Ground-based investigation of HO_x and ozone chemistry in biomass burning plumes in rural Idaho

Ozone (O₃), a potent greenhouse gas that is detrimental to human health, is typically found in elevated concentrations within biomass burning (BB) smoke plumes. The radical species OH, HO₂, and RO₂ (known collectively as RO_x) have central roles in the formation of secondary pollutants including O₃ but are poorly characterized for BB plumes. We present measurements of total peroxy radical concentrations ([XO₂] ≡ [HO₂] + [RO₂]) and additional trace-gas and particulate matter measurements from McCall, Idaho, during August 2018. There were five distinct periods in which BB smoke impacted this site. During BB events, O₃ concentrations were enhanced, evident by ozone enhancement ratios ($\mathrm{\Delta }{\mathrm{O}}_{\mathrm{3}}/\mathrm{\Delta }\mathrm{CO}$) that ranged up to 0.06 ppbv ppbv⁻¹. [XO₂] was similarly elevated during some BB events. Overall, instantaneous ozone production rates (P(O₃)) were minimally impacted by the presence of smoke as [NO_x] enhancements were minimal. Measured XO₂ concentrations were compared to zero-dimensional box modeling results to evaluate the Master Chemical Mechanism (MCM) and GEOS-Chem mechanisms overall and during periods of BB influence. The models consistently overestimated XO₂ with the base MCM and GEOS-Chem XO₂ predictions high by an average of 28 % and 20 %, respectively. One period of BB influence had distinct measured enhancements of 15 pptv XO₂ that were not reflected in the model output, likely due to the presence of unmeasured HO_x sources. To the best of our knowledge, this is the first BB study featuring peroxy radical measurements.

1 Introduction

Unprecedented wildfire activity has been observed in recent years. For example, Brazil's Amazon rainforest wildfires in 2019 (Cardil et al., 2020) and Australia's bush fires in 2019–2020 (Yu et al., 2020) were both marked by historically high amounts of land burned. The United States has also observed several staggering wildfire seasons characterized by hundreds of fatalities and tens of thousands of homes destroyed. For example, the 2018 season included California's deadliest fire ever (the Camp Fire), which was responsible for 85 deaths (CAL-FIRE, 2021), burned a total of 3.5×10⁶ ha of land, and elicited government spending of USD 3.1 billion for fire suppression (NIFC, 2019). Due to climate change, the impacts associated with wildfire are projected to worsen. Many regions, including much of the western United States, are becoming warmer and drier, causing increases in fuel aridity. These conditions are conducive to increases in land area burned, extended fire seasons, and frequent extreme fires (Spracklen et al., 2009; Yue et al., 2014, 2013; Goss et al., 2020). As a result, the amount of land area burned in the western United States is expected to increase by 24 %–124 % by the mid-21st century (Yue et al., 2014, 2013).

Wildfire smoke degrades air quality both locally and far downwind. Wildfire smoke is associated with respiratory and cardiovascular health risks (Liu et al., 2015; Reid et al., 2016; DeFlorio-Barker et al., 2019; Stowell et al., 2019). These health risks will result in modest increases of respiratory illness hospitalizations as wildfire smoke will more frequently impact densely populated areas (J. C. Liu et al., 2016). Smoke from biomass burning (BB), a term that includes wildfires, changes in composition as it is transported downwind. The initial composition is based on direct BB emissions: particulate matter (PM), NO_x, and organic compounds spanning a wide range of volatilities. These emissions then transform through photochemical reactions that form secondary pollutants such as ozone (O₃) and secondary organic aerosol (SOA). The formation of O₃ within BB smoke is of concern because O₃ poses additional health concerns and is a prominent greenhouse gas. While the large quantities of PM within BB smoke are primarily responsible for BB smoke health concerns (Liu et al., 2015; DeFlorio-Barker et al., 2019; Reid et al., 2016), health risks from O₃ exposure include lung irritation, decreases in lung and cardiac function, higher susceptibility to respiratory infection, and early mortality (Bell et al., 2006; Park et al., 2004; Jerrett et al., 2009; Turner et al., 2016; Silva et al., 2013). Tropospheric ozone is the third most important anthropogenic greenhouse gas after CO₂ and CH₄. Increases in tropospheric ozone between the pre-industrial era and the present, from 1750 to 2011, have accounted for 0.40 W m⁻² of radiative forcing. For perspective, CH₄ and CO₂ increases are responsible for 0.48 and 1.82 W m⁻² of radiative forcing, respectively (Stocker et al., 2013).

Mitigating O₃ pollution will be increasingly challenging as background O₃ concentrations are increasing in some regions. Long-term aircraft observations across the Northern Hemisphere indicate average increases of tropospheric O₃ of 5 % per decade (Gaudel et al., 2020). More specifically, western US O₃ concentrations have been increasing 0.41 ppbv yr⁻¹ despite declining concentrations in the eastern United States (Cooper et al., 2012). As biomass burning accounts for an estimated 3.5 % of global tropospheric O₃ formation (Jaffe and Wigder, 2012), this source of O₃ will be more important with projected wildfire activity (Jacob and Winner, 2009; Yue et al., 2015). Cities in the western US, which often have existing O₃ pollution issues, have their air quality degraded by BB smoke. For example, a study focused on eight predominantly western US cities found wildfire smoke to correlate with 19 % of exceedances to the previous 75 ppbv O₃ NAAQS (National Ambient Air Quality Standard) standard despite smoke only being present 4.1 % of the total days studied between May and the end of September (Gong et al., 2017). This NAAQS O₃ standard is in terms of maximum 8 h daily concentration and has been set to a more stringent standard of 70 ppbv since 2015.

The correlation between O₃ concentrations and BB smoke is not completely straightforward. Depleted O₃ concentrations can be found in freshly emitted smoke plumes due to NO emissions reacting with background O₃. Aged plumes have also been observed with depleted O₃. This has been attributed to meteorological conditions (Wentworth et al., 2018), NO_x sequestration via peroxy acetyl nitrate (PAN) formation (Alvarado et al., 2010), and significant PM emissions that attenuate sunlight and limit photochemistry (Xu et al., 2021). Ozone depletion is further exemplified by the wide variation in reported ozone enhancement ratios ($\mathrm{\Delta }{\mathrm{O}}_{\mathrm{3}}/\mathrm{\Delta }\mathrm{CO}$). While most reported ratios are positive, indicating net O₃ formation, negative $\mathrm{\Delta }{\mathrm{O}}_{\mathrm{3}}/\mathrm{\Delta }\mathrm{CO}$ values have also been reported and indicate depletion. $\mathrm{\Delta }{\mathrm{O}}_{\mathrm{3}}/\mathrm{\Delta }\mathrm{CO}$ ratios in temperate and boreal forest (those typical of the United States) are on average 0.018, 0.15, and 0.22 ppbv ppbv⁻¹ for smoke plumes aged <2, 2–5, and >5 d, respectively (Jaffe and Wigder, 2012). The wide variation in net O₃ production within BB smoke is related to several factors, including fire dynamics, the extent of emissions, and meteorological conditions. Each of these factors influence the underlying RO_x (“RO_x” = OH, HO₂, and RO₂) chemistry that controls oxidation processes and secondary pollutant formation.

