3.1. Aged Smoke
 Reid et al. provides a detailed explanation of the formation of aged biomass burning smoke in the Amazon Basin. After being emitted, the smoke particles disperse and have the potential to be rapidly transported into the lower atmosphere up to the strong trade wind inversion at a height of about 3–4 km as a result of the high air temperatures during emission. Smoke from hundreds of fires mix with biogenic emissions from forests and suspended soil particles (and potentially with urban haze). During transport, smoke undergoes photochemical transformations, gas-to-particle conversion, and particle coagulation. Smoke can be entrained into clouds where increased efficiencies of specific chemical reactions may accelerate the growth of the smoke particles.
 Based on their observations during SCAR-B (Smoke, Clouds and Radiation - Brazil),Reid et al. found that condensation and gas-to-particle conversion of inorganic and organic vapors increase the aerosol mass by 20%–45%. 30%–50% of this mass growth likely occurs in the first few hours. The remaining mass growth is probably associated with photochemical and cloud-processing mechanisms operating over several days. After three days, most of the condensation and gas-to-particle conversion has likely been taken place [Reid et al., 1998]. Coagulation is then left to be the only significant particle growth mechanism over the next days of long-range transport.
 Müller et al. [2007a]investigated the growth of biomass burning particles as a function of travel time. They observed that the surface-area-weighted radius (effective radius) increases from values of 0.1μm (1 day after emission), 0.15–0.25 μm (2–4 days after the emission) to values of 0.3–0.4 μm after 10–20 days of travel time. As shown by lidar observations [Mattis et al., 2003] and subsequent model studies [Damoah et al., 2004], fire smoke can survive over weeks in the free troposphere before it is removed by washout processes.
 The composition of the aerosol in the Amazon Basin can be divided into five possible components according to Reid et al. : primary smoke products, secondary smoke products, other anthropogenic materials, biogenic materials, and soils. The relative contribution of these components to the aerosol mass loading is highly variable. Continuously occurring changes in the particle size distributions and compositions during aging have a large impact on the optical properties of the aerosol. Different burning types (flaming and smoldering fires) increase the complexity of observed microphysical, chemical, and optical properties of smoke plumes.
 Our first case study deals with aged smoke. The lidar observations on 10 and 11 September 2008 are shown in Figure 3. A dense aerosol layer extended up to about 4 km. Cirrus occurred in addition, mainly in the upper troposphere. Fog (in the lowermost few hundred meters) occurred and prohibited high quality lidar observations from 0000–0030 UTC (2000–2030 local time, LT). Fog formation also attenuated the laser beam significantly between 0630 and 0700 UTC. The evolution of low clouds around 2 km height started after 0700 UTC on 11 September 2008.
Figure 3. Aerosol layering observed from 10 September 2008, 2130 UTC (1930 LT) to 11 September 2008, 1500 UTC (1100 LT). The range-corrected signal at 1064 nm wavelength is shown. White features indicate low level clouds (around 2 km height agl) and ice clouds (mostly between 12 km height and the tropopause above 16 km height). The red box indicates the signal averaging period for an in-depth study of particle optical and microphysical properties.
Download figure to PowerPoint
 According to AERONET Sun photometer observations in the late afternoon on 10 September 2008 (2039 UTC), the total and fine-mode 500 nm AOD were 0.45 and 0.43 (fine-mode fraction of 95%). The 500 nm AOD increased to 0.6 in the morning of 11 September 2008 (1223 UTC). Photometer-derived Ångström exponents around 1.2 and effective radii of 0.23–0.26μm on the late afternoon on 10 September 2008 are indicative for aged smoke [Reid et al., 1998].
 Figure 4 (right) shows the 550 nm AOD over the Amazon Basin retrieved from MODIS (Moderate Resolution Imaging Spectroradiometer) [Remer et al., 2005] observations on 10–11 September 2008. According to the satellite measurements, an aerosol plume was located southeast of the lidar site. AOD values up to 0.74 in southern Amazonia and of 0.5 for the Manaus region, respectively, were found.
Figure 4. (left) Fire counts (orange dots) as derived by FIRMS for 7–10 September 2008 and HYSPLIT backward trajectories in 6-h steps (indicated by symbols) for the arrival heights of 1500 (blue), 3500 (red), and 5000 m agl (green) for 11 September 2008, 0100 UTC. (right) MODIS AQUA AOD composite (550 nm) for 10–11 September. The red star indicates the lidar site.
