Non-reversible aging can increase solar absorption in African biomass burning aerosol plumes of intermediate age

Recent studies highlight that biomass-burning aerosol over the remote southeast Atlantic is some of the most sunlight-absorbing aerosol on the planet. In-situ measurements of single-scattering albedo at the 530 nm wavelength (SSA530nm) range from 0.83 to 0.89 within six flights (five in September, 2016 and one in late August, 2017) of the ORACLES (ObseRvations of Aerosols above CLouds and their intEractionS) aircraft campaign, increasing with the organic aerosol to black carbon (OA:BC) mass ratio. OA:BC mass ratios of 10 to 14 are lower than some model values and consistent with BC-enriched source 5 emissions, based on indirect inferences of fuel type (savannah grasslands) and dry, flame-efficient combustion conditions. These primarily explain the low single-scattering albedos. We investigate whether continued chemical aging of aerosol plumes of intermediate age (4-7 days after emission, as determined from model tracers) within the free troposphere can further lower the SSA530nm. A mean OA to organic carbon mass ratio of 2.2 indicates highly oxygenated aerosol with the chemical marker f44 indicating the free-tropospheric aerosol continues to oxidize after advecting offshore of continental Africa. Two flights, for 10 which BC to carbon monoxide (CO) ratios remain constant with increasing chemical age, are analyzed further. In both flights, the OA:BC mass ratio decreases over the same time span, indicating continuing net aerosol loss. One flight sampled younger (∼ 4 days) aerosol within the strong zonal outflow of the 4-6 km altitude African Easterly Jet-South. This possessed the highest OA:BC mass ratio of the 2016 campaign and overlaid slightly older aerosol with proportionately less OA, although the age difference of one day is not enough to attribute to a large-scale recirculation and subsidence pattern. The other flight sampled 15 aerosol constrained closer to the coast by a mid-latitude disturbance and found older aerosol aloft overlying younger aerosol. Its vertical increase in OA:BC and nitrate to BC was less pronounced than when younger aerosol overlaid older aerosol, consistent with compensation between a net aerosol loss through aging and a thermodynamical partitioning. Organic nitrate provided 68% on average of the total nitrate for the 6 flights, in contrast to measurements made at Ascension Island that only found inorganic nitrate. Some evidence for the thermodynamical partitioning to the particle phase at higher altitudes with higher 20 1 https://doi.org/10.5194/acp-2021-1081 Preprint. Discussion started: 3 January 2022 c © Author(s) 2022. CC BY 4.0 License.

14) offshore, where it overlaid an aerosol layer of a lower OA:BC mass ratio (Fig. 2). The upper OA-enriched aerosol is younger than the aerosol sampled at a lower altitude, based on an aerosol age (time since emission) estimated from tracers released within the campaign's aerosol forecast model (shown later). This could be consistent with an anticyclonic circulation recirculating the fresher aerosol aloft while slowly subsiding back to the African continent (see Fig. 21 of Redemann et al.,90 2021), losing organic aerosol along the way. We note a small net OA loss has previously been reported from other African BBA aircraft campaigns (Jolleys et al., 2012). If representative of a larger-scale aerosol-meteorology environment, this would help explain an increase in SSA with altitude anecdotally noted in the field that is separate from that caused by the thermodynamical phase partitioning of inorganic nitrate (Wu et al., 2020). Model-derived estimates for the mean aerosol time since emission (age) of up to two weeks were calculated using the campaign's operational aerosol forecast model, the Weather Research and Aerosol Aware Microphysics (WRF-AAM) Model (Thompson and Eidhammer, 2014). The model releases tracers tagged to carbon monoxide (CO) at the fire source for each day of a two-week forecast. The fire source is taken from a burned area product of 500 m spatial resolution from the Mod-100 erate Resolution Imaging Spectrometer (Giglio et al., 2006) and may miss smaller burned areas. The current analysis takes advantage of the model's prior use for seeking out the smoke layers that are subsequently sampled by the aircraft. The model configuration is similar to that in Saide et al. (2016), with the 12-km spatial resolution regional model encompassing a domain (41 • S-14 • N, 34 • W-51 • E) sufficiently large to capture almost all contributing fires. The model is driven by the National Center for Environmental Prediction Global Forecasting System meteorology, using daily smoke emissions from the Quick Fire Emis-105 sions Dataset (Darmenov and da Silva, 2013) released into the model surface layer. These are advected thereafter according to the model physics, with their spatial distribution constrained near real-time with satellite-derived optical depths. This allows a diurnal cycle representation of the daytime burning. The capabilities of WRF-AAM include photochemistry, particulate matter, optical properties, radiative forcing, aerosol radiation and cloud chemistry feedback, and aerosol-cloud interactions (Grell et al., 2005;Wang et al., 2015). 110

Modified combustion efficiency
CO and CO 2 are used to infer fire emission conditions (Collier et al., 2016;Yokelson et al., 1997)  Aerosol Mass Spectrometer (HR-ToF-AMS, referred to as AMS) during ORACLES, building on previous experience in the southeast Pacific (Yang et al., 2011;Shank et al., 2012) and the Arctic (Howell et al., 2014). The native time resolution is approximately five seconds, with the data interpolated onto a one-second temporal grid to facilitate integration with other datasets. The overall uncertainty in the reported aerosol mass concentrations is estimated at 33% to 37%, at a one-minute time 130 resolution, based on Bahreini et al. (2009). Some analyses were restricted to level legs ranging from 4 to 10 minutes, listed in Table S1, further reducing the uncertainty about the mean to 19%-10 %. Further details on the AMS data treatment and the sampling layout are provided in the Supplement.
