Shortwave absorption by wildfire smoke dominated by dark brown carbon

The study investigates how brown carbon (BrC), particularly a dark, water-insoluble subset emitted by wildfires, contributes to shortwave light absorption and atmospheric warming. Traditional understanding assumes BrC absorbs mainly at short visible wavelengths and bleaches rapidly due to photochemical aging, leading climate models to either neglect or minimize BrC’s warming contribution compared to relatively stable black carbon (BC). Wildfire smoke is composed predominantly (>95%) of organic aerosol with minor fractions of inorganics and BC, yet the optical properties and radiative impacts of organic aerosol, especially BrC, remain poorly constrained. With anticipated increases in wildfire activity, quantifying BrC’s absorption across the visible spectrum, its persistence, and its evolution during atmospheric processing is critical to improve radiative forcing estimates. This work measures ensemble- and particle-scale optical properties of wildfire smoke to determine the contribution, persistence, abundance, and size distribution of dark BrC (d-BrC) tar balls and their dependence on co-emitted BC and oxidation conditions.

The prevailing view holds that soluble BrC predominantly absorbs at shorter visible wavelengths with negligible absorption at longer visible wavelengths, underpinning its designation as “brown” carbon. Conventional BrC measurements rely on solvent extraction of the soluble organic fraction and UV–visible spectrophotometry, yielding imaginary refractive index k typically between 10^-1 and 10^-2 for 380–500 nm. This methodology underrepresents insoluble components and often indicates rapid photobleaching, leading models to downweight BrC’s warming effect. Tar balls, a morphologically spherical subset of BrC observed in combustion plumes, have shown a wide continuum of optical properties with k spanning orders of magnitude (10^3 to 10^-1) at short visible wavelengths in prior studies. Uncertainties persist in tar ball formation mechanisms and their evolution in the atmosphere. Previous aircraft, laboratory, and ground-based studies have highlighted variability in BrC optical properties, potential nighttime chemistry (e.g., NO3) effects, and measurement biases of soluble-only approaches, indicating a need for particle-resolved, in situ methods capturing insoluble BrC.

Study design and platforms: Measurements were conducted during wildfire events in the western United States (Nethker/Shady Creek, 204 Cow, Castle and Ikes). Ensemble-scale absorption was measured aboard NASA’s DC-8 aircraft and the Aerodyne Mobile Laboratory (AML), complemented with particle-scale microscopy and spectroscopy from airborne and ground sampling.

Absorption and BC quantification: Photoacoustic spectrometers (PAS) measured aerosol absorption coefficients at 405, 488/532/561, and 664 nm (platform-dependent). Single-particle soot photometers (SP2) measured refractory black carbon (rBC) mass concentration. Data quality control thresholds excluded PAS b_abs < 5 Mm^-1 and SP2 rBC < 50 ng std m^-3. BC absorption at 550 nm was estimated from SP2 rBC with a BC MAC of 11.25 m^2 g^-1, applying an absorption enhancement factor of 1.5 for aged BC coatings; BC absorption at other wavelengths (405, 488 nm) was extrapolated using AAE = 1: b_abs,BC(λ) = b_abs,BC(550) × (550/λ)^(-AAE). Non-BC absorption at each wavelength was computed as b_abs(total) − b_abs,BC.

Sampling for microscopy: Airborne sampling used an impactor (AS-24W) collecting particles on Formvar-coated TEM grids with 50% cut-offs at 100 nm and 700 nm aerodynamic diameters; sampling per transect lasted ~1–3 min at 1.0 L min^-1. Nine airborne TEM grids (~3,275 particles) from Castle and Ikes and Shady Creek were analyzed using STEM (JEM-1400) with EDS (X-Max 80 mm), at 120 keV and 20 s acquisition; area-equivalent diameter cutoff for STEM-EDS was 0.25 μm. Ground sampling via an MP-3 sampler on the AML deposited particles on Cu TEM grids with lacey-carbon supports; ~43 grids were collected and analyzed using an FEI Tecnai G2 Spirit STEM-EDS and an aberration-corrected, monochromated Nion HERMES STEM.

Particle classification and composition: Particles were categorized by electronic darkness/brightness and sphericity in TEM images; electronically dark, thick, minimally deformed spherical particles were identified as d-BrC tar balls. STEM-EDS provided relative elemental weight fractions (C, N, O, Na, Mg, Al, Si, P, S, Cl, K, Ca, Ti, Mn, Fe, Zn). Volatile/semi-volatile species were acknowledged as potentially lost during sampling and TEM vacuum conditions.

