Ozone chemistry in western U.S. wildfire plumes

While ozone increases rapidly in wildfire plumes, downwind its production rate slows dramatically as nitrogen oxide levels decline.

Abstract

Wildfires are a substantial but poorly quantified source of tropospheric ozone (O₃). Here, to investigate the highly variable O₃ chemistry in wildfire plumes, we exploit the in situ chemical characterization of western wildfires during the FIREX-AQ flight campaign and show that O₃ production can be predicted as a function of experimentally constrained OH exposure, volatile organic compound (VOC) reactivity, and the fate of peroxy radicals. The O₃ chemistry exhibits rapid transition in chemical regimes. Within a few daylight hours, the O₃ formation substantially slows and is largely limited by the abundance of nitrogen oxides (NO_x). This finding supports previous observations that O₃ formation is enhanced when VOC-rich wildfire smoke mixes into NO_x-rich urban plumes, thereby deteriorating urban air quality. Last, we relate O₃ chemistry to the underlying fire characteristics, enabling a more accurate representation of wildfire chemistry in atmospheric models that are used to study air quality and predict climate.

INTRODUCTION

Wildfires emit large quantities of reactive trace species to the atmosphere, including primary pollutants, as well as precursors for the production of O₃ and particulate matter (1, 2). The number and size of wildfires are predicted to increase as a result of historical fire suppression practices and ongoing climate change (3). This threatens to offset some of the improvements in air quality in the United States over the past few decades, particularly during fire season (4).

O₃ formation depends on the mix of initial emissions and the postemission atmospheric processing, both of which are highly variable (Fig. 1). As a result, O₃ formation observed in previous field studies exhibits substantial fire-to-fire variability (5). Numerous studies have investigated O₃ chemistry in wildfire plumes using atmospheric models of different dynamical and chemical complexity (6–11), but accurate simulation of wildfire chemistry has proved challenging. Several hypotheses have been proposed to explain the model deficiencies, such as uncertain emission inventories, inaccurate description of oxidation chemistry, and difficulties in modeling plume dispersion. O₃ production from wildfire emissions remains as a major uncertainty in assessing the tropospheric O₃ burden (12).

Fig. 1. Simplified scheme to illustrate the factors influencing O₃ formation in wildfire plumes.

Wildfires emit oxidant precursors, NO_x, and an enormous diversity of VOCs. In the near field, OH produced via photolysis of HONO initiates VOC oxidation, which proceeds in the presence of NO_x and leads to efficient O₃ formation. After a few hours, the HONO has been consumed and NO_x has been both diluted sufficiently and converted to PANs and ${\text{NO}}_3^{-}$ such that the O₃ formation slows by several orders of magnitude. In this simplified scheme, the width of arrows having the same color represents the relative importance of competing pathways.

The in situ observations of a suite of trace species made during the Fire Influence on Regional to Global Environments and Air Quality (FIREX-AQ) campaign (Supplementary Materials, section S1) enable a detailed diagnosis of key variables controlling O₃ formation, including oxidant sources, volatile organic compound (VOC) emissions, and the chemistry of NO_x and peroxy radicals (RO₂; the sum of hydroperoxy radical and organic peroxy radical) (Fig. 1). These variables depend on fire conditions, undergo rapid transitions in chemical regimes, and hence profoundly influence the O₃ chemistry during smoke transport. Building upon our systematic evaluation of O₃ chemistry, we provide a parameterization to estimate the O₃ formation from temperate wildfires.

During FIREX-AQ, the NASA DC-8 aircraft sampled fires representative of those in the major ecosystems in the western United States in July and August 2019. Figure 2B shows one example flight track that involves multiple crosswind transects of a fire plume at different distances downwind. Previous analyses of aircraft-based observations typically studied the plume evolution in a pseudo-Lagrangian framework. Such analysis is often complicated by the fact that fire conditions change over time and by aircraft navigation artifacts, such as missing the dense plume in some crosswind transects (Supplementary Materials, section S1). Here, to investigate the O₃ chemistry in a way that mitigates some of the challenges associated with fluctuations in fire emissions, we apply single transect analysis (STA) that examines the differences in the plume composition across each crosswind transect. Because of the high aerosol optical extinction in the center of large smoke plumes, the center experiences substantially lower actinic flux and photolysis rates than the edges at a given altitude. This provides a different extent of photochemical processing and, in particular, a range of time-integrated exposure of emissions to hydroxyl radicals (i.e., OH exposure) between the plume center and edges (Fig. 2A as an example). Since a single transect samples smoke emitted at similar times, the assumption of stationary fire conditions is often better satisfied in STA than traditional pseudo-Lagrangian analysis. Spatial variability in fire emissions and complex plume structure can still complicate the STA, so transects suitable for the STA are scrutinized by a set of stringent criteria (Supplementary Materials, section S4).

