Journal of Geophysical Research: Atmospheres

Vertical aerosol profiles from Raman polarization lidar observations during the dry season AMMA field campaign

Authors


Abstract

[1] Lidar measurements of the vertical distribution of optical particle properties were performed in January 2006 in the frame of the dry season field experiment of the African Monsoon Multidisciplinary Analyses (AMMA). The measurements were conducted at a remote site in West Africa in the Sahel region just south of the Saharan desert to investigate the influence of particles on the West African monsoon. The portable Raman polarization lidar system POLIS of the University of Munich was operated each forenoon and during selected nights. These measurements are the first Raman and depolarization lidar measurements taken in the Sahel. The particle depolarization ratio and the extinction coefficient profiles allow the characterization of the particles and the distinction between Saharan dust particles and biomass burning aerosol (BBA). The origin of this aerosol was investigated by trajectory analyses. The profiles show a varying dust layer in the lowermost troposphere during the whole period. Above the dust layer a broad layer of BBA reaching up to an altitude of 5 km was frequently observed. The extinction coefficient in the dust layer was between 0.2 and 0.4 km−1 at 355 nm, and the particle depolarization was around 25%. In the BBA layer the extinction coefficient was lower with values between 0.1 and 0.2 km−1, and the particle depolarization ratio was below 10%. Lidar ratio values around 55 sr were found for the dust layer and around 75 sr for the BBA layer. From previous observations, lidar ratios between 40 and 80 sr were expected for BBA.

1. Introduction

[2] Atmospheric particles have not only a direct impact on the Earth radiation budget by scattering and absorption, they can also serve as condensation nuclei for cloud droplets. Thus, aerosol may have an important influence on cloud formation, cloud lifetime, and precipitation. As precipitation is of extreme relevance for life in the Sahel zone (precipitation has dramatically decreased during the last 30 years [e.g., Lamb, 1983; Fontaine and Janicot, 1996; Le Barbe et al., 2002]) the African Monsoon Multidisciplinary Analyses (AMMA) project was initiated to study the life cycle of the monsoon and to improve the precipitation prediction. Consequently, one significant part of AMMA was devoted to investigations of aerosol and its interactions with clouds. One of the main sources for aerosol over West Africa is the Saharan desert where mineral dust particles are lifted by local convection and transported southward. The second main source is forest fires that produce biomass burning aerosol (BBA). The seasonal biomass burning cycle in Africa is well defined. The burning in the AMMA region takes place in winter and defines the northern most extent of fire in Africa. As the year progresses, the fires move southward to central and southern Africa. The source regions and the horizontal transport of the particles may be identified by satellite imagery, but the vertical extension of the aerosol layer and the altitude at which the particles are transported remain unknown. The microphysical and optical properties of the particles and their vertical distribution, however, need to be known to understand their influence on radiative forcing [Haywood and Shine, 1997] and their indirect effect on clouds [Sherwood, 2002]. As mineral dust particles and BBA differ in their size spectrum and their single scattering albedo it is crucial to provide a tool for their discrimination. Moreover, their vertical distribution affects their residence time in the boundary layer and the free troposphere, and hence governs their long-range transport [Müller et al., 2005]. As the optical properties of the aerosol will change with the distance from their source because of, e.g., chemical transformation, wash out, or sedimentation, it is essential to have observations close to their sources and in the long-range regime as well. Consequently, observational programmes are required in both spatial domains to assess the role of aerosol in the climate system.

[3] In the far range domain, mineral dust or biomass burning aerosol was frequently observed in elevated layers [e.g., Prospero and Carlson, 1972]. Dedicated campaigns to characterize aged dust particles were provided, e.g., in the framework of the Puerto Rico Dust Experiment (PRIDE) [Reid et al., 2003]. Over Europe Saharan dust outbreaks have been characterized by, e.g., the European Lidar Network (EARLINET) [Müller et al., 2003; Ansmann et al., 2003] and during MINATROC (Mineral Dust Aerosol and Tropospheric Chemistry) experiment by Gobbi et al. [2003]. Close to their African sources, atmospheric particles and their optical properties have been studied during various field campaigns since the 1980s. During ECLATS (Étude de la Couche Limite Atmosphérique Tropicale Sèche) [Fouquart et al., 1987] in the vicinity of Niamey (Niger) and later during EXPRESSO (Experiment for Regional Sources and Sinks of Oxidants) [Ruellan et al., 1999] in central Africa the vertical distribution of atmospheric particles was measured by airborne in situ measurements. Optical properties, however, could only be derived by extrapolation from ground-based measurements. During the SHADE (Saharan Dust Experiment) campaign [Tanré et al., 2003] airborne lidar measurements allowed a characterization of mineral dust particles [Léon et al., 2003] between Dakar (Senegal) and Cape Verde Islands. Biomass burning aerosol has been characterized during the SAFARI-2000 (Southern African Regional Science Initiative) campaign [Haywood et al., 2003]: Here vertical profiles of scattering properties were measured using an airborne nephelometer. These profiles show how relevant the knowledge of the particles' vertical distribution is to assess their radiative effect [Keil and Haywood, 2003]. In none of these experiments, active remote measurements by lidars were provided for periods of weeks or months, which could be the nucleus for a local climatology; in particular, in the Sahel region no ground based lidar measurements have been performed so far. Raman lidars can provide aerosol extinction coefficients under ambient humidity conditions, and the polarization lidar technique allows the discrimination of dust from other aerosol types like urban, biomass burning, and marine aerosol. In the framework of the dry season field experiment of AMMA, conducted in early 2006, this technique was used for the first time at a remote site in Niger. As the lidar was operated over three weeks it was possible to investigate its potential for long-term observations as they are required for climate change studies.

