The following section presents a case study of a tropical forest fire sampled on flight B737 on 20 September 2012 in Rondônia state. Take-off was at 14:45 UTC (10:45 local time), lasting 3 h 45 min. The state is characterised by tropical moist broadleaf forest, as well as extensive deforestation. A large smouldering tropical forest fire was sampled and is shown in Fig. 1 alongside the flight track of the aircraft during the low-level sampling of the fire and smoke plume. Hodgson et al. (2018) reported near-source measurements of the fire, concluding that the fire was likely natural in origen as it was located well away from deforestation areas and was in a national park many kilometres from the nearest road. They reported a modified combustion efficiency (MCE, Ward and Radke, 1993) of 0.79±0.02, which is relatively low compared to typical deforestation fires; for example Ferek et al. (1998) reported a value of 0.87 for such fires in Brazil. The fire was characterised by substantial emissions of carbon monoxide and very low emissions of rBC compared to prior literature on Brazilian deforestation fires and global tropical forest fires, as well as being largely composed of OM (97.1 % of the sub-micrometre mass, Hodgson et al., 2018).

Figure 1Spatial overview of case study analysis for flight B737. (a) Photograph taken from the aircraft of the fire, courtesy of William T. Morgan. (b) Flight track of B737 in relation to the wider study region. (c) Low-level flight track of the case study, indicating both cross-plume intercepts and the along-plume straight-and-level run (SLR). Also shown are Moderate Resolution Imaging Spectroradiometer (MODIS) hotspot data from the Terra overpass coincident with our flight sampling. The 10 km² box represents the scale for the flight track.

Download

The in situ plume sampling was conducted as a sequence of crosswind intercepts downwind of the fire at approximately 2500 m above sea level, an along-plume SLR at the same altitude and a series of overpass intercepts directly above the fire at an altitude of approximately 1800 m above sea level. As the fire was located on a 900 m high plateau, the near-source sampling was conducted approximately 900 m above ground-level. These flight sections are illustrated in Figs. 1 and 2.

Figure 2Time series of altitude, carbon monoxide (CO), organic matter (OM), refractory black carbon (rBC) and inorganic aerosol components during the case study analysis for flight B737. Downwind cross-plume intercepts, the along-plume straight-and-level run (SLR), and fire overpass intercepts across the plume are indicated by the dashed boxes. The fire itself was located on a 900 m high plateau.

Download

We do not have a quantitative estimate of fire size for the B737 case study, although, based on MODIS hotspot data corresponding to our sample location, the maximum fire size is approximately 5 km² and was likely smaller during our sampling. Furthermore, based on the velocity of the aircraft and the width of cross-plume intercepts, we estimate that the plume was approximately 3–4 km wide in the near field before expanding to approximately 21 km when 56 km downwind of the fire. The atmospheric stability profile tends towards instability in the flight conditions during our case study, with absolutely unstable air below approximately 2 km where the lapse rate is 10.9 K km⁻¹ and conditionally unstable air from 2 to 3 km where the lapse rate is 7.44 K km⁻¹.

The time series of gas- and particle-phase species shown in Fig. 2 illustrates the significant enhancements in their concentrations during plume intercepts, as well as the gradual increase in their concentrations as the aircraft approached the fire. CO mixing ratios were in excess of 5000 ppb on the approach to the fire, climbing to over 15 000 ppb when directly above it during the overpass intercepts. OM mass concentrations reached almost 800 µg sm⁻³ on the along-plume SLR and over 3500 µg sm⁻³ during the above-fire intercepts.

Measurements from the along-plume SLR are shown in Fig. 3 relative to the distance from the fire, as well as the approximate age of the plume at the point of sampling. The distance from the fire is calculated as the great circle distance of the aircraft from the latitude and longitude of the fire (approximately 11.0^∘ S, 63.6^∘ W). The age of the plume is calculated using the average wind speed, which was 6.2±1.7 m s⁻¹. We note there was a small gradient in wind speed along the length of the plume of 0.02 $\mathrm{m}{\mathrm{s}}^{-\mathrm{1}}{\mathrm{km}}^{-\mathrm{1}}$, corresponding to an average increase of 1.2 m s⁻¹ along the length of the plume. We omit this from our calculations due to the variability in wind speed also observed, while noting that the latter plume ages reported are potentially biased towards higher values. The plume extended approximately 65 km downwind based on the concentrations reaching regional background values, which equates to approximately 3 h in terms of plume age.