Fig. 2. Single transect analysis (STA) examines the differences in plume composition across individual transects of the wildfire plumes.

In (B), the flight track on 3 August 2019 is colored by OH exposure, which is lower in plume center than edges, as a result of high aerosol optical extinction in plume center. In (A), the dilution-corrected O_x formation (i.e., ΔO_x/ΔCO) is illustrated in one near-field transect.

The STA is combined with a conceptual model (fig. S13 and Supplementary Materials, section S5) to investigate the daytime chemical closure of odd oxygen [O_x = O₃ + NO₂ + HNO₃ + particulate nitrate + peroxyacylnitrates (PANs)]. O_x accounts for the interconversion between O₃ and other O_x species (13). The instantaneous production rate of O_x can be expressed by the product of three terms: VOC reactivity (VOCR), OH concentration, and the fraction of peroxy radicals that react with NO (f_{RO2 + NO}) (i.e., Eq. 1). VOCR is a condensed parameter summarizing several properties of individual VOCs (Eq.2), including the VOC concentration ([VOC_i]), the reaction rate coefficient of the VOC with OH (k_{OH + VOCi}), the number of peroxy radicals produced from the oxidization of each VOC_i molecule to its first-generation closed-shell products (γ_i), and the alkylnitrate branching fraction of the VOC_i-derived RO₂ + NO reaction (α_i). More details about VOCR are described in the Supplementary Materials, section S7.

Integrating Eq. 1 from the fresh (i.e., lowest OH exposure) to the aged portion (i.e., highest OH exposure) across each plume transect (i.e., Eq. 3) reflects the predicted O_x formation based on the observationally constrained VOCR, OH exposure, and RO₂ chemistry. To account for dilution and background contributions, excess mixing ratios (i.e., the difference between smoke and background air, denoted as Δ in Eq. 3) were normalized to Δ[CO], which is a stable plume tracer. The predicted O_x production can be compared to the direct measurement of the same transect (i.e., left hand side of Eq. 3), providing a diagnostic of chemical closure, enabling constraints on the sources and sinks of O_x. This analysis is denoted as O_x chemical closure analysis.

$$\frac{d[{\mathrm{O}}_x]}{\mathit{dt}}=\text{VOCR}·[\text{OH}]·f_{\mathrm{R}{\mathrm{O}}_2+\text{NO}}$$ (1)

$$\text{VOCR}={\displaystyle {\sum }_{i=1}^{i=\mathrm{n}}}{k}_{\text{OH}+\text{VO}{\mathrm{C}}_i}·[\text{VO}{\mathrm{C}}_i]·{\mathrm{\gamma }}_i·(1-{\mathrm{\alpha }}_i)$$ (2)

$${\left(\frac{\mathrm{\Delta }[\mathrm{O}x]}{\mathrm{\Delta }[\text{CO}]}\right)}_{\text{aged}}-{\left(\frac{\mathrm{\Delta }[\mathrm{O}x]}{\mathrm{\Delta }[\text{CO}]}\right)}_{\text{fresh}}={\displaystyle {\int }_{\text{fresh}}^{\text{aged}}}\frac{\text{VOC}{\mathrm{R}}_{[\text{OH}]t}}{\mathrm{\Delta }{[\text{CO}]}_{[\text{OH}]t}}·f_{\mathrm{R}{\mathrm{O}}_2+\text{NO}}·d([\text{OH}]t)$$ (3)

RESULTS

Variables influencing O_x formation

OH exposure

The OH exposure is estimated from the observed ratio of phenol to benzene (eq. S6 and Supplementary Materials, section S3), both of which are emitted in high yields in wildfires. Phenol reacts with OH ∼20 times faster than benzene, so their ratio serves as a measurement of photochemical processing in the absence of other substantial sinks or sources. The OH exposure is highly correlated with the nitrous acid (HONO) loss. Figure 3A shows the measurements on the 3 August 2019 flight as an example. Before 90% of HONO is lost, the OH exposure correlates with the lost HONO whose photolysis accounts for >50% of the total HO_x production rate (Supplementary Materials, section S3). HONO photolysis is thus a critical OH source in wildfire plumes, consistent with a recent study by Peng et al. (14). After HONO is depleted, the OH exposure continues to increase because of the photolysis of O₃ and aldehydes, albeit at a much slower rate, indicating lower [OH] (figs. S5 and S6).

