We use the GEOS-Chem (Bey et al., 2001) chemical transport model version 12.7.1 to examine the effects of aerosols on O₃–NO_x–VOCs chemistry. We run nested simulations over California regions (27–47° N, 110–130° W) with a resolution of 0.25° (latitude) × 0.3125° (longitude) and 47 vertical levels for the year 2020. The model is driven by the Goddard Earth Observation System Forward Processing product (GEOS-FP) assimilated meteorological field, with a 3 h temporal resolution for three-dimensional variables and 1 h resolution for surface variables. Boundary conditions for the simulations are generated from a global simulation at a resolution of 2° × 2.5° with a 1-year spin-up. The standard tropospheric chemical scheme includes detailed O₃–NO_x–VOCs–aerosol–halogen chemistry. Additionally we have incorporated the ethene and ethyne chemistry as introduced in GEOS-Chem version 13.3.0 (Kwon et al., 2021). Hourly anthropogenic emissions in the US are based on the EPA 2011 National Emission Inventory (NEI) and are scaled to 2020 using national interannual emission trends (US EPA, 2025). Fire emissions are sourced from the Global Fire Emissions Database (GFED, Version 4.1), with emissions categorized by fuel types, including tropical forest, temperate forest, boreal forest, savanna, peat and agricultural waste (Randerson et al., 2015). We allocate 65 % of these fire emissions within the boundary layer (Fischer et al., 2014), so our findings primarily reflect fires that predominantly impact the boundary layer.

Photolysis rates in GEOS-Chem are calculated using the fast-JX scheme (Bian and Prather, 2002). The influence of aerosols on the photolysis rates are considered (Martin et al., 2003), with the adjustments for aerosol size distribution and optical properties in response to relative humidity changes. GEOS-Chem treats black carbon (BC) as externally mixed, making it challenging to explicitly simulate the lensing effect, where BC exhibits larger absorption when coated by a non-absorbing shell. To incorporate this effect, we apply an absorption enhancement factor (the ratio of mass absorption efficiency (MAE) with and without coating) of 1.5 to hydrophilic BC and 1 for hydrophobic BC (Wang et al., 2014).

The heterogeneous uptake of HO₂ is represented by a reaction probability parameterization as shown in Eq. (1), with the loss rate limited by diffusion or free molecular collision (Martin et al., 2003). 1k=aDg+4νγ-1A The first-order rate constant k for the chemical loss of the gas (i.e., HO₂) is calculated based on the mean molecular speed (ν), gas-phase molecular diffusion coefficient (Dg), aerosol radius (a), reaction probability upon impacting the aerosol surface (γ) and aerosol surface area per unit volume of air (A). Consistent with numerous modeling studies (Ivatt et al., 2022; Jacob, 2000; Li et al., 2019; Martin et al., 2003), we adopt a uniform value of 0.2 for γHO2, aligning with the field measurements (Taketani et al., 2012; Zhou et al., 2020, 2021). GEOS-Chem assumes the same γHO2 for all aerosol types, including organic carbon (OC), BC, sulfate-ammonium-nitrate, sea salt separated in two size bins and mineral dust in seven size bins.

To examine the aerosol effects on O₃, we conduct one BASE simulation and five perturbation simulations in GEOS-Chem, as summarized in Table 1. The difference between BASE and BASE_NO_RAD is considered as the radiative effect of all aerosols, and the difference between NO_FIRE and NO_FIRE_NO_RAD represents the radiative effect of aerosols other than fire smoke aerosols. The radiative effect of fire smoke aerosols is therefore calculated as BASE – BASE_NO_RAD – (NO_FIRE – NO_FIRE_NO_RAD). Similarly, the chemical effect of smoke aerosols is calculated as BASE – BASE_NO_CHEM – (NO_FIRE – NO_FIRE_NO_CHEM). Hourly species concentrations, meteorology, photolysis rates and reaction rates for the bottom five layers of the model (approximately 0–550 m) are averaged to investigate aerosol effects on near-surface O₃ and perform regime calculations.

GEOS-Chem's Eulerian framework does not explicitly resolve individual plume pathways or their detailed evolution. We identify about 1633 fire plumes in 2020 that show clear plume patterns with an identifiable plume source and use the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) dispersion model to calculate plume trajectories and plume age. The plume identification method is described in the work of Jin et al. (2023). Fire centers are identified using the Moderate Resolution Imaging Spectroradiometer (MODIS) Active Fire products and subsequently used as starting points for calculating 1 d plume dispersion using the HYSPLIT model with meteorological fields from North American Regional Reanalysis (NARR). The HYSPLIT model is run at an injection height of 1000 m and initialized at the same time of the day (18:00 UTC). In the absence of strong wind variability, the predicted plume trajectories should reasonably represent the progression from the near to far field of fires. The locations of the fire plumes are matched to GEOS-Chem grids to demonstrate changes in aerosol effects along the plumes. In this study, we define plume age as the physical age of the plume, determined as the time required for the plume to reach designated smoke-affected areas. We did not explicitly isolate fire plumes from urban influence in order to examine aerosol effects across a range of background NO_x levels.

