Measurement report: Emission factors of NH₃ and NH_x for wildfires and agricultural fires in the United States

During the 2019 Fire Influence on Regional to Global Environments and Air Quality (FIREX-AQ) study, the NASA DC-8 carried out in situ chemical measurements in smoke plumes emitted from wildfires and agricultural fires in the contiguous United States. The DC-8 payload included a modified proton-transfer-reaction time-of-flight mass spectrometer (PTR-ToF-MS) for the fast measurement of gaseous ammonia (NH₃) and a high-resolution time-of-flight aerosol mass spectrometer (AMS) for the fast measurement of submicron particulate ammonium (${\mathrm{NH}}_{\mathrm{4}}^{+}$). We herein report data collected in smoke plumes emitted from 6 wildfires in the Western United States, 2 prescribed grassland fires in the Central United States, 1 prescribed forest fire in the Southern United States, and 66 small agricultural fires in the Southeastern United States. Smoke plumes contained double to triple digit ppb levels of NH₃. In the wildfire plumes, a significant fraction of NH₃ had already been converted to ${\mathrm{NH}}_{\mathrm{4}}^{+}$ at the time of sampling (≥2 h after emission). Substantial amounts of ${\mathrm{NH}}_{\mathrm{4}}^{+}$ were also detected in freshly emitted smoke from corn and rice field fires. We herein present a comprehensive set of emission factors of NH₃ and NH_x, with ${\mathrm{NH}}_x={\mathrm{NH}}_{\mathrm{3}}+{\mathrm{NH}}_{\mathrm{4}}^{+}$. Average NH₃ and NH_x emission factors for wildfires in the Western United States were 1.86±0.75 g kg⁻¹ and 2.47±0.80 g kg⁻¹ of fuel burned, respectively. Average NH₃ and NH_x emission factors for agricultural fires in the Southeastern United States were 0.89±0.58 and 1.74±0.92 g kg⁻¹, respectively. Our data show no clear inverse correlation between modified combustion efficiency (MCE) and NH₃ emissions. The observed NH₃ emissions were significantly higher than measured in previous laboratory experiments in the FIREX FireLab 2016 study.

1 Introduction

Ammonia (NH₃) is an important trace gas in the Earth's atmosphere that is mostly emitted from agriculture, traffic, the oceans and biomass burning. In the presence of acids, NH₃ rapidly partitions to aerosol particles, which in turn impact air quality and climate (Seinfeld and Pandis, 2016). In much of the atmosphere, NH₃ exhibits a major influence on particle acidity (pH), which is a major controlling parameter for many important aerosol physical and chemical processes (e.g., Pye et al., 2020; Nault et al., 2021). NH₃ is also the largest contributor to deposition of nitrogen from the atmosphere to soil and vegetation, causing surface water eutrophication, soil acidification, and ultimately biodiversity loss (e.g., Bobbink and Higgs, 2014).

Fires emit NH₃ predominantly during smoldering combustion, which occurs at low temperatures (e.g., Lobert et al., 1990; Yokelson et al., 1996, 1997; Goode et al., 1999; McMeeking et al., 2009; Burling et al., 2010; Roberts et al., 2020). NH₃ is typically the third most abundant nitrogen compound (after N₂ and NO) and the most abundant reduced nitrogen compound emitted from fires (Lobert et al., 1990; Roberts et al., 2020; Lindaas et al., 2021).

An important parameter for investigating the atmospheric impact of NH₃ is the emission factor, ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$, which is the mass of NH₃ (in g) that is emitted per mass of fuel burned (in kg). Several literature reviews (Andreae and Merlet, 2001; Akagi et al., 2011; Andreae, 2019; Prichard et al., 2020) report ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ values for different types of fire fuels. A closer look at the literature reveals that emissions from fuels that are typical of the United States have mostly been studied in the laboratory (e.g., Yokelson et al., 1996; McMeeking et al., 2009; Burling et al., 2010; Stockwell et al., 2015; Selimovic et al., 2018; Roberts et al., 2020). Previous work has shown that laboratory fires may not realistically simulate fires occurring in the real world due to different burning conditions and the lack of heterogeneity in fuels (e.g., Yokelson et al., 2013; Hodshire et al., 2019). Only very few studies have reported ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ derived from measurements carried out in the field (e.g., Lindaas et al., 2021).