In addition the uncertainty in the OA:BC mass ratios are expected to be smaller when aerosol concentrations are larger, because of improved signal-to-noise but also towards minimizing effects from dilution, by which OA evaporates through 135 mixing with cleaner environmental air (e.g., Hodshire et al., 2021), and from model-observational disparities in the smoke plume locations, which are likely to be more noticeable at the smoke plume edges. We focus on those OA:BC mass ratios that have become stable with increasing OA mass (Fig. 5). The OA:BC mass ratio is significantly less for air with OA>3 µg m −3 than for air with OA>20 µg m −3 , particularly for younger aerosol (Fig. 3a). Fig. 3b indicates that the OA:BC mass ratio stabilizes at OA mass concentrations ≥ 20 µg m −3 , establishing the threshold we apply throughout this study. We note that 140 Fig. 5a also indicates OA:BC mass ratios can increase again 10 days after emission, but do not pursue this as the model skill in the smoke plume locations is likely to become less over time.
Measurements of f44, the fraction of the OA mass spectrum signal at m/z 44 relative to the total OA mass concentration, indicates the presence of the CO + 2 ion, or oxidation resulting from chemical aging (Canagaratna et al., 2015). f44 serves as a robust physically-based proxy for BBA age and forms one of two metric for continuing aging. Elemental analysis, yielding https://doi.org/10.5194/acp-2021-1081 Preprint. Discussion started: 3 January 2022 c Author(s) 2022. CC BY 4.0 License. calibrated using fullerene soot effective density estimates from Gysel et al. (2011). Further details can be found in Schwarz et al. (2006).

Ratios of ∆BC
∆CO (shortened to BC:CO) serve to assess homogeneity of the aerosol composition at the source emission. The ratios are non-dimensionalized by using the ideal gas law at standard temperature (273K) and pressure (1000 hPa) to convert 155 the CO concentrations from ppb to ng m −3 . Farmer et al. (2011) provide an approach for estimating the contribution to the total nitrate signal from organic nitrate (ON) using the N O + : N O + 2 ratio, building on the observation that organic nitrates typically fragment into larger proportions of NO + than do inorganic nitrates (in their study, organic NO + ratios vary between 1.8 to 4.6 for different organonitrates, compared 160 to 1.5 for NH 4 NO 3 ). Their Equation 1, reproduced below, provides an estimate of the ON fraction that can be readily applied to the ORACLES AMS data, assuming that the ON fraction can be resolved. The success of this approach also assumes that the inorganic nitrates capable of providing a large NO + ratio, such as mineral nitrates, are not present. Both assumptions are justified for the SEA free troposphere.