We employ F0AM (Wolfe et al., 2016) version 4.3 to assess the effectiveness of GEOS-Chem in resolving the aerosol effects on O₃ within fire plumes. We use the Master Chemical Mechanism (MCM) version 3.3.1 (Jenkin et al., 2015), which features a near-explicit chemical mechanism with detailed gas-phase chemical processes. Additionally, we incorporate the heterogeneous uptake of HO₂ by aerosols as described in Eq. (1) and assume a monodisperse size distribution for each aerosol type.

We first evaluate whether the aerosol effects resolved in GEOS-Chem are reproducible in F0AM by initializing F0AM with output from GEOS-Chem. The fire plumes are modeled with a pseudo-Lagrangian style in F0AM, where we set the initial chemical concentrations based on GEOS-Chem grids with a plume age of 1 h and allow them to evolve over the subsequent 5 h. Species used to initiate F0AM include CO, O₃, reactive nitrogen species and some VOCs (Table S2 in the Supplement). Meteorological variables and photolysis-relevant parameters are constrained at each model step and held constant during the integration time of 1 h. We adopt the F0AM's hybrid method for J values calculations, which uses Tropospheric Ultraviolet and Visible (TUV)-calculated solar spectra but does not include explicit aerosol effects. J values of HONO and HCHO from GEOS-Chem are applied to scale box model-calculated J values. CO is an approximately conservative tracer (Robinson et al., 2021); we calculate the first-order dilution rate in F0AM at each model step using the temporal changes in CO concentrations along the fire plumes (Müller et al., 2016), as determined by GEOS-Chem. Configuration details of the F0AM setup are provided in Table S2. Chemical species, meteorological and photolysis variables from GEOS-Chem are matched to those in the MCM. To exhibit the aerosol effects on O₃, we run one base simulation and two perturbation simulations in F0AM: one eliminating the chemical effect and another removing the radiative impact of fire-related aerosols.

We further assess whether the resolution of GEOS-Chem can resolve the in-plume O₃ chemistry by focusing on fresh plumes in F0AM. Unlike the previous setup using GEOS-Chem outputs, here we initiate F0AM with gas-phase pollutants and aerosols (primarily OC and BC) for various fire types according to the GFED emission factors. We adopt aerosol effective radii of 0.035 µm for BC and 0.1 µm for OC, values that closely match GEOS-Chem averages over California in 2020 at 13:30 local time (LT), and assume a particle density of 1.3 g cm⁻³. We convert the emission factors (grams of species per kilogram dry matter burned) to concentrations (parts per billion for gases and µg m⁻³ for aerosols) using a fixed ratio of biomass burned per cubic meter of air. We then scale all pollutants to achieve aerosol concentrations ranging from 1 to 300 µg m⁻³ at the time of emission, allowing us to explore how aerosol effects vary with fire intensity. In this approach, we set only the initial chemical and physical parameters and run the model for 1 h, focusing specifically on the characteristics of fresh plumes. Photolysis rates, which we cannot directly constrain in scenarios with and without fires, are estimated based on the relationship between photolysis rate reduction and PM_2.5 mass as derived from GEOS-Chem (Fig. S2). To prevent the buildup of secondary species, we set a 1 d lifetime for all species by applying a first-order dilution rate of 1/86400 s⁻¹ and background concentrations at zero. Aerosol effects are calculated following the same method as in the F0AM–GEOS-Chem comparison.

We use daily ground-based measurements of O₃ and PM_2.5 from the EPA Air Quality System (AQS) (EPA AQS, 2020) to evaluate the GEOS-Chem simulations. In addition, we analyze the decay of PM_2.5 and NO₂ within fire plumes using observationally derived datasets. Surface PM_2.5 data are from Wei et al. (2023), featuring a daily, 1 km resolution, gapless PM_2.5 dataset spanning 2017–2022. This dataset was generated using a 4-Dimensional Space-Time Extra-Trees (4D-STET) model, which reconstructs missing satellite AOD, establishes AOD-PM_2.5 relationships and predicts high-resolution surface PM_2.5 concentrations. This observation-based 1 km product improves upon earlier 10 km datasets, providing finer spatial detail for plume analysis. Tropospheric NO₂ column data are sourced from TROPOspheric Monitoring Instrument (TROPOMI) retrievals provided by Jin et al. (2023), which incorporate a priori profiles from GEOS-Chem simulations and explicitly account for smoke aerosols during retrieval. Both the surface PM_2.5 and tropospheric NO₂ column data are also used to identify O₃ regimes from observations (see Sect. 3.5).