Fig. 3. Production and fate of OH.

(A) shows that the OH exposure correlates with the amount of HONO loss [f_HONO,lost = 1 − (ΔHONO/ΔCO)/(ΔHONO/ΔCO)_max] for the 3 August 2019 Williams Flats Fire. The correlation indicates that OH is produced mainly by HONO photolysis in the near field. The color represents the relative contribution of HONO photolysis to total HO_x production rate (denoted as f_jHONO). (B) shows that OVOCs, alkanes/alkenes, and furans are the major contributors to total VOCR based on the average of transects included in the O_x chemical closure analysis.

VOC reactivity

The approximately 80 quantified VOCs are classified into seven structural categories. Figure 3B shows the relative contribution to total VOCR of each category averaged from transects included in the O_x chemical closure analysis. On average, oxygenated VOCs (OVOCs) are the largest contributor, together accounting for about one-third of VOCR. The OVOCs are predominantly small aldehydes, including formaldehyde and acetaldehyde (fig. S21). Alkanes and alkenes are the second largest contributors to VOCR. The historically overlooked furans also play an important role in wildfire plumes, contributing about one-fifth of VOCR, consistent with recent findings from lab studies (10, 15). While oxygenated aromatics, primarily guaiacol, catechol, and creosols, account for only one-tenth of total VOCR, their oxidation contributes a much larger fraction of the secondary organic aerosol (SOA) formed [∼60% as found in (16, 17)].

The relative importance of each VOC category to total VOCR changes with OH exposure. An example transect is shown in fig. S22. Many of the primary emissions, including alkenes, furans, and oxygenated aromatics, are rapidly oxidized, and their importance decreases with increasing OH exposure. In contrast, small aldehydes have substantial secondary sources, and, as a result, their contribution to the total VOCR increases over time. The VOCR of longer lived compounds, such as CO, remains relatively constant.

RO₂ chemistry

O₃ is produced via the reaction of RO₂ with NO. There are, however, a number of processes that can compete with this reaction. Thus, to understand O_x formation in wildfire plumes, knowing the RO₂ fate is critical. With direct measurements of organic hydroperoxides (ROOH) and hydroxynitrates (RONO₂) from the OH-initiated oxidation of small alkenes (i.e., ethene and propene), we are able to provide the first experimental constraint on RO₂ fate in wildfire plumes. We probe the competition between RO₂ + NO and RO₂ + HO₂ reactions and thereby estimate the fraction of RO₂ that reacts with NO (f_{RO2 + NO}). Figure 4 shows the evolution of propene-derived ROOH and RONO₂ in two transects with different NO levels. In the transect shown in Fig. 4A, where [NO] is above 500 parts per trillion by volume (pptv), only RONO₂ is produced, as the RO₂ + NO reaction outruns the RO₂ + HO₂ reaction. In the transect shown in Fig. 4B, [NO] is below 500 pptv and reaches as low as 50 pptv. As a result of the low [NO], both ROOH and RONO₂ are produced, suggesting that RO₂ + HO₂ and RO₂ + NO reactions are competitive. H₂O₂, which is a product of HO₂ + HO₂ reaction, shows a similar trend as ROOH in these two transects (fig. S24).

Fig. 4. The measurements of ROOH and RONO₂ from propene oxidation are used to diagnose the RO₂ fate.

The ROOH is not produced in the transect with high [NO] (A) but produced in the transect with low [NO] (B). The signals of both RONO₂ and ROOH are divided by the branching ratio of the corresponding RO₂ reaction (i.e., α). The ROOH signal is multiplied by a factor of 4 to be shown in the same scale as RONO₂. The shaded area represents the 25th to 75th percentile. ppb, parts per billion.

