Interactive comment on “ African biomass burning plumes over the Atlantic : aircraft based measurements and implications for H 2 SO 4 and HNO 3 mediated smoke particle activation ” by V .

This paper describes the observation of biomass burning layers above west Africa during the AMMA project in 2006. The data describe the gas and particle phase characteristics of the plume and are used together with a model to predict CCN properties along a forward trajectory as the air advected across the Atlantic Ocean. The results show that H2SO4 in particular, but also HNO3 can deliver sufficient hygroscopic inorganic material to the biomass burning particles to ensure that they are effective CCN, even in BL cloud, after transport over the Atlantic Ocean.


Introduction
Biomass burning (BB) is a global phenomenon, which has an impact on the environment and climate (Crutzen et al., 1979;Andreae, 1983;Crutzen and Andreae, 1990;Houghton et al., 2001).BB plumes contain elevated concentrations of pollutants, including smoke particles and primary as well as secondary combustion gases.Savanna fires represent the single most important BB-type worldwide (Crutzen and Andreae, 1990;Andreae, 1991;Koppmann et al., 2005).Africa contains about two thirds of the world's savanna regions and 90% of the African savanna fires are believed to be human induced (Koppmann et al., 2005).Since BB plumes can be transported over thousands of kilometers, their impact on the environment and climate may occur far away from BB regions.For example, elevated O 3 present over the South Published by Copernicus Publications on behalf of the European Geosciences Union.
V. Fiedler et al.: African biomass burning plumes over the Atlantic Atlantic has been attributed to secondary O 3 formation in BB plumes originating from Africa (see review by Koppmann et al., 2005;Real et al., 2010).
BB releases primary pyrogenic gases (Koppmann et al., 2005) and primary smoke particles (Table 1, see recent review by Reid et al., 2005), whose characteristics and relative emission rates depend on various factors including particularly the type of bio material combusted and the burning conditions (flaming, smoldering).
Primary pyrogenic gases include, besides the major combustion products H 2 O and CO 2 , numerous minor gases, particularly CO, hydrocarbons, NO, NH 3 and SO 2 (Andreae and Merlet, 2001).Interaction of NO and organics leads to the formation of secondary ozone, which represents a greenhouse gas, an important atmospheric oxidant, and a precursor of OH radicals.
Primary pyrogenic particles contain solid cores, mostly soot (elemental carbon (EC)) and ash, and a semi-volatile coating composed of low vapor pressure organics (organic carbon (OC)), which is formed by rapid OC-condensation.The resulting internally mixed smoke particles have initial median diameters of about 125 nm and a mass ratio OC/EC of about 5-10 (Reid et al., 2005).
As a BB burning plume ages and dilutes, chemical transformations of primary pyrogenic gases take place leading to secondary gases including particularly O 3 .Some secondary gases undergo gas-to-particle conversion leading to chemical processing and additional size growth of primary smoke particles.Of these secondary gases, sulfuric acid (H 2 SO 4 ) and nitric acid (HNO 3 ) are particularly important.Sulfuric acid is formed by OH-induced conversion of the primary pyrogenic gas SO 2 (Reiner andArnold, 1993, 1994).Nitric acid is formed via OH-induced conversion of NO 2 , which results from rapid conversion of primary pyrogenic NO.Due to its very low saturation vapor pressure, sulfuric acid condenses on smoke particles and, due to its very large hygroscopicity, tends to increase smoke particle hygroscopicity.Nitric acid has a much higher saturation vapor pressure than sulfuric acid and would condense on smoke particles only in conditions of temperatures below about 200 K, which are occasionally found only at the tropical tropopause and in the polar lower stratosphere.However, HNO 3 may react with gases possessing large proton affinities.A key candidate is the primary pyrogenic trace gas ammonia (NH 3 ), whose BB release is about ten times larger than that of SO 2 but only about half of that of NO y .HNO 3 reacts with NH 3 yielding ammonium nitrate (NH 4 NO 3 ), which is thermally stable at temperatures typical of the middle and upper troposphere.Therefore, after BB plume ascent to the middle troposphere NH 3 may undergo conversion to NH 4 NO 3 , which could then condense on smoke particles.Gas-phase sulfuric acid, which is more slowly formed than HNO 3 , may convert some fraction of the NH 4 NO 3 , leading to aerosol-phase ammonium sulfate ((NH 4 ) 2 SO 4 ).The HNO 3 not converted to NH 4 NO 3 , would remain in the gas-phase.Hence, the difference of NO y and HNO 3 abundances sets an upper limit to the NH 4 NO 3 abundance.It is also conceivable that primary and secondary organic acids may neutralize NH 3 yielding ammonium salts.In addition, NH 3 may also react with other pyrogenic gases, leading to amides and nitriles.Also, in the very early plume, NH 3 may undergo chemical conversion to NO x increasing with BB fire temperature (Hegg et al., 1988).
However, smoke particles as well as the gases HNO 3 , NH 3 and SO 2 , the precursor of H 2 SO 4 , may experience substantial loss by cloud processes and deposition.Therefore, their concentrations in an aged BB plume and the effects on smoke particle processing by H 2 SO 4 and HNO 3 are difficult to predict.Removal of SO 2 by cloud processes is only moderate since SO 2 dissolution in cloud droplets is only moderate.Importantly, ice clouds do not scavenge gas-phase SO 2 .During droplet freezing dissolved SO 2 may even be released to the gas-phase (Clegg and Abbatt, 2001).However dissolved SO 2 may undergo H 2 O 2 mediated liquid-phase conversion to sulfate, which remains in the aerosol-phase after cloud water evaporation.
By contrast to SO 2 , the trace gases HNO 3 and NH 3 are highly soluble in cloud droplets and therefore may undergo very substantial removal by cloud processes.But, after convective ascent and cloud dissipation, additional HNO 3 can be photochemically formed from NO x , which is not significantly removed by cloud processes.Efficient formation of NH 4 NO 3 requires transport of NH 3 to the middle troposphere where temperatures are sufficiently low to allow thermal stability of NH 4 NO 3 (see above).Therefore, efficient NH 4 NO 3 formation in a middle troposphere air mass lifted by convection probably requires "dry convection".
Considering typical emission factors (Table 1) a biomass burning event releases more NO y than NH 3 .If one neglects removal by deposition and cloud processes, in an air mass which has experienced dry convection to the middle troposphere, the following highly simplified picture evolves: Ultimately, more HNO 3 molecules may be formed than NH 3 molecules are present.Therefore, HNO 3 may neutralize NH 3 , forming NH 4 NO 3 .Only the HNO 3 and NH 3 corresponding to the equilibrium partial vapor pressures of NH 4 NO 3 will remain in the gas-phase.NH 4 NO 3 will condense on the pyrogenic primary particles.The formation of gas-phase sulfuric acid from SO 2 is much slower than HNO 3 formation and the SO 2 released from biomass burning is about ten times less than the NH 3 released.Therefore, gasphase sulfuric acid formed in the ageing middle troposphere plume will convert only some fraction of the particle-bound NH 4 NO 3 to (NH 4 ) 2 SO 4 .Nevertheless, as mentioned above, the fraction of pyrogenic NH 3 reaching the middle and upper troposphere is highly uncertain.
Sulfuric acid, due to its large hygroscopicity, may have a particularly large effect on the ability of smoke particles to take up water molecules from the gas-phase in conditions with relative humidity RH < 100%, and to act as water vapor condensation nuclei (CCN = cloud condensation nuclei)  Andreae and Merlet, 2001) and the measured excess molar emission ratio dx/dCO 2 .The last column denoted with R gives the ratio between dx/dCO 2 and Ex/ECO 2 .