We determine the photochemical regimes by assessing the chain termination rates of HO_x radicals, similar to the method described in Ivatt et al. (2022). The radical termination pathways include (1) loss via NO_x as indicated by the reactions NO₂ + OH → HNO₃ and RO₂ + NO → alkyl nitrate (RONO₂), (2) HO_x self-reactions and (3) heterogenous uptake of HO₂ by aerosols. A predominance of NO_x as the sink for HO_x characterizes a NO_x-saturated regime. Dominance by HO_x self-reactions indicates a NO_x-limited regime. When the rate of HO₂ uptake to aerosol dominates, it indicates a heterogeneous chemistry-inhibited regime. The radiative effect of aerosols, however, has not been considered in the regime calculations. To address this issue, we account for the aerosol radiative effect on O₃ production by using the difference in total HO_x termination rates between BASE and BASE_NO_RAD simulations (ΔRHOx) as a proxy. Notably, ΔRHOx is not an actual chemical pathway; instead, it serves as an indicator of light availability and its influence on the photochemical activities. If ΔRHOx exceeds any of the aforementioned three pathways, it suggests a light-limited regime. We use the reaction rate output from GEOS-Chem to calculate the chain termination rates and ΔRHOx in each grid box at 20:30 UTC (around 13:30 LT) and identify the corresponding regime based on the maximum term. We focus on 13:30 LT because it coincides with a period of strong solar radiation that drives ozone photochemistry and aligns with typical satellite overpass time, facilitating integration of satellite-based observations to identify chemical regimes. Monthly mean regimes are determined by averaging the magnitudes of four terms rather than counting the occurrences of each regime to reflect the cumulative influence of these processes over time.

We further investigate how PM_2.5 levels influence O₃ photochemical regimes using GEOS-Chem. Specifically, we identify all fire-affected grid cells (those with PM_2.5 enhancement larger than 10 µg m⁻³) at 20:30 UTC during 2020. For these grid cells, we calculate the HO_x termination rates, determine the corresponding O₃ regimes, and then group the regimes by PM_2.5 concentrations to derive the probability of each regime at various PM_2.5 levels.

We first evaluate GEOS-Chem-predicted O₃ with daily ground measurements from the EPA AQS, as presented in Fig. S1. The comparison is conducted between AQS sites and the corresponding GEOS-Chem grid cells for the year 2020 around 13:00 LT. The modeled average O₃ levels in California for 2020 are approximately 48 ± 4 ppb, in good agreement with ground observations of 44 ± 9 ppb (R2 of 0.64).

Next, we assess the aerosol effects and the overall impact of fires on O₃ in GEOS-Chem in both the near and far field of fires (Fig. 2). Fire pixels are categorized based on PM enhancement (ΔPM_2.5), calculated as the difference in PM_2.5 mass between the BASE and NO_FIRE simulations for each individual grid cell. Specifically, ΔPM_2.5 values of < 50, 50–100, 100–200 and > 200 µg m⁻³ are used to classify small, medium, large and extreme fire pixels, respectively. It reveals that for fire pixels with small to large PM enhancements, which represent the majority of fires, fires increase O₃ concentrations in both near and far fields, indicating the influence of fires through the emissions of substantial quantities of O₃ precursors outweighs the aerosol effects. Generally, fire pixels with larger PM enhancement are associated with larger increase in O₃ concentrations. In contrast, pixels affected by extreme fires see suppressed O₃ levels in their immediate vicinity, suggesting that the aerosol effect overshadows the emission effect. Furthermore, this O₃ suppression is likely driven by the strong aerosol radiative effect associated with dense plumes near the centers of fires. In the near field of the fires, the average radiative impact on O₃ concentrations for extreme fire pixels is about 60 times that observed in the others. Other factors contributing to the decreased O₃ concentrations may be NO_x titration or sequestration of NO_x into peroxyacetyl nitrate (PAN) in the near field of fires (Jaffe and Wigder, 2012). For extreme fires, O₃ suppression by aerosols is stronger in the near field and weakens downwind, leading to a net increase in O₃ concentrations in the far field (Fig. 2).