Measurement imprecision precludes the estimate of a pointwise f_{RO2 + NO} across each transect, so we apply Eq. 4 to calculate transect-averaged f_{RO2 + NO} using the transect-integrated production of RONO₂ (i.e., P_RONO2; eq. S29) and ROOH (i.e., P_ROOH; eq. S30). f_{RO2 + NO} is calculated from both ethene and propene systems, and they are consistent within 10% (fig. S25). Figure 5A shows the evolution of f_{RO2 + NO} for the Williams Flats Fire sampled on two different days. On both days, the f_{RO2 + NO} decreases with downwind distance, illustrating the transition of RO₂ fate from an RO₂ + NO–dominated regime to a mixed regime with increasing importance of RO₂ + HO₂. The change rate of f_{RO2 + NO} varies between fires. On 7 August 2019, the f_{RO2 + NO} decreases from 1 to 0.7 after the smoke travels from 25 to 100 km. On 3 August 2019, the f_{RO2 + NO} decreases more rapidly with downwind distance, and it reaches ∼60% at 45 km (estimated transport time ∼3 hours). Such difference is likely caused by fire strength and fuel consumption. The fire on 7 August 2019 is the most intense fire sampled during FIREX-AQ, with the fire radiative power (FRP) up to 4.4 × 10⁴ MW and 72.3 km² daily area burned. The fire on 3 August 2019 has lower intensity (i.e., peak FRP ∼1.5 × 10⁴ MW) and smaller daily burned area (43.2 km²). It takes more time for the NO_x concentration in intense fires to decline to a level where RO₂ + HO₂ reactions can become competitive. Note that over 90% of fires around the world have FRP <100 MW (18), so that the transition of f_{RO2 + NO} can occur rapidly. More importantly, a large fraction of wildfire VOCs is oxidized in the mixed regime. As shown in Fig. 5B, for both fires, ∼70% of the VOCR remains when f_{RO2 + NO} decreases to 0.6

$$\begin{array}{cc}\hfill f_{\mathrm{R}{\mathrm{O}}_2+\text{NO}}& =\frac{k_{\mathrm{R}{\mathrm{O}}_2+\text{NO}}·[\text{NO}]}{k_{\mathrm{R}{\mathrm{O}}_2+\text{NO}}·[\text{NO}]+k_{\mathrm{R}{\mathrm{O}}_2+\mathrm{H}{\mathrm{O}}_2}·[\mathrm{H}{\mathrm{O}}_2]}\hfill \\ & =\frac{\frac{P_{\text{RON}{\mathrm{O}}_2}}{{\mathrm{\alpha }}_{\text{RON}{\mathrm{O}}_2}}}{\frac{P_{\text{RON}{\mathrm{O}}_2}}{{\mathrm{\alpha }}_{\text{RON}{\mathrm{O}}_2}}+\frac{P_{\text{ROOH}}}{{\mathrm{\alpha }}_{\text{ROOH}}}}\hfill \end{array}$$ (4)

Fig. 5. The RO₂ fate transitions from an RO₂ + NO–dominated regime to a mixed regime with increasing importance of RO₂ + HO₂.

(A) The f_{RO2 + NO} decreases as smoke transports in the William Flats Fire sampled on 2 different days. The data points are colored by the fire radiative power (FRP) measured at the estimated time of smoke emission. (B) A large fraction of VOCs is oxidized in the mixed regime. The max ΔVOCR/ΔCO is represented by the average ΔVOCR/ΔCO of observations with the top 1% [CO] during the fire sample. The downwind distance is estimated on the basis of the aircraft position and the burned area. The dashed lines are provided as a visual aid. The ellipses represent the uncertainty range.

This regime transition is a result of [NO_x] decrease, which is caused primarily by dilution with ambient air and by chemical loss of NO_x. The major NO_x oxidation products are PAN and nitrate ($\mathrm{N}{\mathrm{O}}_3^{-}$ = HNO₃ + particulate nitrate). Together, they account for nearly all of NO_x oxidation products, NO_z (= NO_y − NO_x− HONO) (fig. S27). The fractions of PAN and nitrate in total reactive oxidized nitrogen (NO_y) increase with OH exposure as a result of NO_x conversion (Fig. 6A), consistent with previous studies (6, 19, 20).

Fig. 6. The evolution of the partitioning of NO_y species.

(A) shows measurements of the 3 August 2019 Williams Flats Fire. As smoke ages, the NO_x and HONO emitted from fires are converted to PAN and NO₃⁻. (B) shows that the fraction of NO_x loss to PAN (f_PAN) across each transect increases with smoke age, which results from evolving CH₃CHO/NO₂ as discussed in the text. Each data point represents one transect, and the transects from the same fire sampling patterns have the same color. The black line is provided as a visual aid. The numbers in parentheses represent the index of a set of crosswind transects in a flight.