RO_x chemistry has rarely been studied within BB plumes. The peroxy radicals HO₂ and RO₂ (R = organic group) oxidize nitric oxide (NO) to form NO₂ (Reactions R1 and R2). During the day, the resulting NO₂ is converted to O₃ by photolysis (Reactions R3 and R4).

$$\begin{array}{}\text{(R1)}& {\displaystyle }{\mathrm{HO}}_{\mathrm{2}}+\mathrm{NO}\to \mathrm{OH}+{\mathrm{NO}}_{\mathrm{2}}\text{(R2)}& {\displaystyle }{\mathrm{RO}}_{\mathrm{2}}+\mathrm{NO}\to \mathrm{RO}+{\mathrm{NO}}_{\mathrm{2}}\text{(R3)}& {\displaystyle }{\mathrm{NO}}_{\mathrm{2}}+\mathit{hv}\to \mathrm{NO}+\mathrm{O}{(}^{\mathrm{3}}\mathrm{P})\text{(R4)}& {\displaystyle }\mathrm{O}{(}^{\mathrm{3}}\mathrm{P})+{\mathrm{O}}_{\mathrm{2}}+M\to {\mathrm{O}}_{\mathrm{3}}+M\end{array}$$

Here, M represents a collision partner such as N₂ or O₂. Within these smoke plumes, the concentration and composition of RO_x species depend on fire emissions, photochemical conditions, and smoke age. BB emissions include direct HO_x precursors, but individual emission factors and ratios are highly variable between fires. Fire dynamics alone have a significant effect on emissions. Flaming conditions have greater combustion efficiencies and are characterized by smaller emission factors of volatile organic compounds (VOCs) and PM, while reactive nitrogen emissions are dominated by HONO, NO, and NO₂ (Burling et al., 2010; Roberts et al., 2020). Smoldering fire conditions have greater VOC emission ratios and PM emissions but lower NO_x emissions (Yokelson et al., 1996). Different fuel sources, such as specific tree species, shrubs, grasses, and crops, have unique emission profiles (Koss et al., 2018). Direct HO_x precursor emissions of formaldehyde (HCHO), acetaldehyde (CH₃CHO), and nitrous acid (HONO) are greater than those from more typical urban combustion sources. HONO within BB plumes is short-lived, with a chemical lifetime ranging 10 to 30 min due to photolysis. This rapid photolysis yields OH and NO and has caused HONO to act as the dominant RO_x source in some freshly emitted smoke plumes (Peng et al., 2020; Yokelson et al., 2009). Ozone formation in young BB plumes is, in almost all cases, initially NO_x-saturated (VOC-limited) but transitions to being NO_x-limited as the NO_x is photochemically processed to nitric acid and organic nitrates (Xu et al., 2021; Alvarado et al., 2015; Müller et al., 2016; Folkins et al., 1997). In particular, NO₂ can be efficiently sequestered in the form of peroxyacetyl nitrate (PAN) due to the importance of the acetaldehyde emissions and the relative importance of the CH₃C(O)OO radical (Peng et al., 2021). The subsequent thermal decomposition of PAN into air masses in which ozone production is NO_x-limited can potentially lead to sustained O₃ formation far downwind of a fire. BB emissions also include unique VOCs that are typically unaccounted for by chemical mechanisms employed by models. For instance, the importance of furanoids for model predictions of secondary pollution formation has only recently been studied (Müller et al., 2016; Coggon et al., 2019; Decker et al., 2019; Salvador et al., 2021; Robinson et al., 2021).

For the best understanding of the composition and concentration of radicals, direct measurements and improved models are necessary. Measurements of any RO_x compounds within BB plumes are rare. Only one study has reported direct measurements of OH in a BB smoke plume, which found that freshly emitted BB plumes (22–43 min aged) had OH concentrations 5 times greater than those of background air (Yokelson et al., 2009). Our understanding of RO_x chemistry in BB smoke has historically relied on calculations and models (Mason et al., 2001; Hobbs et al., 2003; de Gouw et al., 2006; Akagi et al., 2012; X. Liu et al., 2016; Müller et al., 2016; Parrington et al., 2013). Elevated OH concentrations have been suggested for freshly emitted smoke (Hobbs et al., 2003; Akagi et al., 2012), while low OH concentrations were calculated for aged plumes of ∼4 d in western Canada (de Gouw et al., 2006). There are few studies that have focused on peroxy radical chemistry within BB studies, and it appears there have been no direct measurements in BB smoke. Model-suggested peroxy radical concentrations have exhibited a wide range of values (Mason et al., 2001; Parrington et al., 2013; X. Liu et al., 2016; Baylon et al., 2018), in some cases reaching unrealistically high values ([HO₂ + RO₂] ≫200 pptv for wildfires in Nova Scotia, Canada, as presented in Parrington et al., 2013).

This paper focuses on smoke observations collected in McCall, Idaho, in the Pacific Northwest – a region particularly prone to wildfire – as part of the joint NCAR WE-CAN (Western Wildfire Experiment for Cloud Chemistry, Aerosol Absorption, and Nitrogen) and NOAA FIREX (Fire Influence on Regional to Global Environments Experiment) study. Increases in wildfire activity are anticipated for parts of this region, including much of Idaho, because of climate change (Halofsky et al., 2020). Presented are possibly the first measurements of total peroxy radicals in biomass burning plumes, enabling a unique investigation into the impacts of biomass burning on photochemistry and ozone production and the accuracy of commonly used atmospheric chemistry models.

Figure 1 The McCall, Idaho, site is indicated by the red star, and wildfire locations are indicated by orange circles. Panel (a) shows detected wildfires between 15 and 24 August 2018. Panel (b) includes bihourly HYSPLIT back trajectories for the representative smoke impacted day of 17 August. Shown here are active wildfires of 15 through 17 August, as back trajectories end 24 h back. Fire locations were retrieved using the NOAA Hazard Mapping System (HMS) (NOAA, 2021).

2 Methods

2.1 Campaign description

Measurements were collected in McCall, Idaho, during the WE-CAN/FIREX 2018 campaign. McCall (elevation ∼1.5 km) is a rural town in Valley County, Idaho, approximately 160 km north of Boise, Idaho (Fig. 1), within the West Mountains of Idaho. While much of the local area is used for cattle grazing, the town attracts tourists year-round for outdoor recreation due to the presence of mountains, surrounding forests, and a large lake.

The McCall field site included two mobile laboratories and one building. Each were outfitted with instrumentation for gas- and particle-phase measurements. The largest mobile lab was the Aerodyne Mobile Laboratory (AML) (Herndon et al., 2005). The AML split its time between stationary sampling in McCall and mobile measurements in other parts of Idaho and surrounding states. Unlike the AML, the second mobile lab, known as the Miniature Aerodyne Mobile Lab (minAML), was permanently stationed at the McCall site. Our analyses focus on the date ranges 16–18 and 21–24 August when the AML was stationed at the McCall site.

2.2 ECHAMP

ECHAMP (Ethane CHemical AMPlifier) is a chemical amplification-based instrument that was used to measure total peroxy radical concentrations ([HO₂] + [RO₂], or simply [XO₂]). ECHAMP was stationed within the minAML and sampled on a 2 min time base. The sampling and calibration methods have been described in detail elsewhere (Anderson et al., 2019; Wood et al., 2017), and only a brief summary including details specific to the McCall deployment is described here.

ECHAMP “amplifies” each sampled XO₂ molecule into a greater number of NO₂ molecules. The enhancement in NO₂ concentration from “amplification” is then divided by an amplification factor to determine [XO₂]. Amplification is achieved by mixing sampled air with elevated concentrations of NO and C₂H₆ to take advantage of a radical propagation scheme (Reactions 1–2 and 5–8). Since these reactions can proceed multiple times, each XO₂ produces up to 20 NO₂ molecules. There are two sampling channels: in the amplification channel there is an immediate addition of C₂H₆ and NO, whereas in the background channel the C₂H₆ addition is delayed to avoid Reaction (5), and all peroxy radicals are converted into HONO (Wood et al., 2017). The NO₂ within the two channels is transported to and measured by respective cavity attenuated phase shift (CAPS) NO₂ monitors (Kebabian et al., 2008).

$$\begin{array}{}\text{(R5)}& {\displaystyle }\mathrm{OH}+{\mathrm{C}}_{\mathrm{2}}{\mathrm{H}}_{\mathrm{6}}\to {\mathrm{H}}_{\mathrm{2}}\mathrm{O}+{\mathrm{C}}_{\mathrm{2}}{\mathrm{H}}_{\mathrm{5}}\text{(R6)}& {\displaystyle }{\mathrm{C}}_{\mathrm{2}}{\mathrm{H}}_{\mathrm{5}}+{\mathrm{O}}_{\mathrm{2}}+M\to {\mathrm{C}}_{\mathrm{2}}{\mathrm{H}}_{\mathrm{5}}{\mathrm{O}}_{\mathrm{2}}+M\text{(R7)}& {\displaystyle }{\mathrm{C}}_{\mathrm{2}}{\mathrm{H}}_{\mathrm{5}}{\mathrm{O}}_{\mathrm{2}}+\mathrm{NO}\to {\mathrm{C}}_{\mathrm{2}}{\mathrm{H}}_{\mathrm{5}}\mathrm{O}+{\mathrm{NO}}_{\mathrm{2}}\text{(R8)}& {\displaystyle }{\mathrm{C}}_{\mathrm{2}}{\mathrm{H}}_{\mathrm{5}}\mathrm{O}+{\mathrm{O}}_{\mathrm{2}}\to {\mathrm{CH}}_{\mathrm{3}}\mathrm{CHO}+{\mathrm{HO}}_{\mathrm{2}}\end{array}$$