Both aerosol chemical and radiative effects are shown to decrease O₃ in the fire plumes. For grid cells affected by small to large fires, the aerosol chemical effect outweighs the radiative effect. Contrary to the consistent behaviors observed in both the near-field and far-field regions for these fire pixels, those experiencing extreme PM enhancement exhibit pronounced variations. In the proximal areas of fire origins, the radiative effect on O₃ concentrations is much higher than the heterogeneous chemical effect for these extreme fire pixels. Yet, this radiative effect represents a temporary suppression of O₃ production, with its influence decaying rapidly – on average, the effect on O₃ concentrations diminishes by about half within 5 h (Fig. S3). Moving further from the fire centers, the chemical effect starts to dominate over the radiative effect on O₃. The aerosol impacts on O₃ concentrations, through both chemical and radiative pathways, tend to intensify as ΔPM_2.5 increases. The aerosol effects on O₃ concentrations mirror those on O₃ net production (Fig. S4). However, a notable difference exists when comparing large and extreme fire pixels: while their chemical effects on O₃ production are similar in the near field (Fig. S4), extreme fires exert a stronger suppression on O₃ concentrations (Fig. 2). This discrepancy likely stems from differences in transport and mixing. In the near field of extreme fires, HO_x levels are low due to limited photochemical activity, making HO₂ uptake less influential on O₃ production. Nevertheless, extreme fires may cause greater suppression of O₃ concentrations near the source. As O₃ is transported downwind, this initial suppression can lead to a greater reduction in O₃ concentrations despite similar local chemical production. Additionally, extreme fires may experience slower mixing with background air, reducing dilution of ozone-suppressed air and further enhancing the decrease in O₃ concentrations. Overall, aerosol effects resolved in GEOS-Chem highlight the significant heterogeneous chemical influence on O₃ for fires and an exceptionally critical radiative effect for extreme fires.

We first use F0AM to conduct similar experiments with GEOS-Chem output for fire plumes of different scales. We select 12 fire plumes spanning small, medium, large and extreme cases, and comparisons for each individual fire plume are shown in Fig. S5. We find that the overall fire impacts on O₃ concentrations and the aerosol chemical and radiative effects simulated in F0AM exhibit good agreement with those resolved in GEOS-Chem across fire plumes of different scales. Although F0AM does not explicitly account for atmospheric processes such as vertical mixing, turbulent diffusion, and dry and wet deposition, these factors appear to have a negligible impact (beyond their representation as dilution) on the several-hour timescale examined here. The comparison suggests that chemistry and to a lesser extent dilution are the leading factors explaining most variations in aerosol effects. It should be noted that although furanoid compounds markedly influence biomass-burning plume chemistry under both daytime and nighttime conditions (Decker et al., 2019; Xu et al., 2021), their reactions are not represented in either the GEOS-Chem version or the MCM mechanism used in this study.

Our results indicate relatively consistent aerosol effects resolved by different numerical simulation schemes. GEOS-Chem is a global Eulerian model, which solves continuity equations on a geographically fixed frame of reference (Liu et al., 2023; Long et al., 2015), whereas in F0AM plumes are simulated in a pseudo-Lagrangian approach that follows the movement of air parcels. However, the Eulerian model struggles with an unrealistic dilution of small plumes. In our comparison, the initial chemical concentrations used in F0AM are adopted from GEOS-Chem where dilution of initial subgrid plumes has occurred. Consequently, although both GEOS-Chem and F0AM exhibit comparable results, the near-field behavior of subgrid plumes may not be accurately solved by either model.

Next, instead of initiating F0AM using GEOS-Chem simulations, we explore the aerosol influence on O₃ in fresh plumes by initiating F0AM with emission data from GFED. Our analysis reveals that the aerosol influence on O₃ depends on PM mass concentrations (Fig. 3), which is consistent with findings from GEOS-Chem (Fig. 2). Furthermore, at the same PM enhancement, the influence of aerosol chemical and radiative pathways on O₃ concentrations appears to vary distinctly among various fuel types, suggesting underlying factors beyond PM concentrations play a role in controlling aerosol influence. PM enhancement thresholds where the radiative effect outweighs the chemical effect vary by fuel type, being highest for boreal forest fires, followed by peat and temperate forest, and lowest in deforested/tropical forest, agricultural waste and savanna. In the case of temperate forest fires, even small plumes could exhibit a more pronounced aerosol radiative effect than the chemical effect in the near field.

As we control the PM magnitude, the various patterns across fuel types are due to variations in emission factors of O₃ precursors, particularly NO_x. According to GFED, emissions from the boreal forest fires exhibit the highest PM-to-NO_x ratio, followed by those from peat, temperate forest and tropical forest fires. The lowest ratios are observed in agricultural waste and savanna burning. These results highlight that the aerosol influence on O₃ is not only dependent on the abundance of PM but also modulated by NO_x concentrations. Higher NO_x levels can suppress the chemical effect of aerosols by altering HO_x loss pathways; under high-NO_x conditions, more HO_x is consumed by reactions with NO_x, leaving less HO_x for heterogeneous uptake by aerosols. On the other hand, larger PM concentrations enhance HO_x loss through aerosol uptake. The interplay between these two factors largely accounts for the variations in aerosol impacts on O₃ within fire plumes.