[4] In parallel to these campaigns, that were all limited in time, the network of ground based Sun photometers, AERONET [Holben et al., 2001], provides vertically integrated optical aerosol properties over West Africa at several stations since 1995, i.e., primarily the optical depth and the Ångström coefficient. This information is very valuable for the detection of changes in the overall aerosol load. However, vertical information on the aerosol layering is missing.

[5] The objective of this paper is the description and discussion of the above mentioned Raman and depolarization lidar measurements in the framework of AMMA. The derived optical properties, extinction and backscatter coefficient and depolarization ratio, allow the distinction between mineral dust and biomass burning aerosol. The measurements took place in the Sahel region at Banizoumbou (Niger) close to the sources of mineral dust particles and BBA, the Saharan desert and forest fires in western and central Africa. In section 2, the experimental background with respect to the field campaign and site, the instruments and the data retrieval is presented. Section 3 describes three particular cases of observed aerosol profiles. Trajectory analysis gives an indication of the origin of the observed air masses and supports the interpretation of the lidar results. In section 4 an overview over some key optical properties of the particles over the whole campaign is given. They allow the linking of the short-term lidar measurements into a longer time range perspective. A discussion concludes the paper.

2. Experimental Background

[6] The purpose of the African Monsoon Multidisciplinary Analyses project is to investigate the preconditions of the development of precipitation during the West African monsoon over a time frame of 5 years. Experimental efforts were concentrated in 2006 as a 1-year enhanced observation period. Within this year several special observation periods (SOP) were embedded. These SOPs lasted for a few weeks each. The first special observation period, SOP0, has been carried out during the dry season, before the monsoon season set in, and lasted from mid-January to mid-February 2006. In the focus of SOP0 were field experiments to study the distribution of mineral dust and biomass burning aerosol, and their radiative, optical and microphysical properties. The measurements included ground based in situ sampling, and remote sensing techniques applying lidars and Sun photometers. Different combinations of these measurements were implemented at several stations in West Africa forming an east-west and a north-south transect. One of the three primary stations is Banizoumbou (13.5°N, 2.6°E, 250 m above sea level) in Niger, located in the countryside at a distance of 70 km east of the capital Niamey. The two other primary stations are located near Djougou in Benin and M'Bour in Senegal. Secondary measurement sites were established close to Bamako in Mali and Tamanrasset in Algeria.

[7] In Banizoumbou the ground based portable Raman and polarization lidar system POLIS of the University of Munich was operated during SOP0. At the same time a wide range of in situ particle sampling, optical, and radiative measurements were performed. An AERONET Sun photometer was located in 2 km distance from the Banizoumbou station. Radio soundings were launched at Niamey airport four times a day and the ARM mobile facility [Slingo et al., 2006] was stationed at the airport during the whole enhanced observation period 2006. Additionally, airborne in situ measurements of particle properties were performed on board the British research aircraft FAAM operating from Niamey Airport during SOP0. Another small lidar was operated on an ultra light aircraft [Chazette et al., 2007], however, during the last week of SOP0 only.

2.1. Portable Raman and Polarization Lidar POLIS

[8] POLIS is a small, rugged, two channel lidar which is easy to set up in the field and can be operated under logistically difficult conditions. Therefore, it was selected for operation at Banizoumbou where a limited infrastructure is available only. POLIS was set up just outside on the terrace of the small station building while the electronic equipment was sheltered inside.