Optical Instruments: Nephelometers and Particle Soot Absorption Photometers
This study primarily focuses on the SSA at 530 nm wavelength, but also examines the absorption Angstrom exponent (AAE) and mass absorption cross-section (MAC) measurements for evidence of brown carbon. Scattering from all particles is measured continuously by a TSI nephelometer (model 3563) at the (470, 550, 700) nm wavelengths (λ), with a linear regression in log-log space used to estimate the scattering at 530 nm wavelength. The spectral light absorption coefficients (σ a ) of the 175 total aerosol are measured by a Particle Soot Absorption Photometer (PSAP; Radiance Research) at the 470, 530, and 660 nm wavelengths. The values are an average based on measurements from two PSAPs in 2016, and only one PSAP functioned in 2017. Both filter-based measurements are corrected according to Anderson and Ogren (1998). The SSA values are based on the wavelength-averaged (as opposed to wavelength-length-specific) corrections of Virkkula (2010). The use of the average wavelength-corrected values reduces a potential high bias at the shortest wavelength introduced by multiple scattering within 180 the PSAP signal (Pistone et al., 2019), and reduces spurious effects from filter changes . Although similar to Pistone et al. (2019), a stricter aerosol threshold (OA>20 µg m −3 rather than scattering at 530nm > 10 Mm −1 ) is applied and no arithmetic weighting by extinction is done. SSA values at 530 nm are at standard temperature and pressure.
The nephelometer measurements occurred at 40-50% relative humidity, while the PSAP measurements measured at a lower ∼ 20% RH, brought about by heating the PSAP optical block to approximately 50 • C (Pistone et al., 2019). Ambient RH 185 measurements ranged up to 80%, with higher RH data samples excluded by construction. Two other single-wavelength (550 nm) nephelometers (Radiance Research, M903) measured at two different relative humidities, one at 80% and the other at below 40% RH (Howell et al., 2006). The impact on light scattering, estimated from the ratio of the ambient to dry RH measurements, is estimated to be less than 1.2 for 90% of the time within Shinozuka et al. (2020). The 20% increase in scattering by the ambient RH is an upper bound, as the ambient RH was typically <80%. Aerosol absorption can also increase  Should brown carbon contribute to the solar absorption, this should affect the MAC BC+OA .

Determination of Aging
Two independent measures of aerosol age with differing merits are invoked. Model tracers of CO released at the emission source provide a precise but potentially uncertain time estimate while the oxidation information provided by the chemical 200 marker f44 is a clearer indicator of chemical aging. Model-derived trajectories become more uncertain over longer time spans and distances as the small differences in the model meteorology from nature accumulate. In addition, mixing at the aerosol plume edges can appear as an older aerosol age. Importantly, the measurements of f44 correlate well with the younger modelderived age estimates (shown later), supporting the use of both metrics as aerosol age indicators.

205
Flight selection is based on the availability of at least 20 minutes of organic aerosol (OA) data exceeding >20 µg m −3 within the free troposphere (altitudes above 1.5 km) at relative humidities (RH) < 80%, and must possess aerosol model ages between September 2016 and 31 August 2017), zonal easterly winds exceeded 6 m s −1 along ∼ 10 • S at altitudes between 3-5 km. A wind isotach, known as the African Easterly Jet-South (Nicholson and Grist, 2003;Adebiyi and Zuidema, 2016), explains the 220 more zonally-elongated, meridionally-constrained aerosol plume spatial structure extending over the ocean evident in Fig. 5.
Some recirculation back to land around the south Atlantic anticyclone also apparent for these three days (see also Ryoo et al. (2021) for more synoptic detail). Days in which the aerosol is constrained closer to the coast ( Fig. 4) occur when a mid-latitude disturbance passes by to the south, altering the circulation to its north (Diamond et al., 2018;Kuete et al., 2020;Zhang and Zuidema, 2021).

225
The different flights intersect air of different ages, but none with model-estimated ages of less than 4 days. This is older aerosol than the oldest aerosol sampled during the Southern African Regional Science Initiative (SAFARI) campaign , estimated at 2-3 days of age, but is younger than that sampled by the CLoud-Aerosol-Radiation Interaction and the climatological AEJ-S sets in, as this is responsible for its geographic distribution of aerosol at higher altitudes. The AEJ-S, though weaker than climatology in August 2017 Ryoo et al., 2021), was present on 31 August, 2017.

Chemical composition and age distribution within the six flights
Mean f44 values exceed 0.175 after 4 days of model-estimated age since emission (Fig. 6). These values are on par with f44 240 values from Asian/Siberian smoke transported to Alaska over two weeks (Cubison et al., 2011) and indicate highly-oxidized aerosol. f60 values are relatively constant and below 0.005. Both are consistent with chamber studies reporting lifetimes of f44 and f60 of approximately 20 days and 10 hours, respectively (George and Abbatt, 2010;Hodshire et al., 2019). Some continuing oxidation is indicated between model-derived days since of emission of 4-5 to days 5-6 (mean f44 values of 0.175 increasing to 0.21), but no further change in f44 is evident after 6 days since emission.