Substance x Ex
in cloud formation in conditions with RH > 100%.Smoke particle processing by H 2 SO 4 and NH 4 NO 3 is particularly important for atmospheric conditions with only small water vapor supersaturations WSS of only about 0.05%, which are typical for the maritime boundary layer and for maritime stratiform cloud formation (Seinfeld and Pandis, 1998).
The larger the H 2 SO 4 mass fraction of a smoke particle, the smaller will be the activation water vapor supersaturation (WSSa) required for activation.The H 2 SO 4 formed in the ageing plume increases with time until precursor SO 2 is exhausted.Therefore, also the H 2 SO 4 mass fraction of smoke particles increases with time.As a consequence, the ability of a smoke particle of a given size to act as CCN (at a given water vapor supersaturation) increases with time as the H 2 SO 4 mass fraction increases due to H 2 SO 4 uptake (see also model simulations below).In addition, coagulation contributes to increase the smoke particle diameter, which also contributes to decrease the water vapor supersaturation required for activation.However, as the plume ages, the number concentration of smoke particles decreases strongly, due to coagulation and plume dilution.Therefore, the number concentration of smoke particles which can be activated, at a given WSS, is expected to have a maximum at a certain plume age.The larger the rate of SO 2 -conversion to H 2 SO 4 , the smaller will be the plume age at which this maximum occurs and the larger will be the maximum concentration of smoke particles which can be activated.Therefore, the rate of SO 2 conversion to gas-phase H 2 SO 4 in the ageing and diluting plume is crucial in determining the H 2 SO 4 mass fraction of smoke particles, and thereby the evolution of their activation potential.This rate is determined by the OH concentration and its time variation in the plume.In a BB plume, the OH concentration is thought to be controlled mostly by OH-loss via the reaction of OH with NO 2 leading to HNO 3 and by OH-formation via processes involving organic plume gases (preferably acetone-photolysis leading to about 3.2 HO x radicals per acetone molecule) (Singh et al., 1994;Folkins et al., 1997).While elevated NO x tends to decrease OH, increased acetone tends to increase OH.Previous measurements of OH in an aged BB plume at 9000-10 000 m altitude have indicated OH concentrations of about 0.1 pmol mol −1 (Folkins et al., 1997).These, were not much different from ambient OH concentrations outside the BB plume.This led to the conclusion that the additional NO xinduced OH loss was approximately offset by an additional acetone-induced OH formation.
The present paper reports on airborne measurements of SO 2 and HNO 3 along with other gases and smoke particles in two aged savanna fire plumes over the East Atlantic, off the west coast of equatorial Africa (Gulf of Guinea).At the time of our measurements one plume was located in the middle troposphere (MT) and one in the upper troposphere (UT).From the trace gas data we infer the formation of H 2 SO 4 , HNO 3 and NH 4 NO 3 in the plumes and discuss implications with regard to their influence on the smoke particle activation potential.

