Volatile and intermediate-volatility non-methane organic gases (NMOGs) released from biomass burning were measured during
laboratory-simulated wildfires by proton-transfer-reaction time-of-flight mass spectrometry (PTR-ToF). We identified NMOG
contributors to more than 150 PTR ion masses using gas chromatography (GC) pre-separation with electron ionization,
Biomass burning, including wildfires, agricultural burning, and domestic fuel use, is a large source of non-methane organic gases (NMOGs) to the atmosphere (Crutzen and Andreae, 1990; Akagi et al., 2011). These compounds can be directly harmful to human health (Naeher et al., 2007) and contribute to the formation of secondary pollutants including ozone and secondary organic aerosol (SOA; Alvarado et al., 2009, 2015; Yokelson et al., 2009; Jaffe and Wigder, 2012). Because NMOGs from biomass burning are a complex mixture of many species that can change considerably depending on fuel and fire characteristics, many modeling and inventory efforts have had difficulty capturing subsequent chemistry in fire plumes (Alvarado et al., 2009; Grieshop et al., 2009; Wiedinmyer et al., 2011; Heald et al., 2011; Müller et al., 2016; Reddington et al., 2016; Shrivastava et al., 2017). Additionally, a substantial portion of gas-phase carbon may be missing from many field measurements (Warneke et al., 2011; Yokelson et al., 2013; Hatch et al., 2017) and the gas-phase precursors of SOA are not sufficiently understood (Jathar et al., 2014; Alvarado et al., 2015; Hatch et al., 2017). For these reasons, it is important to develop and understand analytical techniques that quantify a large number of biomass burning NMOGs.
Gas chromatography (GC) techniques have been used to identify NMOGs emitted by biomass burning in high chemical detail and provide exact isomer identifications (Hatch et al., 2015, 2017; Gilman et al., 2015). However, online GC techniques do not provide continuous measurement and are limited to certain classes of NMOGs depending on the column(s) selected and required sample preconditioning steps. This makes them nonideal for some important compounds or situations in which fast, continuous measurements are necessary. Whole-air sampling followed by GC can improve the time resolution but is affected by artifacts from canister storage (Lerner et al., 2017).
Proton-transfer-reaction mass spectrometry (PTR-MS) is a complementary technique widely used in atmospheric chemistry,
both stand-alone and with a GC interface (de Gouw and Warneke, 2007; Yuan et al., 2017). This chemical ionization technique
uses
Several recent papers have reported the use of high-resolution PTR-ToF to measure biomass burning NMOGs in the laboratory (Stockwell et al., 2015; Bruns et al., 2017) and the environment (Brilli et al., 2014; Müller et al., 2016). Hatch et al. (2017) suggest that PTR-ToF measures a substantial fraction (50–80 %) of total NMOG carbon mass. The mass spectra resulting from PTR-ToF detection of biomass burning NMOGs are complex, and many peak assignments are tentative. However, it is clear that PTR-ToF can provide detailed NMOG measurements relevant to studying the effects of fire emissions on human health and ozone and secondary organic aerosol formation.
A PTR-ToF instrument (Yuan et al., 2016) was deployed during the Fire Influence on Regional and Global Environments Experiment (FIREX) 2016 intensive at the US Forest Service Fire Sciences Laboratory in Missoula, Montana. This experiment burned a series of natural fuels and characterized the gas- and particle-phase emissions with a range of instrumentation (Selimovic et al., 2017). The aging of these emissions was explored with additional chamber experiments (described elsewhere). In this paper we describe the PTR-ToF instrument operation and interpretation of measurements. The focus is on direct emissions. Building on work by Stockwell et al. (2015), Brilli et al. (2014), and others, we provide new, more detailed, and more highly time-resolved chemistry of NMOG emissions from biomass burning than previously available.
