Journal of Geophysical Research: Atmospheres

Vertically resolved separation of dust and smoke over Cape Verde using multiwavelength Raman and polarization lidars during Saharan Mineral Dust Experiment 2008



[1] Multiwavelength aerosol Raman lidar in combination with polarization lidar at Praia (14.9°N, 23.5°W), Cape Verde, is used to separate the optical properties of desert dust and biomass burning particles as a function of height in the mixed dust and smoke plumes over the tropical North Atlantic west of the African continent. The advanced lidar method furthermore permits the derivation of the single-scattering albedo and microphysical properties of the African biomass burning smoke. A case study is presented to discuss the potential of the technique. The observations were performed during the Saharan Mineral Dust Experiment (SAMUM) in January and February 2008. The height-resolved lidar results are compared with column-integrated products obtained with Aerosol Robotic Network Sun photometer. Good agreement is found. Furthermore, the findings are compared with lidar and aircraft observations recently performed in western Africa and with our previous lidar observations taken in tropical and subtropical regions of southern and eastern Asia. The SAMUM case study represents typical aerosol layering conditions in the tropical outflow regime of western Africa during winter season. Above a dense desert dust layer (with an optical depth of about 0.25 at 532 nm) which reached to 1500 m, a lofted layer consisting of desert dust (0.08 optical depth) and biomass burning smoke (0.24 optical depth) extended from 1500 to 5000 m height. Extinction values were 20 ± 10 Mm−1 (desert dust) and 20–80 Mm−1 (smoke) in the lofted plume. The smoke extinction-to-backscatter ratios were rather high, with values up to more than 100 sr, effective radii ranged from 0.15 to 0.35 μm, and the smoke single-scattering albedo was partly below 0.7.

1. Introduction

[2] Mineral (desert) dust and biomass burning smoke are major components of the atmospheric aerosol system [Prospero et al., 2002; Andreae and Merlet, 2001]. Africa is one of the largest sources of these aerosols. Aerosol outbreaks continuously transport Saharan dust, urban haze, and fire smoke over the tropical Atlantic Ocean toward South and North America [Prospero et al., 1981; Andreae, 1983]. The Saharan dust layer and the western African smoke plumes merge in the winter season and show a complex vertical layering. A height-resolved characterization of the African aerosol column is required to improve our understanding of the mixing of these climate-relevant aerosol components, the long range transport of the mixed plumes over the tropical North Atlantic, and the impact of the aerosol on regional climate [Johnson et al., 2008a; Myhre et al., 2008]. Such observations also provide valuable scenarios for the validation of atmospheric models applied to simulate the long-range transport, the radiative impact, as well as the feedbacks between the different atmospheric processes [Heinold et al., 2009].

[3] The lidar observations presented here were performed in the Atlantic Ocean west of Africa in the framework of the SAharan Mineral dUst experiMent (SAMUM) [Heintzenberg, 2009]. SAMUM can be regarded as an activity to characterize the western African aerosol over the tropical Atlantic at the beginning of the intercontinental long-range transport. In this respect, SAMUM is complementary to the efforts conducted in western Africa in the framework of the African Monsoon Multidisciplinary Analysis (AMMA) and the Dust And Biomass-burning EXperiment (DABEX) projects [Haywood et al., 2008]. Another important issue of our SAMUM lidar study with focus on dust-smoke separation addresses the interpretation of the well-known time series of AErosol RObotic NETwork (AERONET) [Holben et al., 1998] Sun photometer observations at Sal, Cape Verde. These column-integrated extinction measurements provide unique information about the spectral aerosol optical depth over the eastern tropical Atlantic for 15 years but lack a proper interpretation in terms of vertical aerosol layering, especially during winter and spring season [Holben et al., 2001].

[4] Several remote sensing techniques have been developed to separate the aerosol profiles of particle backscattering and extinction related to the fine-mode aerosol (biomass burning smoke, urban haze) and the coarse mode fraction (sea salt, desert dust) [Kaufman et al., 2003a; Léon et al., 2003; Sugimoto et al., 2003; Shimizu et al., 2004; Sugimoto and Lee, 2006; Nishizawa et al., 2007; Huneeus and Boucher, 2007]. The proposed algorithms of Kaufman et al. [2003a], Léon et al. [2003], Nishizawa et al. [2007], and Huneeus and Boucher [2007] are based on two-wavelength backscatter lidar observations and predetermined bimodal lognormal aerosol models consisting of one fine mode and one coarse mode with fixed size distributions and refractive index characteristics but varying concentration throughout the vertical column. The main task in the data analysis after Kaufman et al. [2003a], Léon et al. [2003], and Huneeus and Boucher [2007] is to determine a certain aerosol model (out of 20) which is in best agreement with the measured lidar profiles (at 532 and 1064 nm) and the spectral radiances directly measured with MODIS (MODerate resolution Imaging Spectroradiometer) from 0.55 to 2.1 μm [Kaufman et al., 2003a, 2003b] or with the particle optical depths at 530 and 1060 nm and the effective particle radius retrieved from the MODIS data [Léon et al., 2003]. The scheme after Kaufman et al. [2003a] was designed to analyze data measured with the spaceborne dual-wavelength (532 and 1064 nm) lidar aboard CALIPSO (Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation) which flies in formation with the MODIS instrument. The final products are height profiles of fine-mode- and coarse-mode-related backscatter and extinction coefficients. Both Kaufman et al. [2003a] and Léon et al. [2003] applied their technique to a Saharan dust plume measured close to Cape Verde with airborne lidar in September 2000 in the framework of the SHADE (SaHAran Dust Experiment) [Tanré et al., 2003] campaign. Kaufman et al. [2003b] also analyzed lidar observation of a mixture of dust and smoke over the South Atlantic off the coast of Namibia.

[5] Nishizawa et al. [2007] use two-wavelength lidar observation at 532 and 1064 nm as well and, in addition, the measured particle depolarization ratio at 532 nm to specify the aerosol type of the coarse mode (sea salt or desert dust). According to their forward algorithm, height profiles of the extinction coefficients at 532 and 1064 nm are computed using the assumed two aerosol models consisting of a fine mode (water-soluble urban haze) and a coarse mode (sea salt or desert dust). Again, size distribution and refractive index properties are assumed to be height-independent. The optimum solution matches the observed lidar profiles. The particle depolarization ratio (definition in section 2) is used to select the most appropriate model. If the particle depolarization ratio is larger than 0.1, the aerosol model with the desert-dust coarse mode is assumed to be valid, otherwise the aerosol model containing the sea-salt coarse mode is used to compute the final products, i.e., the fine-mode- and coarse-mode-related backscatter and extinction profiles.