Experiment
Our airborne BB plume measurements were part of the AMMA (African Monsoon Multidisciplinary Analyses) campaign and took place on 13 August 2006, off the western coast of Tropical Africa (Ghana).The measurements were made by various instruments on board the DLR (Deutsches Zentrum für Luft-und Raumfahrt, German aerospace center, Oberpfaffenhofen) research aircraft Falcon, when it dived into and cruised in two plumes at altitudes between about 3900 and 5500 m and between 10 800 and 11 200 m.
The AMMA project aims at a better understanding of the West African Monsoon, its influence on the processing of chemical emissions and its associated regional-scale and vertical transports.For this purpose an airborne campaign was conducted in July/August 2006 with special interest on biomass burning emissions.Further objectives were the characterization of the impact of mesoscale convective systems www.atmos-chem-phys.net/11/3211/2011/Atmos.Chem.Phys., 11, 3211-3225, 2011 on the ozone budget in the upper troposphere and the evolution of the chemical composition of these convective plumes as they move westward toward the Atlantic Ocean.Another objective was to discriminate the impact of remote sources of pollution over West Africa, including transport from the middle East, Europe, Asia and from southern hemispheric BB fires.Sulfur dioxide (SO 2 ) was measured by a chemical ionization mass spectrometry (CIMS) method with continuous inflight calibration using isotopically labeled SO 2 .The CIMSinstrument, which has been developed by MPI-K (Max-Planck-Institute for Nuclear Physics, Heidelberg) in collaboration with DLR, is equipped with a powerful ion trap mass spectrometer.A comprehensive description of the measurement system can be found in Speidel et al. (2007) and Fiedler et al. (2009a,b).The method is based on gas-phase ion molecule reactions in a flow reactor.These reactions involve reagent ions CO − 3 which react with atmospheric SO 2 ultimately leading to SO − 5 product ions.By measuring the abundance ratio of product and reagent ions with the ion trap mass spectrometer, the SO 2 mole fraction can be determined.The SO 2 measurements have a time resolution of 1 second and a detection limit (2 sigma level) of about 20 pmol mol −1 .The relative error is about plus or minus 12% for SO 2 mole fractions larger than 100 pmol mol −1 and increases close to the detection limit to plus or minus 40% (Speidel et al., 2007).
Nitric acid (HNO 3 ) measurements have been carried out with the same CIMS instrument.HNO 3 can be detected using the gas-phase ion molecule reaction of CO − 3 with HNO 3 .This reaction leads to (CO 3 HNO 3 ) − cluster ions, which again are detected by the mass spectrometer.The time resolution of the measurements is 1 s, the detection limit is 0.1 nmol mol −1 in our case, the estimated relative error plus or minus 50%, as we did not use a special HNO 3 calibration, but applied the SO 2 calibration instead.A detailed description of the further developed HNO 3 measurement method with HNO 3 calibration can be found in Jurkat et al. (2010).
Simultaneous measurements of other trace gases (CO 2 , CO, NO, NO y , H 2 CO, O 3 ) were carried out on the Falcon by DLR (see Table 2).
Carbon monoxide (CO) was detected using vacuum resonance fluorescence in the fourth positive band of CO (Gerbig et al., 1999).The accuracy of the CO measurements is ±10% for a time resolution of 5 s and with a detection limit of 3 nmol mol −1 .Carbon dioxide (CO 2 ) was measured with a differential nondispersive infrared instrument (NDIR), the detection limit was 0.1 µmol mol −1 , the sampling rate 1 s and the accuracy ±0.1% (Schulte et al., 1997).
Nitric oxide (NO) and the sum of reactive nitrogen compounds (NO y ) were measured using a chemiluminescence technique (Schlager et al., 1997;Ziereis et al., 2000).The NO y compounds are catalytically reduced to NO on the surface of a heated gold converter with addition of CO.The accuracy of the NO and NO y measurements is ±8% and ±15%, respectively.The time resolution is 1 s and the detection limit 5 pmol mol −1 and 15 pmol mol −1 , respectively.The nitrogen-bearing trace gas NH 3 is not detected by the NO y -instrument.Also ammonium nitrate is not measured, but the NO y -instrument measures HNO 3 released from NH 4 NO 3 by thermal decomposition, when atmospheric air is passed through the flow tube section containing the hot NO y -converter of the NO y -instrument.
Ozone (O 3 ) was measured using an UV absorption photometer (Schlager et al., 1997;Schulte et al., 1997).The accuracy of the ozone detection is ±5%, the detection limit is 1 nmol mol −1 and the time resolution 4 s.Formaldehyde H 2 CO has been measured using a Hantzsch reaction instrument (Kormann et al., 2003).The detection limit of this instrument is 84 pmol mol −1 , the time resolution is 180 s and the uncertainty ±30% at a mixing ratio of 300 pmol mol −1 .
Table 2 compiles the measured atmospheric substances, measurement techniques, specifications of the techniques and corresponding references.
Number concentrations and size distribution of aerosol particles in the size range between 4 nm and 20 µm were measured with a combination of condensation particle counters and a differential mobility particle sizer mounted in the cabin, as well as two wing-mounted optical aerosol spectrometer probes in a similar setup described by Weinzierl et al. (2009).The instruments deployed on the DLR Falcon during AMMA are listed in more detail in the supplement of Reeves et al. (2010).The cabin instruments sampled air through the forward facing DLR Falcon aerosol inlet, which is operated close to isokinetic sampling conditions and has no significant sampling losses for particles up to 1.5 µm particle diameter.The size range of particles in the accumulation and coarse mode above approximately 0.15 µm particle size is covered by measurements of the PCASP-100X and the FSSP-300 wing probes, two instruments which in principle detect the amount of light scattered by single particles.In order to infer the particle size distribution, knowledge on the complex refractive index of the aerosol particles is required (Schumann et al., 2010).In this study, we used for simplicity an refractive index of 1.54 + 0.0i, commonly used to represent an aged ammonium sulfate type aerosol, for all size distribution data discussed.In particular for particles falling into the PCASP-100X size range (0.15-1.0 µm) the possible error introduced by this simplification is estimated to be below natural variability due to the instrument being relatively insensitive to variations in the actual refractive index.Particle concentrations in this manuscript are reported as ambient concentrations.satellites TERRA and AQUA (Justice et al., 2002;Giglio et al., 2003; NASA/GSFC) and detects hotspots/fires as a thermal anomaly using data from the middle infrared and thermal infrared bands.In most cases, this thermal anomaly is a fire, but sometimes it is a volcanic eruption or the flare from a gas well.The minimum detectable fire size is a function of many different variables (scan angle, sun position, land surface temperature, cloud cover, amount of smoke and wind direction etc.), so the precise value slightly varies with these conditions.Results of validation measurements indicate that the minimum flaming fire size typically detectable at 50% probability with MODIS is on the order of 100 m 2 .Under ideal conditions performance is somewhat better and the smallest detectable fire size is approximately 50 m 2 .As can be seen from the figure, fires had been active in a large region covering the southern hemispheric African continent mostly south of the tropical rainforest belt, which suggests that most of the fires were savanna fires.The core of the BB region with the largest density of fire spots was located between about 20-30 • East and 5-15 • South.Also shown in  1).During that dive, the polluted air mass was even visible as a thick brownish layer.When looking downward, surface details were not visible.

Plume localization and trajectories
Figure 2 shows an image of light absorbing aerosol particles measured on 13 August (day of our airborne measurements) by OMI (ozone monitoring instrument) aboard the AURA satellite (Levelt et al., 2006a,b).Plotted is the aerosol index AI, a product determined from the difference between the backscattered UV wavelength in a polluted atmosphere and a pure atmosphere (a positive AI values means absorbing aerosols).The OMI instrument can distinguish between aerosol types, such as smoke, dust and sulfates, and measures cloud pressure and coverage, which provide data to derive tropospheric ozone.The instrument employs hyperspectral imaging to observe solar backscatter radiation in the visible and ultraviolet.The instrument is a contribution of the Netherlands's Agency for Aerospace Programs (NIVR) in collaboration with the Finnish Meteorological Institute (FMI) to the Earth Observing System (EOS) Aura mission.The AI image (Fig. 2) reveals the presence of an extended pollution plume rich in light absorbing aerosol particles, mostly soot particles.The plume of light absorbing particles is present mainly over the Tropical East Atlantic and also over Tropical Africa and covers an area of at least 4 million km 2 .Unfortunately, the height of the soot plume cannot be obtained from the satellite image.The plume exhibits a horizontally inhomogeneous distribution and the dive of the Falcon into the plume took place in one of the denser plume regions (dive region is marked by a cross in Fig. 2).
CALIPSO (Cloud Aerosol Lidar and Infrared Pathfinder Satellite Observations, see also http: //www.nasa.gov/calipso)lidar data also confirm the presence of the middle troposphere plume (hereafter MT-plume) over the Gulf of Guinea, on 13 August.For the region of the Falcon dive into the MT-plume, they indicate a top altitude of about 5000 m and a bottom altitude of about 3000 m (Real et al., 2010).
To investigate the origin of the MT-plume, we have made back-trajectory simulations using the LAGRANTO model (Wernli and Davies, 1997).Figure 3 shows a typical 10day back-trajectory of the lower plume superimposed on a map.The altitude of the trajectory is indicated by the color code.The time span between tick-marks (filled circles) is 24 h.Also shown on the map are the northern edge of the fire region as detected by MODIS between 1 and 10 August 2006 (red dots) and the African copper belt region (blue area).The so-called African copper belt is the region in Zaire and Zambia, where major copper smelters are located, which represent major SO 2 sources.
On 4 August, 10 days prior to our measurements, the air parcel, which was intercepted by the Falcon at 3900 m on 13 August, passed at about 1200 m altitude over the core of the fire region, about 500 km west of the copper-smelters.Hereafter, during 7-10 August, the air parcel traveled at about 4000 m altitude over the north western region of the fire belt.Hence, it seems that the air parcel took up pyrogenic gases mainly on 4 August when it passed at low altitudes over the core of the fire belt.This would imply a time span of 10 days for transit from the core of the BB region to the measurement region.On 5 August, while leaving the core of the BB region, the air parcel ascended from about 1200 m to about 3000 m and on 8 August it reached about 4000 m altitude.
We also investigated air mass trajectories starting from the copper-smelter region during the days of interest.We found out, that all air mass trajectories starting in the copper belt led to the North or even to the East.So an uptake of copper smelter SO 2 is not likely.
To investigate the fate of the MT-plume, after 13 August, we have also made LAGRANTO forward-trajectory model simulations.Therefore, Fig. 3 shows additionally to the 10day back-trajectory a 10-day forward simulation of the trajectory, starting on 13 August at 3900 m altitude at 12:00 UTC.This simulation indicates that, after our measurements, the plume parcel traveled westward over the Atlantic and reached the coast of northern Brazil on 20 August.Hereafter it traveled northwards again.While reaching the Brazilian coast, the plume parcel descended to a lowest height of about 1300 m altitude.