EELS and refractive index retrieval: Low-loss EELS was performed on 33 samples using a Nion spectrometer (60 kV, probe convergence 30 mrad, collection angle 25 mrad, dispersion 20 meV channel^-1, dwell ~50 ms pixel^-1, ~70×70 pixels, ZLP FWHM ~100–150 meV). Grids were vacuum-baked at 160 °C (~465 °C equivalent at 1 atm) for ~6 h to remove surface contaminants. ZLP subtraction used a reflected-tail method; Fourier-log deconvolution removed plural scattering to yield single-scattering distributions S(E). Surface loss S_s(E) was included under no-coupling assumption. Kramers–Kronig analysis (KKA) via HyperSpy (FFT method, thin-film approximation) retrieved the dielectric function ε(E); thickness was set to particle diameter. Spectra were acquired from particle centers (>10 nm from edges for ≥50 nm particles) to minimize surface plasmon contributions. Complex refractive index n + ik was computed from ε_r and ε_i via standard relations. Method validity was cross-checked on co-sampled BC against model values.

Water-soluble BrC characterization: Filter punches from quartz-fiber filters were water-extracted (800 μL, 1 h sonication, 0.22 μm PTFE filtration). UV–Vis absorbance (350–800 nm, 1 nm resolution) provided solution absorption coefficients and mass absorption efficiency; k(λ) was derived via k = (α/ρ) × λ/(4π), with ρ = 1.4 g cm^-3 and absorbance normalized by A(700) to correct instrument drift.

Photochemical aging experiments: Plumes dominated by tar balls with negligible (<0.5%) or undetectable rBC were oxidized in a photochemical reactor to simulate daytime OH and nighttime NO3 conditions, using elevated oxidant concentrations to achieve up to ~84 equivalent hours (3 equivalent days/nights). PAS measured downstream absorption changes. For OH aging, MAC at 561 nm changed within ±2% over 3 equivalent days; for NOx/NO3 aging, MAC increased by 1.47 ± 0.01 over 3 equivalent nights. A Mie theory-based retrieval (PyMieScatt) inferred aged refractive indices using observed b_abs, b_scat (from nephelometry), and TEM-derived lognormal size distributions (diameters ~40–300 nm); inverse calculations were subsampled (one in ten spectral points) for computational feasibility.

Derived optical quantities: Mass absorption cross-sections (MACs) were computed by dividing modeled b_abs by tar ball mass concentration (density 1.6 g cm^-3) from TEM size distributions. Single-scattering albedo (SSA) was computed via Mie theory using EELS-derived complex refractive indices and size distributions. Absorption Ångström exponent (AAE) was calculated between wavelength pairs as AAE = ln[b_abs(λ1)/b_abs(λ2)] / ln[λ2/λ1].

Mass fraction estimation: Using cavity ring-down extinction (indicating refractory PM is 10–50% of total PM), counts/mass shares from TEM and supplementary tables (>85% by number and >88% by mass are organic within refractory PM), and the d-BrC tar ball fraction of refractory organics (0.58 ± 0.06), the d-BrC tar ball mass fraction was estimated at 5–26% of total PM.

• Dark brown carbon (d-BrC) tar balls dominate visible light absorption in wildfire smoke: approximately three-quarters of short-visible absorption and one-half of long-visible absorption in sampled plumes.
• d-BrC tar balls are water-insoluble, thermally stable, resist photobleaching during daytime OH oxidation over three equivalent days, and increase in absorptivity with nighttime NO3 oxidation (MAC enhancement factor 1.47 ± 0.01 over three equivalent nights at 561 nm).
• d-BrC tar balls are abundant: the number ratio of d-BrC tar balls to BC was ~4:1 across altitudes from surface to 10 km, remaining approximately constant; they comprised 5–26% of total PM mass, had mean area-equivalent diameters of 140–200 nm (geometric σ 1.4–1.6), and constituted a substantial fraction of refractory organic aerosol.
• Non-BC absorption fractions from aircraft (DC-8) tropospheric datasets: at 405 nm, Shady Creek 0.86 ± 0.10 and Castle 0.65 ± 0.12; at 664 nm, Shady Creek 0.70 ± 0.14 and Castle 0.46 ± 0.25, after accounting for a BC lensing factor of 1.5.
• Particle-scale optical properties from EELS: imaginary refractive index k decreased with wavelength following power laws sensitive to co-emitted BC mass fractions. For high BC mass fraction (>1.5%), k ∝ λ^(-0.65 to -3 reported in text), and for low BC fraction (<1.5%), k ∝ λ^(-~1). Extended Data Table 1 provides fitted relations: for high-BC fraction, Y0 = 79 ± 2.3, β = −0.65 ± 0.05; low-BC fraction, Y0 = 57 ± 27, β = −1.08 ± 0.08; nighttime oxidation, Y0 = 80 ± 180, β = −1 ± 0.4; daytime oxidation, Y0 = 20 ± 50, β = −0.89 ± 0.32. The real part n was wavelength-invariant at 1.31 ± 0.03 across fires.
• Dependence on burn conditions: at 550 nm, k = 0.13 ± 0.04 during high-temperature flaming (Shady Creek, high BC fractions) and decreased to 0.06 ± 0.03 during mixed/smoldering phases with lower BC mass, likely reflecting reduced graphitization.
• Imaging and spectroscopy: HAADF-STEM and EELS showed d-BrC tar balls with high k at 450, 550, and 650 nm, uniform composition and refractive indices across particle cross-sections, and strong thermal stability (160 °C in vacuum ≈ 465 °C at 1 atm). EELS features indicated lower graphitization (broadened π–π* at ~6 eV) than BC and graphene.
• Water-soluble BrC underestimates absorption: water-extractable BrC exhibited order-of-magnitude lower k than d-BrC tar balls across the spectrum, indicating that solvent-based measurements miss the dominant absorbing organic fraction.
• Modeled SSAs and MACs derived from EELS-based refractive indices and size distributions showed substantial absorption by d-BrC across visible wavelengths, with SSA decreasing with higher k and MAC increasing under nighttime ageing.