[6] Sugimoto et al. [2003] and Shimizu et al. [2004] make use of the measured 532-nm particle depolarization ratio in a more quantitative way. By assuming an Asian dust depolarization ratio of 0.35 and a value of 0.02–0.05 for polluted eastern Asian air (during periods free of desert dust) the contribution of the dust backscattering to the volume backscatter coefficient of particles as a function of height is obtained.

[7] Sugimoto and Lee [2006] use measurements of the particle depolarization ratio and the volume backscatter coefficient, both at 532 and 1064 nm. By assuming dust depolarization ratios at 532 and 1064 nm, for example, of 0.35 at both wavelengths, the dust and nondust backscatter coefficients and the spectral dependence of the backscatter coefficient (backscatter-related Ångström exponent) for both dust and nondust aerosol components can be retrieved.

[8] Our approach is similar to the techniques presented by Sugimoto et al. [2003] and Shimizu et al. [2004]. We use the 532-nm particle depolarization ratio to separate dust and nondust contributions to the 532-nm backscatter coefficient in the first step. Our technique is unique because, for the first time a multiwavelength Raman lidar, providing us with particle backscatter coefficients at 355, 532, and 1064 nm and particle extinction coefficients at 355 and 532 nm, is employed. In contrast to a Raman lidar, elastic-backscatter lidars as used in all of the mentioned approaches, only allow for the determination of backscatter coefficients. We will show that a separation of nondust (i.e., smoke) and dust particle extinction coefficients is possible, and furthermore that a set of smoke backscatter and extinction coefficients at 355, 532, and 1064 nm and at 355 and 532 nm, respectively, can be separated with sufficient accuracy so that a trustworthy inversion of these data in terms of microphysical parameters such as the effective radius and the single-scattering albedo (scattering-to-extinction ratio) is possible. The method is of value, since it can be applied to observations of next-generation space lidars such as high spectral resolution lidars measuring particle extinction at 355 and 532 nm, particle backscatter at 355, 532, and 1064 nm and the depolarization ratio. The prospects of such lidars are currently assessed by space agencies.

[9] Two campaigns were conducted within the SAMUM project. The field site of the first campaign (SAMUM-1, May and June 2006) was close to the Sahara so that the optical properties of pure desert dust could be measured [Tesche et al., 2009; Freudenthaler et al., 2009]. The second field campaign (SAMUM-2) took place in January and February 2008 (phase A) and in May and June 2008 (phase B) at Cape Verde in the outflow regime of the African continent. The method introduced here makes use of the pure-dust (SAMUM-1) and dust-smoke data sets (SAMUM-2, phase A) to separate the contributions of smoke and dust to the observed optical properties over Cape Verde in the winter of 2008.

[10] In section 2, the SAMUM lidars are briefly described. The methodology applied to separate dust and smoke plumes is outlined in section 3. In section 4, the data analysis scheme is applied to a measurement case, and the results are compared with AERONET Sun photometer observations at Praia, Cape Verde. The SAMUM-2 results are further compared with findings from the AMMA 2006 campaign [Johnson et al., 2008a; Osborne et al., 2008; Heese and Wiegner, 2008; Pelon et al., 2008] as well as with our previous observations performed in South Asia during the Indian Ocean Experiment (INDOEX) [Ansmann et al., 2000; Franke et al., 2003; Müller et al., 2003] and the Program of Regional Integrated Experiments of Air Quality over Pearl River Delta (PRD) 2004 (PRIDE-PRD2004) [Ansmann et al., 2005; Müller et al., 2006; Tesche et al., 2008]. A summary and concluding remarks are given in section 5.

2. Experiment

[11] The main goal of SAMUM is the vertically resolved characterization of microphysical, chemical, optical, and radiative properties of pure desert dust and mixed dust-smoke plumes. The observational data of both SAMUM campaigns are used to support atmospheric modeling of desert dust transport and the direct impact of dust on climate, and to improve radiative transfer codes for nonspherical desert dust particles. The optical properties of aerosols in the tropospheric column were monitored with three Raman lidars and an airborne high spectral resolution lidar [Althausen et al., 2000; Tesche et al., 2009; Freudenthaler et al., 2009; Esselborn et al., 2009; Heese et al., 2009]. All lidars also permit the measurement of the depolarization ratio to identify dust layers. The ground-based systems were deployed at Praia Airport (14.9°N, 23.5°W), Santiago island, Cape Verde, during the SAMUM-2 campaign, which lasted from 15 January to 14 February 2008. In addition, a Doppler wind lidar [Engelmann et al., 2008], several Sun photometers, and a radiosonde station (Vaisala RS80, RS92) were run at the Praia site. An AERONET Sun photometer (travel version) was kindly provided by NASA. The SAMUM photometers are described by Toledano et al. [2009].

[12] Our study is based on observations with the multiwavelength Raman lidar BERTHA (Backscatter Extinction lidar-Ratio Temperature Humidity lidar Apparatus) [Althausen et al., 2000; Tesche et al., 2009] and the polarization lidar MULIS (Multiwavelength Lidar System) [Freudenthaler et al., 2009]. BERTHA measures profiles of the elastic backscatter signal Pλ(z) at wavelengths λ of 355, 400, 532, 710, 800, and 1064 nm, the depolarization ratio at 710 nm, and nitrogen Raman signals image at 387 and 607 nm. We use these measured signals to compute the volume backscatter coefficients βλ(z) at 355, 532, and 1064 nm and the volume extinction coefficients σλ(z) at 355 and 532 nm [Ansmann et al., 1992; Ansmann and Müller, 2005], listed in Table 1. The relative error of the volume backscatter coefficients range from 5% to 15%. The relative uncertainty in the extinction coefficients is about 10% and 10–30% for extinction values above and below 100 Mm−1, respectively.

Table 1. Basic Particle Optical Properties, Required Input, and Retrieval Methodsa
  • a

    Required input is signal profiles or computed optical values. Retrieval methods are Raman lidar method (RL) [Ansmann et al., 1992; Tesche et al., 2009] and polarization lidar method (PL) [Murayama et al., 1999; Freudenthaler et al., 2009]. Subscripts indicate wavelength, and elastic backscatter signals (without polarization index) are defined as P = P + equation image. The Rayleigh depolarization ratio δ532,m is obtained from radiosonde profiles (RS) of temperature T(z) and pressure p(z) considering instrumental characteristics [Bucholtz, 1995].