The limited availability of field data is mostly because NH₃ is difficult to measure. NH₃ is a “sticky” molecule that easily adsorbs onto inlet and instrumental surfaces. This makes fast airborne measurements of NH₃ particularly challenging. Müller et al. (2014; see Supplement) have shown that a proton-transfer-reaction time-of-flight mass spectrometer (PTR-ToF-MS) can be used for airborne NH₃ measurements, although with some limitations tied to a relatively slow time response and a poor detection limit due to a large intrinsic background. The conventional PTR-ToF-MS instrument of the University of Innsbruck has measured NH₃ with such constraints in previous airborne studies (Sun et al., 2015; Kelly et al., 2018; Guo et al., 2021; Pan et al., 2021).

The Fire Influence on Regional to Global Environments and Air Quality (FIREX-AQ) study was a joint NOAA and NASA effort to investigate the atmospheric impact of wildfires and agricultural fires in the contiguous United States (Warneke et al., 2022). In summer 2019, the NASA DC-8 Airborne Science Laboratory performed in situ measurements in smoke plumes emitted from wildfires in the Western United States and agricultural fires in the Southeastern United States. The aircraft payload included a modified PTR-ToF-MS instrument that was optimized for the fast measurement of NH₃. It also included an aerosol mass spectrometer (AMS) for fast measurement of submicron particulate ammonium (${\mathrm{NH}}_{\mathrm{4}}^{+}$). This allowed us to measure and report a set of emission factors of NH₃ and NH_x, where ${\mathrm{NH}}_x={\mathrm{NH}}_{\mathrm{3}}+{\mathrm{NH}}_{\mathrm{4}}^{+}$, for different types of fires.

2 Methods

2.1 FIREX-AQ

The FIREX-AQ experiment has been described in detail by Warneke et al. (2022). During the 2019 field campaign, NH₃ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ were measured aboard the NASA DC-8 in smoke plumes emitted from six wildfires in the Western United States (Shady, Williams Flats, Castle, Ridge Top, Mica and Lick Creek, Horsefly), two prescribed grassland fires in the Central United States (Hickory Ridge State Wildlife Management Area, Tallgrass Prairie National Preserve), and one prescribed forest fire in the Southern United States (Black Water River State Forest) (Fig. 1).

Figure 1 Location of the fires that were sampled by the NASA DC-8 during the 2019 FIREX-AQ campaign and for which NH₃ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ data were obtained.

Several downwind transects were typically flown in the smoke plumes emitted from the wildfires. In addition, the NASA DC-8 sampled smoke plumes from a large number of agricultural fires in the Southeastern United States. These small plumes were typically sampled twice in perpendicular direction. We successfully measured NH₃ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ in plumes emitted from 66 agricultural fires.

Vegetation and fuel type information is summarized in Table S1 in the Supplement. Information about fuel types was obtained from the 30 m Fuel Characteristic Classification System (FCCS; Ottmar et al., 2007), the 30 m 2019 dataset of the Cropland Data Layer (CDL) classification, and ground intelligence.

2.2 Instrumentation

A modified PTR-ToF-MS instrument was used for fast-response measurements of NH₃ aboard the NASA DC-8 during FIREX-AQ. The conventional airborne PTR-ToF-MS analyzer has been described in detail by Müller et al. (2014). Only the modifications pertinent to the fast measurement of NH₃ are thus described here.