Because nitrate is a permanent NO_x sink but PAN is a temporary NO_x reservoir, the NO_x loss pathways affect O₃ formation in the long-range transport of wildfire plumes. To investigate the competition between NO_x loss pathways, we use STA. ΔPAN/ΔCO and ΔNO_z/ΔCO correlation slopes (fig. S28) give the relative fraction of NO_x loss to PAN (denoted as f_PAN) as the smoke chemically evolved from the photochemical condition in plume center to that in plume edge across individual transects. f_PAN is different from ΔPAN/ΔNO_z, as the latter is an accumulative property that depends on initial emissions and the integral of NO_x loss over time. Figure 6B shows f_PAN for each transect of several fires as a function of smoke age. Despite fire-to-fire variability, f_PAN is 0.2 to 0.4 at a smoke age of 0.5 hour and rapidly increases to 0.8 to 1 at 2 hours. This trend suggests that the major NO_x oxidation product transitions from $\mathrm{N}{\mathrm{O}}_3^{-}$ to PAN after ∼2 hours of transport.

This transition is mainly driven by the change in [CH₃CHO]/[NO₂], which increases with smoke age (fig. S30) and reflects the fact that NO₂ is chemically lost to other NO_y species, but CH₃CHO has substantial production from VOC oxidation. Larger [CH₃CHO]/[NO₂] favors the PAN formation by producing more acetyl peroxy radical (Supplementary Materials, section S8). Therefore, fire conditions that affect the [CH₃CHO]/[NO₂], or broadly the [VOCs]/[NO_x], alter the partitioning between NO_y species and, as a result, downwind O₃ formation. Figure S33 shows that the plateau value of ΔPAN/ΔNO_y from different fires negatively correlates with the modified combustion efficiency (MCE). This observation is consistent with the finding from STA that higher emission ratios of [CH₃CHO]/[NO₂] (associated with lower MCE; fig. S38) favors NO_x loss to PAN.

O_x chemical closure analysis

We now return to the conceptual model (Eq.3) to test the chemical closure of O_x in wildfire plumes. The O_x production (denoted as P_Ox) across each transect is predicted on the basis of the three key chemical variables: OH exposure, VOCR, and RO₂ fate. Then, this prediction is compared to the measured P_Ox calculated as a sum of the measured individual O_x species (Eq.3). On the basis of a set of stringent criteria (Supplementary Materials, section S4), 25 transects, for which the P_Ox and RO₂ fate can be quantified with high confidence, are selected for this comparison. As shown in Fig. 7, the correlation between the observed and predicted P_Ox is quite strong (r² = 0.64). On average, the predicted P_Ox is higher than the measured P_Ox by 12%, well within the analysis and measurement uncertainties (Supplementary Materials, section S9). Overall, the use of the conceptual model and the comprehensive measurements of VOCs in FIREX-AQ enables remarkably good prediction of O_x production. Such agreement suggests that the majority of VOCs contributing to O_x formation are quantified during FIREX-AQ, at least in the early stage of the wildfire plumes. This provides confidence in the characterization of fire emissions during FIREX-AQ, which will serve as a foundation for future use in chemical transport models (CTMs). Furthermore, as the conceptual model solely based on gas phase chemistry is sufficient to account for the measured O_x production here, we suggest that the role of heterogeneous loss of O₃ and HO₂ is likely minor in wildfire plumes, a hypothesis often invoked when models overpredict the measured O₃ (5, 21).

Fig. 7. The predicted and measured O_x production show reasonable agreement.

The ellipses represent the uncertainty range (Supplementary Materials, section S9). The slope and intercepts are obtained from a York fit.