Download figure to PowerPoint
 A very pronounced fire activity in the south, southeast, and east of the lidar site obviously caused this aerosol plume over Amazonia (see Figure 4, left). Fire counts for 7–10 September 2008 as obtained from MODIS measurements (via Fire Information for Resource Management System (FIRMS) at University of Maryland [Giglio et al., 2003]) and 3-day backward trajectories for the arrival heights of 1500, 3500, and 5000 m from HYSPLIT (Hybrid Single Particle Lagrangian Integrated Trajectory Model,http://ready.arl.noaa.gov/HYSPLIT.php) are shown in Figure 4 [Draxler and Hess, 1998; Draxler et al., 2009]. Meteorological fields from the archived model assimilation data sets of GDAS (NCEP Global Data Assimilation System) were used. The backward trajectories indicate an air-mass flow from easterly directions. The air masses crossed fire-active regions 1–2 days before the arrival at the lidar site. Freshly emitted smoke was added to the obviously already aged biomass burning aerosol (regional haze).
 The vertical profiles of the optical and microphysical properties of the smoke aerosol as observed with our lidar are presented in Figure 5 for the observation period between 0100 and 0200 UTC (indicated by a red box in Figure 3). Particle backscatter and extinction coefficients for the transmitted laser wavelengths, respective particle lidar ratios, Ångström exponents, effective radii, and SSA values (532 nm) determined for this one-hour period are shown. The vertical profiles of relative humidity and potential temperature derived from observations with radiosonde launched at the Manaus military airport at 11 September 2008, 0000 UTC, are given in addition.
Figure 5. Vertical profiles of particle backscatter coefficient, extinction coefficient, and lidar ratio for several wavelengths, Ångström exponents, effective radius (reff), and single-scattering albedo (SSA) observed on 11 September 2008, 0100 UTC–0200 UTC (2100–2200 LT). The 532 nm AOD is 0.44. Potential temperature (Tpot) and relative-humidity (RH) profiles were measured with Manaus radiosonde launched on 11 September 2008, 0000 UTC. Different aerosol layer heights (AL top,HAOD95, Haer, ML top) are indicated by horizontal lines in the backscatter panel. Before the computation of the optical properties, lidar signals are vertically smoothed with window lengths of 270 m (backscatter), 750 m (extinction), and 990 m (lidar ratio). Layer mean values of effective radius and SSA are determined by inversion. Vertical bars indicate the layer depth.
Download figure to PowerPoint
 AL top height, HAOD95, and the optical-depth-related scale heightHaer are 4.5, 3.9, and 2.1 km, respectively. The maximum mixing layer height ML top was about 1.6 km on 10 September 2008. AL top coincides with a strong temperature inversion (trade wind inversion layer). A moist atmosphere with relative humidities of 60%–80% within the lowermost 2.5 km of the troposphere, and 40%–50% from 2.5–4.5 km height was observed. 532 nm particle extinction coefficients were 150–200 Mm−1 for heights <2.5 km in the moist air, and 100–150 Mm−1 in the drier air. The similarity of the relative humidity and the extinction profiles indicate water uptake by the particles. The extinction coefficient for Amazonian smoke roughly increases by a factor of 1.5 when the relative humidity increases from 40% to 80% [Rissler et al., 2006]. All in all, an almost vertically homogeneous haze layer was observed up to 4 km height. Ångström exponents are of the order of 1 and are thus in agreement with the Sun photometer observations.
 It is worth to mention that the observed particle depolarization ratio [Baars et al., 2011] was always very low (<0.03) throughout the dry season. This observation corroborates the assumption that aged water-containing biomass burning particles can be regarded to be spherical.
 The particle effective radius of about 0.3 μm in the main and humid aerosol layer below 2.5 km is roughly a factor of 2 larger than respective values found by Reid et al.  for dry Amazonian smoke particles. Water uptake effects are responsible for these large effective radii on 11 September 2008. The effective radius decreases with height and is 0.2 μm in the drier air in upper part of the smoke layer.