Measurement results
Figure 4 shows in its left panel the altitude profiles of SO 2 and temperature (both descents in red, ascents in blue).The altitude profile of the SO 2 mole fraction indicates the presence of a SO 2 -rich plume in the mid troposphere (hereafter termed MT plume) with a sharp top at about 5200 m (descent) to 5000 m (ascent) and two altitude regimes with different degrees of SO 2 pollution, including an upper plume regime and a main plume regime (below about 4200 m).The Falcon has spent about 11 min inside the MT-plume, below 5200 m, corresponding to a horizontal distance of about 150 km.On that horizontal length scale the top altitude of the MT plume was quite similar, differing only by 200 m.As the Falcon ascended further, a pronounced local SO 2 maximum of about 90 pmol mol −1 was observed at 12:24 UTC.This indicates the presence of a second much less polluted plume in the UT at about 10 800-11 200 m altitude (hereafter termed UT-plume, Fig. 4).Its height extension was only about 400 m which is much less than that of the MT-plume (about 2000 m).
In the right panel of Figure 5, first and second panel, show the time series of measured trace gases (plotted are CO, CO 2 and O 3 , NO y , NO, HNO 3 , and H 2 CO).CO 2 , which is widely used as biomass burning marker, is strongly enhanced inside the MTplume.All trace gases are also markedly increased in the MT-plume.Within the MT-plume, at constant altitude of the Falcon of 3900 m, the trace gases HNO 3 , NO and CO in- crease, whereas the other measured trace gases remain almost constant.
Figure 5, third panel, shows the time series of flight altitude and the SO 2 mole fraction measured by the CIMSinstrument.As the Falcon dived from 9000 to 5500 m, SO 2 increases slightly from about 30-40 pmol mol −1 .After a short cruise just above the top of the visible plume at 5500 m, the Falcon dived further to 3900 m.During that dive, between 5500 m to about 5000 m, SO 2 increases abruptly by a factor of about 10 to about 400 pmol mol −1 , and below about 5250 m, SO 2 further increases abruptly to 1400 pmol mol −1 .During the following short cruise at 3900 m, SO 2 varies between 1400 and 1250 pmol mol −1 .After that 5 min cruise at 3900 m, as the Falcon ascended, SO 2 decreases abruptly to 430 pmol mol −1 , and above 4700 m, SO 2 decreases further to the previous atmospheric background value of 30-40 pmol mol −1 , reached at 5700 m.
Aerosol particle time series data are shown in the forth panel of Fig. 5. Plotted are ambient number concentrations of aerosol particles possessing diameters between 160 and 270 nm (N 160−270 ) and between 270 and 1000 nm (N 270−1000 ).Also given is the outside air temperature during the measurements.As the Falcon dived in the MT-plume both aerosol particle concentrations increase very steeply and are correlated with SO 2 and the other trace gases.Pronounced maxima of particle concentrations were observed at 4-5 km altitude, just shortly before the Falcon had dived to the lowest altitude of 3900 m.
Figure 6 shows the corresponding altitude profiles of measured trace gases and particles.The shape of the altitude profile of NO y is similar to SO 2 , clearly showing an increase in the MT-plume.At 3900 m, NO y reaches up to about 8 nmol mol −1 and the mole fraction ratio dNO y /dCO 2 is about 6×10 −4 (see Table 1).In the MT-plume at 3900 m, NO x /NO y is only 0.13 whereas the measured HNO 3 /NO y ranges between about 0.63 and 1.1.The trace gases CO, H 2 CO and O 3 are markedly increased in the MT-plume.The trace gas H 2 CO can be a primary pyrogenic gas and a secondary gas formed in the plume from primary gases.The gas O 3 represents a secondary gas formed in the plume via NOoxidation by organics.For example, reaction of NO with the PA-radical (peroxyacetyl radical) leads to NO 2 .Photolysis of NO liberates a single O-atom, which combines with O 2 to form O 3 .The mole fraction ratio dCO/dCO 2 , measured in the MT-plume at 3900 m is 0.043.The mole fraction ratio dH 2 CO/dCO 2 measured in the MT-plume at 3900 m is 7.7 ×10 −5 .
Aerosol particle number concentration altitude profiles are included in Fig. 6.Plotted are concentrations of particles possessing diameters of 4-1500 nm (N 4 ), 160-270 nm, 270-1000 nm, 500-1000 nm and 1000-5000 nm.In addition particles with diameters larger than 10 nm, which have been heated to 250 • C during passage of the thermo-denuder, are plotted and denoted N 10 non-volatile.Hereafter, these particles will be termed "non-volatile (nv) particles".They most likely contain black carbon and ash and perhaps also certain organic carbon species with very low vapor pressures.
The N 4 -altitude profile increases in the MT-plume with decreasing altitude, similar to SO 2 .Above the MT-plume, N 4 increases with increasing altitude and reaches maximum values of about 1300 cm −3 .In the MT-plume, N 10 (nv) is nearly identical to the total N 4 .This indicates that in the plume almost all particles contain non-volatile cores.Above the MT-plume, most particles do not contain non-volatile cores.
The N 270−1000 are also very substantially increased in the MT-plume and during dive are as large as the N 4 .This indicates that most particles had diameters larger than 270 nm.Of all measured trace substances, N 270 exhibits the largest increase in the MT-plume at 3900 m (almost a factor of 1000).
The N 500−1000 profile also increases in the MT-plume (by a factor of about 100).The ratio N 500−1000 /N 270−1000 is about 0.01 at 3900 m altitude.
N 1000−5000 is very small (0.1 cm −3 ) at 3900 m.This indicates that clouds were absent, which is consistent with the relatively low RH (<27%) in the MT-plume at 3900 m altitude (see Fig. 4).
Summarizing the results of the altitude profiles, for SO 2 , NO y , H 2 O, O 3 , H 2 CO and aerosol particles in the MT-plume, there were no notable differences between the ascent and descent data.In contrast, for CO, HNO 3 and NO, descent and ascent profiles are markedly different, ascent data (after the penetration of the plume) are larger.This increase of HNO 3 , NO and CO during the cruise at the constant flight-level of about 3900 m altitude in the plume is puzzling.It contrasts the almost constant behavior of SO 2 , NO y , O 3 and CO 2 .The increase of the abundance ratio of excess CO and excess CO 2 (dCO/dCO 2 ) could indicate a decrease of BB fuel carbon oxidation during combustion.Regarding HNO 3 , pyrogenic ammonia (NH 3 ) may have reacted with HNO 3 , yielding ammonium nitrate (NH 4 NO 3 ), which condenses on primary particles.NH 4 NO 3 formation may have occurred during adiabatic cooling of the ascending air mass and/or after ascent.Thermal decomposition of NH 4 NO 3 in the sampling line of the CIMS instrument represents a potential source of gasphase HNO 3 .By contrast to HNO 3 , NO y (a major fraction of which is HNO 3 ) may not be affected by severe sampling line losses since in the NO y instrument HNO 3 is rapidly converted to NO, which is much less sticky than HNO 3 .Furthermore, the transfer line of the NO y instrument is heated to a much higher temperature compared to the CIMS sampling line, which reduces wall losses.
Figure 7 shows two aerosol particle size distributions measured above the MT-plume in background conditions of the free troposphere at 4500 m (528 hPa) and in the MT-plume at 3900 m (645 hPa).The two size distributions are very different with the Aitken mode dominating in the background case and the accumulation mode inside the plume.Above the MTplume the distribution is peaking at 40 nm.In the densest of the MT-plume at 3900 m, the size distribution is peaking at about 250 nm.This resembles the presence of fewer small particles and many more larger particles.The large particles are most likely aged smoke particles, which are substantially larger than typical primary smoke particles (median diameter: about 125 nm Reid et al., 2005).The relatively low particle concentrations below 100 nm particle diameter inside the plume indicate that entrained background aerosol particles must have experienced removal via coagulation with smoke particles, since coagulation with smoke particles becomes more efficient for smaller particles.It also implies that new particle formation is suppressed probably due to scavenging of gaseous sulfuric acid by aged primary smoke particles.The volume and mass concentrations of smoke particles can be inferred from the measured size distribution.Considering particles with diameters of up to 800 nm and assuming a specific weight of smoke particles of 1 g cm −3 , one obtains a smoke particle mass concentration of 6.5 ×10 −11 g cm −3 (at 3900 m, 13 August, 12:04 UTC).These findings also compare very well to the results by de Reus et al. (2001).For HNO 3 , NO, and CO reascent data are markedly larger than descent data.For aerosol particles averaged profiles combining descend and ascend data using the median concentration of data in 100 m altitude bins are presented (no significant differences between descend and ascend observed).