β355,pP355(z), P387(z)RL
β532,pP532(z), P607(z)RL
β1064,pP1064(z), P607(z)RL
S355,pσ355,p, β355,p 
S532,pσ532,p, β532,p 
åβ,p(355, 532)β355,p, β532,p 
åβ,p(532, 1064)β532,p, β1064,p 
åσ,p(355, 532)σ355,p, σ532,p 
δ532,vP532(z), equation imagePL
δ532,mT(z), p(z)RS
δ532,pδ532,v, δ532,m, β532,p 

[13] We use measurements of the volume depolarization ratio at 532 nm with MULIS in our study (see Table 1). This system offers the highest quality of all depolarization measurements performed in the framework of SAMUM. Since the particle depolarization ratio is the key parameter in separating dust and smoke profiles, the retrieval of this quantity is briefly explained. Following the notation of Cairo et al. [1999], the basic lidar equation, which permits us to compute the number of detected photons in each of the measurement channels for different laser wavelengths λ as well as for different polarization states with respect to the plane of polarization of the emitted laser light, indicated by ⊥ or ∥, can be written as

equation image

P0 is the number of transmitted photons, c denotes the velocity of light, and z is the height above ground in case of a vertically pointing ground-based lidar. C describes the overall system efficiency of a given channel. O(z) specifies the overlap of the laser beam with the receiver field of view, which changes from zero at the lidar to one (complete overlap) several hundred meters from the lidar in the case of MULIS; equation image and β(z) are the parallel- and cross-polarized components of the backscatter coefficient β, respectively. The transmission term Tλ(z) = exp{−equation image[σλ,p(z′) + σλ,m(z′)]dz′} is a function of the particle extinction coefficient σλ,p(z) and the Rayleigh extinction coefficient σλ,m(z). equation image is the sum of the contributions of particles,equation image, and of molecules, equation image.

[14] The volume depolarization ratio is defined as [Sassen, 2005]

equation image

For simplicity we drop the dependence on range z in the following derivations. Equation (2) assumes equal overlap characteristics for both channels so that Oλ = equation image.

[15] In case of the MULIS measurements, the volume depolarization ratio is calculated from the measured ratio of the cross polarized to the parallel polarized signal according to the method described by Freudenthaler et al. [2009]. Regular calibration measurements of the well-characterized instrument allow for a highly accurate determination of this quantity without assuming a reference value in the atmosphere. The relative uncertainty of the retrieval of the linear depolarization ratio is estimated to be about 5%.

[16] By using the computed profile of the particle backscatter coefficient βλ,p the particle depolarization ratio

equation image

is determined [Freudenthaler et al., 2009]. The Rayleigh depolarization ratio

equation image

describes the polarization properties of molecules and can be calculated from radiosonde data according to Bucholtz [1995].

3. Method to Separate Dust and Nondust Contributions

[17] The procedure to separate dust and smoke (or to be more general nondust) profiles of backscattering starts from the equation for the volume depolarization ratio written as follows:

equation image

In the first step the contribution of particles and molecules to the volume depolarization ratio are separated using the relationships

equation image
equation image
equation image

The index x = m or p denotes the contribution of molecules or particles, respectively. We replace equation image, equation image, βλ,m, and βλ,p in equation (5) by respective expressions after equations (7) and (8) and obtain

equation image

Simple conversion yields

equation image

and after further rearrangements we obtain

equation image

In the second step, the same approach (from equation (5) to equation (10)) can be repeated to separate mineral dust and nondust contributions. Terms describing the optical properties for the volume (index v), particles (index p), and molecules (index m) in the first step, are now replaced by terms for particles (index p), dust (index d), and nondust particles (index nd), respectively. By starting from

equation image

we finally obtain

equation image

After substituting βλ,nd by βλ,pβλ,d, we solve the resulting equation to obtain a solution for βλ,d,

equation image

[18] To determine the backscatter coefficient of dust particles βλ,d following equation (14), we first have to compute the particle depolarization ratio δλ,p with equation (11). From the profiles of the particle backscatter coefficient and the dust backscatter coefficient we obtain the profile of the nondust backscatter coefficient as βλ,pβλ,d. For the computation after equations (11) and (14) we need to estimate the nondust depolarization ratio δλ,nd and the dust depolarization ratio δλ,d. Values found during the SAMUM-1 campaign are listed in Table 2. The particle depolarization ratio of Saharan dust of 0.31 ± 0.03 at 532 nm [Freudenthaler et al., 2009] is in good agreement with the one of Asian dust of 0.35 as found from long-term observations over China and Japan [Sugimoto et al., 2003; Shimizu et al., 2004]. According to the literature, nondust depolarization ratios can vary from 0.02 to 0.15 with an accumulation around 0.05 [Murayama et al., 1999; Fiebig et al., 2002; Murayama et al., 2004; Müller et al., 2005; Sugimoto et al., 2003; Sugimoto and Lee, 2006; Chen et al., 2007]. For aged Siberian and Canadian forest fire smoke, Müller et al. [2005] found particle depolarization ratio values of 0.02–0.03 at 532 nm over Leipzig, Germany. However, as a result of lifting of soil material within the fire areas, considerably higher particle depolarization ratios (>0.1) are possible [Fiebig et al., 2002; Murayama et al., 2004]. Heese and Wiegner [2008] measured particle depolarization ratios of 0.05 to 0.1 at 355 nm over Banizoumbu, Niger (West Africa), in an almost 4-km-deep biomass burning smoke layer that was advected from central Africa in January 2006. For background aerosol conditions (periods without desert dust, but including maritime cases), Chen et al. [2007] measured volume depolarization ratios of 0.02–0.03 at 532 nm during 124 days at Taiwan in spring of 2004 and 2005.

Table 2. Input Parameters for the Separation of Optical Properties of Dust and Smokea
ParameterValueRange of Values
  • a

    Given are climatological dust values (and uncertainties, 1 standard deviation as used in the error analysis) and full range of observed values. Data are from the SAMUM-1 campaign (Morocco, May–June 2006) [Freudenthaler et al., 2009; Tesche et al., 2009], except for the nondust (smoke) depolarization ratio which is taken from the literature (see references given in section 3).

S355,d53 ± 7 sr35–70 sr
S532,d55 ± 7 sr45–65 sr
δ532,d0.31 ± 0.030.27–0.35
δ532,nd0.05 ± 0.010.02–0.15
åβ,d(355, 532)0.2 ± 0.2−0.5–0.7
åβ,d(532, 1064)0.3 ± 0.20.0–0.7
åσ,d(355, 532)0.0 ± 0.2−0.7–0.7

[19] In the next step, we multiply the dust backscatter coefficients by the dust lidar ratio S532,d to obtain the dust particle extinction coefficient

equation image

S532,d was extensively measured during SAMUM-1 in 2006 [Tesche et al., 2009] (see Table 2). By using the backscatter-related Ångström exponent åβ,d(λ1, λ2) and the extinction-related Ångström exponent åσ,d(λ1, λ2) (as measured for pure dust during SAMUM-1 in 2006 [Tesche et al., 2009]) we obtain

equation image

with λ being 355 or 1064 nm, and

equation image

[20] Finally, we obtain the nondust backscatter and extinction coefficients

equation image
equation image

and, in addition, estimates of the nondust lidar ratio Sλ,nd = σλ,nd/βλ,nd and the nondust Ångström exponents åβ,nd(355, 532), åβ,nd(532, 1064), and åσ,nd(355, 532).