Studies generally report dilution rates in fire plumes on the order of 10⁻⁵–10⁻⁴ s⁻¹ (Decker et al., 2021; Peng et al., 2021; Rickly et al., 2022), with some studies observing rates as high as 10⁻³ s⁻¹ (Robinson et al., 2021). In Fig. 3, we adopt a dilution rate of approximately 10⁻⁵ s⁻¹ and we further assess its impact by increasing this rate by factors of 10 and 100 in our F0AM simulations. Under the 10 × scenario (Fig. S6a), the overall fire effects and aerosol effects on O₃ remain comparable, with similar thresholds at which the radiative effect exceeds the chemical effect. In the 100 × scenario (Fig. S6b), however, these effects diminish substantially. This is likely because an e-folding timescale of 17 min leaves limited time for ozone production before ozone precursors and aerosols are diluted, thereby weakening the influences of fire emissions. Nevertheless, the PM enhancement threshold at which the radiative effect exceeds the chemical effect still decreases following the order boreal forest, peat, temperate forest, tropical forest, agricultural waste and savanna (from > 300 µg m⁻³ down to about 20 µg m⁻³). The sensitivity test supports our findings that both PM and NO_x are key factors controlling aerosol effects on O₃.

The dependence of aerosol effects on NO_x is also evident in GEOS-Chem. Figure S7 suggests that the radiative effect tends to surpass the chemical effect at high NO_x levels. However, GEOS-Chem also indicates that the aerosol chemical effect consistently dominates the radiative effect for regular fires, a phenomenon not reproduced in F0AM. This discrepancy may arise because GEOS-Chem does not accurately resolve the aerosol effects on O₃ for the subgrid-scale young plumes. But for plumes that are not in the immediate vicinity of the fire source, where mixing with background air has occurred, or in the case of large-scale fires that exceed the size of a grid cell, GEOS-Chem should be capable of resolving the aerosol impacts. Additionally, for the range of PM enhancement examined here (within 300 µg m⁻³), F0AM suggests that fire generally enhances O₃ concentrations, aligning with our findings from GEOS-Chem.

Observations of PM_2.5 and NO₂ within fire plumes reveal that NO₂ columns decay more rapidly than PM_2.5 (Fig. 4). An even steeper decline is expected for surface NO₂, as surface measurements are more sensitive to local emission sources compared to large-scale satellite observations (Lamsal et al., 2014). This observational finding implies that as plumes age, the aerosol heterogeneous chemical effect becomes increasingly important, as reflected by the higher PM-to-NO₂ ratio in the far field compared to near sources. This also accounts for why, in GEOS-Chem simulations, the chemical effect tends to outweigh the radiative effect away from fire origins. By integrating GEOS-Chem and box model with observational constraints, our study provides a detailed and comprehensive depiction of aerosol effects within fire plumes and the potential underlying mechanisms.

To summarize, for most fires, there is generally a net positive effect on O₃ concentrations. Near the source, heterogeneous chemical or radiative effects may outweigh each other depending on NO_x levels. As the plumes age, NO_x is rapidly consumed in the plumes, and the aerosol chemical effect tends to be increasingly important. In contrast, extremely large fires are dominated by the aerosol radiative effect, leading to an overall suppression of O₃ in the near field that can extend further from the fire sources. Even for these fires, the radiative effect diminishes rapidly with dilution and is eventually surpassed by the chemical effect downwind.

The importance of aerosol effects on O₃, especially the heterogeneous chemical effect, has been the subject of significant debate. Xu et al. (2021) found that the conceptual model based on gas-phase chemistry adequately explains the O₃ chemistry in western US wildfire plumes (R2 of 0.64) and thus aerosol heterogeneous chemical processes are likely minor. Conversely, Li et al. (2019) and Ivatt et al. (2022) highlighted a significant role of the heterogeneous chemical effect on the near-surface O₃ formation in eastern China and the Indo-Gangetic Plain during the mid-2010s. Even among studies that supported the importance of the aerosol chemical effect, some emphasized its significance in environments with high aerosol loadings, while others pointed to its relevance in clean suburban areas (Li et al., 2022; Xue et al., 2014).

