2.1 Site description

During the week of 29 April 2018, a total of seven small plots (450 to 900 m²) were burned using prescribed fire at Ft. Jackson, South Carolina (SC; 34.05^∘ N, 80.83^∘ W), approximately 10 km east of Columbia, SC. The fort lies entirely within the Sandhills ecosystem in the SC coastal plain, which runs approximately parallel to the Atlantic Ocean coast, 175 km inland. The Sandhills region thus forms a belt that tracks southwest–northeast across sands of varying depth with a high content of pure silica (Porcher and Rayner, 2001). The deep sands support an overstory vegetation that has significant amounts of turkey oak (Quercus laevis Walter) and two native pine species relatively unique to the southeastern US: longleaf pine (Pinus palustris Mill.) and slash pine (Pinus elliottii Engelm.). The understory has substantial quantities of immature turkey oak as well as longleaf and slash pine, along with sparkleberry (Vaccinium arboreum Marshall), and a heterogeneous organic layer of woody material, litter, duff, and cones atop the mineral soil. The longleaf ecosystem depends on fire for maintenance (Cary, 1932). Site details for the seven burn plots, all with a 2-year rough (i.e., burned 2 years prior), are summarized in Table 1. Eight prefire and postfire 1.0 m² biomass clipped plots were established at each 160 m² research block within the larger burn plots where organic vegetative material was collected before and after each fire. Shrub, grasses, forbs, down woody material (0–0.6, 0.6–2.5, 2.5–7.6, 7.6–22.9 cm in diameter), litter, and duff are the major fuel bed components that were targeted. Fuel moisture samples for each major component were collected before ignition to determine fuel moisture content for each fuel bed component. Figure 1 shows photographs of site 16, plot 1 before, during, and after the burn as well as a thermal image of the flame interacting with the fuel.

Figure 1Photographs of site 16, plot 1 on 3 May 2018 between 14:00 and 14:40 local time. The plot (a) before the flame, (b) while the flame interacted with the fuel at 14:33, and (c) smoldering combustion of the fuel. The primary species seen in the understory for this burn plot are sparkleberry and a litter layer of pine needles. (d) Thermal image of the flame interacting with the fuel at 14:33.

Download

Table 1Plot name, date and time, fuel description, atmospheric conditions (all clear-sky days), and area for the prescribed burns.

Download Print Version | Download XLSX

2.2 Collection device

Our approach to sampling used an extractive collection device whose tube inlet sampled air and emissions directly ahead of the flame. This simple solution is similar to other canister methods often used with gas chromatographic analysis (Young et al., 1997) and also conceptually similar to the land-based FTIR used to sample emissions as described by Akagi et al. (2013, 2014) and Burling et al. (2011). The canister sampling package, mounted on a metal frame, contained a set of evacuated canisters that were carried to the individual burn plots. The sampling package consisted of a 12 V swing-piston KNF Neuberger pump (NPK09DC) plumbed with stainless-steel tubing and a pressure-relief valve to regulate the pressure of the system and ultimately the fill pressure of the canisters. The flow rate to fill the canisters was 15 L min⁻¹. A sampling probe (2.5 m of 6 mm stainless-steel tubing plus 2 m of flexible stainless-steel line) was attached to the inlet of the package to collect pyrolysis gases from point sources of vegetation within the burning plots. The device had an in-line two-way valve to control the sampling interval. To capture a pyrolysis sample, the probe was placed near the base of the flame, immediately above the fuel where the pyrolysis gases should be emitted at maximal levels. A total of 7 to 10 aliquots of gas sample were added to a single canister as the device was moved in front of the flame to capture pyrolysis gases. Each 3 L Summa canister was filled to approximately 138 kPa (20 psia) for the FTIR analysis.

2.3 FTIR spectrometer and spectral analysis

Experimental details regarding FTIR measurement and ensuing spectral analysis procedures have been previously reported (Scharko et al., 2019) but are briefly summarized as follows: gases in the canisters were returned from the field to the laboratory and analyzed the same day or the following day using an 8 m White cell (Bruker A136/2-L) and FTIR; canisters were connected to the gas cell via $\mathrm{3}/{\mathrm{8}}^{\prime \prime }$ stainless-steel tubing with both the tubing and gas cell heated to 70 ^∘C to prevent analyte adhesion. The cell was coupled to a purged FTIR spectrometer (Bruker Tensor 37) equipped with a glow bar source, KBr beam splitter, and liquid-N₂-cooled mercury cadmium telluride detector. Spectra were collected from 4000 to 500 cm⁻¹ at 0.6 cm⁻¹ resolution. Spectral analysis was carried out using the MALT5 program (Griffith, 2016) and 50 ^∘C reference spectra from the PNNL database (Sharpe et al., 2004; Johnson et al., 2010) as well as absorption lines from HITRAN (Gordon et al., 2017). MALT5 fits the assigned reference spectral lines to the measured spectrum by optimizing the fit of all gases ascribed to the spectral window and minimizing the residual. The calculation involves input parameters such as path length, resolution, and apodization accompanied by reference absorption cross sections and the measured spectrum with its associated temperature–pressure values. Both H₂O and CO₂ had peaks that were saturated; these regions were eliminated from analysis. In some instances, peaks for the gases of interest were also saturated, in which case the pressure in the gas cell was reduced and the measurement repeated.