Ambient air is first sampled at 2.5 SLPM into a 0.635 cm PFA tee and immediately diluted with 0.8 SLPM O₂. The sampled air then flows through 10 cm of 0.476 cm i.d. PFA tubing protruding out of an inlet box. This ECHAMP inlet box is a weatherproof container with dimensions of 39 cm × 44 cm × 16 cm and was mounted 3 m a.g.l. on scaffolding. The sampled air then entered a glass cross that was internally coated with halocarbon wax. Connected orthogonally from the sampling lines were two 15.2 cm long (0.476 cm i.d) reaction chambers. The reaction chambers each subsampled at a flow rate of 0.83 L min⁻¹ . The remaining flow rate of 1.64 SLPM traveled by a Vaisala HMP60 probe that measured temperature and humidity. The reaction chambers included additions of NO, C₂H₆, and N₂. In amplification mode, 20 sccm of 39.3 ppmv NO in N₂ and 50 sccm of 25 % C₂H₆ in N₂ were added at the beginning of the reaction chamber, while 50 sccm of N₂ is added in a downstream position. In the background mode, the locations for the N₂ and C₂H₆ additions are reversed. The final concentrations of NO and C₂H₆ in the reaction chambers were 0.827 ppmv and 1.32 %, respectively. These reaction chambers alternate in 1 min intervals between background and amplification modes leading to the overall 2 min sampling time. The resulting mixtures from both channels were transported in approximately 23 m of their respective tubing to the minAML that housed both CAPS monitors.

Dilution of sampled air with O₂ is a new addition to ECHAMP and is similar in some ways to its use by the perCIMS method (Hornbrook et al., 2011). The elevated O₂ concentration (40 %) increases the ratio of the rate of the propagation reaction (Reaction R8; C₂H₅O + O₂ → CH₃CHO + HO₂) to the rate of ethyl nitrite formation (C₂H₅O+NO + M → C₂H₅ONO + M), which is a termination reaction. Dilution also dries the sampled air, leading to lower and less variable relative humidity. This is beneficial as the amplification factor decreases with increasing RH (Anderson et al., 2019; Wood et al., 2017). The humidity dependence of the amplification factor is due to humidity-dependent reactions of HO₂ and the HO₂–H₂O adduct reacting with NO to form HNO₃ (Reichert et al., 2003; Butkovskaya et al., 2009) and HO₂ wall losses (Mihele and Hastie, 1998; Reichert et al., 2003).

ECHAMP was calibrated to the methyl peroxy radical CH₃O₂ over a range of relative humidities (RHs) using the CH₃I photolysis method as described in Anderson et al. (2019) six times at McCall. The calibrant was prepared by mixing humidified zero air (ZA) with trace amounts of CH₃I from a permeation source. The resulting mixture flows to a quartz tube, where CH₃I is photolyzed at 254 nm by an O₃-free Hg lamp. The resulting CH₃O₂ calibrant is then added in excess flow to the ECHAMP inlet. To quantify the CH₃O₂ produced, both reaction chambers were initially operated in background mode and the CH₃O₂ source modulated on and off by alternating the flow between a UV photolysis cell and a bypass chamber. An improvement over the prior version of this calibration method was the elimination of dead volume in the chamber. Further details regarding this calibration technique including recent improvements are provided in the Supplement. Calibrations were also conducted using the H₂O photolysis method (Anderson et al., 2019). Unfortunately, inconsistent results were obtained, and at the end of the project we discovered that the quartz photolysis cell was broken. Therefore, calibrations using the H₂O photolysis method were disregarded, and only the CH₃I calibrations were used. Based on the uncertainties in the individual calibration points, the variability among individual calibration points, and uncertainties regarding sampling losses, we ascribe an uncertainty of 34 % (2σ) to the measurements. See Sect. S1 for more information.

2.3 Additional measurements

On board the AML, quantum cascade tunable infrared laser direct absorption spectrometers (QC-TILDAS; Aerodyne Research Inc.) (McManus et al., 2015) were used to measure (1) NO and NO₂; (2) CO, N₂O, and H₂O; (3) HCHO and HCOOH; (4) CH₄ and C₂H₆; and (5) HCN. Ozone was measured by a 2B-Tech UV absorption instrument (model 205). While particulate matter and VOCs can positively interfere with photometric O₃ measurements (Huntzicker and Johnson, 1979; Long et al., 2021), comparison to a separate O_x measurement revealed minimal interferences in the 2B-Tech O₃ observations (see Sect. S2). VOC measurements were made by an ARI Vocus (proton-transfer-reaction high-resolution time-of-flight (PTR-HR-ToF) mass spectrometer) (Krechmer et al., 2018). Measured VOCs include isoprene, acetaldehyde, acetone, acetic acid, benzene, toluene, C₂ benzenes, C₃ benzenes, methanol, total monoterpenes, and the sum of methyl vinyl ketone (MVK) and methacrolein (MACR). BB-related VOCs measured with the Vocus include furan, methyl furan, furfural, the sum of methyl furfural and catechol, and guaiacol. Chemically resolved measurements of particulate matter mass concentrations were made by an ARI soot particle aerosol mass spectrometer (SP-AMS) (Canagaratna et al., 2007). Additional particle-phase measurements were made with an ARI aerosol chemical speciation monitor (ACSM), permanently stationed at the McCall site building (Ng et al., 2011).

On board the minAML, NO₂ was measured with a CAPS monitor and VOCs with the Berkeley Comprehensive Thermal Desorption Aerosol Gas Chromatograph (cTAG) (Wernis et al., 2021). The cTAG measures concentrations of VOCs, intermediate volatility organic compounds, and semi-volatile organic compounds, spanning an alkane-equivalent volatility range from C₅ to C₃₀ every hour via pre-concentration followed by thermal desorption and gas chromatography–time-of-flight mass spectrometry (GC-TOF-MS). This paper uses isoprene, speciated monoterpenes, 2-methyl-3-buten-2-ol (MBO), styrene, benzene, and toluene measurements taken by cTAG.

Meteorological measurements were made both on the AML and permanently at the McCall site. Temperature, wind speed, and wind direction were collected by a 3D R.M. Young (Model 81000RE) sonic anemometer stationed permanently at the McCall site at a height of 10 m. Additional wind was measured with a 3D R.M. Young (Model 81000RE) sonic anemometer mounted to the AML rooftop and corrected for speed and truck orientation with data from a Hemisphere GPS compass (model Vector V103). Temperature, RH, and wind data are shared in the Supplement (Fig. S3). Daily maximum temperatures ranged 22 to 28 ^∘C, while minimum temperatures ranged 4 to 13 ^∘C. Solar irradiance was measured by a permanently stationed ARISense air quality sensor system (Cross et al., 2017). This was used to derive photolysis frequencies of interest, such as $J_{{\mathrm{NO}}_{\mathrm{2}}}$, by scaling measured irradiance to outputs from the National Center for Atmospheric Research (NCAR) Tropospheric Ultraviolet and Visible (TUV) radiation model. This process for deriving photolysis frequencies is described in greater detail in the Supplement.

For most calculations and chemistry analyses, measured concentrations of other compounds were synchronized to the 2 min ECHAMP timescale. This was achieved by averaging greater frequency measurements and linearly interpolating lower frequency measurements. Most measurements fall in the former category and were measured at a 1 Hz sampling rate. VOCs from cTAG were obtained at an hourly rate and were therefore linearly interpolated.

Figure 2 Time series of smoke tracers. Tan shading represents smoke presence, while red shading signifies significant smoke periods. Significant smoke periods are defined by HCN concentrations greater than 1 ppbv. This smoke-defining concentration is indicated by the dashed purple line.

Figure 3 Time series of O₃, NO, NO₂, total peroxy radicals (XO₂), and aldehydes (HCHO and CH₃CHO). Modeled $J_{{\mathrm{NO}}_{\mathrm{2}}}$ frequencies and calculated parameters of P(O_x) and P(RO_x) are also provided. The red shading indicates periods of smoke influence, while somewhat smoky periods are shaded tan. P(O_x) is shown in both 16 min averages (blue circles) and 2 min data (gray points). The dashed $J_{{\mathrm{NO}}_{\mathrm{2}}}$ trace is entirely simulated by NCAR TUV model, whereas the rest of the data are derived from ARISense solar irradiance measurements.