For reducing the instrumental NH₃ background, 12–25 sccm (standard cubic centimeters per minute; “standard” herein means referenced to a temperature of 273.15 K and a pressure of 101 325 Pa) of ultra-pure helium (6.0; Praxair Inc., Danbury, CT, USA) was introduced into the source drift region between the drift tube and the ion source. This reduced the backflow of nitrogen into the plasma region and suppressed NH₃ formation in the plasma (Müller et al., 2020, and references therein).

For improving the instrumental time response to NH₃, all stainless-steel parts in the drift tube were surface-passivated with a functionalized hydrogenated amorphous silicon coating (Piel et al., 2021), and the drift tube was heated to 120 ^∘C. Surface passivation and heating significantly reduces the adsorption of NH₃ to instrumental surfaces, lowering the instrumental response time to ∼2 s (see Fig. 4 of Piel et al., 2021).

A series of inlet configurations were tested during the initial phase of the FIREX-AQ campaign. The fastest response to NH₃ was achieved when air was sampled at a flow rate of ∼60 slpm (standard liters per minute) through a heated Teflon PFA tube (length: ∼2 m, inner diameter: 3.96 mm, wall temperature: 60 ^∘C). Evaporation of ammonium nitrate particles in the main sampling line was not investigated, but it is believed to be small due to the short sample residence time (<25 ms). For inertially separating particles from the analyte air, a small flow was subsampled from the main inlet line in rearward direction and directed into the drift tube through a Teflon PFA tube (length: ∼10 cm, outer diameter: 3.175 mm, temperature: 120 ^∘C). The subsampling flow was set to ∼250 sccm via a pinch valve applied on the PFA tube. An NH₃ time response of a few seconds was ultimately achieved (see Sect. 3).

We performed three types of calibrations in the field: (i) periodic in-flight calibrations using a dynamically diluted calibration standard in a pressurized cylinder (10 ppm NH₃ in N₂; Praxair Distribution Inc., Lancaster, CA, USA), (ii) a ground-based calibration using a dynamically diluted calibration standard in a pressurized cylinder (2.7 ppm NH₃ in N₂; provided by NOAA's Chemical Sciences Laboratory) and (iii) a ground-based calibration using an NH₃ permeation source (provided by NOAA's Chemical Sciences Laboratory). While results from the cylinder-based calibrations were in good agreement, the permeation-tube-based calibration yielded an instrumental response factor that was a factor of 2 higher. For resolving this inconsistency, we carried out an extensive post-mission NH₃ calibration in the laboratory. For that purpose, an artificial atmosphere (NH₃ in air) was generated in a 250 L environmental (“smog”) chamber equipped with a Fourier transform infrared (FTIR) spectrometer (Bruker IFS 66v/S). The concentrations of NH₃ (accuracy: ±5 %) were determined from the FT-IR spectra (120 m path length, 0.125 cm⁻¹ spectral resolution) in a global nonlinear least squares spectral fitting procedure (Griffith, 1996) employing the absolute cross sections of NH₃ (Gordon et al., 2017). The instrumental response factor derived from the post-mission calibration agreed to within ±15 % with the response factors obtained during the in-flight calibrations. We concluded that the permeation rate determined in pre-campaign laboratory measurements had changed in the field. The estimated accuracy of the reported NH₃ mixing ratios is ±15 %. We note that this accuracy estimate is not valid when NH₃ mixing ratios abruptly changed and inlet and instrument surfaces were not equilibrated.

Figure 2 Time traces of (a) mixing ratios of NH₃, ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and CO and of (b) cumulative mixing ratios of NH₃, ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and CO as measured during a transect of the Williams Flats Fire plume on 7 August 2019.

