Quantification of organic aerosol and brown carbon evolution in fresh wildfire plumes

The relative contributions of dilution-driven evaporation of BBPOA versus BBSOA formation from either primary emitted gases or the evaporated BBPOA have remained elusive in wildfire studies. Many studies have hypothesized a serendipitous balance between the two, leading to little to no net change in BBOA relative to carbon monoxide (CO) (considered a conserved tracer of dilution) as fire plumes age, including from WE-CAN (17) and others (e.g., refs. 12, 14, 16, 18, 19). The relative contributions of primary and secondary sources to BrC evolution in wildfire plumes have been even less clear (16, 29, 35–37). To evaluate the various processes, we use WE-CAN observations made in fresh wildfire plumes at different extents of plume dilution and chemical aging, as summarized in Fig. 1.

We evaluate the changes in BBOA and BrC light absorption at 405 nm (BrC Abs₄₀₅) using the normalized excess mixing ratio (NEMR) concept, where the difference between the measured in-plume and out-of-plume (“background”) mixing ratio (defined as enhancement) is normalized to that of CO (called CO enhancement; SI Appendix, Eq. S1). We then use the difference in the OA or BrC Abs₄₀₅ NEMR, i.e., ΔNEMR, within a smoke plume between any given transect and that of the densest part of the plume, which is defined as the plume transect with the highest average CO enhancement. The densest plume transect through each smoke plume was always one of the physically closest transects to the fire source but not always the absolute closest, in part due to the challenges of navigating the aircraft through the central portions of the plumes in the fire’s vicinity. Plume density varies significantly in the vertical, so the assignment of the densest transect depends as much on the vertical location of the aircraft as well as the horizontal proximity to the fire. The densest (and physically freshest) transects were typically sampled at a physical age of ∼20 to 50 min. Any secondary chemistry that occurs in the short time between emission and the densest transect cannot be accounted for with the observational analysis here but is expected to be relatively minor because of the short reaction time. We thus operationally define the OA sampled in the densest plume transect as consisting only of BBPOA, while any OA that is determined to be formed due to oxidative plume aging, we label as BBSOA. Our estimates of total BBSOA concentrations may be underestimated due to these assumptions.

The effects of plume dilution can then be investigated by comparing these ΔNEMR values as a function of an observationally constrained plume dilution factor, defined as the ratio of the CO enhancement in the densest part of a sampled plume to the CO enhancement in each of the other transects of that plume. We use the average CO enhancements across each plume transect to smooth over small-scale plume structures. The dilution factor accounts for dilution due to eddy mixing in both the horizontal and vertical directions but does not separate between them. For example, the dilution factor on the edge of a plume near the fire could be equivalent or higher than that in the core of a plume sampled physically further downwind, despite the downwind plume having spent relatively more time in the atmosphere after emission. In other words, physical age and dilution factor are not necessarily correlated.

In the absence of oxidation chemistry, the ΔNEMR of OA and BrC Abs₄₀₅ in a diluting plume are expected to either be invariant or decrease with further dilution as intermediate-volatility organic compounds (IVOCs) and semivolatile organic compounds (SVOCs) evaporate to maintain equilibrium with decreasing vapor concentrations (38–40) or chromophores become photobleached (16, 37, 41, 42). However, emissions of radical precursors such as nitrous acid (HONO) and the presence of increasing solar radiation as the plume dilutes promote oxidation of vapors that were coemitted with BBPOA or that result from BBPOA evaporation, causing potential BBSOA and sBrC formation. The extent of potential BBSOA chemistry can also be described quantitatively using measured reactive losses (i.e., dilution-corrected decreases) of known BBSOA precursors. Observations of these quantities across a diverse range of wildfire plumes provide the opportunity to examine the competing roles of dilution and potential BBSOA- and sBrC-formation chemistry. In particular, the WE-CAN sampling of aged and fresh plumes during midday with active oxidation chemistry as well as plumes that were emitted and sampled closer to sunset, when the rate of oxidation chemistry had slowed, provided a unique set of observational constraints from rarely probed fire plume conditions.