MT-plume
In the following, measured dX/dCO 2 and expected dx/dCO 2 (expected from savanna biomass burning) will be compared.The comparison will focus on the core of the MT-biomass burning plume penetrated by the Falcon at 3900 m around 12:06 UTC.
Molar ratios dX/dCO 2 of excess trace substance X and excess CO 2 (excess means measured value minus atmospheric background value), measured at 3900 m the MT-plume, are given in Table 1.Also given is the mass ratio dTPM/dCO 2 (here TPM denotes total particle matter).Furthermore, for comparison, the corresponding published emission ratios for savanna fires are also given.
Concerning CO, the measured dCO/dCO 2 is almost equal to the molar emission ratio.The most important CO-loss is due to reaction with OH (rate coefficient: about 2 ×10 −13 cm 3 s −1 at p = 500 hPa).Using the above OH eff = 3.0 ×10 5 cm −3 inferred from SO 2 , one obtains a mean CO-lifetime (against OH-reaction) of about 193 d.This implies that, after 10 days, about 95% of the initial CO should still be present, which is consistent with the observation, when uncertainties of the measured dCO and dCO 2 and the emission factors are considered.
The HNO 3 measured in the plume peak is almost equal to the measured NO y .This implies that PAN and NH 4 NO 3 were not very abundant.The measured dNO y /dCO 2 represents about 46% of the molar emission ratio.The missing NO y may have experienced loss by deposition.Probably, NO y losses occurred mainly via HNO 3 .
For the TPM concentration at 3900 m altitude, our inferred value of 6.5 ×10 −11 g cm −3 (see above) implies a mass ratio dTPM/dCO 2 = 4.2 ×10 −3 g g −1 .The ratio of the emission factors for primary smoke particles and CO 2 is 5.1 ×10 −3 g g −1 .Hence it seems that the measured TPM concentration represents about 74% of the expected primary TPM concentration.Therefore, it seems that some loss of smoke particles by deposition and cloud processes may have occurred.
The observed H 2 CO must be of secondary origin since, in clear sky conditions, H 2 CO is rapidly lost by photolysis.This seems to be consistent with dry convection.The 1/e-lifetime is only about 3 h, which is much smaller than the plume age.For example, H 2 CO can be formed by O 2 -reaction of the PAradical.

UT-plume
The upper tropospheric plume (UT-plume) was probably uplifted by wet convection in the north-western Republic of Congo and the Central African Republic about 10 days prior to our measurements.This contrasts the transport of the MTplume, which was uplifted south of the equator between 10 and 23 degrees south, where wet convection is much less probable and much less intense (see Real et al., 2010).
Regarding SO 2 , 14-38% of the expected SO 2 have been observed (expected from the emission ratios), which is smaller compared to the MT-plume.This may indicate that a substantial fraction of the released SO 2 experienced loss by cloud processes.
The observed HNO 3 represents only about 3.4% of the expected NO y .However, at the much lower temperature of the UT-plume, PAN is thermally stable and may contribute about half of the NO y while HNO 3 may contribute the other half.If so, about 6.8% of the expected HNO 3 would have been present.Compared to the lower plume, this fraction of residing NO y is still markedly smaller.This suggests more efficient HNO 3 removal by wet cloud processes.This is, at least qualitatively, consistent with the above mentioned wet convection induced uplift of the UT-plume.
Regarding aerosol particles, dN 4 is about 2 times dN 10 (nv), which indicates that half of the enhanced particles contain non-volatile smoke particle cores and are not new particles formed in the UT by nucleation.This is at least qualitatively consistent with the relatively low total SO 2 (80 pmol mol −1 ), which hardly allows sufficient H 2 SO 4 formation required for H 2 SO 4 -H 2 O nucleation (see for example Fiedler et al., 2009a).