(aged) low-and semi-volatile oxygenated OA Huffman et al., 2009;Hodzic et al., 2020). The slope of the H:C to O:C mass ratios ( Fig. 7; based on vanKrevelen (1950) describes the evolution of oxygenated organic aerosol. Downward movement along the slope of -1 is consistent with continuing oxidation through the formation of carboxylic acid (C 6 H 5 OH) 250 formation, a process that ultimately replaces one hydrogen atom with one oxygen atom (Heald et al., 2010). Different SOA precursors from different sources or burning conditions may contribute to the range of the observed H:C ratios Ng et al., 2011). Data from the 31 August, 2017 flight suggests more highly oxidized aerosol than any of the 2016 flights, with a mean O:C ratio of 0.81. The higher O:C mass ratio compared to the values from 2016 is not explained through the change in calibration constants between the two years, and related (we presume) to different fuel sources and conditions. Overall, the 255 average (± standard deviation) plume values of H:C, O:C, and the organic-aerosol-to-organic-carbon mass ratio (OA:OC) are 1.2 ±0.1, 0.7 ±0.1, and 2.2±0.1, respectively, over all six flights. The OA:OC mass ratio, a measure of the oxygen content that is useful for model evaluation (Hodzic et al., 2020;Lou et al., 2020), is higher than common model-applied values of 1.4-1.8 for SOA:OC (Aiken et al., 2008;Tsigaridis et al., 2014;Hodzic et al., 2020) and primary OA:OC ratios measured near-source of 1.6 (Andreae, 2019), but are on par with measurements from the Atmospheric Tomography (ATom) campaign made in the 260 same region (Hodzic et al., 2020).  (Collier et al., 2016;Zhou et al., 2017) that are typical for grasslands and savannahs (Janhäll et al., 2010;Vakkari et al., 2018). Mean non-dimensionalized BC:CO ratios vary between 0.008 to 0.011, with a minimum on 24 September. The model-derived age estimates indicate most of the aerosol was emitted at least six days previously, with that on 9/24/2016 being the youngest, corresponding to the lowest f44 values. The 8/31/2017 flight, for which Fig. 7 suggests the highest oxidation, has one of the higher mean f44 values, but also one of the highest 270 mean MCE and a higher mean BC:CO value, suggesting potential fuel type differences as well. OVerall, BC:CO ratios do not increase with increasing MCE as expected based on Kondo et al. (2011), but this likely reflects this study's small range of MCE values, for which Vakkari et al. (2018) also do not find a correlation. The mean values hint at a decrease in BC:CO over the 2016 BB season, consistent with the speculation within Eck et al. (2013) that the more combustible fuel may be ignited earlier in the BB season, but the trend is statistically-insignificant. The BC:CO ratios from the ORACLES and CLARIFY campaigns 275 (Wu et al., 2020) are the highest of those shown for the surveyed literature in Table 1, exceeding those reported for southern Africa in Andreae (2019), and (Formenti et al., 2003). We note that Andreae (2019) indicate a higher emission factor for black carbon from savannah fires than Andreae and Merlet (2001), and, that the Akagi et al. (2011) emission factors, incorporated into many of the CMIP6 models (AR6, WG1, sec 6.2.2.6), produce higher BC:CO values for crop residue than for savannahs, in contrast to Andreae (2019).

280
MCE varies inversely with the moisture content for grasses (Korontzi et al., 2003), with leaf litter and woody fuels tending to dry more slowly than do grasses, and to burn by smoldering, as opposed to flaming. Thus we interpret the high MCE values to reflect a large contribution from dry and dead grasses, as opposed to green grass and woodlands. The high BC:CO values for ORACLES and CLARIFY are also consistent with the burning of dry grass, which produces relatively low emissions of carbon monoxide . That the BC:CO values measured at offshore locations exceeds those measured previously 285 over land suggests a possible coincidence between when the burning of the savannahs occurs with when the free-tropospheric winds capable of transporting the aerosols far to the west are more pronounced. Differences between aerosols in the free troposphere versus the boundary layer may also not be fully accounted for. Overall the mean submicron mass fractions of the six flights combined are 66% OA, 10% nitrate (NO 3 ), 11% sulfate (SO 4 ), 5% ammonium (NH 4 ), and 8% BC. SO 4 :BC remains constant at approximately 1.6-1.7 (not shown), indicating its formation as a secondary inorganic aerosol has ended. 290 5 Evidence for loss of organic aerosol with further oxidation Given that secondary aerosol formation is expected to proceed more quickly when BC:CO ratios are lower (Vakkari et al., 2018), because the precursor gases may be more available (Yokelson et al., 2009), it is important to control for the BC:CO ratio if wishing to attribute OA:BC changes to aging rather than source differences.