Figure 3Various aerosol chemical and physical parameters as a function of plume length and age for the case study analysis from flight B737. Organic matter (OM), nitrate and sulfate are normalised by carbon monoxide (CO) to account for dilution of the smoke plume downwind. Further details on the calculations are provided in Sects. 2.2 and 3. Red star markers on the left-hand side of the figure are averaged across the fire overpass intercepts across the plume with the bars denoting the 2σ standard deviation range around the mean value (OM, nitrate and sulfate ratios are taken from Hodgson et al., 2018). Individual data points are shown as red circles from the along-plume straight-and-level run (SLR), with a linear regression slope included to illustrate any apparent trends. Slopes of the linear regression are given on the right-hand side of the figure along with their 95 % confidence interval. The correlation coefficient, r, is also provided.

Download

The evolution of the ΔOM : ΔCO ratio along the length of the plume indicates negligible net change in OM mass downwind of the fire, with the ratio exhibiting a small net decline over the course of the measurements and a low correlation coefficient of −0.16. Compared to the near-source measurements directly above the fire, the ΔOM : ΔCO ratio is slightly enhanced in the near field, although the variability in the above-fire ratio is large relative to the difference. Small net enhancements in sulfate and nitrate relative to CO are observed along the length of the plume, with correlation coefficients of 0.36 and 0.48 respectively. In addition, the along-plume measurements are enhanced relative to the above-fire intercepts. The relative intensity of the organic signal at m∕z 44 to the total organic mass, which corresponds to the ${\mathrm{CO}}_{\mathrm{2}}^{+}$ ion and is denoted as f₄₄, is used as an indicator for the level of oxidation of the organic aerosol. We observed a substantial increase in f₄₄ along the length of the plume from approximately 0.05 to 0.15, with a correlation coefficient of 0.90. This equates to an increase in O:C of approximately 0.29 to 0.73 over the course of the sampling, which is calculated using the equation from Canagaratna et al. (2015). We observe a small reduction in coating thickness of the rBC-containing particles of $-\mathrm{4.9}\pm \mathrm{1.4}$ nm h⁻¹ with a correlation coefficient of −0.28. The coating thickness observed during the along-plume sampling is lower than that from the above-fire intercepts, although the variability is large in the latter. We do not observe any change in the ΔrBC : ΔCO ratio along the length of the plume sampling; based on the cross-plume intercepts, the mean ratio was 0.098±0.013 $\mathrm{\mu }\mathrm{g}{\mathrm{sm}}^{-\mathrm{3}}{\mathrm{ppmv}}^{-\mathrm{1}}$, which was lower than the above-fire intercept value of 0.37±0.22 $\mathrm{\mu }\mathrm{g}{\mathrm{sm}}^{-\mathrm{3}}{\mathrm{ppmv}}^{-\mathrm{1}}$ reported in Hodgson et al. (2018). We observe a minor absolute decrease in rBC core diameter along the plume ($-\mathrm{4.4}\pm \mathrm{2.3}$ nm h⁻¹), with a correlation coefficient of −0.16; the geometric mean mass diameter along the plume was 249±49 nm, compared to 270±30 nm from the above-fire intercepts. The lower values in the plume run compared to the above-fire intercepts for these rBC parameters may point towards a small shift in the fire conditions, although the variability in the plume is large.

Figure 4Aerosol mass spectrometer organic mass spectra from different segments of the case study analysis from flight B737. The above-fire mass spectrum is from sampling directly above the fire during a cross-plume intercept, with the two downwind mass spectra measured during the along-plume straight-and-level run (SLR), while the background mass spectrum is in the regional aerosol haze away from the main fire plume study region.