Our findings reconcile seemingly contradictory studies by showing that the aerosol effects on O₃ are determined by both aerosol loading and NO_x concentrations. The study by Xu et al. (2021) focused on relatively fresh plumes, which are usually associated with high NO_x concentrations, where the inhibitive effects of smoke aerosols may be secondary. However, as plumes age and both NO_x and PM concentrations decay, the longer-lived accumulation mode aerosols (lifetime of 5–7 d, compared to hours to a day for NO_x) (Jin et al., 2021; Seinfeld and Pandis, 2016) can become more influential in O₃ production. The shift in the relative importance of aerosols vs. NO_x may differ in urban/suburban settings, where PM and NO_x can originate from different sources and possibly lead to more varied concentration patterns. O₃ production can be significantly impacted by heterogeneous chemistry in conditions ranging from heavily polluted areas with high aerosol loadings to cleaner areas with moderate aerosol loadings but low NO_x.

Our findings emphasize that both the heterogeneous chemical and radiative effects can significantly influence O₃ production depending on fire conditions. Driven by these insights, we propose a novel O₃ production regime, termed the “light-limited regime”, which is identified through a sensitivity test in which the radiative effect is turned off and the resulting reduction in HO_x availability outweighs any of the three termination pathways. Figure S8 illustrates the O₃ production regime over California from July to December under a no-biomass-burning scenario. In the absence of fire impacts, most of the areas are in NO_x-limited regimes during the summertime, with a NO_x-saturated regime in the urban cores of Los Angeles and San Francisco. During the cooler months, a large number of regions shift to a VOC-limited regime.

Accounting for the impacts of fires on O₃ reveals significant changes in the O₃ production regimes during the fire season, as shown in Fig. 5 (and Fig. S9). Details about significant fire events and emissions during the 2020 wildfire season in California are summarized in Text S1. It is evident that numerous areas transition to either the heterogeneous chemistry-inhibited regime or the light-limited regime, which we collectively term “aerosol-dominated regimes”.

From August to October, the monthly mean proportions of grid boxes in California entering the aerosol-dominated regimes were 8.9 %, 75 % and 43 %, respectively (Fig. 5). Specifically, 8 %, 60 % and 41 % corresponded to the heterogeneous chemistry-inhibited regime, and 0.9 %, 15 % and 1.7 % were classified as the light-limited regime. The impact of fires on these regimes was minimal for November, when most wildfires were contained. Furthermore, the episodic nature of wildfires caused large daily variations in the O₃ production regime; the heterogenous chemistry-inhibited regime had an average ± standard deviation of 19 ± 13 %, 48 ± 16 % and 33 ± 24 % for the periods of 16–31 August, September and October, respectively. Similarly, the light-limited regime showed 1.6 ± 1.4 %, 13 ± 9.6 % and 3.2 ± 5.9 % for the same periods.

The newly defined light-limited regime extensively reflects the central areas of megafires (Fig. 1). The August Complex, SCU Lightning Complex, Creek, LNU Lightning Complex and North Complex were ranked as the top five fires by burned areas in 2020 (CAL FIRE, 2020b). Notably, during September, the August Complex, Creek and North Complex fires peaked, leading to extensive areas falling under the light-limited regime due to these large-scale wildfires, with the peripheral zones exhibiting heterogeneous chemistry-inhibited regime (Fig. 5). The period from 8 to 10 September, during the fire season, experienced the most extensive coverage of the light-limited regime across the state (32 %–42 %, Fig. S10), coinciding with significant wildfire events. Notably, despite the exceptionally large scale of the SCU and LNU Lightning Complex fires, their impacts on the light-limited regime were much less pronounced compared to the other three fires based on both daily and monthly average. A NO_x-saturated regime was predominant under the impact of these two wildfires. We attribute the difference in regimes to the distinct environments where fires occurred. Contrary to the fires in forest areas, the SCU and LNU fires occurred in the Bay area, an urban region characterized by significant higher background levels of NO_x. Elevated NO_x concentrations enhance HO_x termination through reactions with NO_x, necessitating higher aerosol concentrations for the light effect term to surpass this termination pathway. The observed reduction in the light-limited regime under high NO_x conditions further corroborates our earlier findings on the interactions among aerosol effects, PM and NO_x.