Thermodynamical repartitioning of aerosol versus chemical loss
An additional consideration could be that if the older aerosol is also situated lower within the atmosphere, that a smaller OA:BC mass ratio at a lower altitude could in fact reflect a thermodynamical particle-to-gas repartitioning, as opposed to an irreversible loss of OA. This is investigated further for the same two flights (8/31/2016 and 9/24/2016). The selected aircraft profiles sam-310 pled air with different vertical structures in aerosol age, which we contrast in search of insights. The thermodynamic behavior would be most apparent in the inorganic nitrate fraction (e.g., Wu et al., 2020) and we examine the inorganic nitrate partitioning within the two individual profiles at the locations indicated on Fig. 1. Nitrate only contributes 10% to the total sampled free-tropospheric aerosol mass, and inorganic nitrate even less so, but an examination of IN's thermodynamic partitioning, a reversible gas->particle phase transition favored at higher altitudes because of the lower temperatures/higher relative humidities 315 (Nenes et al., 1997), can also illuminate if some of the OA mass loss may also be thermodynamically reversible.
The 24 September, 2016 profile at 12.3 • S, 11 • E ( Fig. 11a; southernmost profile in Fig. 3, top row) is broadly comprised of one main aerosol layer centered on 5 km aged ∼4 days since emission, and a slightly older smoke layer of ∼5 days in age, centered on 3 km (Fig. 11b). The younger aerosol aloft is consistent with the stronger upper-altitude winds. These also transport moisture, consistent with climatological expectations (Adebiyi et al., 2015;Pistone et al., 2021), generating relative humidities 320 exceeding 80% above 4 km when combined with the cooler high-altitude temperatures (Fig. 11a). The water vapor mixing ratio and statically-stable potential temperature profiles indicate little mixing of air between different altitudes (Fig. 12a). Both the NO 3 :BC and OA:BC mass ratios increase with altitude (Fig. 11b). The inorganic nitrate fraction is approximately 20%, also increasing slightly with altitude. This profile, also shown in , spawned the initial speculation that a large-scale recirculation and subsidence pattern could explain the reduced OA:BC mass ratio at lower altitudes, although the 325 age difference of only one day cannot explain subsidence of more than approximately 500 m.
The 8/31/2016 profile, occurring further south (16.4 • S, 6.5 • E), sampled aerosol that was constrained near the coast (Fig. 2, top row) by an impinging mid-latitude disturbance (Ryoo et al., 2021), a meteorological condition that is common in September when the AEJ-S is less strong . The associated synoptically-driven ascent mixes moisture upward, generating a linear increase in water vapor mixing ratio (Fig. 12b). The relative humidity increases with height (Fig. 11c), 330 reaching 80% and capable of generating mid-level clouds elsewhere (Adebiyi et al., 2020). The aerosol aloft is older, at approximately 9 days in age above 3.5 km, overlying younger aerosol aged between 5-6 days below 3 km (Fig. 11d). Winds are weak below 4 km, and primarily eastward throughout the full profile (Fig. 11c). For this profile, the OA:BC and NO 3 :BC mass ratios also increase with altitude, but not as strongly as on 9/24. The inorganic nitrate fraction is slightly higher in the mean than on 9/24 (∼25% versus ∼20%). This difference between the two days is consistent with a loss of organic nitrate with 335 age, although the smoke emissions at the source could have also contained more inorganic nitrate initially on 8/31/2016 than on 9/24/2016. For the 9/24 profile, the increase in the inorganic nitrate fraction with height of approximately 0.05 can only be because the lower temperatures and higher relative humidities above 4 km favor the particle phase (Nenes et al., 1997;Zhang et al., 1999), as the increase is inconsistent with younger aerosol aloft being organic-nitrate-enriched. Although the older aerosol aloft on 340 8/31 should support an even more pronounced vertical structure in IN:NO 3 than on 9/24, this is not the case. More consistent with the hypothesis that continuing aerosol aging favors a net aerosol loss, is that the overall increase in the NO 3 :BC and OA:BC mass ratios with height, which could contain some thermodynamical repartioning, is less pronounced on 8/31 than on 9/24. It may be that the signal-to-noise ratio is too small to resolve the IN vertical structure well within these individual profiles. This analysis does clarify that most of the nitrate contained within the sampled BBA is organic in nature, in contrast 345 to more aged aerosol sampled at Ascension (Wu et al., 2020).