Parameterization of the O₃ + NO₂ production

The chemistry and dynamics described in this study occur on spatial scales smaller than those used in even modestly high-resolution CTMs. Thus, there is a need to parameterize the near-field chemistry to properly capture the oxidation chemistry. Here, we focus on O₃ and NO₂, as they are critical air pollutants. The production of O₃ and NO₂ across individual transects, which is represented by the difference in Δ(O₃ + NO₂)/ΔCO between aged and fresh smoke, is denoted as P_{O3 + NO2}. P_{O3 + NO2} ranges from 0 to 0.06 and exhibits a positive relationship with the span of OH exposure (ΔOH exposure) across individual transects (r² = 0.47; Fig. 8A). This trend implies more O₃ + NO₂ production as plumes age in the near field, consistent with previous observations (5). In addition to OH exposure, the P_{O3 + NO2} positively correlates with MCE (r² = 0.23; Fig. 8B). Higher MCE indicates more flaming combustion, which usually leads to higher NO_x emissions and lower VOC emissions, together leading to a higher NO_x/VOCR (5, 22, 23). The P_{O3 + NO2} does increase with NO_x/VOCR, as shown in fig. S34. Overall, the positive relationship between P_{O3 + NO2} and MCE suggests that the formation of O₃ + NO₂ in fresh wildfires in the western United States is generally NO_x limited.

Fig. 8. Parameterization of the O₃ + NO₂ production.

The measured production of O₃ + NO₂ (P_{O3 + NO2}) across individual transects exhibits positive correlation with the span of OH exposure (ΔOH exposure) and MCE, as shown in (A) and (B), respectively. Thirty-nine transects are selected for this analysis (Supplementary Materials, section S4). (C) Comparison between predicted and measured P_{O3 + NO2} for individual transects. (D) The emission ratios (ERs) of NO₂ to CO derived from the field [i.e., a + b×max(0,MCE-c)] and measured in the 2016 FIREX FireLab are plotted as a function of MCE.

As the O₃ + NO₂ formation depends on several variables, we develop a statistical model based on multivariate adaptive regression splines (24) to attribute such dependence (Supplementary Materials, section S10). We examine the relationship between P_{O3 + NO2} of each transect and a number of variables (MCE, ΔOH exposure, VOCR, NO_x/VOCR, and RO₂ fate) using stepwise forward selection. The final model form is Eq. 5 [the units of P_{O3 + NO2} and OH exposure are parts per billion (ppb)/ppb and 10¹⁰ molecules cm⁻³ s, respectively]. The model captures 56% of the measurement variance (Fig. 8C)

$$\begin{array}{c}{P}_{{\mathrm{O}}_3+\mathrm{N}{\mathrm{O}}_2}=a+b\times \text{max}(0,\text{MCE}-c)+d\times (\text{OH exposure})\\ a=0.0036\pm 0.0028;b=0.46\pm 0.16\\ c=0.916\pm 0.002;d=0.014\pm 0.0019\end{array}$$ (5)

The terms a + b × max(0, MCE − c) in Eq. 5 are interpreted as the MCE-dependent primary emission ratio (ER) of NO₂ to CO, i.e., ER(NO₂), because O₃ + NO₂ is essentially all NO₂ when there is no chemical aging of fire emissions. To examine this interpretation, we compare the field-derived ER(NO₂) to that measured in the FIREX FireLab 2016 study, where fuel complexes important for western U.S. ecosystems were burned. Figure 8D compiles the ER(NO₂) from lab fuel types that are relevant to FIREX-AQ fires (table S7). The empirical parameterization reasonably predicts the nearly constant ER(NO₂) when MCE is <0.92 and slightly overpredicts the rising ER(NO₂) as MCE increases above 0.92. One factor that complicates this comparison is the fuel dependence of ER(NO₂), which shows larger variability as MCE increases. In comparison to individual fuel types (fig. S36), the empirical parameterization reasonably predicts the ER(NO₂) of douglas fir, Engelmann spruce, and subalpine fir, but slightly overpredicts for fuels like ponderosa pine and manzanita. Among all 253 transects in FIREX-AQ, more than 90% of transects have MCE less than 0.92 (fig. S2), a range where the field-derived parameterization performs accurately, and the ER(NO₂) is largely independent of fuel type (fig. S36). Therefore, this field-derived parameterization is a reasonable approximation of the subgrid scale O₃ + NO₂ production for CTMs without an accurate emissions inventory and fuel characteristics.