The purposes of this work are to improve our understanding of the complex NMOG emissions from biomass burning by
interpreting PTR-ToF measurements of biomass burning emissions, provide emission factors and emission ratios to CO for
many NMOGs, link PTR-ToF measurements to GC, Fourier transform infrared spectroscopy (FTIR), and iodide CIMS (
Controlled biomass combustion experiments were conducted in a large
(12.5
Two types of combustion experiments were conducted. In the first set of experiments, the “stack burns”, emissions were
entrained into the ventilation stack and measured from the 17
Instrumentation details.
An overview of the instruments referenced in this work is given in Table 1.
The PTR-ToF instrument is a chemical ionization mass spectrometer typically using
The instrument used in this work is very similar to that described by Yuan et al. (2016), with two relevant
differences. The PTR-ToF instrument described by Yuan et al. (2016) includes two RF-only segmented quadrupole ion guides
between the drift tube and time-of-flight mass analyzer, while the current version has only one ion guide. The effects of
this are that the sensitivities are slightly higher (
The PTR-ToF instrument is equipped with a switchable reagent ion source that allows for
In
The PTR-ToF instrument transfer inlet was
The gas chromatograph (GC) instrument cryogenically pre-concentrates 4 min samples of NMOGs before separation on one of
two capillary columns (Lerner et al., 2017). The sample stream is separated into two channels that are optimized to
reduce water and carbon dioxide before cryogenic trapping of NMOG. The first channel (trapping at
The eluent from the GC columns was directed to either an electron ionization (EI) quadrupole mass spectrometer (Agilent
model 5975C) or to the PTR-ToF instrument. The quadrupole mass spectrometer has unit mass resolution and was operated in full ion
scan mode from
The GC inlet for stack burns was
Two stack experiments (both Douglas fir) were measured with both GC-EI-MS and GC-PTR-ToF; one stack experiment (Engelmann
spruce duff) and three room experiments (Douglas fir, subalpine fir, and sage) were each measured with GC-EI-MS,
GC-PTR-ToF, and GC-
A number of trace gases measured by the PTR-ToF instrument were also measured by other instruments (Table 1), and in Sect.
Glyoxal, methylglyoxal, and HONO were measured with the NOAA Airborne Cavity-Enhanced Spectrometer (ACES) instrument, which
uses broadband cavity-enhanced spectroscopy. Wavelength-resolved gas-phase extinction was measured in two spectral
regions, one in the UV (361 to 390
The
The calibration factor (units of
The calibration factors of 37 species were determined experimentally by introducing a known concentration of an NMOG from a standard cylinder, a permeation source (Veres et al., 2010a), a diffusion cell for isocyanic acid and methyl isocyanate (Roberts et al., 2010), or a liquid calibration unit (Ionicon Analytik). The calibration factors of these species have an error of 15 % (details in Table S1).
It is unrealistic to experimentally determine calibration factors for all NMOG species detected in biomass burning. Many compounds are highly reactive and cannot be purchased from a commercial supplier. Several methods to estimate calibration factors have been previously used by PTR-MS operators. For example, both Warneke et al. (2011) and Stockwell et al. (2015) estimated calibration factors for uncalibrated species based on ion mass-to-charge ratio and chemical formula in the latter case.
Comparison of measured and calculated calibration factors for several NMOGs. The nine
compounds used to determine the calibration proportionality constant are highlighted as red squares.
The shaded area shows an uncertainty of
Sekimoto et al. (2017) recently developed an improved method of estimating calibration factors. The instrument
calibration factor is linearly proportional to the kinetic capture rate constant of the
If an identified ion mass has only one NMOG contributing, as is the case for 65 % (102) of the ion masses with signal
in the fire, we used the calibration factor from direct calibration or the Sekimoto et al. (2017) method. If an
identified ion mass has more than one NMOG contributing, we used a weighted average of the calibration factors of all NMOG
contributors to this ion mass (Eq. 1). The determination of relative NMOG contributions to the total ion signal of each
individual mass was based on GC-PTR-ToF measurements, comparisons to other instruments, time series analysis, and reported
values from the literature and will be described in Sect.