2.4 Calculation of emission ratios and emission factors

A convenient quantity to compare emissions is the emission ratio (ER). This ratio is calculated by computing the change in the concentration of the analyte of interest relative to that of a co-emitted, long-lived gas, typically CO or CO₂. For the present study, the chosen long-lived gas is CO:

It is important to note that these are the changes in the analyte and CO relative to the background atmosphere (i.e., relative to ambient “clean air” conditions). The background levels of CO and CO₂ were measured using an open-path gas analyzer (OPAG 22) prior to the series of burns. The initial CO₂ level was measured to be 409 ppm (this value agrees with the global averaged CO₂ for May 2018 of 408.7 ppm; Dlugokencky and Tans, 2012), whereas the CO level was often below the OPAG detection limit. (Due to challenging experimental conditions and moderate CO band intensities, the OPAG detection limit was poor for these CO measurements and could not achieve 200 ppb.) Without an instrument to measure ambient CO with sufficient sensitivity we chose 200 ppb for an estimated background level, which is within the range for a typical CO mixing ratio (Seinfeld and Pandis, 2012). We note that the 200 ppb value is sufficiently small that it has a negligible effect on the calculated Δanalyte ∕ ΔCO ratios. Emission ratios can be calculated for a single point in time during the fire, or they can incorporate the full length of the fire. The present ERs were calculated based on the contents of the individual canisters, which represent multiple aliquots, all from the early fire stages. Other studies have obtained fire-integrated ERs, which integrate over the entire duration of the fire (Koss et al., 2018), or fire-averaged ERs determined from the slope of the regression with the intercept set to zero (Yokelson et al., 1999).

Another useful quantity is the emission factor (EF), defined as the number of grams emitted of a given analyte per kilogram of dry fuel consumed and estimated using the following equation (Yokelson et al., 1999; Ward and Radke, 1993):

where F_carbon is the mass fraction of carbon in the fuel, MW_analyte and MW_carbon are the molar masses of the analyte and carbon, respectively, $\frac{\mathrm{\Delta }\mathrm{analyte}}{\mathrm{\Delta }{\mathrm{CO}}_{\mathrm{2}}}$ is the emission ratio of the analyte relative to CO₂, $\frac{\mathrm{\Delta }{\mathrm{C}}_j}{\mathrm{\Delta }{\mathrm{CO}}_{\mathrm{2}}}$ is the emission ratio of species j relative to CO₂, and NC_j is the number of carbons in species j. Note that ΔCO₂ cancels out in Eq. (2). Elemental analysis of similar southeastern fuels was reported in a previous study (Safdari et al., 2018), and the average carbon content by mass for longleaf pine foliage and litter as well as sparkleberry was 0.52, which was the value used for F_carbon. One assumption in Eq. (2) is that all of the carbon in the fuel is released and accounted for in the measurements of the j carbon species. Most carbon emissions are in the chemical form of CO₂, CO, or CH₄. It should be noted that the EF quantities reported here include only compounds measured by the FTIR, and EF values may be overestimated by 1 % to 2 % for most fuels due to undetected carbon species (Akagi et al., 2011).

3.1 Estimating the contribution from high- and low-temperature processes

Modified combustion efficiency (MCE) uses the excess CO and CO₂ values to quantify the degree of combustion that has occurred in a given smoke plume, and the MCE will be discussed below. However, in a recent study Sekimoto et al. (2018) suggested that MCE may not be the best quantity to adequately describe pyrolysis, but rather that emissions of volatile organic compounds (VOCs) from biomass burning may be correlated with high- and low-temperature pyrolysis factors obtained by carrying out positive matrix factorization (PMF) analysis on the emission profiles. The authors further suggested that the ratio of acetylene (C₂H₂) to furan (C₄H₄O) could be used to estimate the high- and low-temperature pyrolysis factors. They used the emission profiles from the analysis of 15 different fuels to calculate the following ratio that estimates the high- and low-temperature VOC emissions:

We have adopted this estimation approach and have used the acetylene-to-furan ratio to assess the relative contributions from high- and low-temperature processes. The average results are displayed in Fig. 2 alongside the results from Koss et al. (2018), Gilman et al. (2015), and Akagi et al. (2013). For comparison purposes, the values displayed in Fig. 2 were determined using average ERs for acetylene and furan. The present results (right-most bar) are approximately an order of magnitude greater than all previous studies, likely due to the timing of collection and the sampling probe's proximity to the flame. The juxtaposed values from the previous studies were obtained using (a) fire-integrated ERs, (b) discrete ERs sampled every 20 to 300 s, or (c) fire-averaged ERs, all of which incorporate several different phases of the fire compared to the present flame-front measurements. Using the Sekimoto et al. (2018) estimation approach, higher acetylene-to-furan ratios indicate a greater contribution from the high-temperature processes. The markedly high ratio observed in this study suggests that samples were collected when high-temperature pyrolysis was indeed the dominant process. This observation is consistent with the time profile for the contribution of the high-temperature pyrolysis factor presented by Sekimoto et al. (2018), which demonstrates that the contribution from high-temperature pyrolysis [high-T ∕ (high-T + low-T)] can easily exceed 0.95 in the early stages of fire but is reduced to smaller fractions (≤0.3) in the latter stages. Another key difference is that the sampling probe used at Ft. Jackson was positioned to extract gases directly before the flame front, yet in close proximity to it, in order to limit further reactions. In particular, if the highly flammable acetylene molecules were captured prior to subsequent oxidation reactions, this would explain the enhanced ratio of high- to low-temperature VOC emissions as seen in Fig. 2.

Figure 2Ratio of acetylene (C₂H₂) ∕ 0.0393 to furan (C₄H₄O) ∕ 0.0159 to predict the ratio of high- to low-temperature VOC emissions as outlined by Sekimoto et al. (2018). Error bars represent 1σ. For the present study average results were determined from the 10 collected samples preceding the flame front for acetylene and furan. Koss et al. (2018) values were fire integrated, while Gilman et al. (2015) used 20–300 s integrations. Akagi et al. (2013) reported EFs that were used to calculate ERs for acetylene and furan.

Download

3.2 CO₂ and CO emissions and MCE values

As expected, other than H₂O vapor, CO and CO₂ were the predominant gases observed as emissions. Table 2 displays the EF (g kg⁻¹) and ER (ppb ∕ ppm_CO) values averaged for the 10 field measurements. Most prescribed burn studies have focused only on the flaming and smoldering stages. The flaming stage is characterized by more oxidized products and a higher modified combustion efficiency (e.g., 0.85–0.97, even 0.99 for pure flaming) (Ward and Hao, 1991), where the MCE is defined as

The smoldering phase, with lower MCE values (typically ranging 0.65–0.85) (Urbanski, 2013), displays more nonoxidized (or less-oxidized) species but with a greater fraction of OVOCs observed. The arithmetic mean MCE and standard deviation for all 10 measurements at Ft. Jackson was 0.83±0.04. Such MCE values would normally characterize data gathered during smoldering combustion wherein a combination of processes such as pyrolysis along with glowing combustion of char take place (Yokelson et al., 1997). Since the present study was aimed at the collection of pyrolysis gases preceding the flame front, characterizing the results in terms of MCE values may not be appropriate: the lower MCE values do not represent the fire being in the smoldering stage but rather suggest that pyrolysis and other early-phase process gases were captured (at least in part) prior to the onset of combustion. As noted, the methodology used with this collection device ideally extracts the pyrolysis gases before they are combusted. Due to the proximity of these gases to the flame, some entrainment of ambient air and air from the flame region was likely unavoidable. As we were sampling a moving zone, some combustion products were also likely to be sampled.

Table 2Study averages of EF (g kg⁻¹) and ER (ppb ∕ ppm_CO) for the 10 pyrolysis measurements along with standard deviation (SD). The SD represents the variation for the 10 nonidentical measurements. For the study averages, the arithmetic mean MCE was found to be 0.83±0.04; n/a – not applicable.