In these two examples, the younger aerosol occupies a more humid environment on 9/24, and a drier environment on 8/31. A compositing of OA:BC, NO 3 :BC and aerosol age by RH for all six flights reveals younger aerosol is more likely to occupy a more humid environment than does older aerosol (Fig. 13). The mean NO 3 :BC ratio decreases by almost 50% as the RH decreases from 70% to 30% (Fig. 13a), consistent with a thermodynamic repartitioning. For the same data samples, the mean 350 OA:BC mass ratio decreases from 10.5 ± 2.6 for RH values between 60-80% to 9.9 ± 2.1 for 20% < RH < 40%. This indicates that a thermodynamical repartition can only explain a relative decrease in OA:BC with age of less than 10%. We conclude that most OA loss in the southeast Atlantic free troposphere is irreversibly lost after 4 days, with thermodynamic repartitioning providing a minor contribution (dilution and wet removal processes are already excluded by construction). This finding is not new in qualitative terms, with prior field campaigns in remote areas also highlighting a net OA loss as BBA ages beyond a day 355 (e.g., Capes et al., 2008;Jolleys et al., 2012Jolleys et al., , 2015Hodzic et al., 2015;Konovalov et al., 2019). Causes may be an evaporation of the organic material (Jolleys et al., 2015) as part of a continuing oxidation. The 9/24 example indicates how the advection of relatively fresher aerosol at higher altitudes can accentuate the observed vertical structure in nitrate and organic aerosol to black carbon, through a combination of both the thermodynamical repartitioning and continuing loss of organic nitrate and aerosol, whereas the vertical structure in OA:BC and NO 3 :BC in the 8/31 profile is less pronounced, which can be interpreted 360 as a compensation between the thermodynamical repartitioning and chemical loss of OA over time.

Radiative implications and inferences
The low OA:BC ratio of southern African BBA emissions in September, combined with a subsequent loss of OA as the smoke plume advects westward, has implications for the single scattering albedo (SSA, the ratio of the scattering efficiency to extinction efficiency).  Formenti et al., 2003), and slightly less than Dubovik et al. (2002) 370 based on AERONET-derived column-average measurements of BBA over continental Africa. They are on par with AERONET September-mean values at Mognu (Eck et al., 2013). SSA values reported at Ascension Island, further offshore Wu et al., 2020), are lower. The aerosol outflow on 24 September 2016 in the middle free troposphere registered both the highest OA:BC mass ratio (14.5) and the highest SSA (0.89) of the ORACLES-2016 campaign. An SSA parameterization on OA:BC provides a straightforward connection between the BBA chemical properties to the modeling of the direct aerosol 375 radiative effect for this region, with the caveat that incorporation of the remaining ORACLES data would further enhance the robustness of this relationship. The relationship is not anticipated to hold for the more polluted northern hemisphere, for which SOA production is expected to be more significant (Jolleys et al., 2012). In addition, such a parameterization is only effective if the model OA:BC mass ratios are realistic. GIven that OA:BC mass ratios are often too low in models, their absorption of sunlight will also be overestimated (Brown et al., 2021) until the chemical composition is correctly modeled.

380
Of further interest is whether any brown carbon absorption of sunlight is occurring, meaning a wavelength-dependent contribution to the total solar absorption that is typically small but potentially significant. Brown carbon is associated with primary organic aerosol, as SOA is typically considered to only scatter sunlight (Laskin et al., 2015). By 5 days, the primary organic aerosol that may contribute to wavelength-dependent solar absorption is expected to be gone, with SOA mostly scattering light (Bond and Bergstrom, 2006). Nevertheless, some studies suggest oxidation can result in new chromophrores (O'Brien and should be present east of the prime meridian (Carter et al., 2021). The AAE values (470-660nm), calculated from the level legs occurring within the individual flights, reach approximately 1.2 closer to the coast and in several locations further west (Fig. 15).