 The lidar-derived SSA of the water-containing smoke particles of around 0.9 in the main layer below 2.5 km height are consistent with the values found byReid et al.  for dry smoke particles (0.8–0.9 for 550 nm) in southeastern Amazonia. The mass fraction of black carbon in the Amazonian smoke aerosol is in the range of 5%–10% [Reid et al., 1998, 2005], but may vary from 2%–30% [Reid and Hobbs, 1998; Reid et al., 2005]. Tesche et al.  presented SSAs of, on average, even <0.8 (at 532 nm) for highly absorbing African smoke (at relative humidities below 60% in the smoke layers). Dubovik et al. reported AERONET-photometer-derived values from 0.9–0.94 for Amazon forest-fire smoke and 0.86–0.92 for South American savanna smoke (at ambient humidity conditions).
 In the drier layer between 3 and 4 km in Figure 5, the SSA slightly increased to values around 0.93. Different burning characteristics (smoldering versus flaming fires), differences in the burning material and thus composition of the smoke particles, and transport time may be responsible for the differences in the smoke optical properties, effective radius, and single-scattering albedo observed with the lidar above and below 2.5 km height.
 Finally, we estimated the mass-specific extinction coefficients and smoke mass concentrations based on combined photometer-lidar observations [Ansmann et al., 2012]. In the retrieval, a density of the smoke particles of 1.35 g/cm3 is assumed [Reid and Hobbs, 1998; Reid et al., 2005]. The specific extinction coefficients are around 4 m2/g (AERONET photometer, evening of 10 September 2008) and 3.5 m2/g (lidar, moist layer below 2.5 km height) and 4.5 m2/g (lidar, dry layer above 2.5 km height). With increasing water content (and thus decreasing particle density toward 1 g/cm3) the mass-specific extinction coefficient increase (e.g., toward 6 instead of 4.5 m2/g). Reid derived mass-specific extinction coefficients of 4 ± 1 m2/g (at 550 nm) for dry Amazonian smoke particles. By using a specific extinction coefficient of 4 m2/g we obtain particle mass concentrations of 30–40 μg/m3 in the main part of the smoke haze layer below 2.5 km height, and a value around 15 μg/m3 for the dry layer from 3–4 km height.
3.2. Young Smoke
 A case dominated by freshly emitted smoke was observed in early evening of 15 August 2008 (1835–1935 LT). Lidar profiles of optical and microphysical properties are shown in Figure 6. Reid et al.  mentioned that smoke emissions in Brazil have a strong diurnal cycle. Fires are generally ignited in the late morning through late afternoon. Thus haze sampled in the early evening are most likely to contain a large fraction of young smoke.
Figure 6. Same as Figure 5 except for 15 August 2008, 2235–2335 UTC (1835–1935 LT). AOD (532 nm) is 0.15. Potential temperature (Tpot) and relative humidity (RH) profiles were measured with Manaus radiosonde launched on 16 August 2008, 0000 UTC.
Download figure to PowerPoint
 The optical properties show a distinct layering of particles. AL top was close to 4.5 km and the optical-depth scale heightHaer at about 1450 m, almost coinciding with the maximum ML top in the afternoon of 1600 m. 532 nm particle extinction coefficients ranged from 20–120 Mm−1 in the lowermost 3 km of the atmosphere. The 532 nm optical depth was 0.15. AERONET photometer observations are not available for this case because of persistent cirrus layers.
 The Ångström exponents were significantly higher than on 11 September 2008 (aged smoke case) with values of 1.5–2 for the wavelength range from 355–532 nm. Correspondingly, the effective radius was small with values around 0.13 μm. If we take the water-uptake effect into account (relative humidities ranged from 60–90% in the lowermost 3.5 km), the dry particle effective radius was certainly clearly below 0.1μm. According to Reid et al. , the high Ångström exponents of 1.5–2 and the very low effective radii point to freshly emitted smoke.
 The lidar ratios showed surprisingly low values for fresh smoke. We again expected highly absorbing particles and thus values >70 sr. The lidar ratio increases not only with increasing particle absorption but also with decreasing particle size. Such low values of 30–60 sr together with the high Ångström exponents indicate weakly absorbing particles. The negligible wavelength dependence of the lidar ratio is another characteristic for fresh smoke. The reason for these unusually low lidar ratios remains unclear. However, according to Reid and Hobbs  and Reid et al. , the black carbon content can vary from 2%–30%. Müller et al.  presented statistics for Canadian and Siberian forest fire smoke (after travel times of >6 days) and also found lidar ratios spanning a large range from 30 to 90 sr. In agreement with the rather low lidar ratios, the SSA is high with values of 0.92–0.95 (see Figure 6, bottom).