[21] The profiles of the nondust backscatter coefficients at 355, 532, and 1064 nm and of the extinction coefficients at 355 and 532 nm are finally used as input in inversion schemes which allow for the retrieval of microphysical particle properties [Müller et al., 2001].

4. Results

4.1. Case Study: 31 January 2008

[22] Complex aerosol layering was observed over Praia during the SAMUM-2 campaign on all days from 18 January to 14 February 2008 except on 28–29 January (dust only) and on 9 February 2008 (clean maritime conditions, no smoke, no dust). Figure 1 provides an overview of the situation on 31 January 2008 based on the aerosol lidar observations at 710 nm and radiosonde profiles of meteorological parameters. The measurement is representative of most of our SAMUM-2 observations. A two-layer system consisting of a pure dust layer below 1500 m and a lofted aerosol layer consisting of a mixture of dust and smoke is clearly visible in the height time display of the 710-nm volume depolarization ratio (Figure 1d). Owing to the comparably long wavelength the difference between the volume and the particle depolarization ratio is much smaller compared to the 532-nm values. Thus, at 710 nm the volume depolarization ratio already is a good indicator of the nonsphericity of the measured particles. The African aerosol plume extends to about 5 km height. Daytime measurements of Sun photometer as well as nighttime lidar measurements (for the time period shown in Figure 1) show particle optical depths of 0.6 at 532 nm.

Figure 1.

Complex stratification of dust and smoke observed with BERTHA over Praia, Cape Verde, on 31 January 2008. (b) Range-corrected 710-nm signal (arbitrary units) and (d) volume depolarization ratio at 710 nm are shown with 15-m vertical and 10-s temporal resolution. Radiosonde profiles of (a) relative humidity RH, temperature T, and virtual potential temperature θv (evening sonde, Vaisala RS80, launch at 2123 UTC) and of (c) horizontal wind speed v and direction dir (morning sonde, Vaisala RS92, launch at 1110 UTC) are shown in addition.

[23] The radiosonde humidity and temperature profiles (Figure 1a) indicate a shallow maritime boundary layer up to 350 m, a dry, almost well-mixed Saharan dust layer (400–1400 m), and a lofted aerosol layer with enhanced and varying relative humidity above 1.5 km height. The lofted layer actually can be separated in two layers (1.5–3.0 and 3.0–5.0 km height) according to the profiles of the relative humidity and the potential temperature. Thin cumulus clouds developed above 4.6 km after 2230 UTC (not shown). Thus, swollen aerosol particles are present in the moist layer above 3.8 km height in Figure 1.

[24] We launched a Vaisala RS92 sonde in the morning, providing us with temperature, pressure, humidity, and wind information and an old RS80 radiosonde in the evening measuring temperature, pressure, and humidity only. Northeasterly, southerly, and southwesterly winds prevail in the dust layer and in the lower and upper parts of the lofted aerosol layer, respectively, on 31 January. Wind speeds are 5–10 m/s and 5–15 m/s below and above 1.5 km height, respectively.

[25] Figure 2 provides information about the possible origin and sources of the observed African plume. During the dry season in northern hemispheric winter, fire sources in western and central Africa are located in a latitudinal band stretching from about 3°N to 13°N. These fire sources (red spots) are indicated in the underlying map of western Africa in Figure 2. MODIS satellite observations of fires within the period from 21 to 30 January 2008 are accumulated ( HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory) [Draxler, 1988] backward trajectories at 500 and 1000 m height (orange in Figure 2) indicate pure dust transported from desert areas north of a belt with strong fire activity, whereas backward trajectories at heights from 1500 to 4000 m (green, blue, and black in Figure 2) arrive within the lofted aerosol layer. All these latter trajectories (except the one arriving at 4 km height) crossed areas with strong biomass burning about 3–6 days before the arrival over Cape Verde, and desert regions 6–10 days prior to the lidar observation. As a result, the aerosol above 1500 m is expected to consist of a mixture of mineral dust and smoke.

Figure 2.

Ten-day HYSPLIT backward trajectories ending at Praia on 31 January 2008, 2200 UTC. The underlying image is a MODIS 10-day fire map that shows all locations of fires (red spots) detected during the 21–30 January period.

[26] Figure 3 shows profiles of the dust and smoke backscatter coefficient as well as of the volume and particle depolarization ratio at 532 nm. The separation procedure described in section 3 is applied with biomass burning smoke representing nondust particles. The signal averaging period is from 2133 to 2232 UTC, as shown in Figure 1. The particle depolarization ratio required for the separation between dust and smoke particles is computed from the measured volume depolarization ratio after equation (11) by using the primary retrieval products listed in Table 1. The Rayleigh backscatter coefficients βλ,m (not listed in Table 1) are computed from radiosonde temperature and pressure profiles and known Rayleigh scattering and backscattering coefficients for Standard Atmosphere conditions [Bucholtz, 1995]. By applying equation (14), we yield the dust backscatter coefficient in Figure 3a, and by subtracting the dust from the total particle backscatter coefficient, the nondust backscatter coefficient (mainly biomass burning smoke) remains.

Figure 3.

(a) Separation of dust (red) and smoke (green) particle backscatter coefficients at 532 nm, (b) relative contribution of dust backscatter and smoke backscatter to the total backscatter coefficient (black profile in Figure 3a), (c) uncertainties in the retrieved dust (red) and smoke (green) backscatter coefficients, and (d) 532-nm volume and particle depolarization ratio. Unsmoothed 1-h mean profiles for the time period in Figure 1 are shown.

[27] Figure 3b shows the relative contributions of dust and smoke to the total particle backscatter coefficient. An almost pure, and well-mixed Saharan dust layer is identified up to 1400 m height. Laser-beam receiver-field-of-view overlap effects prohibit a trustworthy retrieval of backscatter coefficients below a height of about 500 m on that day. A plume containing dust and smoke particles extends from 1.2 to 4.8 km height. Smoke particles contribute about 60–70% to the observed total particle backscatter coefficient in the range from 1.4 to 3.8 km height.