The quantitative relationships between these three quantities, Δ(OA NEMR) or Δ(BrC Abs₄₀₅ NEMR), extent of BBSOA-formation chemistry, and plume dilution across multiple wildfire plumes are summarized in Fig. 1 B and C. We separate the observations into three broad categories of dilution, ranging between dilution factors of 1 to 3, 3 to 6, or 6 to 13 relative to the densest sampled portion of each fire plume (referred to as low, medium, and high dilution), where the larger the dilution factor, the more dilute the plume has become. For each of these categories, we plot the observed ΔNEMR versus the fraction of measured phenolic compound vapors (described below in Evaporated BBPOA Dominates among BBSOA Sources) that have reacted relative to the densest sampled portion of the plume. That fraction is an approximate measure of the chemical age of the plumes and thus is a better proxy for the extent of BBSOA-formation chemistry than the physical plume age. Both the chemical age and the dilution rate in each plume will proceed nonlinearly relative to physical age, and the relative rates will also be variable across different plumes, as variables such as time of day and fire size change. The use of this chemical age proxy is based on the assumption that SOA precursors, whether directly emitted as gases or formed as evaporated primary OA (POA), will have similar reactivities toward hydroxyl (OH) radicals and other oxidants as emitted phenolic compounds. The processes of dilution and SOA formation shown in Fig. 1 occurred rapidly within the first few hours of physical age.

The dilution of the plume leads to a certain amount of net evaporation of OA and BrC or other losses (e.g., photobleaching of chromophores), and thus ΔNEMR < 0 when dilution has occurred but BBSOA-formation chemistry has not proceeded. However, as the extent of BBSOA-formation chemistry increases, the ΔNEMR in all dilution regimes increase. Trend lines of the ΔNEMR versus extent of potential BBSOA formation are fit for each category of dilution using least orthogonal distance regression (uncertainties quoted here are ±1 SD). In the absence of BBSOA-formation chemistry, the y intercepts of the trend lines define the magnitude of BBPOA or BrC Abs₄₀₅ evaporation. BBSOA that formed from oxidation in the plume prior to sampling the densest transect could also be evaporating, but evaporation of BBPOA is likely the dominant process since BBSOA is expected to be a minor component in the densest transects. Little to no evaporation is measured with low dilution. A medium dilution factor causes evaporation of ∼25 ± 9 micrograms per standard cubic meter per part per million (µg sm⁻³ ppm⁻¹) OA and 10 ± 5 per megameter (Mm⁻¹) ppm⁻¹ BrC Abs₄₀₅, while a high dilution factor causes evaporation of 70 ± 9 µg sm⁻³ ppm⁻¹ OA and 32 ± 7 Mm⁻¹ ppm⁻¹ BrC Abs₄₀₅. With a typical OA emission ratio (ER) of 250 ± 30 µg sm⁻³ ppm⁻¹ (17), this corresponds to evaporation of 10 ± 4% and 28 ± 5% of total BBPOA. For BrC Abs₄₀₅, these values similarly correspond to evaporation or other losses of 8 ± 4% and 26 ± 8% based upon a typical ER of 125 ± 30 Mm⁻¹ ppm⁻¹.

Superimposed on top of evaporation are the consistent trends of increasing ΔNEMR values with increasing extent of potential BBSOA-formation chemistry, indicated by the positive slopes of the trend lines. The magnitude of change in ΔNEMR OA and BrC Abs₄₀₅ between no BBSOA-formation chemistry and the full measured extent of BBSOA-formation chemistry then defines the magnitude of BBSOA and sBrC Abs₄₀₅ formation from all sources and, while >0 in all dilution regimes, varies as a function of dilution factor. With a low dilution factor, while a smaller range of chemical processing was sampled, the slope is also smaller suggesting relatively suppressed BBSOA formation. With medium dilution, the observed change in ΔNEMR of OA and BrC Abs₄₀₅ at the full extent of BBSOA-formation chemistry requires production of ∼35 ± 18 µg sm⁻³ ppm⁻¹ BBSOA and 25 ± 10 Mm⁻¹ ppm⁻¹ sBrC Abs₄₀₅. At high dilution, ∼75 ± 16 µg sm⁻³ ppm⁻¹ BBSOA formation and 45 ± 12 Mm⁻¹ ppm⁻¹ sBrC Abs₄₀₅ formation are required to explain the observed trends.