2.4 Smoke events

Time periods impacted by smoke were identified with observations of the biomass burning tracers HCN, CH₃CN, organic aerosol (OA), and CO. HCN was used as the primary tracer for BB smoke. Nitriles are commonly used as tracers for BB, and HCN is particularly useful in the absence of nearby vehicle sources. Emissions of HCN are essentially inert within BB plumes with an atmospheric lifetime of 2 to 4 months (Li et al., 2000). HCN emission ratios are dependent on the biomass burning fuel type (Koss et al., 2018; Coggon et al., 2016) and can vary with fire dynamics (Roberts et al., 2020). Figure 2 presents McCall site observations and indicates smoke-impacted time periods when ECHAMP was actively sampling and the AML was present. The lowest concentrations of HCN, CH₃CN, CO, and OA were all observed on 24 August, suggesting the air sampled up until then was always somewhat affected by BB emissions. We experienced five distinct smoke-impacted periods evident by clear enhancements of HCN along with organic PM and CO. The periods of greatest smoke influence (red shaded regions in Fig. 2) were identified by sustained periods of HCN concentrations greater than 1 ppbv. General smoke presence (tan shaded regions in Fig. 2) was identified before and after each significant smoke period when background smoke tracer concentrations remained elevated compared to stable background air. There were distinct smoke periods in the early evenings of 16 and 17 August. These occurred from 18:48 to 21:55 MDT on 16 August and 15:27 to 18:15 MDT on 17 August. While appreciable enhancements in CO were observed during both of these periods, a significant OA enhancement of ∼20 µg m⁻³ was only observed during 17 August. For the period 21 to 24 August, smoke events occurred at earlier times. Periods of significant smoke influence on 22, 23, and 24 August were from 12:55 to 16:34, 12:26 to 17:07, and 07:42 to 15:00 MDT, respectively. The 22 August smoke event had minimal enhancement of OA. The smoke event on 23 August was quite distinct with the greatest HCN concentrations observed for the entire campaign. After this 23 August significant smoke period highlighted in Fig. 2, concentrations of HCN, CO, and OA decreased but remained at levels above background concentrations until the 24 August event. Following the 24 August significant smoke period, CO and OA decreased to their lowest observed concentrations indicating that we were sampling an air mass with minimal smoke influence.

We identify the sources of the observed smoke by pairing NOAA Air Resources Laboratory HYbrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) (Stein et al., 2015) model back trajectories with satellite-detected wildfire locations (Lindaas et al., 2017; Rogers et al., 2020). Figure 1a shows all wildfires detected for the date range we focus our analyses on. Three well-documented fires are indicated: the Mesa Fire, the Rattlesnake Creek Fire, and the Rabbit Foot Fire. The Mesa Fire, located 38.3 km southwest of McCall, was responsible for burning 34 700 acres (∼ 14 000 ha) and started on 26 July 2018 (FWAC, 2021b). The Rattlesnake Creek Fire started on 23 July 2018 and burned over 8000 acres (>3300 ha) (FWAC, 2021c). The Rabbit Foot Fire began on 2 August 2018, burned 36 000 acres (∼15 000 ha), and was not contained until November (FWAC, 2021a). Figure 1b shows 17 August as a representative smoke-impacted day. Shown are bihourly HYSPLIT 48 h long back trajectories initialized from the McCall site at 10 m a.g.l. using archived Global Data Assimilation System (GDAS) 1^∘ meteorological data. Based on fire tracer observations (see Fig. 2), the 17 August significant smoke period started at ∼ 15:30 MDT and persisted until 18:15 MDT. This is consistent with the 16:00 and 18:00 MDT back trajectories that show that air sampled during that period was transported past an active fire approximately 60 km southwest of the McCall, Idaho, site. Though an uncontrolled portion of the Mesa Fire may have also contributed to these smoky conditions, the HYSPLIT trajectories would suggest earlier smoke influence possibly starting near 14:00 MDT. The smoke likely traveled between 3–5 h to get to the McCall site from the unspecified active fire for the 17 August example shown. Figures showing HYSPLIT trajectories for other smoke-influenced days are included (Sect. S4). The likely sources and smoke ages for other events are detailed here. The 16 August smoke event was likely sourced from the same unspecified wildfire ranging in age between 4–10 h. The smoke on 22 August was likely 12–18 h or longer and sourced from an unspecified fire located east between the McCall site and the Rabbit Foot Fire. The 23 August smoke was likely sourced from the southwest from another unspecified fire at an age of 12–18 h, though the Mesa Fire may have contributed here. The smoky conditions of the morning of 24 August were likely from wildfires in central Oregon aged 18–30 h until especially clean air was sourced from northern Oregon beginning at approximately 12:00 MDT.

2.5 Calculations

Ozone enhancement ratios $\mathrm{\Delta }{\mathrm{O}}_{\mathrm{3}}/\mathrm{\Delta }\mathrm{CO}$ from smoke influence were determined for the most distinct BB events of 16, 17, and 24 August. For the 17 and 24 August events, the ozone enhancement ratio was determined using the York bivariate linear regression method (York et al., 2004) using a continuous section of O₃ and CO data that includes 60 min of background air, a transitional smoke period (tan shaded regions in Fig. 2), and 60 min of significant smoke period data (red shaded regions in Fig. 2). Enhancements in NO₂ were typically under 0.2 ppbv, and so the difference between considering ΔO_x (O_x = O₃ + NO₂) and ΔO₃ was negligible. The linear regressions are included in the Supplement (Fig. S8). $\mathrm{\Delta }{\mathrm{O}}_{\mathrm{3}}/\mathrm{\Delta }\mathrm{CO}$ for the 16 August event was determined using Eq. (1) with O₃ and CO data collected during a stable period at the start of the significant smoke period and a stable background prior to smoke presence.

$$\begin{array}{}\text{(1)}& \begin{array}{rl}\mathrm{\Delta }{\mathrm{O}}_{\mathrm{3}}/\mathrm{\Delta }\mathrm{CO}=& \left(\right[{\mathrm{O}}_{\mathrm{3}}{]}_{\mathrm{Smoke}}-{\left[{\mathrm{O}}_{\mathrm{3}}\right]}_{\mathrm{Background}})/\left([\mathrm{CO}{]}_{\mathrm{Smoke}}\right.\\ -& {\left[\mathrm{CO}\right]}_{\mathrm{Background}}\left.\right)\end{array}\end{array}$$

This event had a temporary depletion in O₃ by ∼20 ppbv for the start of smoke significance and then returned to near-background levels of O₃. $\mathrm{\Delta }{\mathrm{O}}_{\mathrm{3}}/\mathrm{\Delta }\mathrm{CO}$ values were not calculated for the remaining 22 and 23 August smoke events. These events had less distinct O₃ enhancements and occurred at times when O₃ also increased during non-smoky time periods.

The gross instantaneous O_X production rate P(O_X), often referred to as P(O₃), is the rate at which NO is converted to NO₂ by reaction with HO₂ or RO₂ (Eq. 2). Since we measure the sum of HO₂ and RO₂, we calculate P(O_x) with Eq. (3) using ECHAMP XO₂ and AML-based QC-TILDAS NO measurements. Noting that the differences between $k_{{\mathrm{HO}}_{\mathrm{2}}+\mathrm{NO}}$ and most $k_{{\mathrm{RO}}_{\mathrm{2}}+\mathrm{NO}}$ rate constants are small (Anderson et al., 2019), we use an effective rate constant (k_eff) equal to $k_{{\mathrm{HO}}_{\mathrm{2}}+\mathrm{NO}}$. P(O_x) from box model results were calculated using Eq. (2) and the model HO₂ and speciated RO₂ concentrations.

$$\begin{array}{}\text{(2)}& {\displaystyle }\begin{array}{rl}P\left({\mathrm{O}}_x\right)=& k_{{\mathrm{HO}}_{\mathrm{2}}+\mathrm{NO}}\left[{\mathrm{HO}}_{\mathrm{2}}\right]\left[\mathrm{NO}\right]\\ +& \sum k_{\mathrm{RO}\mathrm{2}\left(i\right)+\mathrm{NO}}\left[{\mathrm{RO}}_{\mathrm{2}\left(i\right)}\right]\left[\mathrm{NO}\right]\end{array}\text{(3)}& {\displaystyle }P\left({\mathrm{O}}_x\right)=k_{\text{eff}}\left[{\mathrm{XO}}_{\mathrm{2}}\right]\left[\mathrm{NO}\right]\end{array}$$