[28] Uncertainties in the retrieved backscatter profiles in Figure 3c are computed by applying the law of Gaussian error propagation to the retrieval equations. The accuracy in this calculation is checked by means of a sensitivity study in which each of the individual input parameters in Table 2 is varied within the indicated range of values while the other input parameters were set to the climatological mean value. The error curves in Figure 3c include the statistical uncertainty (signal noise) and signal calibration uncertainties in the retrieval of the particle backscatter coefficient of 5–10%. The relative errors of the dust and smoke backscatter coefficients are about 20% and 10–15%, respectively.

[29] In the next step, we separate the dust and smoke volume extinction coefficients as illustrated in Figure 4. By using the backscatter coefficient profile of desert dust (Figure 3) and the Saharan dust lidar ratio of 55 sr (Table 2) the dust extinction coefficient at 532 nm is obtained with equation (15) (red profile in Figure 4). The total particle extinction coefficient (black profile in Figure 4) is directly determined from the observation of the nitrogen Raman signals (see Table 1). The difference between these two extinction profiles yield the profile of the smoke extinction coefficient after equation (19). As can be seen in Figure 4, 70–90% of light extinction in the lofted layer above 1600 m height is caused by smoke.

Figure 4.

(a) Dust (red), smoke (green), and total (black) particle extinction coefficient at 532 nm, which correspond to the backscatter coefficient profiles in Figure 3, and (b) relative contribution of dust and smoke extinction to total particle extinction. The 1-h mean signal profiles for the time period shown in Figure 1 are smoothed with 300-m window length before further processing. Error bars (1 standard deviation) indicate the total retrieval uncertainty.

[30] Overlap correction uncertainties below 1400 m are too large for a trustworthy retrieval of particle extinction profiles from Raman signals on that day. Since the dust extinction coefficient is determined from the dust backscatter coefficient (multiplied by the dust lidar ratio), it is reliable down to heights of about 500 m. The backscatter coefficient is calculated from the ratio of the 532-nm signal to the 607-nm Raman signal. In the case of signal ratios the overlap effect widely cancels out.

[31] The uncertainty in the calculation of the total and smoke particle extinction coefficient is mainly a function of signal noise (statistical error). The uncertainty in the estimated dust extinction coefficient is about 25% taking the uncertainties in the computation of the dust backscatter coefficient and in the dust lidar ratio (13% uncertainty) into account. Since dust extinction values are on the order of 20 ± 10 Mm−1 in the smoke-dust layer above 1400 m, the influence of uncertainties in the dust extinction estimates of about 5 Mm−1 are of minor importance in the estimation of the smoke light-extinction profile. The relative error of the smoke extinction coefficient ranges from 20 to 40% within the lofted layer.

[32] Smoke extinction values around 80 Mm−1 (green profile in Figure 4) indicate strong pollution. Heese and Wiegner [2008] measured total particle extinction coefficients of 100–200 Mm−1 in western Africa's Sahel region just south of the Saharan desert on 30 January 2006 during the AMMA campaign, when a thick lofted plume (2–5 km height) was advected from areas with strong fire activity in central Africa. Values of up to 800 Mm−1 were measured in aerosol plumes over fire areas [Johnson et al., 2008a].

[33] In the third step we compute the full set of smoke backscatter and extinction coefficients to obtain backscatter- and extinction-related Ångström exponents for the smoke particles and the wavelength dependence of the smoke lidar ratio. The result is shown in Figure 5. According to equation (17) and the extinction-related Ånsgtröm exponent in Table 2, the 355-nm dust extinction coefficients are equal to the 532-nm dust extinction coefficients. The shown smoke 355-nm extinction values are obtained from the difference of the total 355-nm extinction coefficient (see Table 1) and the dust 355-nm extinction coefficient by the use of equation (19). Correspondingly, the dust backscatter coefficients at 355 and 1064 nm are computed using equation (16) and the backscatter-related Ångström exponents in Table 2. The shown smoke backscatter coefficients at 355 and 1064 nm are then obtained by the use of equation (18). Uncertainties (error bars in Figure 5) are computed by applying the law of error propagation and mainly depend on signal noise and the uncertainties in the input parameters in Table 2.

Figure 5.

Smoke (a) backscatter coefficients, (b) extinction coefficients, (c) lidar ratios, and (d) backscatter- and extinction-related Ångström exponents retrieved from the BERTHA observations on 31 January 2008. The 1-h mean signal profiles are smoothed with 660-m window length to reduce the statistical uncertainty. Error bars (1 standard deviation) indicate the total retrieval uncertainty.

[34] The profiles of the lidar ratios and Ångström exponents also indicate the two-layer structure of the lofted aerosol plume as suggested by the radiosonde profiles of potential temperature and relative humidity in Figure 1. Smoke lidar ratios range from 60 to 80 sr in the lower part and from 80 to 100 sr in the upper part of the lofted aerosol layer at both 355 and 532 nm. Ångström exponents computed from the 355-nm and 532-nm backscatter and extinction coefficients are roughly 1–2 in the lower part and 1–1.5 in the upper part of the pollution plume. The backscatter-related Ångström exponents calculated from the 532-nm and 1064-nm backscatter coefficients drop to low values in the upper layer. This indicates the presence of a bimodal aerosol distribution with a significant coarse mode fraction consisting of strongly swollen aerosol particles or the presence of a few drizzle droplets.

[35] Ansmann et al. [2002] showed that the extinction-related Ångström exponent is equal to the sum of the backscatter-related Ångström exponent and the lidar-ratio-related Ångström exponent (not shown). Thus, almost equal backscatter and extinction Ångström exponents (355, 532 nm) in the upper part of the lofted layer indicate that the wavelength dependence of the lidar ratio must be small as it is the case in Figure 5. The only exception is visible in the lowest part of the lofted layer below 2.5 km height.

[36] In the final step of the separation procedure we estimate the microphysical properties of the smoke aerosol. The inversion algorithm developed by Müller et al. [1999] (latest version: Kolgotin and Müller [2008]) is applied to the spectrally resolved backscatter and extinction coefficients in Figure 5. Height profiles of the effective radius (surface-area weighted mean radius) and the single-scattering albedo (scattering-to-extinction ratio) of the smoke particles are shown in Figure 6 together with backscatter coefficients (total, dust, smoke). The uncertainty of the inversion products considers errors introduced by the inversion procedure itself and 20% uncertainty in each of the five optical input parameters (three backscatter and two extinction coefficients).

Figure 6.

(a) Backscatter coefficients of particles, dust, and smoke at 532 nm and (b) effective radius and single-scattering albedo (SSA, 532 nm) retrieved with an inversion algorithm [Müller et al., 1999] applied to the profiles of backscatter and extinction coefficients in Figure 5.