 O'Neill et al. presents a Sun photometer study of aerosol properties at ambient conditions of boreal forest fires in western Canada. Several photometers were close to the fire sources (30–600 km), and others far way (>2000 km). For the small distances, the smoke-related fine-mode Ångström, effective radii, and single-scattering albedo, were 1.5–2.5, 0.13–0.17μm, and mostly >0.95, respectively. For the large distances (aged smoke), they found lower Ångström exponents (1–1.5) and lower single scattering albedos (mostly <0.95). The effective radii were similar for the both data sets. These findings for boreal forest fires are qualitatively in good agreement with our lidar observations of aged and young smoke.
3.3. Pristine Condition in the Wet Season
 The main results of the lidar observations performed during the wet season are discussed by Baars et al. . For the first time, clear and unambiguous indications for a significant long-range transport of African biomass burning aerosol to the Amazon Basin were documented [Ansmann et al., 2009; Baars et al., 2011].
 Here, we present one case obtained for rather pristine conditions to emphasize the strong contrast between natural aerosol conditions and man-made haze and smog situations in the Amazon Basin.Figure 7shows the aerosol layering measured with lidar on 23 and 24 April 2008. Rain associated with strong washout effects was not observed during the complete lidar measurement session starting at 1800 UTC (1400 LT). In the beginning of the analyzed time period (at 2200 UTC), low-level clouds were present at around 1 km agl which prohibited the penetration of the laser beam to higher altitudes. At about 2300 UTC, the low-level clouds dissolved and the vertical extent of the aerosol layer of 2 km became visible to the lidar.
Figure 7. Temporal evolution of aerosol layering on 23 April 2008, 2200–2400 UTC (1800–2000 LT), in terms of the range-corrected 1064 nm signal. The mean 532 nm AOD for the 2300–0000 UTC period was estimated to be 0.019.
Download figure to PowerPoint
 The vertical profiles of the particle backscatter coefficient at 532 and 1064 nm and the respective Ångström exponent for the cloud-free period after 2330 UTC are shown inFigure 8. One aerosol layer near to the surface and a second aerosol layer centered at around 1 km height were observed. In the higher layer the low level clouds occurred. The particle backscatter coefficients multiplied with a lidar ratio of 60 ± 20 sr provide estimates for the 532 nm extinction coefficients. Values of 5–15 Mm−1 are rather low and indicate pristine conditions. The 532 nm AOD of 0.019 ± 0.008 was estimated from the extinction profile. The Ångström exponents of around 1.5 are typical for accumulation mode particles. Because of instrumental problems, no information from the UV channels was available on that day.
Figure 8. Vertical profiles of the 532 nm (green) and 1064 nm particle backscatter coefficient (red) and corresponding Ångström exponent measured on 23 April 2008, 2330–0000 UTC (1930–2000 LT). Extinction coefficient values (upper scale in the left panel) are obtained by multiplying the backscatter coefficient with a lidar ratio of 60 sr. The 532 nm AOD is estimated from the backscatter profile to be 0.019. Temperature (T) and relative humidity (RH) profile were measured with Manaus radiosonde on 24 April 2008, 0000 UTC. Different layer heights are indicated by horizontal lines in left panel.
Download figure to PowerPoint
 The maximum ML top on that day was calculated to be 760 m. Thus, it was slightly higher than the observed top of the first aerosol layer. The AOD scale height Haer was 960 m. The AL top and the height HAOD95 coincide at 1750 m.
 The low observed AOD value is even more remarkable when taking the high relative humidity of >80%–90% into account. Natural Amazonian aerosol was classified as moderately hygroscopic in previous experiments [Zhou et al., 2002; Rissler et al., 2004] so that hygroscopic growth should enhance the aerosol light scattering. In summary, one can conclude that the observed aerosol conditions on 23 April 2008 with an AOD of 0.02 represent background or natural aerosol conditions over the Amazon rain forest.
 Such pristine conditions with an AOD < 0.05 at 532 nm were observed in about 50% out of all measurement cases during the wet season. Aerosol was then trapped in the lowermost 2.5 km of the troposphere. However, in about one third of all measurements advection of smoke and dust aerosol from Africa was observed as discussed by Baars et al. . AODs ranged then typically from 0.07–0.25.