The aerosol effects and regime calculations derived from modeling are subject to uncertainties, primarily associated with the HO₂ uptake coefficient (γHO2) and fire emission inventory. Due to the challenges of directly observing or constraining the aerosol heterogeneous uptake through measurements, we rely on model simulations to estimate the chemical effect. Consequently, the results are influenced by the γHO2values used in the analysis, a parameter that varies with aerosol types and relative humidity. A summary of γHO2 reported in previous laboratory measurements and field studies is provided in Table S1. Organics constitute a major fraction of biomass-burning aerosols. Laboratory studies measuring the uptake coefficient from single-component organics have reported values of 0.007–0.09 for humic acid (Lakey et al., 2015), < 0.01–0.13 for levoglucosan (Taketani et al., 2010) and 0.02–0.18 for dicarboxylic acids (Taketani et al., 2013), across a variety of relative humidity levels. In comparison, field studies generally report higher values (0.08–0.40) (Taketani et al., 2012; Zhou et al., 2020), likely due to the presence of copper and iron ions in the particles that are known to enhance HO₂ uptake (Mao et al., 2013). To our knowledge, no studies have specifically measured γHO2 for biomass-burning aerosols in field settings, but Taketani et al. (2012) reported values of 0.2–0.37 for samples strongly affected by biomass burning. To assess the impact of γHO2 on our results, we conduct sensitivity tests using γHO2 of 0.1 and 0.02 for a 1-month simulation during September. Under the γHO2= 0.1 scenario, aerosol effects across fire sizes are similar to Fig. 2: the aerosol chemical effect outweighs the radiative effect for small to large fire pixels, while extreme fire pixels show a pronounced radiative effect (Fig. S11a, b). Although the overall fire effect reduces O₃ net production rate, its influence on O₃ concentrations is minimal. The spatial pattern of photochemical regimes remains largely unchanged under this scenario (Fig. S12a).

Given that γHO2 measured for single-component organics likely underestimates values for ambient aerosols, the γHO2=0.02 case is tested as a conservative lower bond. Under this assumption, aerosol chemical and radiative effects on O₃ concentrations become comparable for most fire pixels, whereas extreme fire pixels continue to exhibit a pronounced radiative effect (Fig. S11c, d). Although this strong radiative effect suppresses O₃ production in near-field extreme fire pixels, O₃ concentrations still increase. With this substantially reduced uptake coefficient, the spatial extent of heterogeneous chemistry-inhibited regimes decreases markedly. Nevertheless, overall aerosol influences remain important, with 31 % of California falling into aerosol-dominated regimes (Fig. S12b). Future research measuring γHO2for smoke aerosols is needed to better constrain this parameter.

Furthermore, we evaluate GEOS-Chem simulations of PM_2.5 with ground-based measurements from EPA's AQS. We find that GEOS-Chem tends to overestimate PM_2.5, simulating 2020 daily average PM_2.5 levels at 24 ± 23 µg m⁻³, compared to 12 ± 5.5 µg m⁻³ from ground-based observations. During the fire season, modeled PM_2.5 concentrations are about 1.2, 4.1 and 2.4 times higher than the ground observations in August, September and October, respectively. Outside the peak fire months, the agreement improves, with modeled PM_2.5 concentrations being 0.6, 1.4 and 0.9 times the observed values in July, November and December, respectively. The overestimates of PM_2.5 is likely driven by overestimated fire emissions in GFED (Qiu et al., 2024). These comparisons, however, are limited by factors such as the sparse ground observations (∼ 72 sites for PM_2.5), the potential unrepresentativeness of a single site for the coarse grid in GEOS-Chem and the GEOS-Chem modeled decay of PM further from the fires. To assess the potential impacts of model overestimates on our analysis, we perform additional simulations by scaling monthly biomass-burning emissions based on the model–observation comparisons. Specifically, GFED fire emissions are adjusted by dividing total emissions by 0.6, 1.2, 4.1, 2.4, 1.4 and 0.9 for July through December, respectively. Despite the substantial reduction in overall fire emissions, the aerosol and total fire effects on O₃ concentrations for most fires remain consistent with Fig. 2, whereas the radiative effect for extreme fires declines markedly in both the near and far field due to reduced aerosol loading (Fig. S11e, f). Aerosol-dominated regimes still accounted for about 7 %, 54 % and 17 % of the total area in August, September and October, respectively (Fig. S13). Notably, aerosol-dominated regimes remain dominant in September during the 2020 fire season.

Recognizing that the regime classification discussed above may be affected by model inputs and performance, we further explore how these model-based findings can be applied to observational data, with a primary focus on identifying aerosol-dominated regimes. We first investigate whether PM_2.5 as an indicator of aerosol concentrations can be used to identify the regime shift. Figure 6a shows the average fractional contribution of each HO_x termination pathway at various PM_2.5 levels. As PM levels increase, HO_x loss via self-reaction declines, while aerosol heterogeneous uptake and photolysis reduction effects become increasingly dominant. Figure 6b exhibits the probability of each regime at various PM_2.5 levels. Low PM_2.5 levels are usually associated with a NO_x-limited regime. The heterogeneous chemistry-inhibited regime is more likely to occur as PM_2.5 levels increase until the light-limited regime overshadows it at extremely high PM_2.5 concentrations. At a PM_2.5 concentration of 30 µg m⁻³, O₃ production already transitions to the heterogeneous chemistry-inhibited regime in most areas under the impact of fires. A considerably higher PM_2.5 concentration (∼ 500 µg m⁻³) is required to enter the light-limited regime. We observe a similar pattern of HO_x losses and regime shifts when reducing the γHO2 value to 0.1, as shown in Fig. S14. In this calculation, the PM_2.5 threshold for shifting to a heterogeneous chemistry-inhibited regime increases slightly from 30 to 40 µg m⁻³.