[37] As can be seen, the effective radius shows values around 0.2 μm in the lower part of the lofted plume (dry aerosol particles) and values up to 0.35 μm in the upper part (swollen aerosol particles, some activated particles). The single-scattering albedo ranges from 0.9 (less absorbing particles) in the lower part of the lofted plume to rather low values of 0.60–0.65 in the upper part, indicating highly absorbing smoke particles. The latter finding is in agreement with the rather high lidar ratios of 80–100 sr indicating highly light-absorbing particles.

4.2. Comparison With AERONET

[38] We checked the quality of our multiwavelength lidar observations by comparing them with respective AERONET Sun photometer observations. This comparison is shown in Table 3. The day of 31 January 2008 was cloudless so continuous Sun photometer observations from 0900 UTC (0800 local time) to 1800 UTC were possible. The optical properties remained fairly constant from morning to the evening hours so favorable conditions are given for the lidar-photometer comparison.

Table 3. Comparison of Lidar Observations and Respective AERONET Sun Photometer Retrieval Productsa
Lidar (2132–2232 UTC, Mean)AERONET (0900–1800 UTC)
Parameter0–1.5 km1.5–5.0 kmColumnParameterColumnRange of Values
  • a

    Lidar observations are given as 1-h mean values. AERONET data are measured on 31 January 2008 from 0900 to 1800 UTC; daytime mean values and range of values observed over the day are presented. Column mean extinction values are computed by combining optical depth and layer depth (5 km) information. In the case of the lidar data, uncertainties (±1 standard deviation) are given in addition. Signal noise and uncertainties in the dust (d) and smoke (s) separation and in the overlap correction (missing data in the lowermost 500 m) are taken into account. Indices p, f, and c indicate the contributions of all particles and the fine- and coarse-mode fraction, respectively.

σ355,p (Mm−1)180 ± 50128 ± 15143 ± 25σ380,p (Mm−1)148130–160
σ532,p (Mm−1)173 ± 4588 ± 10114 ± 20σ500,p (Mm−1)124110–135
σ532,d (Mm−1)162 ± 4523 ± 565 ± 15σ500,c (Mm−1)6255–68
σ532,s (Mm−1)12 ± 1065 ± 1549 ± 20σ500,f (Mm−1)6255–68
τ355,p0.27 ± 0.060.45 ± 0.040.72 ± 0.10τ380,p0.710.68–0.76
τ532,p0.26 ± 0.060.31 ± 0.030.57 ± 0.10τ500,p0.590.56–0.65
τ532,d0.24 ± 0.080.08 ± 0.020.32 ± 0.09τ500,c0.290.25–0.33
τ532,s0.02 ± 0.010.23 ± 0.050.25 ± 0.07τ500,f0.290.25–0.33
åτ,p(355, 532 nm)0.09 ± 0.30.93 ± 0.30.58 ± 0.2åτ,p(380, 500 nm)0.660.56–0.77
åτ,d(355, 532 nm)0 ± 0.20 ± 0.20 ± 0.2   
åτ,s(355, 532 nm)1.18 ± 0.31.18 ± 0.31.18 ± 0.3   

[39] On the basis of the lidar-derived extinction profiles at 355 and 532 nm and the dust and smoke extinction profiles at these two wavelengths, the optical depths of the dust layer, of the elevated smoke-dust layer, and of the tropospheric column were calculated (shown in Table 3). For simplicity, we assume that the particle extinction coefficient in the maritime boundary layer (≤400 m; see Figure 1) equals values retrieved for heights >500 m (about 160–170 Mm−1; see Figure 4). From that approach we estimate the particle optical depth of the lowermost 500 m to be about 0.08 at 532 nm. An uncertainty of this estimation of 0.03 particle optical depth is considered in Table 3. This estimation of the optical depth below 500 m is in agreement with observations of other SAMUM lidars whose results in the lowest heights are less affected by the overlap effect.

[40] As can be seen from the column values (0- to 5-km column) a very good agreement is obtained in terms of the mean (total) particle extinction coefficients, the respective particle optical depths, and the Ångström exponents for the short wavelength range (355–532 nm, 380–500 nm). From the AERONET sky radiance and optical depth observations a bimodal lognormal volume size distribution was determined with a fine mode peaking at about 0.11 μm mode radius and a coarse mode peaking at 2.0–2.5 μm mode radius. The AERONET inversion algorithm determines the minimum of the size distribution somewhere within the size interval from 0.194 to 0.576 μm. The size distribution is split in two at the retrieved minimum to calculate the fine and coarse mode effective radius and the contribution to total optical depth, respectively. Thus, AERONET provides information of the 500-nm optical depth caused by the coarse-mode fraction (particles with radii >0.25 μm on 31 January 2008) and the fine mode particle fraction (radius <0.25 μm), respectively [O'Neill et al., 2003]. On 31 January 2008 the fine mode contribution to the total optical depth is 0.5 (mean value from 0900 to 1800 UTC) which is close to the lidar-derived ratio of smoke-to-total optical depth of 0.44 at 532 nm. The effective radius of the fine mode is 0.12 μm which is much smaller than the values we retrieved for biomass burning smoke presented in section 4.1. While the lidar-derived smoke effective radius represents the whole size distribution of particles with a low particle depolarization ratio the AERONET fine mode is cut off at 0.6 μm. Therefore, fine mode effective radii have to be smaller than values obtained from the inversion of dust-screened lidar data.

[41] AERONET also provides a column value for the single-scattering albedo at 440 and 670 nm. On 31 January 2008 these values range from 0.80 to 0.85 (441 nm) and 0.83 to 0.93 (675 nm) in the time period from 0900 to 1800 UTC. The extinction-weighted integration of the lidar-derived single-scattering albedo at 532 nm yields a column value of 0.89 according to the single-scattering albedo values above 1500 m height in Figure 6b and by assuming a single-scattering albedo of 0.97 for dust below 1500 m height as estimated from Sun photometer observations (at 441 and 675 nm) on a pure dust day (29 January 2008).