AAE increases weakly with OA:BC (correlation coefficient r of 0.27). Although not conclusive (little absorption is expected by brown carbon at 470 nm), the correlation and slightly elevated values closer to the coast are consistent with some brown carbon absorption close to the coast. To be consistent with Bond and Bergstrom (2006) would require some chromophores 400 to be present in the African aerosol outflow after 4 days since emission, perhaps as a byproduct of oxidation (O'Brien and Kroll, 2019). The model-derived age estimates indicate most of the aerosol was emitted at least six days previously, with that on 9/24/2016 being the youngest at 4-5 days corresponding to the lowest f44 values. At this time since emission, all primary organic aerosol should be consumed, as well as brown carbon (Laskin et al., 2015).
Another assessment of brown carbon absorption can be done using the mass absorption coefficients at 470 nm relative to 405 the sum of the BC and OA mass concentration (Carter et al., 2021), shown for the same level legs in Fig. 16. These range from 0.94 closer to the coast to 1.73 m 2 g −1 further to the northwest, decreasing with increasing OA:BC (r=0.86) as expected.
These values are less than the median value of approximately 1.75 m 2 g −1 reported in Carter et al. (2021) for the 2016 ORACLES deployment. The MAC BC+OA values closest to the coast, range between 0.94 to 1.2 m 2 g −1 . This is significantly less than the median value of 1.51 m 2 g −1 reported in (Carter et al., 2021) based on the Saleh et al. (2014) parameterization, 410 suggesting that the Saleh et al. (2014) parameterization may be overestimating brown carbon absorption over the southeast Atlantic. The discrepancy could be explained by different data treatments, with this study only selecting data from level legs with relatively homogeneous aerosol characteristics, so as to increase the signal-to-noise ratio and not convolve results with differences attributable to vertical structure. The level-leg mean mass absorption coefficient at 660 nm relative to the black carbon mass concentration alone is 10.8 m 2 g −1 , slightly higher than the median value of 9.3 m 2 g −1 reported for ORACLES-2016 by Carter et al. (2021) and less than the CLARIFY median value of 11.5 m 2 g −1 . This is likely because the BC-enriched 31 August 2017 flight contributes strongly to the mean value we report here. The 31 August 2017 flight occurred during the CLARIFY time frame, suggesting that the CLARIFY MAC BC,660nm may also have been elevated because the OA:BC mass ratio of the source emissions was lower compared to ORACLES values, as opposed to more continued aging. More analysis will be required to come to a more 420 definitive conclusion. More significant is that all of these MAC values exceed the Bond and Bergstrom (2006) value of 6.25 m 2 g −1 reported for strongly light-absorbing carbon.
As has been previously noted (Taylor et al., 2020), lensing, by which absorption increases through a Mie effect generated by the OA coating, is in theory able to both decrease the SSA and maintain a constant AAE (Lack et al., 2012;Cappa et al., 2012). Mie calculations, which require spherical shapes, may overestimate lensing effects, but aged BBA do compact from 425 the fractals that can define soot upon emission (Taylor et al., 2020). There is some indication that the co-emitted sulfate can contribute to the lensing (Christian et al., 2003), as can the enhanced humidity present within the aerosol layers (Redemann et al., 2001). Combined with other absorptive coating characteristics (Denjean et al., 2020), and photo-bleaching (Taylor et al., 2020), this may explain why the BBA over the remote southeast Atlantic is more light-absorbing than noted elsewhere without requiring the presence of brown carbon. We have not examined the impact of particle size and this may also contribute to the 430 explanation (smaller particles scatter less light according to a size 4 dependence).

Conclusions
In this study we attempted to place on firmer footing an early interpretation made during the ORACLES 24 September 2016 research flight, in which an increase in the single-scattering albedo with height was associated with a smoke plume enriched in organic aerosol mass relative to black carbon mass . The OA-enriched aerosol layer was sampled close 435 to the coast at an altitude of 4-6 km within strong westward winds, and overlaid a lower, older aerosol layer of a lower OA:BC mass ratio. In the field, the hypothesis was made that loss of OA through chemical aging while the aerosol was recirculated anticyclonically to the African continent, subsiding slowly in the meanwhile, could explain the pronounced vertical structure in aerosol composition and thereby in the single-scattering albedo . If a net OA loss is characteristic of BBA over the southeast Atlantic once the aerosol has advected away from the continent, this contributes to the explanation for 440 why the BBA over the SEA is so absorbing of sunlight in September.