Aerosol model simulations
To investigate the evolution between 4 and 24 August, of the smoke particle size, number concentration, H 2 SO 4 -mass fraction, and the critical (minimum) water vapor super saturation required for smoke particle activation (WSSa), we have made model simulations (AEROFOR model, Pirjola, 1999;Pirjola and Kulmala, 2001;Pirjola et al., 2003).The model assumes smoke particles with an initial uniform diameter of 125 nm and an initial number concentration of 2.2 ×10 5 cm −3 .This number concentration was calculated considering an expected dTPM/dCO 2 = 5.1 ×10 −3 g g −1 , a dTPM = 6.5 ×10 −11 g cm −3 measured on 13 August in the MT-plume at 3900 m, and an initial dTPM calculated from the measured dTPM , a specific weight of 1 g cm −3 , and taking an 1/e-time of 6.4 days for plume dilution.This dilution time was taken from Real et al. (2010).The plume parcel trajectory and meteorological data were taken from the LAGRANTO model.The trajectory is composed of two segments, a back-trajectory segment (4-13 August) and a forward-trajectory segment (13-24 August).
The model treats the following processes: mutual smoke particle coagulation (coagulation with entrained background particles is neglected); OH-induced SO 2 -conversion to H 2 SO 4 ; uptake of H 2 SO 4 (plus associated H 2 O) by smoke particles.The model is tuned to reproduce the SO 2 concentration (1.4 nmol mol −1 ) measured on 13 August at 3900 m.Uptake of H 2 SO 4 and sulfate entrained into the plume from the background atmosphere is neglected.For the period 4-13 August, the OH eff = 6.2 ×10 5 cm −3 , inferred from the observation on 13 August (at 3900 m), of 45% of the expected maximum SO 2 (see above) was considered.The corresponding noon-time OH is about 2.0 ×10 6 cm −3 .For the period tp = 0-24 d (13-24 August), OH was taken from Logan et al. (1981).The model does not consider NH 4 NO 3 .
Figure 8a shows, for the entire simulation period of 20 days (4-24 August), time sequences of the atmospheric pressure and pressure altitude of the air parcel intercepted by the Falcon on 13 August at 3900 m.
Figure 8b shows for the same case as Fig. 8a the outside air temperature T and RH.T varies between about 275 and 297 K. T measured aboard the Falcon (290 K) markedly exceeds the modeled value (280 K).This discrepancy may be due to additional heating via light absorption by soot containing smoke particles.RH varies between about 8 and 50%.The largest local maxima of about 50% are reached during the initial ascend at tp = 2 d and during the final ascend at tp = 20 d.Water vapor supersaturation (RH > 100%), according to the model, was never reached.
Figure 9 shows time sequences of the modeled molecular number concentrations of OH, SO 2 , and gas-phase H 2 SO 4 .Also given is a curve representing an inert plume dilution tracer having the same initial concentration as SO 2 .Both, OH and gas-phase H 2 SO 4 exhibit a pronounced diurnal variation.The SO 2 concentration curve exhibits a weak diurnal variation and decreases with time much more steeply than the plume dilution tracer.This indicates that SO 2 -depletion was preferably due to OH-induced SO 2 -conversion to gas-phase H 2 SO 4 .
Figure 10 shows a time sequence of number concentrations of smoke particles (N sp ) in the plume parcel (left axis of Fig. 10).N sp decreases with increasing plume age tp, due to coagulation and plume dilution.Initially coagulation dominates and later, as N sp has decreased sufficienly, coagulation becomes slow and plume dilution dominates.For tp = 10 days (13 August), the model N sp is 1000 cm −3 , which is smaller than measured N sp (2000 cm −3 ).After 20 days, N sp has decreased to about 200 cm −3 .
Figure 10 additionally shows a time sequence of the smoke particle wet diameter D sp for two cases: without and with binary H 2 SO 4 -H 2 O condensation (right axis of Fig. 10).Without binary H 2 SO 4 -H 2 O condensation, D sp increases to 450 nm (tp = 10 d) and 470 nm (20 d), due to mutual smoke particle coagulation.With binary H 2 SO 4 -H 2 O condensation, as tp increases, D sp increases to 490 nm (tp = 10 d) and 545 nm (20 d), due to mutual smoke particle coagulation plus binary H 2 SO 4 -condensation.For tp = 10 d (13 August), the modeled D sp = 450 nm (without binary H 2 SO 4 -H 2 O condensation) is close to the peak dry D sp (about 400 nm) of the experimental aerosol volume size distribution.
Figure 11 shows a time sequence of the mass concentrations of the primary smoke particle components (EC + OC), and the secondary components H 2 SO 4 , and H 2 O (left axis of Fig. 11).The EC + OC curve decreases by a factor of about 20, due to plume dilution.On day 1, the H 2 SO 4 curve increases steeply and on days 3-6 reaches a maximum of about 6000 ng m −3 .Hereafter until tp = 13 d the H 2 SO 4 curve decreases only moderately, and ultimately it decreases more steeply.The H 2 O mass fraction of smoke particles varies, mostly in response to the variability of RH.
Figure 11 additionally shows a time sequence of the H 2 SO 4 -mass fraction of smoke particles (right axis of Fig. 11).It exhibits a slight diurnal variation but generally increases throughout the simulation period.At tp = 10 d (13 August) it is about 9% and at tp = 24 d (24 August) it is about 14%.