[42] Because lidar permits vertical profiling we can use the opportunity to discuss to what extend the AERONET photometer observations alone permit us to identify the presence of smoke layers (above the dust layer) and to estimate the smoke optical properties within the lofted plume. Table 3 indicates rather different optical characteristics in the dust layer and in the lofted smoke-dominated layer so that the column values do not describe the situation properly. The differences in the optical properties are highlighted in Figure 7. Height-resolved optical properties (symbols, 500 m vertical resolution) are compared with respective layer-mean values (green lines) and the column values (blue lines), the latter would be observed by a Sun photometer or from space (e.g., with MODIS). The column lidar ratio would be retrieved from a combination of a Sun photometer and a standard backscatter lidar such as the CALIPSO lidar. For simplicity, values of extinction coefficient, lidar ratio, and Ångström exponent within the lowermost 1000 m are assumed to represent pure dust conditions. Figure 7 indicates rather different optical characteristics in the different layers. Thus, the interpretation of the column values alone can be rather misleading. For example, a column lidar ratio of 64 sr does not indicate, in this case, the presence of highly absorbing particles uniformly distributed on the vertical since we find very large lidar ratios of 100 sr in the lofted plume at 3.8–4.2 km height (see Figure 7).

Figure 7.

Profile (lidar, green symbols) versus column-integrated values (passive remote sensing, blue line) of (a) particle extinction coefficient (532 nm), (b) lidar ratio (532 nm), and (c) Ångström exponent (355–532 nm). Same case as in Figures 3 and 4 and Table 3 is shown. Values of the lowermost 1000 m are estimated from Figure 4 by assuming pure dust conditions (Table 2). Column values (vertical lines and numbers) are calculated from the height-resolved values (circles, given with a resolution of 500 m) for the dust layer (green line, 0–1.5 km height) and the lofted dust-smoke layer (green line, 1.5–5.0 km height) and the entire column (blue line, 0–5 km height).

[43] Figure 8 shows the full time series of AERONET observations from 23 January to 14 February 2008. In addition to the 500-nm optical depth and the 380- to 500-nm Ångström exponent, the ratio of the optical depth of the column above 1.5 km height to the total particle optical depth (from individual lidar measurements) and the fine-mode optical depth (from AERONET) are shown. Assuming usual wintertime conditions of a dust layer near ground level topped by a lofted smoke layer, the Ångström exponent can be used as an indicator for the presence of a lofted smoke layer while the fine-mode optical depth can be interpreted as the optical depth of the lofted plume (adding a coarse-mode optical depth of roughly 20%).

Figure 8.

(a) Daytime mean particle optical depth (solid green circles, AOD, 500 nm, AERONET Sun photometer, level 2.0 data) and relative contribution of the particle optical depth above 1.5 km height to the total particle optical depth (open green circles, from lidar evening observation) and (b) daytime mean Ångström exponent (solid blue circles, computed from 380 and 500 nm optical depth, from AERONET) and fine mode fraction retrieved from the AERONET data (open blue circles, taken from the AERONET web side, assigned as provisional). Blue line at 0.3 in Figure 8b marks the threshold value; Ångström exponents >0.3 indicate smoke. The grey shaded area marks the date of the presented case study of 31 January 2008.

[44] In Figure 8 the 380- to 500-nm Ångström exponent is frequently higher than 0.3. In all of these cases a smoke layer was present. According to our lidar observations, dust dominated on 27–29 January, a lofted plume was absent and the Ångström exponent was close to zero. On 9 February, westerly wind prevailed at all heights and rather clean maritime conditions with an optical depth of about 0.05 and a 380- to 500-nm Ångström exponent of 0.15 was determined. Figure 8 contains the fine-mode fractions for days with an Ångström exponent clearly higher than 0.3. The fine-mode fraction of the optical depth is roughly equal to the 500-nm optical depth in Figure 8a multiplied by the optical depth ratio also shown in Figure 8a. Thus, the fine-mode optical depth can be regarded to approximate the extinction properties in the lofted layer. The optical depth ratio is comparably high in February owing to the absence of dust advection in the lowermost troposphere during the second half of the SAMUM campaign.

4.3. Comparison With AMMA, INDOEX, and PRIDE-PRD

[45] A similar aerosol stratification as presented above was frequently observed with aircraft over western Africa during the AMMA/DABEX campaign in January–February 2006 [Johnson et al., 2008a, 2008b; Osborne et al., 2008]. The average profile of the 550-nm extinction coefficient of all AMMA/DABEX aircraft flights shows a dust dominated layer up to 1.5 km height with peak extinction values of 200 Mm−1 and a mixed aerosol layer with extinction values from 40–110 Mm−1 up to 5 km height [Johnson et al., 2008a]. The campaign-mean dust and smoke extinction values in the center region (2–4 km) of the lofted layer are 30 ± 20 Mm−1 and about 35 ± 10 Mm−1, respectively. Both dust and smoke particles contribute to about 50% to light extinction in the lofted plumes. Ångström exponent for the mixture and smoke alone, derived from the mean 450- to 550-nm smoke extinction coefficients, are about 0.8 and 1.6. Most values of the dust-smoke mixture were found to be in the range from 0.5 to 1.5 (450- to 700-nm range [Johnson et al., 2008b]). Single-scattering albedo values were mainly between 0.73 and 0.93 for the lofted dust-smoke mixture [Johnson et al., 2008b] and estimated to be 0.72–0.87 for pure smoke particles [Osborne et al., 2008].

[46] Heese and Wiegner [2008] and Pelon et al. [2008] presented several lidar case studies and observed extinction values of 50–300 Mm−1 at heights above 1000 m. Heese and Wiegner [2008] reported typical lidar ratios of 50 sr (mainly dust) to 90 sr (mainly smoke) at 355 nm in the center of lofted smoke-dust layer.

[47] McConnell et al. [2008] performed a research flight over the Atlantic Ocean (10°N–12°N) very close to the African continent on 15 February 2006, about 450–500 km upwind of Praia, Cape Verde. They observed a near-surface dust layer up to 700–800 m height with dust scattering coefficients from 300–1000 Mm−1 and respective Ångström exponents of around zero, and a lofted aerosol plume from 1200–1400 m to 4200 m height. Scattering coefficients at 550 nm ranged from 100–300 Mm−1 in the main part of the plume. The 550-to 700-nm Ångström exponent mainly ranged from 0.3 to 0.9 in the lofted layer.

[48] Capes et al. [2008] report that the number distributions of fresh biomass burning aerosol were dominated by particles with radii less than 0.1 μm, whereas aged aerosol was characterized by a peak in the number distribution at radii between 0.1 and 0.2 μm which points to effective radii which are frequently >0.2 μm. The freshest aerosol size distribution corresponds to aircraft flight levels within smoke plumes directly above visually observed fires. More aged aerosol size distributions are typical for higher altitudes (2–4 km), far from the source regions.