Importantly, the medium and high dilution trend lines reach amounts of ΔOA NEMR and ΔBrC Abs₄₀₅ NEMR at the higher extent of potential BBSOA formation that are consistent with there being little to no net change in total OA or BrC Abs₄₀₅ with plume aging, at least over the first ∼6 h of aging after emission that were primarily sampled during WE-CAN flights. Thus, we illustrate a quantitative separation of the processes leading to typically small observed changes in BBOA relative to CO as a function of plume age. Our observational analysis suggests that balance is achieved between dilution-driven evaporation (or other losses) and concurrent BBSOA and sBrC formation in wildfire plumes. When a typical western US wildfire plume has diluted by an average factor of ∼10 (relative to the densest plumes sampled here), approximately one-third of total BBOA will have converted from BBPOA to BBSOA, which is consistent with the observed increase in OA oxidation state measured in these same plumes (17). Moreover, our observational insights support dynamic OA repartitioning and oxidation invoked in recent models of BBOA emission and evolution (20, 21).

The trend lines in Fig. 1 B and C emerge because the WE-CAN flights sampled a wide enough range of plume-transect conditions to fill in the Fig. 1 coordinates. The campaign-wide trends allow us to separate the effects of dilution-driven evaporation and BBSOA formation averaged over multiple fires. For practical sampling reasons, the data from transects of a single plume typically do not cover a broad enough extent to robustly determine such trends. In reality, both evaporation and BBSOA formation occur simultaneously in most plumes. In a typical midday plume with rapid photochemistry (i.e., many of the plumes sampled here and during previous campaigns), dilution-driven evaporation appears to be the rate-limiting step. SOA formation occurs rapidly after evaporation of BBPOA, causing observed ΔOA and ΔBrC Abs₄₀₅ NEMRs to remain small as daytime plumes progress from low to high extent of BBSOA-formation chemistry and dilution simultaneously. Regardless of the specific path taken through the Fig. 1 coordinates (i.e., relative rates of dilution and BBSOA formation) for a given plume, our results suggest all of these western US wildfire plumes converge toward maintaining near-constant OA and BrC Abs₄₀₅ NEMRs once both dilution and BBSOA-formation chemistry have occurred.

The above analysis suggests that BBSOA and sBrC formation equivalent to 5 to 30% of BBPOA and BrC emissions occurred over the extent of chemical aging in wildfire plumes sampled during WE-CAN. In what follows, we evaluate the extent to which simultaneous observations of a suite of trace gases and particulate components commonly identified as the dominant BBSOA and sBrC precursors explain these extents of BBSOA and sBrC formation.

Several recent studies have combined laboratory measurements with modeling to investigate the sources of BBSOA from oxidation of smoke from various sources (11, 22–25). Generally, the BBSOA precursors fall into four main categories: oxygenated aromatics (i.e., phenolic compounds), reduced aromatics, biogenics, or heterocyclic compounds (e.g., furans). The emissions, and thus magnitude of BBSOA formation, from each category can depend on various parameters, including fuel type, fuel condition, and flaming versus smoldering burn conditions (e.g., refs. 2, 10, 43). The contributions of these categories in wildfire smoke are also uncertain, in large part due to a lack of measurements, especially for phenolic compounds. Here, we evaluate the ERs in wildfire smoke of compounds in each of the four categories using a combination of simultaneous, high-time resolution observations of volatile organic compounds (VOCs), IVOCs, and SVOCs measured with an I⁻ CIMS and a PTR-ToF-MS. Instrument descriptions are provided in SI Appendix.