Smoke particle activation
Modeled WSSa as function of modeled diameters and H 2 SO 4 -mass fractions of smoke particles, for six time steps tp (1, 2, 4, 6, 10, 24 days) are given in Fig. 12.Here it is assumed that H 2 SO 4 would have the same effect on WSSa as NH 4 NO 3 (for which the figure was originally plotted, see Seinfeld and Pandis, 1998)  decrease with tp from about 0.3% (tp = 1 d) to about 0.033% (tp = 24 d), red crosses.Also included are values for the sum mass fraction of H 2 SO 4 and NH 4 NO 3 as blue crosses.In this case WSSa decreases to about 0.025% (tp = 24 d).
In comparison, WSS involved in maritime cloud formation are about 0.3-0.8%(cumulus clouds) and about 0.05% (maritime stratiform clouds).Hence, H 2 SO 4 -processed smoke particles contained in the plume parcel under consideration have developed a potential for maritime cumulus cloud formation (after 1 day) and for maritime stratiform cloud formation (after 10 days).
However, the modeled RH (Fig. 8b never exceeded 100% in the plume parcel under consideration.It is conceivable that small fluctuations, which are not considered by the model, may have resulted in small WSS, particularly at tp when the modeled RH was large. The critical (minimum) WSS required for smoke particle activation (WSSa) decreases with increasing smoke particle diameter and increasing mass fraction of soluble material contained in the smoke particle (see Fig. 12).Conceivable smoke particle components, which are particularly efficient in this regard, are the secondary species H 2 SO 4 and NH 4 NO 3 .The hygroscopicity of the semi-volatile organic coating of primary smoke particles is not well known.This organic coating may contain water soluble species as for example salts of organic acids.Organic acids may have experienced conversion to salts by reaction with primary pyrogenic NH 3 .However, it is also conceivable that much of the organic coating is hydrophobic.
Sulfuric acid is formed in the plume via OH-induced conversion of SO 2 (see above).The 1/e-lifetime of SO 2 is determined by the OH-concentration which can be quite vari-  able and is difficult to predict, particularly for the MT-plume where attenuation of solar UV-radiation and complex organic chemistry complicate modeling of OH.If particles and SO 2 would not be removed at different rates, the ratio of the sulfur mass concentration and primary particle mass concentration would remain equal to the ratio of the corresponding mass emission factors (mean value: 0.175/8.3= 0.021) g S/g PSP; PSP denotes primary smoke particles).If the released SO 2 would ultimately be completely converted to H 2 SO 4 , the ratio of the H 2 SO 4 -mass concentration and primary smoke particle mass concentration would be about 0.065 g H 2 SO 4 /g PSP.Considering the ranges of expected emission factors for SO 2 and TPM, one obtains an ultimate ratio ranging from 0.025 to 0.153 g H 2 SO 4 /g PSP.
Nitric acid is formed in the plume by OH-induced conversion of NO 2 .The NO 2 -lifetime against OH-reaction is about 4 days when the above inferred OH eff of 3.0 ×10 5 cm −3 is considered.However, NO x is present not only as NO 2 but also as NO.Therefore the NO x -lifetime against conversion to HNO 3 is larger, depending on the abundance ratio NO 2 /NO x .For example, for an assumed NO 2 /NO x = 0.5, one obtains an NO x -lifetime of about 8 days, which is much smaller than the SO 2 -lifetime.This would imply that after 10 days dNO x /dCO 2 decreased to only about 29% of its initial value.
Hence, HNO 3 was formed first and may have converted NH 3 and ammonium salts to NH 4 NO 3 .However, NH 4 NO 3 may become thermally stable only after sufficient cooling of the plume, after it had ascended to about 3900 m, on 8 August.For example, in the plume at 3900 m, (T = 290 K, and relative humidity RH = 25%), NH 4 NO 3 should be solid and the NH 4 NO 3 dissociation equilibrium constant is about 4.9 nmol mol −1 × nmol mol −1 (for an atmospheric pressure of 630 hPa at 3900 m).This implies that the equilibrium mole fractions for each HNO 3 and NH 3 are about 2.4 nmol mol −1 .In comparison, the measured HNO 3 mole fraction is about 8 nmol mol −1 .The upper limit NH 3 mole fraction expected from the NH 3 emission ratio (see Table 1) is about 4 nmol mol −1 , if removal of (NH 3 + NH 4 ) would have been the same as NO y -removal.This implies that the expected NH 3 exceeds the equilibrium NH 3 by about 2 nmol mol −1 .Therefore, about 2 nmol mol −1 of solid NH 4 NO 3 may have been present, containing about 25% of the total NO y .This is not in conflict with our observations when uncertainties of experimental data and BB emission factors are considered.
Considering the above estimated NH 4 NO 3 mole fraction, the mass fractions of dry smoke particle components would become 7.07% (H 2 SO 4 ), 8.3% (NH 4 NO 3 ), and 15.4% (H 2 SO 4 + NH 4 NO 3 ).If this soluble material (NH 4 NO 3 plus H 2 SO 4 ) would have the same decreasing effect on WSSa as NH 4 NO 3 , WSSa would be lowered to about 0.05%, for a smoke particle with a diameter of 480 nm initially not containing soluble material (see Fig. 12).
However, as suggested by the trajectory model, after 13 August, the plume descended again and T and RH increased somewhat, which tends to lower the NH 4 NO 3 associated with smoke particles.On the other hand, ongoing SO 2conversion to GSA (Fig. 9) tends to increase the H 2 SO 4 -H 2 O mass associated with smoke particles.
Hence, it seems that in the MT-plume, smoke particle processing by H 2 SO 4 and NH 4 NO 3 may have had a marked effect on the smoke particle activation potential.It also seems that, in the aged MT-plume on 19 August, H 2 SO 4 was more important than NH 4 NO 3 in increasing the activation potential of smoke particles.

Summary and conclusions
The main findings of the reported airborne BB plume measurements are: a.An about 10 days old BB plume, located at about 3900-5500 m altitude, has been probed above the eastern Atlantic (Gulf of Guinea).
b.The plume originated from BB fires in the Southern-Hemisphere African savanna belt.
c.The plume was lifted by dry convection and had greatly elevated abundances of gas-phase and particle-phase pollutants.
d.The gases SO 2 (precursor of H 2 SO 4 ) and HNO 3 , which have a potential to mediate smoke particle activation had measured mole fractions of up to 1400 and 9000 pmol mol −1 .e. Our data indicate that a large part of NO y experienced loss probably via HNO 3 by deposition.
f. SO 2 did not experience a marked loss.
g.As the plume was ageing and diluting, SO 2 experienced OH-induced conversion to H 2 SO 4 , which induced rapid binary (H 2 SO 4 -H 2 O)-condensation on smoke particles.
h. H 2 SO 4 condensation, besides coagulation size growth, increased the activation potential of smoke particles.Also NH 4 NO 3 formation may have contributed somewhat to increase the activation potential.
i.After 13 August (day of our measurements), the plume traveled over the Atlantic while descending to 1300 m altitude after 8 days.On 19 August it reached the west coast of south America (French Guyana) and hereafter traveled northward over the Atlantic.j.On 19 August, smoke particles had a potential to become activated already at a very small WSS of only 0.05%, which would allow them to act as CCN in maritime stratiform cloud formation.
k. Another much less polluted BB plume observed at 10.8-11.2km altitude was lifted by wet convection.It had experienced more efficient removal of SO 2 , NO y and particles probably by wet cloud processes.