[49] Table 4 summarizes AMMA results and compares them with SAMUM-2 findings, and with further data recorded in south Asia (INDOEX) and subtropical east Asia (PRIDE-PRD). We observed a similar aerosol long-range transport (with the same lidar used at Cape Verde) over the Indian Ocean at the Maldives during INDOEX in 1999 and 2000 [Franke et al., 2003; Müller et al., 2003]. The observed tropical and subtropical south and southeast Asian aerosol mainly consisted of a mixture of urban haze (fossil fuel combustion) and particles that originate from domestic biomass burning (fuel wood, crop waste, dung cake) and accidental and controlled forest and agricultural fires. The haze from the northern parts of the Indian subcontinent crossed the Bay of Bengale and traveled about 6 days before arriving at the Maldives. Above the polluted maritime boundary layer, lofted continental aerosols were typically present from 1 to 3 km height. Extinction coefficients mostly ranged from 50 to 150 Mm−1 in the lofted haze (see Table 4). Aging of the particles as well as convective mixing with maritime aerosols which increase the single-scattering albedo and decrease the lidar ratio may have slightly altered the original aerosol properties. As can be seen, slightly lower lidar ratios were measured during INDOEX compared to SAMUM-2 values for the 31 January case. However, lidar ratios were frequently between 60 and 90 sr (see Table 4). INDOEX values for the single-scattering albedo, the effective radius, and the Ångström exponent (355–532 nm) are similar to the values found for western African aerosol (SAMUM-2, AMMA).

Table 4. Comparison of SAMUM-2 Observations for 31 January 2008 With Respective Results From AMMA 2006, INDOEX 1999/2000, and PRIDE-PRD 2004a
  • a

    AMMA 2006 data from Heese and Wiegner [2008], Johnson et al. [2008a, 2008b], Osborne et al. [2008], and Pelon et al. [2008]; INDOEX data for 1999/2000 from Franke et al. [2003] and Müller et al. [2003]; and PRIDE-PRD 2004 data from Ansmann et al. [2005], Müller et al. [2006], and Tesche et al. [2008]. Except for PRIDE-PRD (well-mixed boundary layer aerosol, 0–3 km height), only properties of aged aerosols in lofted layers above 1 km height are considered here. For SAMUM-2 (1–5 km) the smoke particle extinction coefficient σ532,p, lidar ratio S532,p, single-scattering albedo SSA (532 nm), and Ångström exponent åτ,p(355,532 nm), as well as the effective radius reff, are listed. For AMMA (1–4 km), åτ,p(400,700 nm), σ500–550,p, S355,p, and SSA (550 nm) are listed; for INDOEX (1–3 km), åτ,p(355, 532 nm), σ532,p, S532,p, and SSA (532 nm) are listed; and for PRIDE-PRD (0–3 km), photometer-derived åτ,p(380, 500 nm), lidar-derived S532,p, and photometer-lidar-derived SSA (532 nm) are listed.

σp (Mm−1)20–10050–30050–150200–800
Sp (sr)60–10050–9050–8035–60
rp,eff (μm)0.15–0.3 0.10–0.250.20–0.35
SSA (≈550 nm)0.6–0.90.73–0.930.80–0.950.75–0.85

[50] In addition, Table 4 contains subtropical aerosol characteristics measured in southern China in the framework of a series of PRIDE-PRD observations [Ansmann et al., 2005; Müller et al., 2006; Tesche et al., 2008]. East Asian aerosols show a pronounced coarse mode due to the combustion of coal and dried plants. Low combustion temperatures as well as less strict environmental regulations favor the emission of large particles. Consequently, Ångström exponents are low, single-scattering albedo values are low, and extinction coefficients are high. Sometimes they were as high as 1200 Mm−1 with corresponding particle optical depths of up to about 2 at 532 nm. Surprisingly, east Asian lidar ratios are comparably low with values of 35–60 sr [Ansmann et al., 2005; Tesche et al., 2007]. The same (35–60 sr) was found for aged southeast Asian aerosols during INDOEX [Franke et al., 2003]. Since the single-scattering albedo values for western African, northern Indian, and southern Chinese aerosols are similar, subtropical Chinese aerosol must contain a larger fraction of larger particles with spherical shape (probably large hygroscopic biomass burning particles in the accumulation mode). The lidar ratio is very sensitive to particle size as long as the particles are spherical and decreases with increasing particle radius. The large particles of western African aerosol are mainly nonspherical (mineral dust). Nonsphericity reduces the 180° scattering efficiency by a factor of 2–3 [Mattis et al., 2002] so that the overall lidar ratio remains high even in the presence of large particles.

5. Summary

[51] A new lidar-based approach to separate profiles of optical parameters due to the influence of dust and smoke particles in the west African outflow plume was presented. The method makes use of particle backscatter and extinction profiling at several wavelengths and vertically resolved depolarization ratio measurements. The technique was applied to a SAMUM-2 observation over Praia, Cape Verde, in the outflow regime of the west African continent. Complex aerosol layering prevails over the tropical Atlantic during winter and spring seasons.

[52] The case study of 31 January 2008 revealed a two-layer aerosol system. A desert dust layer with an optical depth of about 0.25 at 532 nm extended from ground level (or from the top of a shallow maritime boundary layer) to 1.5 km height. That layer was topped by a lofted layer (from 1.5 to 5.0 km height) consisting of desert dust (0.08 optical depth) and biomass burning smoke (0.24 optical depth). Extinction values were 20 ± 10 Mm−1 (desert dust) and 20–80 Mm−1 (smoke) in the lofted plume. The smoke extinction-to-backscatter ratios were rather high with values up to more than 100 sr, effective radii ranged from 0.15 to 0.35 μm, and the smoke single-scattering albedo in parts showed values below 0.7.

[53] The height-resolved lidar results were compared with column-integrated products obtained with an AERONET Sun photometer. Good agreement was found. It was further discussed to what extent Sun photometer data permit estimations of the optical properties of lofted smoke from column observations. Good agreement was found with other observations in the west African area recently conducted in the framework of the AMMA/DABEX campaign. The findings were further compared with similar observations of Asian haze performed over the Indian Ocean south of the Indian subcontinent and in the subtropical southern China.


[54] We thank Colonel Antonio Fortes for his efforts in overcoming logistical and organizational problems of any kind. His support enabled our successful campaign in the first place. We express our deepest gratitude to the airport authority of the Republic of Cape Verde. We particularly want to thank the director of Praia airport, Euridice Mascarenhas, and her staff for the warm welcome offered to our science team. E. Mascarenhas supported us in a very extraordinary and unbureaucratic manner, which made our work at Praia airport very pleasant. We would like to acknowledge the great help we received by the airport staff members Daniel Lima and Antonio Pinheiro. We thank Brent Holben and Wayne Newcomb (deceased 18 December 2008) for providing us with the AERONET photometer and solid support. The SAMUM research group is funded by the Deutsche Forschungsgemeinschaft (DFG) under grant FOR 539.