Download Print Version | Download XLSX

3.3 Emissions of lightweight hydrocarbons

Besides CO and CO₂, the second most abundant class of gases generated during the prescribed burns was lightweight hydrocarbons (HCs). The lightweight HCs detected by the FTIR include methane, ethane, ethene, acetylene, propene, allene, 1,3-butadiene, isoprene, and isobutene. Most have been previously identified in fire emissions using FTIR either in laboratory experiments (Burling et al., 2010; Christian et al., 2003, 2004; Gilman et al., 2015; Goode et al., 1999; Hatch et al., 2017; Selimovic et al., 2018; Stockwell et al., 2014; Yokelson et al., 1996, 1997) or field settings (Akagi et al., 2013, 2014; Alves et al., 2010; Burling et al., 2011; Goode et al., 2000; Hurst et al., 1994a, b; Karl et al., 2007; Paton-Walsh et al., 2010), but the present experiment reports the first IR detection of allene (Scharko et al., 2019). Figure 3 shows the individual correlations between these lightweight HCs and excess CO mixing ratios. The analyte vs. ΔCO correlation coefficients range from 0.97 (ethene and allene) to 0.66 (isoprene and isobutene). In all cases, the correlation coefficients were larger with CO than with CO₂. Positive relationships have been observed for CO correlations in previous burning studies (Hurst et al., 1994a, b).

Figure 3Mixing ratios (ppm) for the 10 measurements as a function of excess CO (ppm) for (a) ethane (C₂H₆), (b) ethene (C₂H₄), (c) acetylene (C₂H₂), (d) allene (C₃H₄), (e) propene (C₃H₆), (f) 1,3-butadiene (C₄H₆), (g) excess methane (CH₄), (h) isoprene (C₅H₈), and (i) isobutene (C₄H₈). The dashed lines are a linear fit to the data.

Download

While the observed ER for excess methane was comparable, ERs for ethene and acetylene were considerably greater than previously reported values. Specifically, Fig. 4 shows a comparison of emission ratios for methane, ethene, and acetylene to previously reported values of Gilman et al. (2015) and Akagi et al. (2013). As noted, different sampling methods complicate the comparison. The present data represent a collection of instantaneous grab samples extracted directly before the flame front, whereas the other data represent time-averaged values. Ethene and acetylene have both been observed as pyrolysis products in prior laboratory work (Palma, 2013) but may react further. For example, the addition reaction of acetylene to benzene or naphthalene can produce styrene or cyclopenta-fused polycyclic aromatic hydrocarbons (Ledesma et al., 2002). Alternatively, ethene and acetylene can undergo combustion (Simmie, 2003). Nevertheless, the high ER values for ethene and especially for acetylene in the present study further suggest that the samples were collected when the high-temperature pyrolysis process was dominant; Sekimoto et al. (2018) also observed that high-temperature pyrolysis profiles are often associated with unsaturated hydrocarbons.

Figure 4Average ERs (ppm ∕ ppm_CO) for (a) excess methane, (b) ethene, and (c) acetylene for this study and for previously published laboratory and field (ground and air based) investigations. Error bars represent 1σ. Gilman et al. (2015) present discrete ERs with sample acquisition of 20 to 300 s. Akagi et al. (2013) present fire-averaged EFs calculated using ERs derived by the regression method. The ERs for Akagi et al. (2013) shown above were derived from the ratio of the EFs for the gas of interest and CO multiplied by the molar mass of CO ∕ molar mass of the analyte.

Download

3.4 Emissions of lightweight oxygenated hydrocarbons

The noncyclic oxygenated hydrocarbons detected via FTIR analysis include formaldehyde, methanol, formic acid, acetaldehyde, acetone, acetic acid, and acrolein. On average, acetaldehyde and methanol had the highest ER values in this group, with ERs relative to CO of 0.009 and 0.006, respectively, as seen in both Table 2 and Fig. S2 in the Supplement. For all measurements collected at sites 16 and 24A, acetaldehyde was consistently the highest, with ER values ranging from 0.005 to 0.014. Site 24B (north diamond plot only) followed a different trend with highest ER values for acetic acid, followed by methanol, acetaldehyde, and formaldehyde in decreasing order. The ERs for acetic acid and formaldehyde at site 24B are at least 7.9 and 2.5 times greater, respectively, than the other burn sites. One key difference observed for site 24B was fuel composition, namely the presence and partial consumption of larger logs (i.e., 7.6–20.3 cm diameter woody material). Other differences include the presence of live pine seedlings and less turkey oak compared to other plots. This particular plot had the highest herbaceous and forb prefire loading and consumption with a higher fuel moisture content (205 % compared to the next highest value of 144 %). This high fuel moisture content was reflected in the ER for water, which was at least 4.7 times greater than the other plots. The pyrolysis of cellulose (one of the three primary components of biomass as discussed below) forms levoglucosan. Shen and Gu (2009) outline secondary decomposition pathways for levoglucosan, in which the initial step is rehydration to generate glucopyranose. They demonstrate how glucopyranose can then form formaldehyde, methanol, and acetic acid via secondary decomposition routes. This pathway (or a similar one) may have been favored at site 24B: the greater ERs for acetic acid and formaldehyde observed at plot 24B may have thus been influenced by the greater fraction of woody material and the presence of herbaceous and forb fuels, all with higher moisture contents. This hypothesis warrants further investigation and could be studied in the laboratory.