Submicrometer (50 % cutoff size for a vacuum aerodynamic diameter ∼1 µm, about 850 nm geometric diameter for most fire plumes based on in-field calibrations) ${\mathrm{NH}}_{\mathrm{4}}^{+}$ was measured by an Aerodyne high-resolution time-of-flight AMS instrument (DeCarlo et al., 2006; Canagaratna et al., 2007), with a time resolution of up to 10 Hz. The accuracy (2σ) of the ${\mathrm{NH}}_{\mathrm{4}}^{+}$ data is estimated to be ±34 % (Bahreini et al., 2009), while the detection limit was typically much smaller (25 ppt at 1 Hz in clean air, ∼200 ppt in fire plumes). The inlet flow was optimized to allow for near-real-time sampling (0.3 s residence time) and to minimize particle volatilization in the inlet. We note that, based on the current state of knowledge, the AMS ${\mathrm{NH}}_{\mathrm{4}}^{+}$ data collected in fresh smoke plumes suffer from a minor (≤20 %) positive interference from reduced organic nitrogen compounds, which are known to be abundant constituents of biomass burning particles (e.g., Mace et al., 2003). A general correction is still under development based on a positive matrix factorization (PMF) analysis.

Carbon monoxide (CO) and methane (CH₄) were measured by the differential absorption carbon monoxide measurement (DACOM) instrument (Sachse et al., 1991), which is based on mid-infrared wavelength modulation spectroscopy. The uncertainty in the CO data is 2.1±0.2 ppb; the uncertainty for CH₄ is about 1 %. Carbon dioxide (CO₂) was measured by a LICOR model 7000 analyzer (Vay et al., 2009), which is based on nondispersive infrared absorption spectroscopy. For CO₂ <500 ppm, the accuracy is 0.25 ppm and the precision is 0.1 ppm, while for higher mixing ratios, the total uncertainty is about 2 %.

2.3 Emission factor, modified combustion efficiency

We used the carbon mass balance method for calculating ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ (Yokelson et al., 1996, 1999). The underlying assumption is that the carbon in the fire fuel is predominantly emitted as CO₂, CO and CH₄. ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ (in g kg⁻¹) is thus described by the simplified Eq. (1):

$$\begin{array}{}\text{(1)}& {\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}={\displaystyle \frac{\mathrm{\Delta }{\mathrm{NH}}_{\mathrm{3}}}{\mathrm{\Delta }{\mathrm{CO}}_{\mathrm{2}}+\mathrm{\Delta }{\mathrm{CH}}_{\mathrm{4}}+\mathrm{\Delta }\mathrm{CO}}}\times {\displaystyle \frac{\mathrm{17}}{\mathrm{12}}}\times F_{\mathrm{c}}\times \mathrm{1000},\end{array}$$

where Δ is the above background mixing ratio in the plume of the respective trace gas, 17 is the molar mass of NH₃ (in g mol⁻¹), and 12 is the molar mass of carbon (in g mol⁻¹). The fraction of carbon in the fuel is denoted as F_c, which is typically in the 0.45–0.55 range (Akagi et al., 2011). We assumed F_c to be 0.50 and note here that the resulting 10 % of uncertainty in ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ is small compared to the natural variability of ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$. A problem in the calculation of ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ arises from the fact that NH₃ is a “sticky” compound. When the aircraft first penetrates a smoke plume, NH₃ molecules typically adsorb onto inlet and instrumental surfaces, thereby delaying the signal response of the analyzer. When the airplane exits the plume, the desorbing NH₃ molecules cause a signal tailing (Fig. 2a). For calculating Δ, we thus applied the method described in the Supplement of Müller et al. (2016) and calculated cumulative volume mixing ratios including the immediate period (10 s) before the plume was encountered (background) and the period after the plume encounter (seconds to minutes) when the NH₃ signal tailed off (Fig. 2b). The signal tailing was particularly pronounced during the initial phase of the campaign (before 24 August 2019) when the inlet configuration had not yet been optimized. During the later phase of the campaign, only a few seconds of tailing were observed. NH_x is the sum of NH₃ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$; ${\mathrm{EF}}_{{\mathrm{NH}}_x}$ was calculated as the sum of ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ and ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{4}}^{+}}$. The modified combustion efficiency (MCE) was calculated as $\mathrm{\Delta }{\mathrm{CO}}_{\mathrm{2}}/(\mathrm{\Delta }{\mathrm{CO}}_{\mathrm{2}}+\mathrm{\Delta }\mathrm{CO}$).