We first characterize several regimes of reactive carbon in the gas phase of wildfire smoke plumes during WE-CAN as a function of physical plume aging. The physical age of each plume transect was estimated as the distance from the center of the plume transect to the estimated fire location, divided by the average wind speed measured throughout the plume. A representative mass spectrum from the I⁻ CIMS showing the difference and percentage change of a wide array of measured plume components upon aging of a fresh plume can be found in SI Appendix, Fig. S1 for the Taylor Creek fire plume sampled on 30 July 2018 and in SI Appendix, Fig. S2 for the Sharps fire plume sampled on 31 July 2018. The most prominent compounds in the spectrum were oxidized organic molecules containing five or fewer carbon atoms. Previous research has identified these compounds as likely to be small multifunctional organic compounds such as furans (22, 44) or derivatives thereof. Several phenolic compounds were also present in fresh smoke, although at relatively lower concentrations than many of the smaller oxidation products.

Fig. 2 illustrates the NEMR values as a function of physical age after emission for several groups of compounds in the Taylor Creek plume. As in SI Appendix, Figs. S1 and S2, nearly all of the compounds that depleted with aging in the I⁻ CIMS were likely phenolic in nature, joined also by several likely furans and HONO (45). Several compounds measured by the PTR-ToF-MS (including monoterpenes, reduced aromatics, and several furans) also followed this behavior, suggesting direct emissions followed by rapid reaction early in the plume evolution for the compounds shown in Fig. 2A. Other compounds (e.g., nitrophenolics) exhibit behavior suggesting they were initially formed through oxidation chemistry of emitted precursors (or from evaporation of BBPOA) and were then reacted away or condensed into the particle phase once their precursor sources had been depleted (Fig. 2C). Finally, another group of mostly smaller compounds (one to five carbon atoms) were found in persistently increasing concentrations as the products of oxidative plume chemistry (Fig. 2B). These compounds likely include furan oxidation products and fragmentation products from compounds with six or more carbon atoms. SI Appendix, Figs. S1 and S2 illustrate that while variability is seen in the behavior of individual compounds across different fires, the broader trends outlined in Fig. 2 are consistent for all fire plumes. Similar trends are also seen in the PTR-ToF-MS spectrum for both fire plumes (SI Appendix, Fig. S3). For reference, the Taylor Creek plume is highlighted among all plume transects executed during WE-CAN in Fig. 2D.

For the reactive, directly emitted species in Fig. 2A, the NEMR as a plume ages will depend on the extent of oxidation or other chemical transformation that has occurred between time of emission and time of sampling. For the fastest reacting species, such as some phenolic compounds, the lifetime with respect to oxidation can be as little as tens of minutes or less, depending on conditions. To illustrate this reactivity for several major directly emitted species across all four major emission categories, average NEMRs across each plume transect during the campaign were plotted as a function of estimated physical plume age in Fig. 3. Similar figures for all other species with obvious behavior suggesting direct emissions and rapid decay are shown in SI Appendix, Figs. S4 and S5. For the fastest reacting species C₆H₆O₂ (likely catechol) and C₇H₈O₂ (likely methyl catechol), nearly all of the compound had reacted away within the first 60 min or less of plume aging. Most other compounds typically reacted away within the first 100 to 200 min. The slowest-reacting compounds, such as C₆H₆ (benzene) and C₇H₈ (toluene), typically remained to longer ages than were sampled during WE-CAN (SI Appendix, Fig. S5).

We assume the reactive loss of these species is controlled by plume concentrations of OH, which in turn is likely produced predominantly by HONO photolysis in these plumes (2, 15, 44–47). Consistent with this assumption, much of the scatter of NEMR values in Fig. 3 (and SI Appendix, Figs. S4 and S5) is explained by the measured photolysis frequency of HONO (j_HONO), which is a proxy for the amount of sunlight available to drive photochemistry. During daytime when j_HONO in plumes was large (1 × 10⁻³ to 2 × 10⁻³ s⁻¹; HONO photolysis lifetime of 8 to 17 min), OH concentrations were likely to be relatively higher early in the plume evolution, causing most of the compounds to react away within the first 1 to 2 h following emission. During the several plumes that were sampled as the sun was setting, j_HONO decreased to 1 × 10⁻⁴ to 5 × 10⁻⁴ s⁻¹ (HONO photolysis lifetimes of 33 to 167 min), leading to much lower OH production (relative to midday plumes), allowing many compounds to persist within physically older plumes. HONO also followed this same pattern, as presented in ref. 45 and shown in the SI Appendix, Fig. S6 for reference. Nitrate (NO₃) radicals may have also contributed to the decay of phenolics, especially when j_HONO was relatively low, but contributions were likely minor during most daytime plumes.