Table 3 compares the present ER values with values from Akagi et al. (2013), Stockwell et al. (2014), Gilman et al. (2015), and Koss et al. (2018). The present ERs are comparable to other burn studies except for acetaldehyde, which appears to be marginally greater, and formaldehyde and acetic acid, which both appear to be lower. The higher ratio for acetaldehyde may be due to differences in the sampling approach; i.e., samples collected in the present study may contain species that were generated during an earlier period in the thermal decomposition process. In a controlled laboratory study by Stein et al. (1983), acetaldehyde was observed as one of the initial products emitted from the pyrolysis of glycerol, a product pyrolyzed from levoglucosan. This same study also observed that acetaldehyde would continue to decompose (under pyrolysis conditions) to smaller molecules such as ethene, methane, H₂, and CO (Stein et al., 1983). The greater average ER for acetaldehyde observed in the present study may be due to gases being captured (via the collection device) and removed from heat either between decomposition steps or before combustion. It is also possible that the trends seen for the OVOCs in Table 3, in particular the higher values for acetaldehyde, arise due to (a) differing vapor pressures, (b) differing degrees of the onset of combustion, (c) differing degrees of pyrolysis emissivity as suggested by Stein et al. (1983), (d) differing degrees of IR-spectroscopic sensitivity (i.e., certain other species with limited or no detectability), or (e) some combination of all these effects. Further analysis is warranted.

Table 3Average emission ratios (ppb ∕ ppm_CO) for this study and for previously published fire studies; n/a – not applicable.

Koss et al. (2018) present the fire-integrated ERs. Gilman et al. (2015) present discrete ERs with sample acquisition of 20 to 300 s. Stockwell et al. (2014) present the fire-integrated ERs. Akagi et al. (2013) present fire-averaged EFs calculated using ERs derived by the regression method. The emission ratios for Akagi et al. (2013) were obtained from the ratio of the emission factors for the analyte and CO multiplied by the molar mass of CO ∕ molar mass of the analyte.

Download Print Version | Download XLSX

The slightly lower ERs for formaldehyde and acetic acid may in part be explained by secondary decomposition pathways. Proposed pathways that generate formaldehyde and acetic acid proceed through intermediates formed by the decomposition of levoglucosan (Shen and Gu, 2009). Formaldehyde is generated from a number of intermediates such as hydroxyacetone (acetol) (Lindenmaier et al., 2016) and 5-hydroxymethyl-furfural. While the formation mechanism for acetic acid is via the decomposition of the intermediate hydroxyacetaldehyde (glycolaldehyde) (Johnson et al., 2013), which undergoes a dehydration reaction to a ketene and then a rehydration to acetic acid (Shen and Gu, 2009), it is possible that the present conditions and fuels (save for site 24B) were not favorable for the above chemical pathways.

3.5 Emissions of aromatic compounds

In the present study, furan, furaldehyde, and naphthalene were all detected via FTIR. Previous fire studies have used FTIR to detect phenol and/or furan (Burling et al., 2011; Akagi et al., 2014; Hatch et al., 2017; Christian et al., 2003, 2004; Stockwell et al., 2014; Karl et al., 2007; Selimovic et al., 2018; Yokelson et al., 2013; Burling et al., 2010; Akagi et al., 2013). One of these studies also detected furaldehyde (Selimovic et al., 2018). To the best of our knowledge, however, this is the first burning study that has used IR spectroscopy to identify naphthalene vapor, though it has previously been detected in biomass burning emissions via other methods (Koss et al., 2018; Gilman et al., 2015). Naphthalene has also been detected in tar samples generated from the controlled pyrolysis of similar fuels (Safdari et al., 2018).