Figure 3 (a) NH₃/CO, ${\mathrm{NH}}_{\mathrm{4}}^{+}$/CO, NH_x/CO and cumulative mean NH_x/CO as measured during nine downwind transects of the plume emitted from the Williams Flats Fire on 7 August 2019. NH₃/CO increases during the first three transects which were thus used for determining ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ and ${\mathrm{EF}}_{{\mathrm{NH}}_x}$. (b) Mean MCE as measured for the same nine downwind transects. MCE remains stable at ∼0.91 indicating stable burning conditions. (c) Latitude–longitude plot showing the location of the fire (black cross) and the color-coded flight track in the NH₃ mixing ratio.

Figure 4 Mixing ratios of NH₃, ${\mathrm{NH}}_{\mathrm{4}}^{+}$, and CO as measured when the NASA DC-8 transected (a) the plume emitted from the Williams Flats Fire on 7 August 2019 in downwind direction and (b) the plume emitted from a small corn field fire in the Mississippi River valley on 26 August 2019.

Data from 180 plume transects were included in our analysis of the wildfire emissions. We only used data from plume transects in which CO mixing ratios exceeded 300 ppb for more than 20 s and from plumes in which MCE values were stable (standard deviation of MCE < 0.05). Data from seven plume transects were excluded due to missing NH₃, ${\mathrm{NH}}_{\mathrm{4}}^{+}$ or CH₄ data. Our emission factor (EF) analysis was not based on a single plume transect in closest proximity to the wildfire, as we observed in several plumes that $\mathrm{\Delta }{\mathrm{NH}}_{\mathrm{3}}/\mathrm{\Delta }\mathrm{CO}$ increased during a few initial downwind transects (Fig. 3). The reason for this increase (typically <15 %) is unclear. We included all plume transects in our analysis, up to where $\mathrm{\Delta }{\mathrm{NH}}_{\mathrm{3}}/\mathrm{\Delta }\mathrm{CO}$ reached its maximum and derived an average ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ and ${\mathrm{EF}}_{{\mathrm{NH}}_x}$ (± standard deviation, SD). All SDs reported herein only reflect the measured variability and do not include measurement uncertainties in the underlying variables (NH₃, ${\mathrm{NH}}_{\mathrm{4}}^{+}$, CO₂, CO, CH₄). Data from 164 plume transects were included in our analysis of the agricultural fire emissions. Data from 12 plume transects were excluded due to missing NH₃ or ${\mathrm{NH}}_{\mathrm{4}}^{+}$ data.

3 Results and discussion

3.1 Airborne measurements of NH₃ in smoke plumes

Figure 4a shows the mixing ratio of NH₃ as measured by the PTR-ToF-MS instrument on 7 August 2019 aboard the NASA DC-8. The aircraft flew nine downwind transects at an altitude of 5160 m a.s.l. (above sea level) for sampling the plume emitted from the Williams Flats Fire in Washington, DC. The NH₃ signal increased with CO when the aircraft entered the plume, exhibited a similar time trend as CO within the plume and decreased to background levels outside the plume, although with some tailing (few minutes). NH₃ maxima ranged from 110 to 200 ppb, which were typical maximum NH₃ levels measured in fire plumes throughout the 2019 FIREX-AQ field campaign. Additionally, Fig. 4a shows the time trace of ${\mathrm{NH}}_{\mathrm{4}}^{+}$ as measured by the AMS instrument, with maximum mixing ratios ranging from 42 to 65 ppb. The observation of significant amounts of ${\mathrm{NH}}_{\mathrm{4}}^{+}$ indicates that ${\mathrm{NH}}_{\mathrm{4}}^{+}$ was primarily emitted (as for example observed by Lewis et al., 2009) and/or gaseous NH₃ had been partly converted to particulate ${\mathrm{NH}}_{\mathrm{4}}^{+}$ by the time of sampling (≥2 h after emission). A rapid conversion can be caused by the fast reaction of NH₃ with primarily emitted acids such as hydrochloric acid (HCl), nitric acid (HNO₃) and organic acids or occur more slowly downwind via the reaction of NH₃ with secondary formed acids. As stated in Sect. 2.2, the ${\mathrm{NH}}_{\mathrm{4}}^{+}$ measurement suffered from a minor (≤20 %) positive interference from reduced organic nitrogen compounds.