The uncertainty for calibration factors for identified NMOGs ranges from 15 to 50 % depending on the calibration method used (Table S1). For ion masses for which we were not able to propose a NMOG, a calibration factor was estimated based on the elemental composition of the ion mass (Sekimoto et al., 2017). The uncertainty for calibration factors for unidentified species is within 10 % higher to 50 % lower.
Ambient humidity can change the measured sensitivity of an NMOG species (Yuan et al., 2016). For species whose calibration
factor was measured, a humidity correction factor was also experimentally determined. We currently have no method to
predict the humidity dependence of the sensitivity for other species, so for all other species no humidity correction was
applied. To minimize the error from this omission, we calibrated compounds that were abundant in emissions and that likely
have strong humidity dependence. These include compounds with proton affinities close to water (e.g., HNCO) and compounds
whose ionization mechanism includes loss of water (e.g., 1-propanol). Excluding these compounds, the average measured
humidity correction factor was less than 15 % for the humidity conditions experienced during FIREX
(5–18
During the Fire Lab experiments we measured 574 ions that were enhanced in emissions from one or more fuel types. Of
these, we identified 156 ion masses with a high degree of certainty and for which a calibration factor can be
determined. An additional 12 ion masses were identified as fragments of one or more NMOGs whose main product ion was
already included in the list of 156 ions. Finally, four ions were identified as being a common product of a large number of
structurally dissimilar NMOGs. These 172 ions, their identification, and support for that identification are listed in the
Supplement. The Supplement provides detailed information on the isomer contributions to each mass
(Table S1), sensitivities and calibration uncertainty (Tables S1 and S10), literature references (Table S6 in the
Supplement), GC measurements (Table S7 in the Supplement), and observations from time series correlations (Table S9 in the
Supplement). The Supplement additionally includes quantitative information on OH rate constants (Table S5 in the
Supplement), instrument intercomparisons (Table S8 in the Supplement), and
Identifications of many NMOGs emitted from biomass burning have been previously reported using GC, PTR-MS, and optical methods. We compiled a list of observed NMOGs and identifications to use as a starting point. The papers we referenced included Karl et al. (2007), Warneke et al. (2011), Brilli et al. (2014), Stockwell et al. (2015), Müller et al. (2016), and Bruns et al. (2017), which focus on PTR-MS measurements, and Gilman et al. (2015) and Hatch et al. (2015, 2017), which focus on GC measurements. Gilman et al. (2015) used 1-D-GC and focused on the most volatile species, and Hatch et al. (2015, 2017) used 2-D-GC and included many additional less volatile species. NMOG emission factors of identified compounds and the estimated mass of unidentified species have been reviewed by fire and/or ecosystem type globally (e.g., Akagi et al., 2011; Yokelson et al., 2013), but significant recent measurements have not yet been included in the online updates: e.g., Hatch et al. (2017). Finally, for some compounds, we referenced studies of the pyrolysis products of lignin, cellulose, and hemicellulose, which used GC-MS, X-ray spectroscopy, FTIR, theoretical calculations, and other analytical methods to identify major products and common reaction pathways (Patwardhan et al., 2009; Lu et al., 2011; Zhang et al., 2012; Heigenmoser et al., 2013; Collard and Blin, 2014; Liu et al., 2017a).
We assessed each identification as strongly or weakly supported. Strong identifications include those reported by many separate studies, NMOGs identified using GC methods (especially 2-D-GC-ToF-MS), and those supported by evidence from pyrolysis or other literature. Weak identifications include those with disagreement between different studies, tentative identifications based on only mass-to-charge ratio or elemental formula, and identifications that are inconsistent with reported formula or that are chemically implausible (e.g., highly strained structure). Identifications from the literature and citations are listed in Table S6. Overall we found literature evidence for 68 % of our ion identifications. Our interpretation differs from previously published PTR-MS interpretations for 34 ion masses as noted in Table S6. Forty-eight ion masses have not been previously reported in PTR-MS measurements of biomass burning.