For all of the compounds shown in Fig. 3 as well as other emissions shown in the SI Appendix, Figs. S4 and S5, we calculated ERs in order to estimate the magnitude of all directly emitted BBSOA sources quantified using the available techniques. Since many of these compounds react rapidly after emission, we estimated their ERs as the average of the measured NEMRs in plume transects where both rate of and time allowed for photochemistry were minimized, in order to derive the best estimate of ER at the fire source. Specifically, we used NEMRs from the closest plume transect of each fire that also had average j_HONO <5 × 10⁻⁴ s⁻¹ and was sampled <100 min after emission (except for C₆H₆O₂ and C₇H₈O₂ measured with I⁻ CIMS, for which a 50-min threshold was used to account for their faster reactivity). Several of the highest estimated ERs were 1.8 parts per trillion per part per billion (ppt ppb⁻¹) for C₄H₄O, 1.7 ppt ppb⁻¹ for both C₆H₆ and C₅H₄O₂, 1.1 and 0.42 ppt ppb⁻¹ for C₆H₆O and C₆H₆O₂ measured using I⁻ CIMS (likely phenol and catechol/benzenediols), and 0.68 ppt ppb⁻¹ for C₁₀H₁₆. Even with this restrictive method of estimating ERs, the lifetimes of some of these species were of the same order as the shortest physical plume ages that were sampled, due mainly to practical restrictions of flying close to wildfires. Therefore, the ERs presented here may be underestimated, perhaps by as much as a factor of 2, and the uncertainties are larger for the faster reacting compounds. Generally, the ER values reported here (SI Appendix, Table S1) range from 0.1 to 2 ppt ppb⁻¹ and are of the same order as previously reported values for laboratory measurements of BB, which themselves exhibit large variability (e.g., refs. 44, 48, 49). Indeed, laboratory BB experiments may not be directly comparable to field measurements, due to differences in fuel mixtures, fuel moisture content, relative contributions of flaming versus smoldering versus distillation emissions, etc.

Using these ERs, we calculated a bottom-up estimate of total expected BBSOA formation from all major measured BBSOA precursor gas emissions. For each precursor, we calculate a potential BBSOA mass contribution using SOA yield values from the literature (SI Appendix, Table S2 and refs. therein), together with the observed depletion of the precursor over the range of plume ages sampled during WE-CAN. With this method, we estimate that typical western US wildfire plumes produce a net 9.8 ± 1.4 μg sm⁻³ ppm⁻¹ BBSOA from the near-field oxidation of vapors observed in typical daytime plumes. This accounts for 13 ± 3% of the total 75 ± 16 µg sm⁻³ ppm⁻¹ estimated BBSOA formation. In fractional terms, phenolic compounds were by far the largest measured source of BBSOA from directly emitted gases at 56%, followed by heterocyclic compounds at 18%, aromatic hydrocarbons at 17%, and biogenic compounds at 8% (Fig. 4). Our observations of a suite of phenolic compounds in fresh daytime wildfire plumes allow for this direct comparison against other common BBSOA precursors and suggest that the BBSOA formed may be light-absorbing and thus contribute to sBrC, as we discuss in Phenolic BBSOA and BrC Light Absorption.