Figure 4b shows the mixing ratios of NH₃, ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and CO as observed when the NASA DC-8 crossed a plume emitted from a small cornfield fire in the Mississippi River valley on 26 August 2019 at an altitude of 325 m a.s.l. All data are shown at the frequency they were recorded (5 Hz) which resulted in an increased noise for NH₃. The tailing was however reduced to a few seconds with the improved PTR-ToF-MS inlet. We show the 5 Hz data for demonstrating that we succeeded in measuring such small fire plumes from a jet aircraft. For further analysis, we used the 1 s integrated data. Notably, the AMS instrument detected significant amounts of ${\mathrm{NH}}_{\mathrm{4}}^{+}$ in this very fresh plume, indicating that either direct emission from the fire or a rapid conversion of NH₃ to ${\mathrm{NH}}_{\mathrm{4}}^{+}$ had occurred. The latter could be caused by the fast reaction of NH₃ with HCl, which is emitted in significant amounts from agricultural fires (Liu et al., 2017). Another plausible explanation is the resuspension of recently applied ammonium nitrate fertilizer.

Due to the fact that ${\mathrm{NH}}_{\mathrm{4}}^{+}$ was already present in very fresh smoke (due to direct emission or rapid conversion), we will herein also report ${\mathrm{EF}}_{{\mathrm{NH}}_x}$, as suggested in previous work by Hegg et al. (1990). In Fig. 5a, we plot ${\mathrm{EF}}_{{\mathrm{NH}}_x}$ against ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ for the six wildfires in the Western United States that were investigated during the 2019 FIREX-AQ field campaign. The two EFs are highly correlated (R²=0.96), with the slope of the linear regression curve being close to unity (1.07±0.05). This regression analysis suggests that ${\mathrm{NH}}_{\mathrm{4}}^{+}$ added ∼0.5 g kg⁻¹ (offset of the regression line: 0.47±0.11) to ${\mathrm{EF}}_{{\mathrm{NH}}_x}$ throughout the campaign. The offset may be interpreted as the typical direct ${\mathrm{NH}}_{\mathrm{4}}^{+}$ emission factor (or fast conversion of NH₃).

Figure 5 ${\mathrm{EF}}_{{\mathrm{NH}}_x}$ vs. ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ as derived from in situ measurements in the plumes of (a) 6 wildfires and (b) 66 small agricultural fires. In panel (a), SD bars represent the observed natural variability (measurement uncertainties in NH₃ and NH_x not included). In panel (b), circles represent field-dominated fuels and crosses represent timber-dominated fuels (see Sect. 3.3). The blue lines represent linear regression curves of the average values for wildfires (a) and of all agricultural data in (b). Shaded areas are 95 % confidence bands.

In the case of the agricultural burns, the NASA DC-8 sampled the plumes in very close proximity to the fires. ${\mathrm{EF}}_{{\mathrm{NH}}_x}$ and ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ had again a regression slope of ∼1. The offset was mainly caused by elevated ${\mathrm{NH}}_{\mathrm{4}}^{+}$ emissions and low NH₃ emissions from some of the cornfield fires (Fig. 5b).