Instantaneous RO_x production rates P(RO_x) from measured compounds were calculated using Eq. (4). Each term represents a compound that undergoes photolysis to produce two RO_x radicals, with compound-specific photolysis rate constants (frequencies) indicated by the J variables. Our calculated values for P(RO_x) are limited by the lack of measurements for HONO, which is the dominant HO_x source in freshly emitted BB smoke (Peng et al., 2020; Robinson et al., 2021). While HONO has a lifetime of ∼20 min during daylight hours, dark plume conditions and possible photochemical formation on aerosol particles may lead to sustained HONO concentrations. Ozonolysis of measured alkenes had minimal contribution to daytime P(RO_x) values and was therefore omitted from this calculation but is included in model predictions.

$$\begin{array}{}\text{(4)}& \begin{array}{rl}& P\left({\mathrm{RO}}_x\right)=\\ & {\displaystyle \frac{\mathrm{2}{J}_{{\mathrm{O}}^{\mathrm{1}}\mathrm{D}}\left[{\mathrm{O}}_{\mathrm{3}}\right]{\mathrm{k}}_{({\mathrm{O}}^{\mathrm{1}}\mathrm{D}+{\mathrm{H}}_{\mathrm{2}}\mathrm{O})}\left[{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\right]}{k_{({\mathrm{O}}^{\mathrm{1}}\mathrm{D}+{\mathrm{H}}_{\mathrm{2}}\mathrm{O})}\left[{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\right]+{\mathrm{k}}_{({\mathrm{O}}^{\mathrm{1}}\mathrm{D}+{\mathrm{N}}_{\mathrm{2}})}\left[{\mathrm{N}}_{\mathrm{2}}\right]+{\mathrm{k}}_{({\mathrm{O}}^{\mathrm{1}}\mathrm{D}+{\mathrm{O}}_{\mathrm{2}})}\left[{\mathrm{O}}_{\mathrm{2}}\right]}}\\ & +\mathrm{2}{J}_{\mathrm{HCHO}}\left[\mathrm{HCHO}\right]+\mathrm{2}{J}_{{\mathrm{CH}}_{\mathrm{3}}\mathrm{CHO}}\left[{\mathrm{CH}}_{\mathrm{3}}\mathrm{CHO}\right]\\ & +\mathrm{2}{J}_{{\mathrm{CH}}_{\mathrm{3}}{\mathrm{COCH}}_{\mathrm{3}}}\left[{\mathrm{CH}}_{\mathrm{3}}{\mathrm{COCH}}_{\mathrm{3}}\right]\end{array}\end{array}$$

2.6 Zero-dimensional modeling

The Framework for 0-D Atmospheric Modeling (F0AM) box model (v3.2) (Wolfe et al., 2016) was used to evaluate ECHAMP XO₂ measurements and further investigate BB impacts on instantaneous chemistry. F0AM simulations were conducted for the ECHAMP 2 min time basis for dates on which AML concomitant measurements were present. Modeling was conducted separately for the two date ranges of interest of 16 to 18 and 21 to 24 August and then combined for analysis. We primarily focus on results acquired by employing a subset of the Master Chemical Mechanism (MCM) version 3.3.1 (Saunders et al., 2003; Jenkin et al., 2003, 2015) that included only the relevant chemical species in order to avoid unnecessary reactions and improve model time consumption. The model was constrained with all available measurements (see Supplement for full list), including concentrations of ozone, formaldehyde, acetaldehyde, acetone, isoprene, speciated monoterpenes, MVK, MACR, and MBO. While F0AM allows for total NO_x to be constrained, we instead constrained NO and NO₂ individually. Model results obtained using total NO_x constraints led to nearly identical daytime XO₂ predictions but with unrealistic nighttime XO₂ values. The base MCM mechanism, referred to as “MCM-base” from here on, was augmented for two additional F0AM simulations. First, the MCM-base was expanded by including additional chemistry for BB-related VOCs (referred to as “MCM-BBVOC”) of furan, methyl furan, furfural, methyl furfural, and guaiacol by manually adding the relevant chemical reactions to the MCM as detailed by Coggon et al. (2019). Second, a mechanism referred to as “MCM-BBVOC-het” included heterogeneous chemistry for HO₂ loss on organic aerosols in addition to the previously detailed BB VOC chemistry. The heterogeneous loss rates are dependent on the predicted [HO₂] values, organic aerosol surface area concentration, an uptake coefficient (γ), and mean molecular speed (Tang et al., 2014). Aerosol surface area concentrations were not measured and were instead calculated from mass concentration measurements by applying a specific surface area. The default specific surface area was set to 4 m² g⁻¹. This setting falls slightly below the typical values measured for an urban environment of Tokyo, Japan (Hatoya et al., 2016). The default uptake coefficient was 0.20, as recommended by Jacob (2000). We explored the sensitivity of model results to both the specific surface area and uptake coefficient parameters by varying settings. We also share F0AM results acquired using the GEOS-Chem chemical mechanism. This version of the GEOS-Chem mechanism uses version 9-02 (Mao et al., 2013) with isoprene chemistry updates (Marais et al., 2016; Fisher et al., 2016; Travis et al., 2016; Kim et al., 2015). A small first-order dilution was implemented for all model experiments so that all compounds would have 24 h lifetimes in order to prevent unreasonable accumulation of secondary species with background concentrations for all unmeasured compounds set to 0 ppbv (Wolfe et al., 2016). Initial concentrations of unmeasured species were also set to 0 ppbv for the model results presented in this paper. Minimal changes in model results were observed for additional simulations that included a “spin-up” period in order to determine initial concentrations of unmeasured compounds. As mentioned earlier, HONO is a particularly important RO_x precursor in BB plumes. In addition to it not being measured during this study, the zero-dimensional models utilized cannot be expected to accurately predict HONO concentrations since a portion of the HONO in the sampled air masses was undoubtedly emitted directly by the smoke. Furthermore, there are no HONO formation processes in the chemical mechanisms besides its homogenous formation from the reaction of OH with NO (i.e., there are no heterogeneous formation mechanisms). A complete description of our model setup, including observational constraints and uncertainties, is provided in the Supplement (see Sect. S7).

3 Results and discussion

3.1 Smoke influence on ozone and its precursors

Figure 3 shows site observations of O₃ and relevant measurements of NO_x, total peroxy radicals, and aldehydes. Typically O₃ increased from ∼20 ppbv overnight up to daily maximum concentrations between 50 and 60 ppbv, with the greatest O₃ concentrations of near 70 ppbv observed on 23 August. NO_x concentrations were consistently low, with typical daytime [NO] and [NO₂] values below 0.1 and 1.0 ppbv, respectively, suggesting O₃ production in the NO_x-limited regime (Sillman et al., 1990). Ozone production is further discussed in Sect. 3.2. XO₂ typically reached daily maximum concentrations between 40 and 60 pptv. Distinct increases in [O₃] and [XO₂] were observed during some BB events. The 17 August event had the clearest impact, with [O₃] increasing ∼10 ppbv and [XO₂] increasing ∼15 pptv over 45 min, resulting in the maximum [XO₂] observed of near 70 pptv for the entire campaign. The 16 and 24 August smoke events had distinct [O₃] increases but with less notable change in [XO₂]. For 24 August, smoke impacted the site throughout the morning and persisted into the afternoon. Ozone concentrations were ∼53 ppbv for over an hour until smoky conditions dispersed. Ozone concentrations then decreased and remained near 46 ppbv, demonstrating a 7 ppbv elevation associated with the presence of smoke. No distinct impact on XO₂ was observed for this event. The smoke impact on 16 August was unique as [O₃] was depleted by ∼10 ppbv upon smoke arrival, returned to near-background concentrations, then was again depleted by ∼10 ppbv as smoke exited. Enhancements in [NO₂] during these periods of depleted O₃ were small – only 0.5 and 1.0 ppbv – thus [O_x] was depleted as well. This variation in [O₃] and smoke tracers was likely due to different regions of this smoke plume impacting the site at different times. The edges of the plume may have been sampled at the beginning and end of the smoke period and the plume center sampled during the period with the greatest O₃ concentrations. XO₂ responded similarly to O₃ and was depleted by ∼10 pptv upon the arrival and departure of smoke. The 23 August event was particularly smoky, with the greatest smoke tracer concentrations of HCN, CO, and organic PM, as well as the highest concentrations of O₃ for the entire campaign.