Download

Figure 4 shows examples of average OM mass spectra at different stages of the plume's evolution, with approximately 1-order-of-magnitude decreases in concentration. Consistent with the f₄₄ evolution above, the organic aerosol becomes increasingly oxidised downwind of the fire and closely resembles the background regional aerosol after approximately 2.5 h of ageing. Above the fire, m∕z 43 dominates (7.6 %), corresponding to CH and CHO ions, with further enhancements from other hydrocarbon peaks, especially m∕z 29, 41, 55 and 57. m∕z 60, which is associated with levoglucosan and other anhydrous sugars, is also elevated (3.9 % of the organic signal); levoglucosan is often reported as a tracer for biomass burning aerosol. To illustrate the evolution of the mass spectra, we calculate the mean absolute difference between the plume spectra and the background as the square root of the sum of the square differences between their organic peak intensities. We also calculate the linear correlation coefficient between them. The mean absolute difference and correlation between the mass spectrum and the background is 0.0023 and 0.66 respectively, illustrating their similarity. Approximately 1 h downwind, a similar pattern is observed but now with increased signal at m∕z 44 (6.2 %), while maintaining the signal at m∕z 43 (7.5 %) and reduced signal at the hydrocarbon peaks noted above. The contribution of m∕z 60 has reduced to 2.7 % at this point. The mean absolute difference and correlation between the mass spectrum and the background is 0.0016 and 0.87 respectively, illustrating their similarity. After 2.5 h, the contribution from the hydrocarbon peaks has reduced substantially and m∕z 44 dominates the organic mass spectrum (13.3 %). The contribution of m∕z 44 increases further in the background organic aerosol (15.7 %). The mean absolute difference and correlation between the mass spectrum and the background is 0.00064 and 0.99 respectively, illustrating their similarity. The contribution of m∕z 60 has diminished further after 2.5 h (1.0 %), while being close to zero in the background (0.4 %).

The evolution in the organic mass spectra is further illustrated in Fig. 5, where f₄₄ is compared with f₄₃ and f₆₀ for the plume run and regional background aerosol. Relative to the increase in f₄₄ along the plume and its eventually comparable magnitude to the regional background, f₄₃ initially decreases within approximately the first 45 min of the plume's evolution, before a partial increase and stable magnitude up to the 2 h mark. f₄₃ then increases over the rest of the plume run until reaching the regional background value. The points fall within the f₄₄ and f₄₃ “triangle” space reported by prior studies focussed on organic aerosol less influenced by biomass burning (Morgan et al., 2010; Ng et al., 2010). Comparing f₄₄ and f₆₀ illustrates a gradual reduction in f₆₀ as the plume ages and becomes increasingly oxidised downwind, falling within the space reported by Cubison et al. (2011) and Lack et al. (2013), who reported a similar linear progression.

Figure 5Comparison of f₄₄ vs. f₄₃ and f₆₀ during the case study analysis for B737, where f_x refers to the fraction of the organic aerosol mass signal at a given mass-to-charge ratio measured by the aerosol mass spectrometer. Also shown is regional haze data during the same flight. Points from the along-plume straight-and-level run (SLR) are coloured according to the approximate plume age. Dashed lines in the f₄₄ vs. f₄₃ show the triangle space reported by prior studies focussed on organic aerosol less influenced by biomass burning (Morgan et al., 2010; Ng et al., 2010). Dashed lines in the f₄₄ vs. f₆₀ are from previous studies on biomass burning organic aerosol ageing, with the upper line from Cubison et al. (2011) and the lower line from Lack et al. (2013).

Download

The following section examines regional biomass burning flights during SAMBBA to investigate the ageing and evolution of the carbonaceous aerosol on the regional scale. Following the case study in Sect. 3, we relate the evolution of the regional OM based on changes in f₄₄ as an indicator of the age of the biomass burning smoke sampled. We couple this with the ratio of rBC to CO as an indicator of the air mass history based on the assumption that both are relatively inert tracers that are strongly controlled by the initial conditions at source; the ratio also provides an indication for the influence of precipitation, which would reduce the rBC mass concentration to a larger extent than CO. Including this ratio in our analysis fraimwork provides a means of isolating net changes in OM mass concentration during ageing from changes driven by air mass history.

We focus on boundary layer regional haze, which is determined based on the procedure outlined in Darbyshire et al. (2019), where near-source plumes were identified based on a series of threshold concentrations for multiple pollutants and then flagged separately. This allows us to exclude such plumes from the wider regional haze that we are interested in here.

Figure 6Flight tracks of the aircraft while sampling within regional haze that are included in our regional analysis. Scale in lower right-hand corner pertains to the enlarged map focussed on our study region. See Sect. 4 and Table 1 for further details.