When controlled for variability in the source emissions through the use of the BC:CO mass ratio, a decrease in OA:BC of 25-35% could be attributed to chemical aging. A simple SSA parameterization based on OA:BC of SSA=0.801+0.0055*(OA:BC), indicates that the range of OA:BC of 8 through 14 equates to an SSA variability of 0.83 to 0.89. A 25% change in OA:BC through chemical aging corresponds to a relative SSA change near 0.1, or almost 20% of the observed variability. We attribute 445 the remaining 80% to variations occurring in the production of SOA near the emission source, based on for example an increased availability of precursor gases when BC:CO ratios are lower (Yokelson et al., 2009;Vakkari et al., 2018).
The aerosol sampled during ORACLES-2016 and 31 August 2017 possessed modified combustion efficiencies exceeding 0.975, some of the highest reported in the literature surveyed for Table 1, with the exception of Ascension Island (Wu et al., 2020). Such values are characteristic of savanna grasses (Janhäll et al., 2010;Vakkari et al., 2018) that are also dry (Korontzi 450 et al., 2003), which are known to possess higher BC efficiencies (Andreae, 2019). For this region far removed from urban and industrial sources of pollution, continued production of SOA after 1-2 days is expected to remain minor (e.g., O'Brien and Kroll, 2019), with other work also reporting a small net OA loss with time for smoke from northern Africa (Jolleys et al., 2012(Jolleys et al., , 2015. Recent analysis of filters from August 2017 during ORACLES and CLARIFY also consistent with continued OA loss (Dang et al., 2021). This contrasts with northern hemisphere fire emission sources. Brown carbon production has been 455 linked to low OA:BC ratios (Saleh et al., 2014;McClure et al., 2020). We do not see much evidence for brown carbon in the ORACLES-2016 AAE and MAC values, perhaps because brown carbon is more closely linked to primary than to secondary organic aerosol, although we are limited by optical measurements that do not extend to wavelengths smaller than 470 nm.
At face value, this works further supports the use of optical parameterizations upon chemical parameters for improving the radiative representations of BBA in global aerosol models with increasingly sophisticated SOA schemes (e.g., Lou et al.,460 2020). Such parameterizations place more pressure on producing realistic representations of the BBA composition, however.
This study adds to literature indicating that OA model estimates made by multiplying the organic carbon by a factor of 1.4 will underestimate OA in this (and other) regions (Aiken et al., 2008;Tsigaridis et al., 2014;Shinozuka et al., 2020;Doherty et al., 2021), so that SSA parameterizations based on OA:BC or related chemical parameters will overestimate solar absorption (Brown et al., 2021). This study's OA:OC mass ratios of 2.2 ± 0.1 has also been shown for the Atomic Tomography mission 465 (Hodzic et al., 2020). Modeled OA:BC mass ratios can also be overestimated by over a factor of two over the southeast Atlantic in global models with sophisticated aerosol schemes (Chylek et al., 2019), suggesting the loss of OA with aging or slower SOA production processes (Kroll and Seinfeld, 2008;McFiggans et al., 2019) can also be under-accounted for. This is more important for remote environments containing thick smoke layers lacking the precursor gases for additional SOA production.
Southern Africa produces approximately one-third of the world's carbon emissions through biomass-burning (van der Werf 470 et al., 2010), with the global majority of the absorbing aerosols above cloud occurring above the southeast Atlantic (Waquet et al., 2013), indicating the importance of realistic representations for this radiative climate (Mallet et al., 2021).
September is a unique transition month. This is when the AEJ-S is strong, because the thermal gradient between the Kalahari heat low and the cooler, moist equatorial climate is more pronounced, driving the upper-level winds that transport the aerosol (Adebiyi and Zuidema, 2016;Kuete et al., 2020). The winds occur to the north of the heat low, with only dry convection 475 lofting the aerosols generated by the burning of the flaming-efficient dry grasses. The winds distribute aerosol with low OA:BC mass ratios as far away as south America (Holanda et al., 2020), so that the entire south Atlantic is covered by a blanket of highly-absorptive aerosol. This has a pronounced climate impact, because the south Atlantic stratocumulus deck is also at a peak extent and thickness then (Zuidema et al., 2016a), reinforced in part by the radiative stabilization of the south Atlantic free troposphere. Although IPPC AR5 assessments suggest the ability of smoke to both scatter and absorb sunlight leads to a 480 net compensation globally, this is not the case for the southeast Atlantic.   The best-fit line is SSA= 0.93-0.39*(BC:TC), with a correlation coefficient of -0.79. Times and spatial ranges of the level-legs provided in Table S1.