Figure 4 Correlation between O₃ and smoke tracer HCN for all observations between 09:00 and 22:00 MDT (a), 14:00 and 16:00 MDT (b), and 16:00 and 18:00 MDT (c). Data points are colored by date collected.

For the entire campaign, there is a positive correlation between daytime O₃ (and O_x) concentrations and smoke tracer HCN (Fig. 4a), with the highest values for both observed on 23 August. Most periods of elevated HCN occurred during the times of day when [O₃] was usually high, even in the absence of smoke (afternoon or early evening), so the overall positive correlation between O₃ and HCN may be partially coincidental. The positive correlation remains, however, when the analysis is restricted to 2 h periods of afternoon and early evening data to limit the time-of-day dependence (Fig. 4b and c). These more specific O₃–HCN comparisons remain impacted by day-to-day variability in O₃ from changes in background O₃ values, meteorology, and BB HCN and O₃ precursor emissions. Smoke age also plays a role in this correlation plot. Based on literature trends where $\mathrm{\Delta }{\mathrm{O}}_{\mathrm{3}}/\mathrm{\Delta }\mathrm{CO}$ values increase with smoke age until an eventual plateau (Jaffe and Wigder, 2012; Baker et al., 2016; Xu et al., 2021), young smoke plumes are likely to have smaller O₃ enhancements relative to smoke tracers like HCN compared to aged plumes. Clusters of data points at HCN concentrations below 0.75 ppbv are observed for the 18 and 24 August data sets. For 18 August, there was no distinct BB-influenced period and little variability in [HCN]. This led to the cluster of 18 August data points with [O₃] near 60 ppbv. The 24 August data cluster near 50 pbbv [O₃] captures the stable period after [O₃] is depleted by ∼7 ppbv upon smoke departure followed by a slow build in concentration. A similar figure with O₃ plotted against CO but for times specific to the arrival or departure of smoke is shown in the Supplement (Fig. S8).

Daily maximum P(RO_x) values calculated from measured compounds ranged from 0.45 to 0.65 pptv s⁻¹ (Fig. 3). P(RO_x) was dominated by HCHO photolysis and the reaction of O(¹D) (from O₃ photolysis) with water vapor. RO_x production from photolysis of acetaldehyde and acetone was of minor importance. Aldehydes were enhanced for some of the BB events. Typically, HCHO ranged between 1.5 and 4.5 ppbv, and BB events led to enhancements of near 2 ppbv for the 16, 17, and 24 August events. This led to distinct impacts on P(RO_x) and is most evident for the 24 August event. Distinct enhancements in O₃ for the 17 and 24 August events similarly affected P(RO_x). While measurements of [XO₂] generally reflect P(RO_x) trends, this was not the case for the 17 August event. As XO₂ increased by ∼27 % upon the arrival of the smoke-affected air mass, P(RO_x) from measured compounds increased by at most 5 %. Changes in NO_x were mostly negligible during this period, though [NO] remained below 0.1 ppbv, indicating that small changes in [NO] could have had a large impact on [XO₂]. Increases in both [O₃] and [HCHO] of ∼20 % did not contribute to a significant increase in P(RO_x) due to a ∼30 % decrease in [H₂O] (see Fig. S16) and a ∼10 % decrease in photolysis frequencies. The sudden increase in measured [XO₂] when there were only small changes in P(RO_x), NO_x, and VOCs suggests the prominence of unmeasured RO_x sources such as HONO. Peng et al. (2020) suggested HONO enhancement ratios (ΔHONO/ΔCO) of ∼0.1 pptv ppbv⁻¹ for western US wildfire BB plumes aged 3 h, the likely age of our smoke plume sampled here, causing HONO photolysis to remain a significant RO_x source even after 3 h of aging. On average, HONO accounted for >90 % and 50 % of P(RO_x) in 30 min aged plumes and 3 h aged plumes, respectively (Peng et al., 2020).

3.2 Ozone production

We describe the extent of ozone formation for the 16, 17, and 24 August BB-influenced periods using the commonly used $\mathrm{\Delta }{\mathrm{O}}_{\mathrm{3}}/\mathrm{\Delta }\mathrm{CO}$ metric. These values depict the O₃ produced in transit to the McCall site while accounting for plume dilution or overall smoke influence of the site air sampled. $\mathrm{\Delta }{\mathrm{O}}_{\mathrm{3}}/\mathrm{\Delta }\mathrm{CO}$ values were −0.02, 0.06, and 0.03 ppb ppbv⁻¹ for the 16, 17, and 24 August smoke events, respectively. These calculated values fall within the wide variability and range of literature $\mathrm{\Delta }{\mathrm{O}}_{\mathrm{3}}/\mathrm{\Delta }\mathrm{CO}$ values for boreal and temperate forest fire smoke plumes aged less than 2 d, including numerous examples of ozone depletion for aged plumes (Jaffe and Wigder, 2012). Though the smoke was likely sourced from the same wildfire for the 16 and 17 August events (Sect. 2.4), we observe O₃ depletion on 16 August and O₃ enhancement on 17 August. $\mathrm{\Delta }{\mathrm{O}}_{\mathrm{3}}/\mathrm{\Delta }\mathrm{CO}$ values were not calculated for the 22 and 23 August smoke events as we were unable to attribute the observed increases of O₃ to smoke influence as they occurred at the same time that O₃ usually increased during non-smoky time periods, as mentioned in Sect. 2.5. O_x enhancement ratios are not presented but differed insignificantly from O₃ enhancement ratios as NO₂ concentrations were much lower than O₃ concentrations (see Fig. 3).

Instantaneous O₃ production rates are calculated using NO and XO₂ concentrations (Fig. 3). Gaps in P(O_x) are due to measurement gaps in XO₂ when ECHAMP was offline for calibrations and diagnostic tests. The highest P(O_x) values occurred on 17 and 18 August during non-smoky periods between 10:00 and 12:00 MDT, reaching formation rates slightly greater than 8 ppbv h⁻¹. For the entire campaign, median P(O_x) peaked at 11:00 MDT at 5.8 ppbv h⁻¹. As NO concentrations were low and rarely exceeded 1 ppbv, changes in [NO] had a near-linear impact on P(O_x). Noisy P(O_x) periods, such as the entire afternoon of 16 August, are mainly attributed to the atmospheric variability of and measurement precision for NO. Overall, there is little correlation between P(O_x) and smoke tracers. However, elevated P(O_x) during the 17 August event is somewhat evident. The ∼27 % increase in XO₂ and near-constant value for NO led to this temporary increase in P(O_x). P(O_x) increased from ∼2.5 to 8.9 ppb h⁻¹ during the transition from background air to significant smoke, remained elevated for 34 min, and then returned to near-background P(O_x) rates. The overall lack of impact of BB influence on P(O_x) is further depicted in the P(O_x) diurnal cycle of Fig. 5. Modeled P(O_x) results for the same time period are also presented in Fig. 5a with the green median trend. These model results were acquired using F0AM with the MCM-BBVOC mechanism. Modeled P(O_x) is consistently greater than measured values, with the greatest discrepancy occurring in the 07:45 to 08:15 MDT period. This difference is due to modeled [XO₂] being greater than measured [XO₂]. While we present results acquired using four unique chemical mechanisms, model-predicted P(O_x) was always greater than measurements, though within the combined uncertainties.

Figure 5 P(O_x) results derived from 2 min XO₂ observations. Panel (a) shows the diurnal profile of all calculated P(O_x) values. The data points are colored by smoke tracer HCN. The green and blue traces represent 15 min median values of P(O_x) as determined using measurements and model results, respectively. The provided model results were acquired with F0AM using the MCM-BBVOC mechanism. Panel (b) shows the variation of P(O_x) with NO. Data are colored by P(RO_x) and sized by [HCN].

Figure 5b shows the relationship between P(O_x), [NO], and P(RO_x). For P(RO_x) values above 0.4 ppt s⁻¹, P(O_x) increases almost linearly with [NO] up until at least 400 ppt, consistent with ozone formation being NO_x-limited. The vast majority of NO concentrations are below 400 pptv. P(O_x) at lower P(RO_x) values (less than 0.2 ppt s⁻¹) exhibits much more noise and are typically below 2 ppb h⁻¹.