The BBSOA source estimate of 9.8 ± 1.4 μg sm⁻³ ppm⁻¹ is conservative but still small compared to the measured range of 200 to 300 μg sm⁻³ ppm⁻¹ total BBPOA (17). There are of course significant uncertainties in this bottom-up BBSOA estimate, stemming both from potential low biases in laboratory determined SOA yields with associated vapor wall loss and also unmeasured or unquantified vapors in the wildfire plumes. Moreover, as noted above, we may underestimate the ER of the most reactive BBSOA precursors, especially the phenolic compounds that make up the largest portion of the estimate. Consensus suggests that SOA yields from chamber experiments can be biased low by a factor of 2 to 4 in some cases (20, 50), and the previous laboratory BB studies with more comprehensive speciation and quantification of BB emissions of VOCs and IVOCs suggest the categories of precursors we quantify here are the dominant BBSOA contributors. If we propagate the bottom-up estimate in the SI Appendix, Table S2 but multiply the SOA yields by a factor of 2, we would estimate 19.6 ± 2.8 μg sm⁻³ ppm⁻¹, or 26 ± 7% of the total. Therefore, even taking into account large potential low biases in SOA yields, we conclude that BBSOA formation from gas-phase oxidation of vapors emitted at the fire source plays a minor role in determining total BBOA concentrations and properties. Indeed, even if a 100% SOA yield were applied to all BBSOA precursors in SI Appendix, Table S2, including furans, which is unlikely, these precursors could account for only 62 ± 14% of total 75 ± 16 µg sm⁻³ ppm⁻¹ estimated BBSOA formation.

As this bottom-up estimate from direct gas-phase emissions only accounts for 9.8 ± 1.4 µg sm⁻³ ppm⁻¹ BBSOA out of the total top-down maximum estimate of 75 ± 16 µg sm⁻³ ppm⁻¹ in Fig. 1A, it follows that the remainder of that total (65.2 ± 16.1 µg sm⁻³ ppm⁻¹) is formed from subsequent oxidation of evaporated BBPOA with near unit SOA yield. Thus, evaporated BBPOA is shown here to be the dominant source of BBSOA in aged wildfire plumes, as illustrated schematically in Figs. 1A and 4. The specific identities of evaporated BBPOA compounds remains unclear, for several reasons. This subset of compounds will have relatively low O:C ratios, which is a class of compounds that tend not to cluster strongly enough with I⁻ anions to be sensitively detected in the I⁻ CIMS technique. These compounds are also likely IVOCs or SVOCs with relatively low volatility, which are not typically sampled well in short pulses with the PTR-ToF-MS technique due to inlet wall effects (51, 52). Lastly, and likely most significantly, the sum of evaporated BBPOA is likely spread over many compounds, each with relatively low concentrations maintained by a pseudo-steady state, where evaporation and oxidation with recondensation are approximately balanced, making detection and attribution difficult in a complex BB plume environment.

As noted above, oxidation of phenolic compounds in the fresh wildfire plumes explains a substantial portion of inferred BBSOA. The related nitrophenolic products, which are known to contribute to BrC Abs₄₀₅, were also simultaneously speciated in the gas phase (SI Appendix, Fig. S7) and detected in the particle phase as part of total OA in these plumes. Similar to the trends for OA as a function of dilution factor and extent of reaction, BrC Abs₄₀₅ exhibited both dilution-driven loss and secondary formation (Fig. 1C). To investigate the contributions of nitrophenols to BBSOA and BrC, either as direct emissions or stemming from their in-plume oxidation, we conducted a series of laboratory chamber experiments probing oxidation of phenolic compounds and the resulting SOA formation and aerosol light absorption during the MOONLIGHT campaign (see SI Appendix for more details). Briefly, this campaign probed the OH and NO₃ radical oxidation of phenol, catechol, cresol, and guaiacol under high-nitrogen oxide conditions to mimic wildfire plumes. These experiments allowed direct connections between gas-phase phenolic compounds, their oxidation products, and the corresponding particle composition and light-absorbing properties.

The major advance of this laboratory campaign was the connection of the magnitude of BrC Abs₄₀₅, measured using the PAS, with the concentrations of light-absorbing particulate nitrophenolic products. The nitrophenolic compounds were detected as part of total OA using an AMS, and they were detected as part of the speciated particulate molecular composition measurements made using a Filter Inlet for Gases and Aerosols (FIGAERO) coupled to the I⁻ CIMS (53). The I⁻ CIMS, AMS, and PAS were the same instruments used to generate observations shown in Figs. 1–3 during the WE-CAN aircraft campaign. BrC Abs₄₀₅ was directly measured during MOONLIGHT, while absorption from BC was subtracted from the 405-nm PAS signal during WE-CAN (SI Appendix, Sect. 1.1.1). The FIGAERO-CIMS provided quantitative molecular-level measurements of particulate nitrophenols, which allowed for calibration of real-time particulate nitrophenol parent tracer ions measured in the AMS and thus a direct connection to observations of particulate nitrophenols and BrC Abs₄₀₅ in the wildfire plumes studied during WE-CAN.