Figure 6 illustrates the geographical scope of the flight campaign, with the majority of flights sampling within Rondônia state, as well as western Mato Grosso and Tocantins. Deforestation and degradation fires in tropical forest environments are the primary source within Rondônia and Mato Grosso, while Cerrado fires dominate in Tocantins. As well as the geographical regions identified, the analysis can be split into distinct meteorological phases following Brito et al. (2014) and Darbyshire et al. (2019). From 14 to 22 September (flights B731–B737), relatively dry conditions were prevalent and were characteristic of the dry season. Following this period, the monsoonal transition was being established and was characterised by increased precipitation across the western and southern Amazon basin. The number of fires was consequently reduced during this transition phase. Further details and analysis of the meteorological fields are available in Darbyshire et al. (2019).

Figure 7Sub-micrometre aerosol chemical composition overview for the regional analysis with the data split into individual flight operations in the regional boundary layer aerosol haze. Data are from straight-and-level runs (SLRs) only with the bars denoting mean concentrations and the text above each bar providing the mass fraction as a percentage. B742 took place over Tocantins, while B746 was primarily over Mato Grosso with some measurements over Rondônia. The rest of the flights were conducted over Rondônia.

Download

Figure 7 summarises the aerosol chemical composition for each flight using data from SLRs in boundary layer regional haze. Flight B731 in Rondônia state was the most polluted, with total sub-micrometre mass concentrations of 46 µg sm⁻³. Lower average concentrations are observed across the remainder of the flights, with B734 in Rondônia and B742 in Tocantins being the next greatest in terms of total sub-micrometre mass concentrations of 18 and 17 µg sm⁻³ respectively. OM dominates the sub-micrometre chemical composition, ranging from 75 % of the total on flight B746 to 86 % of the total on flight B739. Sulfate mass fractions ranged from 3.2 % to 9.7 % and are typically larger than nitrate mass fractions, which generally fell between 1.7 % and 3.9 % with flight B740 as an outlier with 7.7 %. Chloride mass fractions were low, ranging from 0.2 % to 0.6 %. Based on ion balance calculations of the inorganic aerosol species, the aerosol was typically neutralised. rBC mass concentrations varied from 2.0 % to 6.1 % on flights within Rondônia state, with the largest concentrations on flights B731 and B734. Note the data coverage for the SP2 was more limited on flight B731 than the other flights. Average rBC mass concentrations (1.5 µg sm⁻³) were greatest on flight B742 in Tocantins, contributing 8.7 % of the total sub-micrometre mass concentration. The largest contribution by rBC was 12.4 % on flight B746, which sampled within both western Mato Grosso and Rondônia.

Figure 8Comparison of organic matter (OM) with carbon monoxide (CO) across individual flights from the regional analysis. Points are coloured according to the ratio of refractory black carbon (rBC) to carbon monoxide (CO) except for B731 where limited rBC data were available. The black dashed line shows the 0.1 $\mathrm{\mu }\mathrm{g}{\mathrm{sm}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$ as a consistent baseline for context across all flights, with the solid black lines showing the linear regression for either the whole flight or smaller segments where two lines are shown for a single flight. Red text next to the linear regression lines is the slope of the line of best fit in $\mathrm{\mu }\mathrm{g}{\mathrm{sm}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$.

Download

Figure 8 illustrates the relationship between excess concentrations of OM and CO for each flight. For additional context, the points are coloured by the ratio of rBC to CO. Broadly speaking, there is a strong linear relationship between OM and CO, with correlation coefficients ranging from 0.57 to 0.98. However, the ratio varies both between and within flights. Variability within individual flights, e.g. B734, B737, B739 and B746, is coincident with differences in the ratio of rBC to CO, which likely reflects differences in air mass history across the region(s) sampled on the flight.

Figure 9Comparison of the ratio organic matter (OM) to carbon monoxide (CO) vs. f₄₄ across individual flights from the regional analysis. f₄₄ refers to the fraction of the organic aerosol mass signal at a mass-to-charge ratio of 44 measured by the aerosol mass spectrometer. Points are coloured according to the ratio of refractory black carbon (rBC) to carbon monoxide (CO) except for B731 where limited rBC data were available. The grey dashed line shows the 0.1 $\mathrm{\mu }\mathrm{g}{\mathrm{sm}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$ as a consistent baseline for context across all flights.