The compounds with new and revised identifications were compared to review values of emission factors in Akagi et al. (2011) and Yokelson et al. (2013). A limited number of species from PTR are included in these reviews, largely because of uncertainty in identification. PTR species that have been detected but not included in review tables of EF include many more highly functionalized and larger molecules, and most of our updated identifications are these species. Yokelson et al. (2013) do include a number species from PTR (ion trap) that were not identified, and the identities of many of these have now been determined in this work.
Compounds that are included in review tables and for which we have updated the assignment are mostly unsaturated hydrocarbons and heteroatom-containing species, for which the identifications have been updated to include other contributing VOCs. For such species whose EF was determined solely from PTR, the actual emission factor should be lower than the reported value.
Gas chromatographic separation before measurement with PTR-MS is a powerful tool that has been widely used in many
environments (Warneke et al., 2003; Karl et al., 2007; Warneke et al., 2011; Yuan et al., 2014). The combination of
measured chromatographic retention time and product ions with GC-PTR-ToF, GC-
The relative intensities of the eluted peaks were used to quantify the relative contribution of each NMOG to each ion
mass. The size of a chromatographic peak is determined not only by the mixing ratio of that NMOG in ambient air and the
mass spectrometer response, but also by the trapping and elution efficiencies of the GC pre-separation unit. As isomers
have the same molecular weight and elemental composition, their volatilities and trapping efficiencies are generally
similar. For example, pyrrole and 3-butene nitrile have similar vapor pressures of 1.1 and 2.5
The same GC methods were used to identify some signals from the
At
Some species measured by the PTR-ToF instrument have several possible isomers, have not been previously reported in the literature, and are not transmittable through the GC. The identifications of these compounds are less certain. For these, we selected several reasonable isomeric structures based on the types of compounds typically seen in biomass burning emissions: substituted furans and aromatics, nitriles, pyridines, terpenes, and carbonyls. Then, we compared the temporal profile of these ion signals during several fires to compounds with more certain identification. Compounds with similar structure and functionality likely have similar behavior. Dissimilar compounds can also sometimes have similar temporal profiles (Yokelson et al., 1996), but it is still likely that time series correlation points to the correct assignment or a species with similar chemical functionality as the true assignment.
An example of how time series correlation is used to identify a species is shown in Fig. 3;
The contribution of isomers to any particular PTR ion exact mass was consistent among the four fuels (Douglas fir,
Engelmann spruce duff, subalpine fir, and sage) sampled with GC-PTR-ToF (Table S7). Comparing all GC-PTR-ToF samples, the
isomeric speciation on a particular exact mass typically varied by only 11 % (the SD of the contribution of each
isomer to total signal on that mass) and therefore the same study-average NMOG contributions to each ion exact mass were
used for all fuel types, regardless of whether or not supporting GC information was available. This is similar to the variation
in
isomer speciation reported by Hatch et al. (2015; 5 % on average), who investigated six diverse fuel types. Compounds
that had larger variability between GC-PTR-ToF samples (and between fuel types) include
The instantaneous speciation of isomers may also change over the course of a fire, especially as the fire shifts between various higher- and lower-temperature chemical processes. We used time series correlation to identify several masses that may have variable NMOG contributors. This analysis was done on Fire 2, which burned representative ponderosa pine forest-type fuels. This fire was selected because ponderosa pine was the most comprehensively measured fuel type during the FIREX 2016 experiment, this particular fire had distinctly different NMOG speciation at the beginning (higher temperature) and end (lower temperature) of the fire, and reagent ion depletion did not affect the results.