The findings directly challenge the prevailing assumption that brown carbon rapidly photobleaches and contributes negligibly to visible light absorption compared to black carbon. By combining ensemble absorption measurements with particle-resolved EELS-derived refractive indices, the study shows that dark, water-insoluble tar balls in wildfire smoke are widespread and persist as strong absorbers across the visible spectrum. Their abundance (four-fold more numerous than BC) and substantial mass fraction (5–26% of total PM) mean they can dominate plume absorption, especially at shorter visible wavelengths. The persistence of absorption under daytime OH oxidation and enhancement under nighttime NO3 oxidation indicate that d-BrC remains climatically relevant for multiple diurnal cycles, contradicting assumptions of rapid bleaching. The strong dependence of k on co-emitted BC mass fraction and burn conditions links combustion regime and carbon structure to optical properties, providing mechanistic insight (reduced graphitization during smoldering reduces k). Because climate models typically parameterize BrC with low k values derived from soluble extracts and often neglect insoluble components, current radiative forcing estimates likely underrepresent smoke-induced warming. Incorporating realistic, particle-scale refractive indices, wavelength dependences (power-law k(λ)), and chemical aging effects for d-BrC should improve model fidelity for wildfire aerosol radiative impacts.

This work identifies dark, water-insoluble brown carbon tar balls as the dominant contributor to shortwave absorption in wildfire smoke, accounting for roughly 75% of short-visible and 50% of long-visible light absorption. d-BrC tar balls are abundant relative to BC, thermally stable, resist daytime photobleaching, and exhibit enhanced absorptivity under nighttime oxidation. Particle-resolved EELS retrievals provide wavelength-invariant n (~1.31) and high k following power laws sensitive to burn conditions and co-emitted BC. Solvent-based methods substantially underestimate BrC absorption, highlighting the need to revise climate model parameterizations to include insoluble d-BrC, realistic spectral k, and diurnally varying oxidation effects. Future research should quantify the global prevalence of d-BrC across fire types and ecosystems, constrain its atmospheric lifetime and transport, refine size-resolved optical properties and mixing states, and integrate these into radiative transfer and Earth system models to improve estimates of smoke aerosol radiative forcing and climate feedbacks.

• BC absorption estimation used assumed parameters (MAC = 11.25 m^2 g^-1, absorption enhancement factor = 1.5, AAE = 1), which introduce uncertainties due to variability in BC mixing state, morphology, and coatings.
• TEM-based analyses may underrepresent volatile and semi-volatile components lost during sampling and under vacuum, potentially biasing composition and number fractions toward refractory particles.
• EELS retrieval relies on thin-film approximation and spectra collected from particle centers; spectra near particle edges (<10 nm) and particles <50 nm require caution due to surface plasmon effects, potentially limiting generalizability to very small particles.
• Aging experiments used elevated oxidant concentrations to simulate equivalent times, which, while standard, may not capture all real-atmosphere processes or multiphase interactions; inverse RI retrievals for aged aerosols employed spectral subsampling, smoothing fine spectral features.
• Water-soluble BrC measurements reflect only the extractable fraction and may not represent the total BrC pool; density assumptions (1.4 g cm^-3 for soluble organics; 1.6 g cm^-3 for tar balls) and lognormal size distributions introduce additional uncertainty into MAC, SSA, and mass fraction estimates.
• Spatial and temporal coverage is limited to specific western US fires sampled during FIREX-AQ; generalization to other regions, fuels, and fire regimes requires further validation.