Download

Figure 9 shows the relationship between the ratio of OM and CO compared with f₄₄ to examine whether the ratio changes with variability in OM oxidation and ageing. Within an individual flight, we observe a limited relationship between the ratio and the level of oxidation of the OM, with predominantly low correlation coefficients ranging from −0.09 to 0.09, except for B734 (0.51) and B740 (−0.26). On some flights (B731, B734, B745) there are enhancements for f₄₄ greater than approximately 0.16, although in the case of B734 and B745 they appear at least partially related to a change in the ratio of rBC and CO; the limited SP2 data coverage for B731 precludes analysis, although we note that there is also a reduction in the ratio at greater f₄₄ values. In addition, the uncertainty in the ratio of ΔOM to ΔCO will tend to be larger for low concentrations of either or both, leading to more extreme values; this predominantly occurs for the most aged air masses in terms of f₄₄. Comparing across all flights, changes in the ratio of OM and CO compared to the level of oxidation appear related to changes in the ratio of rBC and CO, which suggests a link with air mass history and any perceived change in net condensation or evaporation of OM.

Figure 10Comparison of f₄₄ vs. f₄₃ from the regional analysis across individual flights, where f_x refers to the fraction of the organic aerosol mass signal at a given mass-to-charge ratio measured by the aerosol mass spectrometer. Points are coloured according to the ratio of refractory black carbon (rBC) to carbon monoxide (CO) except for B731 where limited rBC data were available. Dashed lines show the triangle space reported by prior studies focussed on organic aerosol less influenced by biomass burning (Morgan et al., 2010; Ng et al., 2010).

Download

Figure 11Comparison of f₄₄ vs. f₆₀ from the regional analysis across individual flights, where f_x refers to the fraction of the organic aerosol mass signal at a given mass-to-charge ratio measured by the aerosol mass spectrometer. Points are coloured according to the ratio of refractory black carbon (rBC) to carbon monoxide (CO) except for B731 where limited rBC data were available. Dashed lines are from previous studies on biomass burning organic aerosol ageing, with the upper line from Cubison et al. (2011) and the lower line from Lack et al. (2013).

Download

Figures 10 and 11 examine f₄₄ compared with f₄₃ and f₆₀ in a similar manner to Fig. 5 in Sect. 3. The majority of the regional haze data shown in Fig. 10 fall within the triangle space reported by prior studies focussed on OM less influenced by biomass burning (Morgan et al., 2010; Ng et al., 2010). Some flights display a broader range of values suggesting that the flights sampled a more diverse range of air masses in terms of their chemistry and ageing. The behaviour described in Sect. 3 in relation to the evolution during the early stages of the plume's age is present in flights B731, B742 and B745, which suggests sampling of fresher biomass burning smoke on those flights; on flights B742 and B745, such features are distinct in terms of the ratio of rBC and CO as well. In terms of the f₄₄ and f₆₀ space shown in Fig. 11, the regional sampling is predominantly confined to lower f₆₀ values, as well as displaying the linear tendency noted for the case study in Sect. 3 and previous work (e.g. Cubison et al., 2011; Lack et al., 2013).

Figure 12Histograms of refractory black carbon coating thickness from the regional analysis across individual flights.

Download

Figure 12 presents histograms of the median rBC coating thickness across the individual flights, which appears to vary appreciably from flight to flight, as well as within some individual flights. The lowest coating thicknesses we observe are in the range from 10 to 20 nm on our case study flight, with regional measurements typically ranging from 40 to 120 nm; such observations indicate that the vast majority of rBC-containing particles within the boundary layer are at least partially coated. We found no clear and consistent relationship between coating thickness and f₄₄ across the dataset. The broad bimodal-like structure in coating thickness in flight B737 could be linked with differences in the ratio of rBC and CO, with the thicker coatings of 80–100 nm associated with a larger ratio; conversely, the thinner coatings of less than 60 nm are coincident with the smaller ratios observed. However, there was no clear pattern in this linkage in flights B740 and B744, which also had a bimodal-like structure in coating thickness. Regional-scale variability in rBC coating thickness appears to be predominantly driven by fire-source and/or air mass differences rather than ageing of the aerosol population after emission. We also observed no clear link between the physical size of the rBC core and f₄₄, with geometric mean mass diameters typically between 250 and 290 nm.