Table 1 ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ and ${\mathrm{EF}}_{{\mathrm{NH}}_x}$ derived from in situ measurements in the plumes of six wildfires in the Western United States. The lines in bold are average values of the above-listed data for the respective fire.

3.2 NH₃ and NH_x emissions from wildfires in the Western US

In situ measurements of NH₃ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ were made in smoke plumes emitted from six wildfires in the Western United States. Table 1 provides a detailed overview of ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ and ${\mathrm{EF}}_{{\mathrm{NH}}_x}$ derived from these measurements. Plumes from the Shady, Williams Flats and Castle fires were sampled multiple times, and we list the data from each of the sampling patterns as well as the average value. ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ and ${\mathrm{EF}}_{{\mathrm{NH}}_x}$ were lowest for the Shady Fire. The low emissions may be caused by the difference in fuels, which in the case of the Shady Fire was modified or managed xeric understory (see Table S1).

Average ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ and ${\mathrm{EF}}_{{\mathrm{NH}}_x}$ values for the six wildfires in the Western United States were 1.86±0.75 and 2.47±0.80 g kg⁻¹, respectively. We compare our results to those obtained in two recent studies. Lindaas et al. (2021) investigated NH₃ emissions from wildfires in the Western United States during the 2018 WE-CAN campaign. We calculated an average ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ of 1.48±0.91 g kg⁻¹ for the WE-CAN data. This is slightly lower than the average ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ reported herein but within the combined uncertainties of the two methods: ±12 % for the quantum-cascade tunable infrared laser direct absorption spectrometer (QC-TILDAS) used during WE-CAN and ±15 % for the PTR-ToF-MS analyzer used during FIREX-AQ. Selimovic et al. (2018) investigated emissions from fires fueled by a wide range of US vegetation types in the FIREX FireLab 2016 laboratory study. We only used the data for fuels relevant to FIREX-AQ (see Table S1) and obtained a significantly lower average ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ of 0.67±0.38 g kg⁻¹ for the FIREX FireLab data. This finding seems to confirm that laboratory fires do not realistically simulate wildfires (e.g., Yokelson et al., 2013; Hodshire et al., 2019) and thereby underestimate real-world emissions of NH₃.

In Fig. 6, we plot the measured ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ values (six wildfires, multiple sampling of three fires) as a function of MCE along with trends from the WE-CAN and FIREX FireLab studies. In the case of the FIREX-AQ data (regression line and confidence band in blue), ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ and MCE correlated poorly, with Pearson's coefficient of determination (R²) being only 0.04. As opposed to the WE-CAN study (regression line and confidence band in black) and the FIREX FireLab experiments (regression line and confidence band in green), we did not find a clear inversion correlation between MCE and NH₃ emissions.

Figure 6 Scatter plot of ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ (as measured for six wildfires in the Western United States during FIREX-AQ and obtained from two literature sources) vs. MCE.

3.3 NH₃ and NH_x emissions from agricultural fires in the Southeastern US

In situ measurements of NH₃ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ were made in smoke plumes emitted from 66 small agricultural fires in the Southeastern United States. Values of ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ varied widely, covering a range from 0.09 to 3.60 g kg⁻¹. The following average values and standard deviations were derived: ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}=\mathrm{0.89}\pm \mathrm{0.58}$ g kg⁻¹, ${\mathrm{EF}}_{{\mathrm{NH}}_x}=\mathrm{1.74}\pm \mathrm{0.92}$ g kg⁻¹, $\mathrm{MCE}=\mathrm{0.92}\pm \mathrm{0.04}$.

Table 2 ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$, ${\mathrm{EF}}_{{\mathrm{NH}}_x}$ and MCE as measured during FIREX-AQ for small agricultural fires in the Southeastern United States burning on different types of fuel. The lines in bold are average values of the above-listed data for field-dominated fuels and timber-dominated fuels, respectively.