The other term in Eq. 5 (d × OH exposure) is interpreted as the O₃ + NO₂ formation during plume aging. This linear dependence of O₃ + NO₂ production on OH exposure is likely confined to the near field of wildfire plumes (i.e., maximum OH exposure used to constrain the parameterization is 2.5×10¹⁰ molecules cm⁻³ s, which is roughly 7 hours transport time) before the RO₂ chemistry transitions to HO₂-dominated reactivity. We compile the literature values of ΔO₃/ΔCO from boreal and temperate wildfires over a wide range of plume ages in fig. S35 and find that the aircraft-based observations of ΔO₃/ΔCO in the free troposphere typically reach a maximum value of 0.1 at 3 to 5 days downwind, which is only about twice the value after 7 hours of aging observed in this study. The ΔO₃/ΔCO is relatively constant afterward and even shows a decreasing trend in some plumes that are ∼10 days old. This observation suggests that the major fraction of O₃ in wildfire plumes in the free troposphere is produced in the near field, consistent with the analysis above that the wildfire plumes quickly run out of NO_x and then the reaction of HO₂ with RO₂ efficiently competes with NO.

DISCUSSION

Uncertainties in emissions characterization and oxidation chemistry are long-standing challenges in understanding O_x production in wildfire plumes. The agreement between the measured and predicted O_x production in this study indicates that the oxidation of VOCs has been accurately captured by the comprehensive suite of analytical instruments deployed here. This chemical closure provides confidence in diagnosing the key chemical variables influencing O_x formation. These variables undergo rapid transition in chemical regimes. HONO photolysis is the major source of OH in the near field. Once the primary HONO is consumed, the rate of photochemistry in the plume decreases quickly. O_x formation also slows because of the changing fate of RO₂ radicals. Given the high VOC/NO_x produced in the fire, the RO₂ fate transitions within a few hours from an RO₂ + NO–dominated regime to a mixed regime with increasing importance of the RO₂ + HO₂ reaction. A large fraction of VOCs is oxidized in the mixed regime. The changing RO₂ fate affects not only O_x formation but also SOA formation. To estimate SOA formation in wildfire plumes, previous studies have used high NO_x SOA yields from chamber experiments (16, 17). The SOA yields of aromatics, which are critical SOA precursors in wildfire plumes, are generally higher under low NO_x condition than high NO_x condition (25, 26). Therefore, the estimated SOA formation in some previous studies may be biased low if the rapid transition to low NO_x chemistry is not represented accurately.

The O₃ chemistry in temperate wildfire emissions is generally in the NO_x-limited regime. Thus, fire conditions that influence the NO_x emissions and sinks critically determine the O₃ formation. Wildfires with higher MCE have higher emission ratios of HONO and NO_x, which tend to increase O₃ formation. On the other hand, higher MCE is associated with lower CH₃CHO/NO_x, which tends to decrease the fraction of PAN in NO_y and the downwind O₃ production. Given that the high concentrations of VOCs are still present in the aged plumes, O₃ formation will be enhanced when the wildfire smoke is provided with additional NO_x, either internally from the PAN decomposition when plumes descend to higher temperature (27) or externally from mixing with NO_x-rich urban plumes (28) or lightning-derived NO_x (29).

The rapid transition of O₃ chemistry within wildfire plumes highlights a known issue in CTMs, which simulate O₃ formation by generally uniformly mixing wildfire emissions into a few large grid cells. This treatment introduces substantial bias in predicting O₃ formation. Representing the near-field subgrid plume evolution using a field-constrained parameterization such as that developed here and subsequently diluting the chemically processed emissions into a larger grid cell may be an efficient approach to improve the prediction accuracy of CTMs. The amount of O₃ produced in these underresolved plumes can be substantial. For example, using a representative value of Δ(O₃ + NO₂)/ΔCO in the near field (i.e., 0.045) and the estimated CO flux from wildfires averaged from 2011 to 2015 in the western United States [i.e., 5240 ± 2240 Gg year⁻¹ (30)], we estimate that O₃ produced in wildfire plumes can sustain a 3-ppb enhancement in boundary layer O₃ concentration over the western United States during fire season (Supplementary Materials, section S10). The episodic nature of wildfires can result in more severe impacts on the occurrence of O₃ exceedances (5).

MATERIALS AND METHODS

Descriptions of the FIREX-AQ campaign and instrumentation; calculation of OH exposure, VOCR, and RO₂ fate; criteria of transect selection for STA; conceptual model to investigate O_x chemistry and associated uncertainty analysis; statistical model to estimate the O_x background level; and parameterization of the O₃ + NO₂ production can be found in the Supplementary Materials.

Supplementary Materials

This PDF file includes:

Supplementary Text

Figs. S1 to S38

Tables S1 to S7

References