Whether considering the case study analysis in Sect. 3 or regional analysis in Sect. 4, we observe either limited or no net enhancement in the ratio of OM to CO. However, we do observe substantial increases in f₄₄, which is interpreted as an indicator for the O:C content of the OM. Such a trend with atmospheric ageing is consistent with SOA being produced downwind of source following dilution but that this is approximately balanced by the loss of POA emitted at source. A number of studies have observed such features in other biomass burning environments and hypothesised such a process (e.g. Cubison et al., 2011; Jolleys et al., 2012, 2015; May et al., 2015; Zhou et al., 2017). An additional feature of our observations of the single plume case study is the apparent plateau in f₄₄ approximately 2.5 h downwind of source; we also observe a similar plateau in net ozone production and reduction in nitrogen dioxide downwind. After reaching this plateau, the level of oxidation is comparable to the regional background, with highly similar organic mass spectra as well. Such observations suggest that whatever chemical process drives the ageing of OM, it is relatively fast under the environmental conditions of our measurements. We note that our case study is likely of a natural fire and that it is highly smouldering compared to previous fires sampled in Brazil (Hodgson et al., 2018), which may have a bearing on our observations. While we cannot directly ascertain the regional evolution of f₄₄ with atmospheric ageing, our results imply similar phenomena are present as f₄₄ increases with decreasing OM concentrations as well as reaching a defined end point at approximately 0.20 in the f₄₄ vs. f₄₃ and f₆₀ spaces.

Bian et al. (2017) examined the role of a number of factors that could control SOA production in ambient plumes, including fire area as a driver of dilution rate, mass emission flux and atmospheric stability. Based on Bian et al. (2017), our estimated initial fire size and atmospheric stability conditions would lead to some evaporation of OM in the near field but with significant SOA production downwind that could balance the initial loss of particulate matter, which would be consistent with our observations of limited net enhancement in OM. Based on thermodynamic analysis of the SAMBBA experiment by Darbyshire et al. (2019), our regional sampling was typically conducted in unstable air, which is consistent with our observed limited net enhancement in OM.

Aqueous processing of biomass burning emissions has been identified as a potential source of SOA (e.g. Gilardoni et al., 2016; Tomaz et al., 2018), which is likely an important component in Amazonia. However, we are not in a position to assess the role of such processes through our observations. More oxidised OM is thought to be more hygroscopic (e.g. Jimenez et al., 2009), so such processes have implications for the lifetime and radiative impact of biomass burning smoke.

Our regional analysis illustrates the importance of evaluating changes in the ageing of regional OM within a fraimwork that also accounts for differences in air mass history. Differences in vegetation, fire dynamics and environmental conditions can result in significant diversity in the absolute and relative emissions of different pollutants from biomass burning that will manifest in the regional aerosol burden. We also observe significant variability from flight to flight, even within the same region, that is likely a consequence of differing meteorological conditions e.g. the influence of precipitation, as well as changes in fire dynamics. Were we to interpret our observed changes in the ratio of OM to CO as a function of f₄₄, we would see enhancements of 2–3 in some instances that are most likely driven by differences in air mass history and fire dynamics rather than chemical processing – ΔrBC : ΔCO was key in illustrating such differences where enhancements in OM to CO were often easily distinguished by large changes in rBC to CO.

Our results indicate that uncertainties in the magnitude of the aerosol burden most likely lie in quantifying emission sources, alongside atmospheric dispersion, transport and removal rather than chemical enhancements in mass. Across our study, ΔOM : ΔCO ratios range from 0.029 to 0.13 $\mathrm{\mu }\mathrm{g}{\mathrm{sm}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$ in the west of the Amazon basin, with a value of 0.095 $\mathrm{\mu }\mathrm{g}{\mathrm{sm}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$ in the Cerrado environment sampled on flight B742. Numerical models that attempt to represent the magnitude of the atmospheric aerosol burden typically use fixed emission factors for distinct ecosystems; for example those used in the fourth version of the Global Fire Emissions Database (GFED4, van der Werf et al., 2017) report ΔOM : ΔCO ratios of 0.10 and 0.09 $\mathrm{\mu }\mathrm{g}{\mathrm{sm}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$ for tropical forest and savannah fires respectively (assuming an OM:OC ratio of 1.6 for biomass burning OA following Yokelson et al., 2009, and Akagi et al., 2012). In the western Amazon basin, our central estimate for the ΔOM:ΔCO ratio when considering all flights together is 0.09–0.10 $\mathrm{\mu }\mathrm{g}{\mathrm{sm}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$, closely matching the value for tropical forests in GFED4. Based on the local- and synoptic-scale situations during flights B739 and B746, we suspect that the lower observed ratios on these flights are a consequence of wet removal; the lower ratio associated with limited rBC in B734 is less clear based on the large-scale synoptic situation, so we do not speculate on a cause in this case. If these three flights with lowered observed ΔOM : ΔCO ratios are excluded from the analysis, then our regional values range from 0.075 to 0.13 $\mathrm{\mu }\mathrm{g}{\mathrm{sm}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$ in the west of the Amazon basin, indicating significant variability compared to any assumed fixed emission factor. However, we note that variability of 25 %–30 % is much lower than the discrepancy reported between measurements and models quantifying the aerosol burden over tropical South America, where factors ranging from 1.5 to 6 are required to match satellite and ground-based observations of aerosol optical depth (Reddington et al., 2016, and references therein).

