3 formation is fueled by HO 2 /CH 3 O 2 +NO reactions.The model indicates that HO 2 radicals are primarily generated in the CO + OH, 2-furfural+OH, and formaldehyde + OH reactions.CH 3 O 2 radicals are primarily formed