Fig. 5A shows the strong correlation between the sum of particulate nitrophenols (C₆H₅NO₃ and C₆H₅NO₄) measured using the FIGAERO-CIMS versus the corresponding tracer ions measured by the AMS during phenol and catechol oxidation experiments. This relationship shows that the AMS tracer ions can provide an estimate of the mass concentration of the sum of C₆H₅NO₃ and C₆H₅NO₄ in ambient OA. The two compounds were summed due to the fact that nitrocatechol fragments into both C₆H₅NO₃⁺ and C₆H₅NO₄⁺ ions in the AMS, and the particulate concentrations of C₆H₅NO₄ measured by the FIGAERO-CIMS were always much larger than C₆H₅NO₃, so it was not possible to independently determine their individual relationships between FIGAERO-CIMS and AMS measurements. The time series of FIGAERO-CIMS particulate nitrophenolic compounds, BrC Abs₄₀₅, and select AMS tracer ions are shown in the SI Appendix, Fig. S8 and indicate that this interpretation extends to other nitrophenolic compounds as well (SI Appendix, Fig. S9).

We found that BrC Abs₄₀₅ was correlated with the sum of particulate nitrophenolic compounds measured with the FIGAERO-CIMS (SI Appendix, Fig. S10) and with the corresponding tracer ions in the AMS (Fig. 5B). Nitrophenolic compounds were the dominant contributors to the initial SOA mass measured in the chamber experiments but decreased as a fraction over time with correlated changes in BrC Abs₄₀₅, and thus we assume they are responsible for the majority of measured light absorption. Under this assumption, the nitrophenolic compounds (predominantly nitrocatechol) have an effective absorption cross-section of 9.0× 10⁻¹⁸ cm² per molecule (SI Appendix, Fig. S10), which is similar to values expected for particulate nitrophenolics (54, 55).

The AMS tracer ions quantified during the MOONLIGHT chamber experiments were also quantified during WE-CAN wildfire plume measurements. Thus, we can apply the relationships of particulate nitrophenolic mass and BrC Abs₄₀₅ as a function of AMS tracer ions (Fig. 5 A and B) to the wildfire measurements to estimate particulate nitrophenolic mass and their contributions to BrC Abs₄₀₅ in the wildfire plumes. Shown in Fig. 6 is a summary of gas-phase phenolics, estimated particulate nitrophenolics, OA, and BrC Abs₄₀₅ observations from nine wildfires during WE-CAN with coherent plumes for which physical age could be estimated. Each of these plumes were sampled at a minimum of three different physical ages. The sum of phenolics exhibited a clear decay, as indicated in Fig. 3 and SI Appendix, Fig. S4 (SI Appendix). Both OA and BrC Abs₄₀₅ NEMRs remain roughly constant over the range of plume ages, although some variability within and across plumes is evident. As explained above in Insights into Dilution-Driven BBPOA Evaporation and BBSOA Formation, this view conceals the balance between dilution-driven evaporation and subsequent secondary formation. Using the MOONLIGHT campaign relationships with the empirical ±50% error bars shown in Fig. 5, we estimate that particulate nitrophenolic compounds, dominated by components with molecular compositions of nitrophenol and nitrocatechol, comprise an average of 9 ± 5 μg sm⁻³ ppm⁻¹ OA NEMR (with an SD of ±3 μg sm⁻³ ppm⁻¹ due to variability between different fires) and can explain an average of 36 ± 18 Mm⁻¹ ppm⁻¹ BrC Abs₄₀₅ NEMR (SD of ±13 Mm⁻¹ ppm⁻¹ between fires). These values correspond to an average of 4 ± 2% of OA and 29 ± 15% of BrC Abs₄₀₅ NEMRs (with SDs between fires of ±1 and ±12%, respectively). Thus, while accounting for only a small fraction of total OA, phenolic-derived OA makes an outsized contribution to BrC Abs₄₀₅, reaching as much as 40% or more in some wildfire plumes. This is consistent with previous literature illustrating the outsized role in BrC absorption of compounds with relatively small mass concentrations such as nitrophenolics (e.g., refs. 33, 56, 57). Given the depletion of particulate nitrophenols and BrC Abs₄₀₅ in the chamber experiments with extended oxidative aging, this molecular connection could therefore explain a significant fraction of the apparent decreases of BrC in aged BB plumes through the continued oxidation of nitrophenols into products with fewer or weaker chromophores. As shown in SI Appendix, Fig. S11, a modest correlation is found between the evolution of BrC Abs₄₀₅ NEMR and the evolution of the estimated contribution of nitrophenols to BrC Abs₄₀₅, although other subsets of BrC are likely contributing as well.