We identified three ions with a high signal whose NMOG contributors may be substantially different between high- and
low-temperature processes in a fire:
These three pairs of identifications in Fig. S2 and their relative contributions to total ion signal are not well
constrained. An additional instrument technique, such as a fast GC capable of separating substituted furans and aromatics
or a better understanding of
Several species detected by the PTR-ToF instrument were also measured by other instruments. The intercomparison is summarized in
Fig. 4. All slopes shown in the figure and discussed in the text are the orthogonal distance regression (ODR) slope of
Fifteen species were compared between the PTR-ToF and FTIR instruments (Fig. S4). Methanol, formaldehyde, formic acid, propene,
acetic acid, ethene, acetylene, furan, phenol, and furfural were calibrated directly on the PTR-ToF instrument and have an uncertainty
of 15 %. For HONO, HCN, and ammonia, we were not able to determine a calibration factor directly and so we set the
calibration factors equal to the slope of the comparison between the FTIR and PTR-ToF instruments during Fire 72 (ponderosa pine with
realistic fuel mixture, selected for early data availability, long burning time of 30 min, and mix of flaming and
smoldering conditions). Sensitivity to HCN has strong humidity dependence (Knighton et al., 2009; Moussa et al., 2016),
and this was experimentally determined and corrected. Glycolaldehyde was calibrated using the method from Sekimoto
et al. (2017) with an uncertainty of 50 %; the PTR-ToF measurement of
Methanol has agreed within stated uncertainties between PTR-MS and FTIR in several previous studies (Christian, 2004; Karl
et al., 2007; Warneke et al., 2011; Stockwell et al., 2015), and this work shows an average slope of 0.99 and
The high degree of correlation between PTR-ToF and FTIR for acetylene and ethene is notable because these two compounds
cannot be ionized by proton transfer from
Other compounds, including 1,3-butadiene, furan, hydroxyacetone, phenol, and furfural, agreed within a factor
of 2
(slopes of 1.6, 1.5, 0.6, 0.7, and 0.6, respectively) and average
Three species were compared between the PTR-ToF and ACES instruments: HONO, glyoxal, and methylglyoxal (Fig. S3). HONO agrees with an
average slope of 1.13 and
The comparison of glyoxal is similarly poor (slope
Some data were compared to
We quantified the emission ratios relative to CO and the emission factors in
Ion exact masses, formulas, and NMOG contributor(s); the emission ratios and emission factors of those contributors.
Continued.
Comparison of emission ratios (
Comparison of emission ratios to Stockwell et al. (2015). The dashed line in each panel shows a
The emission ratios and emission factors of the identified compounds averaged over all fires are reported in Table 2. Emission ratios and emission factors of both identified and unidentified compounds for specific fuel types are given in Tables S2 and S3. The large relative SDs of both emission ratio and emission factor for each NMOG indicate large differences in emission composition between different fires. An analysis of the differences in emission composition between different fuels and combustion processes will be presented in a separate paper. Figure 5 compares the average emission ratios determined in this work to several other studies. Our emission ratios have similar values, ranging from a factor of 1.7 higher on average than Gilman et al. (2015) to 0.7 higher than the average of Stockwell et al. (2015). The differences in slopes and scatter are likely due to different fuel types, fire conditions, and sampling strategies. Stockwell et al. (2015) also reported detailed speciation within particular structural categories (non-oxygenated aromatics, phenols, and furans). We compared our speciation for comparable fuel types – coniferous canopy, chaparral, and peat – and the agreement for coniferous fuels and chaparral is within a factor of 2 despite differences in ion identification and calibration factor (Fig. 6). The ER to CO are likely the easiest way to incorporate this new NMOG data into models since CO emissions from wildfires are relatively well characterized (Liu et al., 2017b).