Phenol and phenolic compounds were not definitively observed in this study due to their IR bands being somewhat weak and obscured by a number of other species, namely acetic acid, carbon dioxide, acetylene, and hydrogen cyanide. However, phenolic compounds have been identified in products generated from the pyrolysis of lignin in controlled laboratory experiments by Kibet et al. (2012). Lignin, one of the three main components of biomass, can account for 10 %–35 % of the biomass, and its chemical structure consists of polymers of various phenolic alkyl side-chain subunits (Shen et al., 2015). When undergoing thermal decomposition, lignin will release volatiles at temperatures between 200 and 400 ^∘C. The proposed mechanism can generate intermediates such as phenoxy radicals that ultimately lead to the formation of phenols (Kibet et al., 2012). In the present study, spectral evidence of phenol was in fact observed in some measurements, but the IR bands at 1176 and 752 cm⁻¹ were weak and were masked by other compound signatures, hindering spectral quantification. Mixing ratios of phenol above the detection limit might have been anticipated since prior controlled pyrolysis investigations of sparkleberry and longleaf pine have detected phenol as a component in the tar (Safdari et al., 2018, 2019; Amini et al., 2019a). While the phenol signal was weak, furan and furaldehyde, however, were clearly detected, and their formation likely stemmed from the thermal degradation of the other main constituents of biomass. Besides lignin, the other primary macromolecular components are cellulose and hemicellulose, which account for approximately 50 % and 15 %–35 % by weight, respectively (Shen et al., 2015). The pyrolysis of cellulose is known to produce furaldehyde, furan, and other low-weight oxygenated compounds (e.g., acetic acid) via the intermediate levoglucosan (Bai et al., 2013). Moreover, furaldehyde and methanol have both been observed as volatile products from the pyrolysis of methyl β-D-xylopyranoside, a model compound for xylan-based hemicellulose (Shafizadeh et al., 1972).

Naphthalene is a polycyclic aromatic hydrocarbon with several sources, including as a biomass burning emission product. It was detected using FTIR for the first time in these studies (Scharko et al., 2019). Its IR detection was not unexpected given that it has been observed in collected tar samples generated by the laboratory pyrolysis of similar fuel types (Safdari et al., 2018), but its identification in an experimental IR spectrum can be challenging as depicted in Fig. 5. Most of its IR bands have only moderate cross sections with the exception of the ν₄₆ band, which has a strong Q branch at 782.3 cm⁻¹ (green trace in Fig. 5). For this band to be observed, however, it needs to be deconvoluted from the acetylene rotational–vibrational lines also present in this spectral domain (red trace in Fig. 5). Better retrievals for naphthalene were obtained using a higher spectral resolution (0.6 cm⁻¹) since the Q branch of the ν₄₆ band is quite sharp (FWHM ∼ 1 cm⁻¹), even at atmospheric pressure (Scharko et al., 2019) .

Figure 5Measured and scaled reference spectra for acetylene (C₂H₂) and naphthalene (C₁₀H₈) as well as residual with and without C₁₀H₈ fit. The measurement is from site 16, plot 6, msmt. 2, and the detected mixing ratio for naphthalene is 7.37 ppm. Spectra are offset for clarity. Reference absorption lines for acetylene are from HITRAN, and the reference spectrum for naphthalene is from PNNL.

Download

Figure 6(a) Mixing ratios (ppm) for naphthalene (C₁₀H₈) as a function of excess CO (ppm) measured by FTIR for each of the 10 canisters. The dashed line is a linear fit. (b) Average emission ratios (ppm${}_{{\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{8}}}$ ∕ ppm_CO) for this study and for previous laboratory studies. Error bars represent 1σ. Koss et al. (2018) present the fire-integrated ERs. Gilman et al. (2015) present discrete ERs with sample acquisition of 20 to 300 s.

Download

Figure 6a plots the mixing ratios (ppm) for naphthalene as a function of excess CO (ppm), while Fig. 6b displays the ERs for naphthalene for this study and previous studies. The average naphthalene ER for this study is substantially greater than the values from both Gilman et al. (2015) and Koss et al. (2018). The average for Koss et al. (2018), however, is in turn an order of magnitude greater than the highest average for Gilman et al. (2015). The higher ER for naphthalene in this study (shown in Fig. 6) clearly suggests that the method to capture pyrolysis gases was (at least in part) quite successful; i.e., we were able to collect naphthalene gas prior to it undergoing further reactions. Besides oxidation, under the right conditions naphthalene can also continue to react in a Diels–Alder type of reaction to form still larger polyaromatics (Fairburn et al., 1990; Richter and Howard, 2000). Sekimoto et al. (2018) also linked naphthalene with the high-temperature profile, and it appears that the samples in the present study were indeed collected when the high-temperature process was dominant. The detection of naphthalene suggests that benzene and/or styrene, which are the main precursors to polycyclic aromatic hydrocarbons, may also be present. Styrene was not detected via FTIR methods, and benzene is challenging for IR analysis since its one strong band (ν₁₁ mode at 673 cm⁻¹) is obfuscated by the CO₂ ν₂ bending mode under such polluted atmospheric conditions.