3.3 Model evaluation

To investigate our understanding of photochemistry in biomass burning plumes we conducted zero-dimensional modeling constrained by our available measurements excluding ECHAMP XO₂ observations. Figure 6 shows a time series of model-predicted HO₂, speciated RO₂, OH, and OH reactivity acquired using the MCM 3.3.1 with added BB VOC chemistry (“MCM-BBVOC”). Additional results included in the Supplement were acquired using the three other mechanisms: MCM-base, MCM-BBVOC-het, and the base GEOS-Chem mechanism (Figs. S11–S13). Model [OH] (Fig. 6) had daily peak values ranging from 0.10 to 0.15 pptv (2–3×10⁶ molec. cm⁻³). The MCM-base [OH] results were typically higher but generally agreed within 5 % of the other two altered MCM mechanisms. Daytime GEOS-Chem OH concentrations were typically ∼75 % higher than the base MCM prediction. OH predictions were generally unaffected by smoke influence for all mechanisms though upon smoke arrival on 17 August [OH] decreased from ∼0.12 to 0.05 pptv for the MCM-BBVOC mechanism, with similar results observed with the other mechanisms. Modeled XO₂ comprises HO₂ (typically ∼45 %–50 %) and CH₃O₂ (∼20 %–25 %) and the remaining portion a combination of CH₃CO₃, RO₂ derived from isoprene oxidation, and other organic peroxy radicals. Measured XO₂, as included in Fig. 3, is overlaid on Fig. 6. Daytime modeled [XO₂] (between 09:00 to 21:00 MDT), using the MCM-base, MCM-BBVOC, MCM-BBVOC-het, and GEOS-chem mechanisms, is consistently greater than measured [XO₂] on average by 28 %, 34 %, 27 %, and 20 %, respectively (Fig. 7), all of which are within the combined 2σ measurement uncertainty (34 %) and predicted model uncertainty (25 %), though there are periods in which model values are greater than measured XO₂ by nearly a factor of 2 (i.e., the afternoon of 18 August). The inclusion of BB chemistry (MCM-BBVOC) increased XO₂ predictions (compared to MCM-base results), and the inclusion of heterogeneous HO₂ uptake (MCM-BBVOC-het) led to slight decreases (compared to MCM-BBVOC results). The rapid increase in [XO₂] observed on 17 August was not captured under any model conditions. As discussed in Sect. 3.1, this discrepancy from 17 August is likely at least partially due to the impact of HONO, which was not measured and unlikely properly accounted for by the zero-dimensional model. Observed XO₂ on 23 August was high compared to all other days observed, while modeled XO₂ is only somewhat greater than other days. Comparison between daytime XO₂ observations and model results (Fig. 7) was determined using the York bivariate linear regression method (York et al., 2004). Both measured and modeled [XO₂] are low at night (below 10 pptv). Model XO₂ predictions with all chemical mechanisms agreed within 10 %.

The OH reactivities from the MCM-BBVOC mechanism ranged from 5 to 25 s⁻¹ (Fig. 6). These OH reactivities are divided into several categories based on direct measurements and an “other” category for non-measured model outputs. The “other” category was often the greatest category, contributing as much as 40 % of the total reactivity at times. Incorporating heterogeneous losses to the BB VOC mechanism (MCM-BBVOC-het results) had nearly no effect on OH reactivity, and removing BB chemistry (MCM-base results) led to a smaller range in OH reactivity – from 5 to 20 s⁻¹. The GEOS-CHEM OH reactivities were lowest, ranging 3 to 14 s⁻¹, due to this mechanism having a limited number of reactions, causing fewer measured compounds to be constrained. The BB VOC category shown (catechol, furan, methyl furan, furfural, methyl furfural, and guaiacol) was often the greatest contributor to the portion of OH reactivity attributed to measured compounds. BB VOC values were typically 2 to 5 s⁻¹, accounting for 10 % to 30 % of measured reactivity. This category played a lesser role in the MCM-base mechanism as only one measured compound was included – catechol – leading to at most 10 % of measured reactivity. Reactivity from carbonyls (HCHO, CH₃CHO, acetone, and methyl ethyl ketone (MEK)) typically contributed 15 % to 20 % of the measured reactivity. The inorganic category typically accounted for 10 % of measured reactivity, or 1 to 3 s⁻¹, and was dominated by CO. The four biogenic categories included were isoprene, monoterpenes (measured monoterpenes sorted into α-pinene, β-pinene, and limonene), MVK/MACR (methyl vinyl ketone and methacrolein), and MBO (2-methyl-3-buten-2-ol). These categories typically account for between 30 % to 50 % of the OH reactivity attributed to measured species. Other measured reactivity categories were minor and included aromatics (benzene, toluene, C₂ benzenes, C₃ benzenes, and phenol), alkanes (CH₄, C₂H₆, and ethyne), and acids (formic acid and acetic acid). The “other” category had greatest contributions from carbonyl compounds. Changes in OH reactivity were observed for some smoke periods. Subtle changes were noted during the 16 and 17 August events due to changes in the inorganic, carbonyl, biogenic, and BB VOC portions. A noticeable decrease in reactivity from 10 to <5 s⁻¹ occurred on 24 August upon smoke departure. Decreases in nearly all reactivity categories contributed with notable contributions due to depletion in aldehyde concentrations and [CO]. This period sustained the lowest reactivities for the entire data set while also having the lowest concentrations of smoke tracers.

Figure 6 Time series of modeled OH, HO₂, speciated RO₂, and OH reactivity (OHR). These results were acquired using the MCM-BBVOC mechanism. Periods of smoke are shaded as per Fig. 2. Measured XO₂ is included as black markers for comparison. Note that the isoprene XO₂ category includes several RO₂ species produced from isoprene oxidation.

Figure 8 shows modeled RO_x production and termination acquired using MCM-BBVOC-het mechanism for the period of 16 through 18 August. RO_x production is sorted into several categories, and only net rates are provided for reversible processes (e.g., net decomposition of peroxy nitrates). The sum of measurement-based P(RO_x), comprising the reaction of O(¹D) with H₂O and the photolysis of HCHO, CH₃CHO, and acetone, is overlaid in Fig. 8 (shown speciated in Fig. 3) and typically accounts for between 50 %–60 % of total modeled values. For modeled P(RO_x), O(¹D) + H₂O and HCHO photolysis both contribute ∼25 % of predicted RO_x production during midday. Photolysis of carbonyls accounts for most of the remaining modeled daytime P(RO_x), though only a fraction (less than 15 %) of this category is from the measured carbonyl compounds acetaldehyde and acetone. Unmeasured carbonyls account for the rest, with methylglyoxal and glycolaldehyde accounting for 26 % and 8 %, respectively. Predicted methylglyoxal concentrations were typically between 0.4 and 0.6 ppbv. These concentrations are about an order of magnitude greater than those measured at mountaintop sites (Mitsuishi et al., 2018; Kawamura et al., 2013) but lower than those observed at a suburban site in China (Liu et al., 2020) and in biomass burning plumes observed in the Amazon (Kluge et al., 2020). Methylglyoxal is largely formed from the oxidation of MVK and MACR, themselves oxidation products of isoprene. A BB-VOC-related dicarbonyl, 4-oxo-2-pentenal (listed as C5DICARB within MCM), which is formed from methyl furan oxidation, accounted for 13 % of P(RO_x) from carbonyl photolysis. Photolysis of acids and alcohols contributes up to ∼10 % of modeled P(RO_x). Nighttime P(RO_x) is typically ∼0.1 pptv s⁻¹ and is primarily from alkene ozonolysis (>80 %) and the reaction of alkenes with NO₃. Net formation of peroxy nitrates RO₂NO₂, mainly PAN, was the dominant modeled RO_x sink from the morning until ∼12:00 MDT most days. Given that neither PAN nor the acetyl peroxy radical (CH₃CO₃) was directly measured and that a zero-dimensional model cannot be expected to accurately model PAN concentrations due to its long lifetime, this result is highly uncertain. Midday RO_x termination was dominated by RO_x self-reactions, with ∼75 % of L(RO_x) from HO₂ + RO₂ and the remainder of L(RO_x) almost entirely from HO₂ + HO₂. Reactions between RO₂ and other RO₂ played a comparatively minor role (<5 % L(RO_x)). Other L(RO_x) categories consisted of reactions of RO₂ + NO, RO + NO₂, and OH + NO₂. Heterogeneous uptake of HO₂ generally had small contributions to L(RO_x) but at times of elevated PM concentrations such as the BB-influenced periods 17, 23, and 24 August accounted for up to ∼10 % of total RO_x termination (see Sect. 3.4 for further discussion). Results from the 21 through 24 August period (see Fig. S17) are similar to the results presented above for 16–18 August, though the unmeasured portion of P(RO_x) is smaller for the former. This resulted from the considerably smaller concentrations of MVK, MACR, isoprene, and BB VOCs measured during this period. The smaller portions for the carbonyls and “alcohols, acids” categories led to smaller P(RO_x) totals that peaked near 0.7 pptv s⁻¹ rather than the calculated 1.2 pptv s⁻¹ for the period shown in Fig. 8.