We present quantitative measurements of the evolution of BBSOA and sBrC Abs₄₀₅ in wildfire plumes, illustrating that the balance of dilution-driven BBPOA evaporation and subsequent BBSOA formation is not merely serendipitous but a result of evaporated BBPOA being the dominant source of BBSOA. We find that phenolic compounds and their oxidation products are substantial contributors to total BrC Abs₄₀₅ in western US wildfire plumes, despite less contribution to total OA mass. Total phenolic emissions were typically 1 to 3 ppt ppb⁻¹, although this may be an underestimate due to the fast reactivity of some. With a combined analysis of simulation chamber measurements and field data, nitrophenolic compounds were estimated to account for an average of 4 ± 2% of total OA mass but 29 ± 15% of total BrC Abs₄₀₅.

By estimating a dilution factor for each plume transect, we illustrated the competing effects of evaporation of primary OA and BrC versus secondary formation from phenolics and other gases. Our observationally constrained analysis indicates that the oxidation of evaporated BBPOA plays a dominant role in determining aerosol properties, with relatively minor contributions from the oxidation of directly emitted precursor gases in such large wildfires in the western United States. The estimate of BBSOA from emitted vapors is admittedly uncertain, but even after accounting for large potential biases, we find that evaporated BBPOA likely explains at least half of the BBSOA that forms in the first 3 to 6 h of wildfire plume evolution.

Essentially all of the wildfire plume data presented herein were taken during the afternoon hours while photochemistry was active. HONO photolysis was the likely driver of phenolic compound evolution (45), but this source of OH would be inactive during nighttime hours. These measurements leave a substantial gap in our understanding of the evolution of phenolic compounds during nighttime hours. Nitrate radical chemistry would likely play an important role in such plumes. Our MOONLIGHT chamber experiments suggest a strong connection between NO₃ oxidation of phenolic compounds and production of BrC Abs₄₀₅ consistent with past studies. Thus, future work, such as nighttime flights or nighttime ground-based stationary/mobile laboratory measurements, should focus on this question and on how the emission rates of phenolic compounds relative to CO may change as the fires tend to become less active during nighttime.

Descriptions of the WE-CAN and MOONLIGHT campaigns and further details of instrument operation, calibration, and analysis and methods of calculating NEMRs and the dilution factor can be found in SI Appendix, along with all SI figures and tables referenced throughout the article.

We sincerely thank everyone involved in the planning and operation of the 2018 WE-CAN campaign and 2019 MOONLIGHT campaign. B.B.P., Q.P., C.D.F., B.H.L., and J.A.T. were supported by US NSF Grant AGS-1652688 and National Oceanic and Atmospheric Administration (NOAA) Grant NA17OAR4310012). L.A.G., M.A.P., S.M.K., and D.K.F. were supported by NOAA Grant NA17OAR4310010. R.P.P., Y.S., and S.M.M. acknowledge support from US Environmental Protection Agency Grant R835883. This research was supported in part by US NSF Grants AGS-1650786 and AGS-1650275. This material is based upon work supported by the NCAR, which is a major facility sponsored by the US NSF under Cooperative Agreement 1852977. The data were collected using NSF’s Lower Atmosphere Observing Facilities, which are managed and operated by NCAR’s Earth Observing Laboratory. The operational and scientific support from NCAR’s Earth Observing Laboratory and Research Aircraft Facility is gratefully acknowledged.