We grouped the agricultural fuels into field-dominated and timber-dominated fuels. The field-dominated fuels include corn, rice, soybeans and grassland pasture. The timber-dominated fuels include evergreen forest, coniferous timber, prescribed understory fire, pile burning, slash burning, pile timber slash mixture burns and pile burning of longleaf pine tree stumps. Table 2 lists ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$, ${\mathrm{EF}}_{{\mathrm{NH}}_x}$ and MCE for the two main categories and 11 subcategories.

The data listed in Table 2 indicate that field-dominated fuels emit more NH₃ and NH_x than timber-dominated fuels. Agricultural areas are usually nitrogen-fertilized, which may cause increased NH₃ emissions. ${\mathrm{EF}}_{{\mathrm{NH}}_x}$ is roughly a factor of 2 higher than ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$, which indicates higher primary ${\mathrm{NH}}_{\mathrm{4}}^{+}$ emissions and/or a very rapid NH₃ to ${\mathrm{NH}}_{\mathrm{4}}^{+}$ conversion in these fresh plumes.

Figure 7 Scatter plot of ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ vs. MCE as measured for 66 agricultural fires (circles: field-dominated fuels, crosses: timber-dominated fuels) and obtained from four literature sources.

Table 3 ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$, ${\mathrm{EF}}_{{\mathrm{NH}}_x}$ and MCE as measured during FIREX-AQ for two prescribed grassland fires in the Central United States and one prescribed forest fire in the Southern United States.

Figure 7 shows ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ as a function of MCE for different fuels (as measured in individual fires), the averages derived for field-dominated fuels and timber-dominated fuels and the results from four previous studies (McMeeking et al., 2009; Stockwell et al., 2015; Müller et al., 2016; Selimovic et al., 2018). Also in this case, ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ and MCE correlated poorly with R² being 0.05. The literature values match the low NH₃ emissions (<1 g kg⁻¹) we observed for most agricultural fires, but the high NH₃ emissions from burning rice and corn residues have not been reported before.

3.4 NH₃ and NH_x emissions from other fires

We also measured NH₃ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ in smoke plumes emanating from two prescribed grassland fires in the Central United States and one prescribed forest fire in the Southern United States. These fires do not fall within the two main categories discussed in the previous two sections and are thus separately presented here. Table 3 lists the fire details, ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$, ${\mathrm{EF}}_{{\mathrm{NH}}_x}$ and MCE for these three fires.

4 Conclusions

During the 2019 FIREX-AQ field campaign, we measured NH₃ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ aboard the NASA DC-8 in wildfire and agricultural fire plumes. We found that ${\mathrm{NH}}_{\mathrm{4}}^{+}$ was either directly emitted from the fire (consistent with past laboratory experiments) and/or NH₃ had already partially partitioned to particulate ${\mathrm{NH}}_{\mathrm{4}}^{+}$ at the time of sampling. We thus also evaluated emissions of NH_x and produced a comprehensive set of ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ and ${\mathrm{EF}}_{{\mathrm{NH}}_x}$ for wildfires in the Western United States and agricultural fires in the Southeastern United States. Our data show no clear inverse correlation between MCE and ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$. Values of ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ measured in plumes of large wildfires were similar to those observed during the 2018 WE-CAN field campaign, but significantly higher than observed in the FIREX FireLab 2016 laboratory study. We also report the first extensive set of field-measurement-derived ${\mathrm{EF}}_{{\mathrm{NH}}_{\mathrm{3}}}$ and ${\mathrm{EF}}_{{\mathrm{NH}}_x}$ values for different types of agricultural fires in the Southeastern United States. Emissions of NH₃ were highest from fires of corn and rice residues. Substantial amounts of ${\mathrm{NH}}_{\mathrm{4}}^{+}$ were detected in freshly emitted smoke from some of the corn and rice field fires, which warrants further investigation.