Figure 7 Comparison of all daytime XO₂ observations (09:00–21:00 MDT) to modeled XO₂.

Figure 8 Modeled RO_x production and termination for 16, 17, and 18 August. The results provided were modeled with F0AM using the MCM-BBVOC-het mechanism. Net rates are provided for reversible processes. RO_x termination L(RO_x) and RO_x production P(RO_x) are separated into several straightforward categories as represented in their respective legends. Heterogeneous HO₂ loss is represented by the HO₂ + AER (AER = aerosol) category. P(RO_x) derived from direct measurements, as presented in Fig. 3, is shown for comparison and is represented by the blue markers. The modeled results are shown in 10 min intervals and were averaged from the 2 min basis, while measurement-based P(RO_x) is shown in 2 min intervals.

3.4 Model sensitivity

There are important fundamental limitations to how well a zero-dimensional model can describe the McCall measurements. Concentrations of several radical precursors such as nitrous acid, glycolaldehyde, methylglyoxal, and glyoxal were not constrained by measurements and were instead determined by the model. For days on which dilute biomass burning plumes arrived suddenly, it is unrealistic to expect that the model can accurately determine the concentrations of these compounds which depend on the history of the air mass. XO₂ predictions are sensitive to several model inputs. Since some secondary compounds' concentrations build up in the model and can drive significant RO_x production, we explore the sensitivity of model XO₂ predictions to the first-order dilution rate constant applied to all compounds. Shortening the dilution lifetime from 24 to 6 h (the minimum value suggested by Wolfe et al., 2016) decreases model XO₂ predictions and reduces the GEOS-Chem XO₂ overprediction from 20 % to 2 % (Fig. S15). The most important unmeasured radical precursors that are affected by this dilution are methylglyoxal, glycolaldehyde, and 4-oxo-2-pentenal due to their collective contribution to P(RO_x). We also investigate model XO₂ sensitivity to NO_x. Increasing NO_x inputs by 50 % decreases peak daily XO₂ predictions by roughly 10 %.

As the 15 pptv increase in [XO₂] observed on 17 August is not captured by model simulations, we include HONO as an additional model constraint in additional GEOS-Chem simulations (see Sect. S8). Constrained HONO concentrations were determined by the product of selected $\mathrm{\Delta }\mathrm{HONO}/\mathrm{\Delta }\mathrm{CO}$ values to the measured CO mixing ratios during BB periods. To achieve a similar ∼15 pptv XO₂ enhancement as measured during the 17 August BB event, a HONO enhancement ratio of near 3 pptv ppbv⁻¹ is required, which provides an additional 0.15–0.60 ppbv HONO throughout the BB period. This $\mathrm{\Delta }\mathrm{HONO}/\mathrm{\Delta }\mathrm{CO}$ value is 30 times larger than observed by Peng et al. (2020) for similarly aged plumes. While this value is likely unrealistic, larger $\mathrm{\Delta }\mathrm{HONO}/\mathrm{\Delta }\mathrm{CO}$ ratios have been reported by Peng et al. (2020). Other unmeasured RO_x precursors were likely present and at least partially responsible for the elevated XO₂ concentrations observed.

Model sensitivity to heterogeneous HO₂ uptake was also investigated. The introduction of heterogeneous losses of HO₂ had overall minimal impacts (see Fig. 7 MCM-BBVOC-het results), even though a fairly high HO₂ uptake coefficient of 0.2 was used (Abbatt et al., 2012). Heterogeneous losses decreased modeled [OH] and [XO₂] by an average of 3.4 % and 2.9 %, respectively, though the impact is more evident for smoke periods with elevated PM. Our analysis of HO₂ heterogeneous uptake is limited by the uncertainty in the HO₂ uptake coefficient and the specific surface area parameter. To investigate the sensitivity of model results to these parameters, these parameters were varied for several GEOS-Chem model simulations (see Figs. S9 and S10). We focus our heterogeneous chemistry sensitivity tests on a 40 min period during the 17 August BB event in which OA concentrations were 30 µg m⁻³. Inclusion of heterogeneous chemistry with standard parameter settings (γ=0.2) leads to a 11 % decrease in [HO₂], whereas use of a much higher and likely unrealistic HO₂ uptake coefficient of 0.5 resulted in a 25 % HO₂ decrease. A similar HO₂ decrease results from using a higher specific surface area of 10 m² g⁻¹. Use of a smaller HO₂ uptake coefficient of 0.02 led to a nearly negligible decrease in [HO₂]. Constraining the OA concentration at 100 µg m⁻³ for the same period – much higher than actually observed – leads to heterogeneous [HO₂] loss of 3 % for an uptake coefficient of 0.02 and 30 % for an uptake coefficient of 0.2. The only conditions in which heterogeneous loss of HO₂ to BB smoke would appear to be important would be for less dilute plumes ([OA]] >100 µg m⁻³) and a high uptake coefficient (γ>0.2).

4 Conclusions

Peroxy radicals were measured during the FIREX 2018 campaign in McCall, Idaho in order to better characterize RO_x chemistry and study O₃ formation within BB plumes. There were five distinct BB-influenced periods that were identified using smoke tracers, primarily HCN. HYSPLIT back trajectories were paired with satellite data to suggest smoke sources and plume age. Most smoke periods had distinct enhancements in O₃, with enhancement ratios of up to 0.06 ppbv ppbv⁻¹$\mathrm{\Delta }{\mathrm{O}}_{\mathrm{3}}/\mathrm{\Delta }\mathrm{CO}$. Zero-dimensional box model results for XO₂ were generally greater than measured XO₂. These simulated results were acquired with F0AM using a variety of chemical mechanisms – GEOS-Chem, MCM, and two expanded versions of MCM. All model iterations followed the general trends observed for [XO₂] measurements, though a measured 15 pptv XO₂ enhancement during a 17 August BB event was not captured by any model iteration. This includes simulated results acquired using expanded versions of the MCM that had additional BB VOC chemistry and heterogeneous HO₂ and OH losses intended to better capture BB influence. Heterogeneous losses overall had minimal impact on XO₂, even though BB smoke periods with elevated organic PM levels often near 30 µg m⁻³ led to decreases in [HO₂] by 10 %. Heterogeneous chemistry was investigated using a variety of HO₂ uptake coefficients and aerosol specific surface areas. At the greatest settings, these variables led to decreases in HO₂ of at most 25 % and 25 %, respectively, during a period with PM above 30 µg m⁻³. Both measured and modeled XO₂ concentrations were used to calculate P(O_x). The presence of smoke had an overall negligible impact on P(O_x) as NO_x enhancements were minimal.

Quantification of P(RO_x) suffered from lack of constraints for several compounds including HONO, methylglyoxal, and glycolaldehyde. HONO photolysis likely contributed to the enhanced concentrations in [XO₂] measured during the 17 August BB period, though a rather large $\mathrm{\Delta }\mathrm{HONO}/\mathrm{\Delta }\mathrm{CO}$ value near 3.0 ppbv ppbv⁻¹ is necessary for a similar XO₂ enhancement. An additional gas-phase process in need of validation by measurements is the role of PAN formation as a RO_x sink during the morning. The importance of unmeasured RO_x precursors was especially sensitive to the first-order dilution rate constant applied to all unmeasured species. Finally, the role of heterogeneous HO₂ uptake as a RO_x sink would benefit from more direct measurements of particle size distribution and knowledge of HO₂ uptake coefficients. Heterogeneous HO₂ uptake was minimal for the dilute BB plumes studied here, and it would appear to only be important in less dilute BB plumes if the uptake coefficient is relatively high (e.g., 0.2).