It is important to note that classifying a regime as “heterogeneous chemistry-inhibited” or “light-limited” does not necessarily imply a net suppression of O₃. The regime classification approach based on HO_x termination rate does not directly compare with the aerosol and emission effects quantified in Sect. 3.1 and 3.2. For example, a “heterogeneous chemistry-inhibited” regime indicates that HO₂ uptake is the largest sink of HO_x, but this does not imply that the combined aerosol chemical and radiative effects outweigh the influence of VOC and NO_x emissions. As shown in Fig. 2, large fires with PM_2.5 enhancement of 200 µg m⁻³ still exhibit net O₃ increases despite strong heterogeneous chemical effects.

Further investigations uncover that the PM_2.5 threshold required for most grid boxes to transition to a heterogeneous chemistry-inhibited regime is highly dependent on NO_x concentrations (Fig. 7a). Here we categorize NO_x concentrations into four classes: 0–1, 1–10, 10–25 and > 25 ppb, and the PM_2.5 thresholds likely to induce aerosol-dominated regimes are approximately 18, 25, 185 and 320 µg m⁻³, correspondingly. We primarily focus on fire plumes in this study, but grid boxes not affected by fires appear to exhibit similar trends in the probability of aerosol-dominated regimes. These results support our earlier findings that in scenarios with high NO_x concentrations, more PM is needed to attain a comparable level of aerosol contribution as observed in low NO_x scenarios. As NO_x concentrations increase, HO_x levels tend to decrease (Fig. 7b), which necessitates higher PM levels for aerosol effects to surpass the emission effects. Since PM_2.5 and NO₂ can be derived from ground-based or satellite observations, we explore how their ratio can be used to imply aerosol-dominated regimes. While a surface PM_2.5 / NO₂ ratio may seem more straightforward based on our analysis, the limited spatial coverage of surface NO₂ measurements poses a challenge. Tropospheric NO₂ column data, which are closely related to surface sources and have been widely used in O₃ sensitivity analyses (Martin et al., 2004), offer a practical alternative. When combined with high resolution and gapless surface PM_2.5 estimates derived from the integration of observations and machine learning, the PM_2.5 / NO₂ column ratio serves as a proxy to constrain aerosol effects on near-surface O₃ production. Indeed, we find a clear relationship between this ratio and the likelihood of aerosol-dominated regimes (Fig. 7c). When the ratio (PM_2.5 / NO₂ column) reaches about 20 µg m⁻³ (10¹⁵ molec. cm⁻²)⁻¹, the aerosol-dominated regimes are likely to prevail and will consistently be dominant at higher ratios.

We therefore adopt a threshold of 20 µg m⁻³ (10¹⁵ molec. cm⁻²)⁻¹ for PM_2.5 / NO₂ to identify aerosol-dominated regimes and apply it to satellite-derived surface PM_2.5 and tropospheric NO₂ column data. The resulting aerosol-dominated regimes are highlighted in red in Fig. 8. Overall, the observation-based aerosol-dominated regimes in Fig. 8 align well with the model-based classification in Fig. 5. Both methods reveal similar spatial distributions: aerosol-dominated regimes were widespread in September, peaked in the northern and central region in October, but exhibited relatively larger discrepancies in August. Spatially, the observation-based method estimates that approximately 20 %, 47 % and 16 % of the state fell within aerosol-dominated regimes from August to October; these values are larger than the model estimates for August but smaller for September. Despite these differences, the general agreement between the two regime classifications underscores the significant role of aerosols in surface O₃ photochemistry under wildfire conditions. This analysis also highlights the utility of satellite-derived PM_2.5-to-NO₂ ratio for pinpointing aerosol-dominated regimes. Applying this metric to other regions and environments may need further investigation. While the comparison of fire and urban plumes is beyond the scope of this study, it is worth noting that fire and urban plumes may differ substantially in emissions and aerosol composition and thus O₃ chemistry. Future research is therefore warranted to incorporate more sophisticated representations of these differences. Additionally, it may be valuable to compare the robustness of this metric with a fully satellite-based indicator, such as the AOD / NO₂ ratio.