The 156 PTR ions for which we have identified the NMOG contributors account for a significant fraction of the instrument
signal and total NMOG detected by the PTR-ToF instrument in each fire. Across all 58 stack fires measured with PTR-ToF, an average
of 90 % of the instrument signal from
Histogram of total emission (quantified as emission ratio to CO) of
identified and unidentified NMOGs sorted by
In terms of NMOG mass detected by PTR-ToF, an average of 88 % and a minimum of 82 % is accounted for by identified
species (Fig. 7b). This is an improvement over Warneke et al. (2011), in which only 50–75 % of the detected mass was
identified, and is comparable to Stockwell et al. (2015), with improved identification of emissions from peat and updated
ion assignments (Table S6). Identifying the NMOG contributors to additional ions will not increase this by much because
the remaining (unidentified) ions each account for only a small part of the remaining signal. The unidentified portion is
a small fraction of the overall detected emissions, but compared to the identified portion, it consists of species that
are heavier, contain more oxygen atoms, and are less volatile (Fig. 8). The average molecular mass of unidentified species
is 120
The detected and identified NMOGs fall into several broad structural categories: furan-type compounds, benzene-type compounds (aromatics), terpenes, non-aromatic molecules containing oxygen, nitrogen, or sulfur, and other hydrocarbons (mostly alkenes). We also included pyrroles, thiophenes, and pyridines as structural categories, but these account for less than 1 % of detected emissions on a molar basis. Terpenes include isoprene, monoterpenes, oxygenated monoterpenes, and sesquiterpenes. Non-aromatic oxygen-containing molecules include alkyl carbonyls, esters, and acids. Non-aromatic nitrogen-containing molecules include HCN, HONO, isocyanic acid, methyl isocyanate, amines (including ammonia), and nitriles. Aromatics and furans include alkyl-substituted and oxygenated derivatives of benzene and furan. On average over all fires, non-aromatic oxygenates were the most abundant, comprising 51 % of detected emissions (Fig. 9a). The compounds in each category include a range of functional groups, of which alcohols and carbonyls were the most abundant (Fig. 9b). Many compounds also include an alkene functional group. Some compounds, such as guaiacol, have several functional groups. In these cases, the NMOG was counted once in each category.
Compared to several previous laboratory studies reporting highly chemically detailed emissions using GC instruments (Hatch
et al., 2015; Gilman et al., 2015; Hatch et al., 2017), we observed a similar range and type of speciation for
non-oxygenated aromatics, thiophenes, pyrroles, pyridines, alkyl nitriles, alkyl ketones, alkyl esters, and small
alcohols. However, this work and a previous PTR-MS study (Stockwell et al., 2015) also observed more highly substituted
oxygen-containing aromatics and furans, such as hydroxymethylfuranone and syringol. These substituted compounds contribute
significant additional reactivity. For example, Gilman et al. (2015), who studied similar fuels, reported OH reactivity of
1.3–5.5
Reaction with the hydroxyl radical (
Volatility of NMOGs during Fire 2 (ponderosa pine). For simplicity, ammonia is excluded from this figure because of its very high concentration (600
The volatility distribution of emitted species also changes over the course of these lab fires. We determined the
saturation vapor concentration (
Gas-phase emissions of NMOGs and some inorganic compounds were measured with a high-resolution PTR-ToF instrument during
the FIREX 2016 laboratory intensive. Using a combination of techniques, including GC pre-separation,
This work provides a guide to interpreting PTR-ToF measurements of biomass burning that is strongly supported by the literature and complementary analytical techniques. This will serve as a foundation for future use of FIREX 2016 PTR-ToF data and interpretation of PTR-ToF field measurements. Finally, this work provides the best available emission factors and emission ratios to CO for many wildfire-generated NMOGs.
Data are available from the
CSD NOAA archive at
The supplement related to this article is available online at:
Joost de Gouw worked as a consultant for Aerodyne Research, Inc. during part of the preparation phase of this paper.
Abigail R. Koss acknowledges funding from the NSF Graduate Fellowship Program. Kanako Sekimoto acknowledges funding from the Postdoctoral Fellowships for Research Abroad from the Japan Society for the Promotion of Science (JSPS) and a Grant-in-Aid for Young Scientists (B) (15K16117) from the Ministry of Education, Culture, Sports, Science and Technology of Japan. Robert J. Yokelson and Vanessa Selimovic were supported by NOAA-CPO grant NA16OAR4310100. Jordan R. Krechmer and Jose L. Jimenez were supported by DOE (BER/ASR) DE-SC0016559. We thank the USFS Missoula Fire Sciences Laboratory for their help in conducting these experiments. This work was also supported by NOAA's Climate Research and Health of the Atmosphere initiative. Edited by: Robert McLaren Reviewed by: two anonymous referees