3.6 Emissions of nitrogen-containing species

Gases such as NH₃, NO₂, NO, HCN, and HONO have been identified using FTIR spectroscopy in fire laboratory experiments multiple times (Selimovic et al., 2018; Gilman et al., 2015; Christian et al., 2003, 2004; Goode et al., 1999; Yokelson et al., 1996, 1997; Stockwell et al., 2014; Hatch et al., 2017; Burling et al., 2010; Karl et al., 2007) as well as in field studies (Yokelson et al., 1999; Burling et al., 2011; Goode et al., 2000; Akagi et al., 2013, 2014; Karl et al., 2007). Multiple other methods have also been used to detect N-containing gases, such as HNCO and CH₃CN (Gilman et al., 2015; Christian et al., 2003, 2004; Yokelson et al., 2009; Akagi et al., 2013; Karl et al., 2007; Roberts et al., 2010). The amount and speciation of N-containing compounds emitted are dependent on fuel type and nitrogen content (Stockwell et al., 2014; Burling et al., 2010; Coggon et al., 2016). Moreover, emissions can usually be linked to a stage of combustion: NO, NO₂, HNCO, and HONO are all associated with the flaming stage, while NH₃ and HCN are primarily associated with smoldering combustion but have also been suggested as pyrolysis gases (Goode et al., 1999; Yokelson et al., 1996; Roberts et al., 2010; Burling et al., 2010; Hansson et al., 2004; Di Blasi, 2008). Biomass pyrolysis experiments carried out in an inert (i.e., oxygen-free) atmosphere have revealed that NH₃, HCN, and HNCO are all generated (Hansson et al., 2004). These compounds are all considered to be NO_x (NO+NO₂) and N₂O precursors because they are oxidized via combustion (Hansson et al., 2004).

It is important to note that ammonia and related amine compounds are often best sampled via open-path techniques such as an open White cell, as these compounds are notorious for adhering to the walls of sampling or analysis devices, including those made from steel, glass, or Teflon (Stockwell et al., 2014; Yokelson et al., 2003). Sampling and passivation techniques have been discussed extensively (Neuman et al., 1999; Roscioli et al., 2015) for these “sticky” molecules, and the present method may thus not be optimal for these compounds. This caveat in mind, the major N-containing compound that was identified in the present pyrolysis study was HCN. This is consistent with previous small-scale and controlled laboratory studies that have shown HCN as the primary N product resulting from the pyrolysis of amino acids (Haidar et al., 1981; Johnson and Kan, 1971). This observation is further evidence that the gas samples were extracted when high temperature was the dominant process; Sekimoto et al. (2018) have associated HCN with the high-temperature pyrolysis profile. Figure 7a shows the correlation between HCN and excess CO (R²=0.89). Previous field fire studies have observed similar trends (Simpson et al., 2011; Stockwell et al., 2016). Figure 7b shows a comparison between the ERs for HCN for this study as well as from previous laboratory and field (both ground and airborne) studies. The present values are comparable to other ground-based measurements (Guérette et al., 2018; Akagi et al., 2013) but differ significantly from a few of the laboratory studies. It should be noted that although conducted at a different time of the year (late October–early November 2011), the studies by Akagi et al. (2013) took place near the same location as the current study (i.e., the same military base), and the ERs for HCN they report are not significantly different from the present measurements. This suggests that the ratio of initial gases released of HCN to CO is consistent with the ratio of these gases over the duration of the fire, or at least the fire-averaged ratio. With regards to ERs for HCN, the major factor that appears to influence these values is fuel type, particularly the fuel's peat content. Both laboratory (Stockwell et al., 2014) and field (Stockwell et al., 2016) studies of Indonesian peat have shown greatly enhanced ERs for HCN compared to the studies represented in Fig. 7b, which consist mostly of pine, grasses, and fuels of non-peat origin. The range of averages of ERs for HCN shown in Fig. 7b is 0.0028–0.0095; the averages for the Indonesian peat in laboratory and field studies were 0.015 and 0.021, respectively (Stockwell et al., 2014, 2016), and interestingly are about an order of magnitude greater than the range of values seen in Fig. 7b.