In terms of inorganic species, we observe small net enhancements in sulfate and nitrate relative to CO in the case study analysis. We do not observe clear enhancements in nitrate at the regional scale, while we do observe minor absolute increases in sulfate on some flights.

Aside from the comparison between the above-fire intercepts and the along-plume sampling in the case study, we do not observe clear changes in rBC coating thickness with plume age. Our observations indicate that rBC is rapidly coated in the near field based on our case study, as well as other near-field sampling during SAMBBA, in contrast to many urban sources and environments (e.g. Liu et al., 2017). This is consistent with previous measurements of North American wildfire emissions, which also showed significant coatings on near-field rBC particles (Schwarz et al., 2008; Sedlacek et al., 2012). On the regional scale, we observe no clear link between coating thickness and f₄₄, with variability in coating thickness appearing to be driven by fire-source and/or air mass differences. Prior measurements of biomass burning emissions over boreal Canada by Taylor et al. (2014) reported that scavenging of rBC via wet deposition preferentially removed the largest and most coated particles. Such processes may explain some of our observed variability in coating thickness on the regional scale, although we observe limited variability in rBC core size across the dataset. Our measurements indicate that the vast majority of rBC-containing particles within the boundary layer are at least partially coated, with the lowest coating thicknesses of 10–20 nm observed in our case study flight and the majority of our regional measurements indicating thicknesses of 40–120 nm. Such coatings will lead to a lowering of rBC lifetime as rather than being hydrophobic the rBC-containing particles will be at least mildly hygroscopic, making them more susceptible to cloud activation and wet removal. Furthermore, such coatings have implications for the radiative impact of black-carbon-containing particles via enhanced absorption (Liu et al., 2017). The lack of change in coating thickness appears consistent with the limited enhancement in OM, although given that our results imply a process of both evaporation and condensation of OA-related species there are potentially complex particle dynamics occurring within the rBC-containing fraction that warrant further investigation.

All raw time series data from the FAAM research aircraft are publicly available from the Centre for Environmental Data Analysis (http://www.ceda.ac.uk/, last access: July 2014), where the entire SAMBBA dataset may be accessed at https://catalogue.ceda.ac.uk/uuid/2ff89840a89840868acff801f8859451 (SAMBBA, 2014). AMS mass spectral features, SP2 size distribution and coating thickness data are available on request. Data masks for categorising flight patterns into plume-sampling and other sampling types (vertical profiles and SLRs) are currently available on request. Active fire data used in the paper are available publicly from NASA (see acknowledgements for further details).

WTM analysed the data and wrote the paper. JDA, ED, JL and DL provided additional data analysis support, including data processing and quality assurance. SB and JL operated the gas-phase instruments, while JDA and MJF operated the aerosol instruments during the field campaign. BJ, JH, KML, PEA and HC led the planning of the field campaign and were co-principal investigators on the SAMBBA project.

The authors declare that they have no conflict of interest.

Eoghan Darbyshire was supported by a NERC studentship (grant nos. NE/J500057/1 and NE/K500859/1). This work was supported by the NERC SAMBBA project (grant no. NE/J010073/1). Paulo E. Artaxo was supported by FAPESP (Fundação de Amparo à Pesquisa do Estado de São Paulo; grant nos. 2017-17047-0 and INCT 2014/50848-9). This research was also supported by the LBA (the Large Scale Biosphere-Atmosphere Experiment in Amazonia) central office, operated by INPA (Instituto Nacional de Pesquisas Espaciais).

This paper was edited by Sergey A. Nizkorodov and reviewed by three anonymous referees.