Figure 1
Figure 1 depicts a MODIS (Moderate Resolution Imaging Spectroradiometer) image of fires in Africa for the period 1-10 August 2006.MODIS is an instrument on the

Fig. 1 .
Fig. 1.Fires in Central Africa detected by MODIS (see text) for Northern Africa between the 1st and the 10th of August 2006.Also shown is the flightpath of the DLR research aircraft Falcon, color coded with flight altitude.

Fig. 1
Fig.1is the Falcon flight path with the flight altitude color coded.To probe the plume, the Falcon took off on 13 August 2006 at Ouagadougou (Burkina Faso, 12.35 • N, −1.51 • W) and flew at 9000-11 000 m altitude in southern direction to the equatorial Atlantic region off the coast of Ghana (western branch of the flight path in Fig.1).Here it dived into the plume to a lowest height of 3900 m where it cruised for about 5 min.Hereafter it climbed out of the plume again and flew back to Ouagadougou (right branch of the flight path in Fig.1).During that dive, the polluted air mass was even visible as a thick brownish layer.When looking downward, surface details were not visible.

Fig. 2 .
Fig. 2. Aerosol Index AI of light absorbing aerosols measured by OMI on Satellite AURA on the 13 August 2006.

Fig. 3 .
Fig. 3. Combination of the LAGRANTO 10-day back-trajectory starting at 13 August 2006 12:00 UTC with the LAGRANTO 10day forward-trajectory starting at 13 August 2006 12:00 UTC.The color bar gives the air trajectory pressure.The red dots mark the northern edge of the fire region as shown in Fig. 1 and the blue area marks the so-called African copper belt.

Fig. 4 .
Fig. 4. Altitude profiles of the measured SO 2 mole fraction and the Temperature (first panel, measurements during the descent of the Falcon in red, during reascent in blue).Altitude profile of the measured water vapor concentration and relative humidity (second panel).
Figure4shows in its left panel the altitude profiles of SO 2 and temperature (both descents in red, ascents in blue).The altitude profile of the SO 2 mole fraction indicates the presence of a SO 2 -rich plume in the mid troposphere (hereafter termed MT plume) with a sharp top at about 5200 m (descent) to 5000 m (ascent) and two altitude regimes with different degrees of SO 2 pollution, including an upper plume regime and a main plume regime (below about 4200 m).The Falcon has spent about 11 min inside the MT-plume, below 5200 m, corresponding to a horizontal distance of about 150 km.On that horizontal length scale the top altitude of the MT plume was quite similar, differing only by 200 m.As the Falcon ascended further, a pronounced local SO 2 maximum of about 90 pmol mol −1 was observed at 12:24 UTC.This indicates the presence of a second much less polluted plume in the UT at about 10 800-11 200 m altitude (hereafter termed UT-plume, Fig.4).Its height extension was only about 400 m which is much less than that of the MT-plume (about 2000 m).In the right panel of Fig.4water vapor concentration (H 2 O) and relative humidity (RH) are shown.In the MTplume, H 2 O and RH are elevated.The absolute water vapor mole fraction reaches up to 7200 µmol mol −1 , in the upper part of the MT-plume in a layer between about 4400 and 5100 m.This indicates upward transport of humid air.Temperature reaches up to 292 K at about 4000 m and RH reaches up to 64%, in the upper part of the MT-plume, at 5000 m.Figure5, first and second panel, show the time series of measured trace gases (plotted are CO, CO 2 and O 3 , NO y , NO, HNO 3 , and H 2 CO).CO 2 , which is widely used as biomass burning marker, is strongly enhanced inside the MTplume.All trace gases are also markedly increased in the MT-plume.Within the MT-plume, at constant altitude of the Falcon of 3900 m, the trace gases HNO 3 , NO and CO in-

Fig. 5 .
Fig. 5. Time sequences of measured atmospheric trace gases and particles.Panel 1: CO, CO 2 and O 3 mole fractions.Panel 2: NO, NO y , H 2 CO and HNO 3 mole fractions.Panel 3: SO 2 mole fraction and flight altitude.Panel 4: Number concentration of aerosol particles possessing diameters from 160 to 270 nm and from 270 to 1000 nm.Also given is the outside air temperature T. Between 11:57 and 12:08 UTC the pollution plume is detected in all trace gases and in the particles.

Fig. 6 .
Fig. 6.Altitude profiles of SO 2 , HNO 3 , NO, NO y , O 3 , CO, CO 2 , HCHO, and aerosol particle number concentrations N x where the index x denotes the size range referred to in nm (N 4−1500 , N 10−1500 (non-volatile particles), N 160−270 , N 270−1000 , N 500−1000 and N 1000−5000 ).For HNO 3 , NO, and CO reascent data are markedly larger than descent data.For aerosol particles averaged profiles combining descend and ascend data using the median concentration of data in 100 m altitude bins are presented (no significant differences between descend and ascend observed).

Fig. 7 .
Fig. 7. Aerosol particle size distributions measured (a) above the MT-plume at 4500 m (528 hPa), and (b) in the MT-plume at 3900 m (645 hPa) altitude.Different symbols represent data points of different instruments (see legend).The line as well as the lower and upper end of the grey shaded area represent log-normal fits to the data (averaged, minimum, maximum concentrations within the integration time period).
Fig. 8. (a) Time sequences of the atmospheric pressure and pressure altitude of the air parcel intercepted by the Falcon on 13 August at 3900 m for the entire simulation period of 20 days (4-24 August).(b) Time sequences of temperature and relative humidity.

Fig. 9 .
Fig. 9. Time sequences of the modeled molecular number concentrations of OH, SO 2 and gas-phase H 2 SO 4 .Also given is a curve representing an inert plume dilution tracer having the same initial concentration as SO 2 (SO 2 only dilution).

Fig. 10 .
Fig. 10.Time sequence of the modeled number concentrations of smoke particles in the plume parcel, with and without considering condensation (left axis).Time sequence of the modeled smoke particle diameter D sp for two cases: without and with binary H 2 SO 4 −H 2 O condensation (right axis).

Fig. 11 .
Fig. 11.Time sequence of the modeled mass concentrations of the primary smoke particle components (EC+OC), and the secondary components H 2 SO 4 and H 2 O (left axis).Time sequence of the modeled H 2 SO 4 wet and dry mass fraction of smoke particles (right axis).

Table 1 .
Emission factors Ex, molar emission ratios Ex/ECO 2 (adopted from

Table 2 .
Compilation of gas phase instruments deployed on the Falcon during AMMA.Also compiled are detection limits and uncertainties.