Figure 7(a) Mixing ratios (ppm) for HCN as a function of excess CO (ppm) measured by FTIR. The dashed line is a linear fit. (b) Average emission ratios (ppm_HCN ∕ ppm_CO) for this study and previous laboratory and field investigations. Error bars represent 1σ. Koss et al. (2018) and Stockwell et al. (2014) present fire-integrated ERs. Gilman et al. (2015) present discrete ERs with sample acquisition of 20–300 s. Simpson et al. (2011) present fire-averaged ERs derived by regression. Guérette et al. (2018) present a single ER from all fires and derived by regression. Akagi et al. (2013) and Liu et al. (2016) present fire-averaged EFs calculated using ERs derived by regression. The ERs for Akagi et al. (2013) and Liu et al. (2016) were derived from the ratio of the EFs for HCN and CO multiplied by the molar mass of CO ∕ molar mass of HCN.

Download

Figure 8(a) Mixing ratios (ppm) for methyl nitrite (CH₃ONO) as a function of excess CO (ppm) as measured by FTIR. The dashed line is a linear fit. (b) Average emission ratios (ppm${}_{{\mathrm{CH}}_{\mathrm{3}}\mathrm{ONO}}$ ∕ ppm_CO) for this study and a previously published study carried out in the laboratory using different fuels representative of three US regions. Error bars represent 1σ. Gilman et al. (2015) present discrete ERs with sample acquisition of 20 to 300 s.

Download

In the present study, trace amounts of HONO were detected, but NH₃ was not observed. The absence of NH₃ was somewhat unexpected since, similar to HCN, it is a known product from the pyrolysis of amino acids (Haidar et al., 1981) and has been observed in prior prescribed fires conducted at Ft. Jackson (Akagi et al., 2013, 2014). There are several possible explanations for the lack of NH₃ in the measurements. First and foremost, experimentally NH₃ is well known to adhere to certain surfaces (e.g., steel), and in this study it may have adhered to the canisters or tubing walls and was thus not detected (Neuman et al., 1999; Roscioli et al., 2015; Stockwell et al., 2014; Yokelson et al., 2003). Second, Sekimoto et al. (2018) observed that NH₃ is more often associated with a low-temperature pyrolysis profile, and it appears that the present samples were extracted during a period when high-temperature pyrolysis was the main process. Third, NH₃ is strongly linked with the smoldering phase (Goode et al., 1999; Yokelson et al., 1996), and samples were not collected during this phase. Fourth, the speciation of the N species emitted is dependent on the fuel composition and amount of oxygen (Ren and Zhao, 2012, 2013a, b), so it is possible that in the present study the conditions favored HCN instead of NH₃.

The IR quantification of other N species, such as NO, NO₂, CH₃NO₂, and HNCO, was obstructed due to interferences from H₂O, CO, and CO₂ as well as the low emission values for some of these N species. Specifically, NO and NO₂ were likely not observed as these species are usually associated with flaming combustion. HNCO has been linked with pyrolysis processes, and its main formation pathway is the cracking of cyclic amides along with HCN, which is also a product of pyrolysis (Hansson et al., 2004). After accounting for the challenges in measuring NO, NO₂, and HNCO, the second most prevalent N-containing species observed in this work was methyl nitrite (CH₃ONO). Methyl nitrite has previously been detected in emissions from biomass burning using other methods (Gilman et al., 2015). Figure 8a shows the plot of mixing ratios for methyl nitrite as a function of excess CO. Unlike HCN (Fig. 7a), methyl nitrite exhibits only minimal correlation with excess CO. As one possible alternative explanation, methyl nitrite is known to be associated with rocket-propelled grenades (RPGs), but the Ft. Jackson military base records did not indicate RPG usage for these burn plots (Scharko et al., 2019). While few fire studies have observed methyl nitrite, Gilman et al. (2015) have detected it using gas chromatography–mass spectrometry (GC-MS). Figure 8b shows a comparison of the results from Gilman et al. (2015), separated by US region, with the present results. It is worthy to note that both studies observed similar ERs and that in the Gilman study, methyl nitrite had the second highest mean ER after HCN for N-bearing species in southwestern fuels. Our observation of methyl nitrite is thus not unprecedented, but this was its first reported detection via FTIR (Scharko et al., 2019). In the present study, three measurements (site 16, plot 1, msmt. 1; plot 24A, msmt. 3; and plot 24B) had higher ERs for methyl nitrite than the others, and it is unclear why this is the case. Other measurements collected at the same location reported lower ER values. If the three highest ER measurements in question are not included in the regression then the correlation between methyl nitrite and CO is stronger, and the average ER is closer to values reported by Gilman et al. (2015) for southeastern fuels. One possible explanation for the three greater ER values is that the fuels may have contained more components such as nitrate esters and isopropyl nitrate, both of which are known to release minor amounts of methyl nitrite under controlled pyrolysis conditions (Boschan et al., 1955; Griffiths et al., 1975).