[9] POLIS consists of a compact, flashlamp pumped, frequency doubled and tripled Nd:YAG laser (Big Sky by Quantel) emitting pulses of linear polarized light at the wavelength of 355 nm only. The repetition rate of the laser is 20 Hz and the pulse energy is 7.5 mJ. By choosing the wavelength in the ultraviolet region, where the eye is less sensitive, the lidar is eye-safe beyond a distance of 30 m from the laser. The telescope is Cassegrainian with an effective diameter of 200 mm. The optical design of the system was optimized by considering, that on the one hand measurements in the lowermost boundary layer are desired. On the other hand signals from the upper free troposphere are required for calibration and the limited dynamic range of the electronics has to be taken taking into account. The best solution results in a measurement range starting at 120 m from the lidar. The data acquisition system (Licel) of POLIS combines both digital and analog detection to increase the dynamic range of the acquired signal. The data acquisition allows a vertical sampling of 7.5 m. More technical details of this lidar system are described by Heese et al. [2002, 2004].

[10] POLIS can be operated in two modes: in the depolarization mode and in the Raman mode. In the depolarization mode the two channels of the lidar are used for measuring the parallel- and cross-polarized component of the backscattered light to determine the depolarization ratio at 355 nm. In the Raman mode, measurements at 355 nm and at the nitrogen Raman shifted wavelength at 387 nm are performed. Two separate detection units exist for depolarization and for Raman measurements. Switching between the two modes is easily done within 30 min by replacing the complete detector block. Because of precise manufacturing of the mechanical components no optical realignment of the system is necessary.

[11] The basic operation of the lidar was done in the depolarization mode, because the main interest of the field experiment was the discrimination between different aerosol types. Those are, in particular, mineral dust and biomass burning particles. Depolarization measurements were taken almost every day in the morning hours until noon. The measurements had to be stopped when the air temperature exceeded 35°C, because the water cooling circuit of the laser became inefficient. Since Raman backscattering is three orders of magnitude lower than elastic backscattering, the signal to noise ratio during daytime is not sufficient to retrieve aerosol extinction profiles. Therefore, the Raman measurements were restricted to nighttime when the sky background is low. However, for logistical reasons, the measurements could only be performed during selected nights. Table 1 gives an overview over the POLIS measurements performed during SOP0. For each day, the place, the operation mode, and the time period of the measurements are given. Note, that all times are given in universal time (UTC), which quite well fits for the interpretation of the diurnal cycle as Banizoumbou's longitude is only 2.6°E.

Table 1. POLIS Measurements During SOP0
DateStart Time (UTC)End Time (UTC)PlaceMode
11 Jan 200608391135BanizoumbouDepol
14 Jan 200608091037BanizoumbouDepol
15 Jan 200607571201BanizoumbouDepol
16 Jan 200607581100BanizoumbouDepol
17 Jan 200608051115BanizoumbouDepol
19 Jan 200608321143BanizoumbouDepol
20 Jan 200608411100BanizoumbouDepol
21 Jan 200607371125BanizoumbouDepol
22 Jan 200608231122BanizoumbouDepol
23 Jan 200608161124BanizoumbouDepol
24 Jan 200608231122BanizoumbouDepol
25 Jan 200608570936BanizoumbouDepol
26 Jan 200608341137BanizoumbouDepol
27 Jan 200608051115BanizoumbouDepol
27/28 Jan 200617100521BanizoumbouRaman
28 Jan 200606291105BanizoumbouDepol
29 Jan 200608191114BanizoumbouDepol
29/30 Jan 200617320613BanizoumbouRaman
30 Jan 200606470900BanizoumbouDepol
30/31 Jan 200618190640Niamey AirportRaman
31 Jan 200607281113Niamey AirportDepol
31 Jan 200617551859Niamey AirportDepol
1 Feb 200614070000Niamey AirportDepol
2 Feb 200600000901Niamey AirportDepol

2.2. Data Evaluation

[12] The quality of POLIS measurements has been approved with the EARLINET reference lidar MULIS, also owned by the University of Munich. A comparison of aerosol backscatter profiles with different aerosol loads in the boundary layer from simultaneous measurements in February 2003 is presented by Heese et al. [2004]. From this comparison an agreement of the backscatter coefficient of 1–3% was found in the planetary boundary layer, where most of the aerosol was present. These values lie well below the EARLINET criterion of less than 20% mean deviation for the aerosol backscatter coefficient at 355 nm, as given by Matthias et al. [2004]. In the following sections we describe how two relevant aerosol properties, the particle depolarization ratio and the extinction coefficient, have been retrieved from our measurements.

2.2.1. Depolarization Ratio

[13] The depolarization measurement takes advantage of the fact, that the emitted laser radiation is linearly polarized. By measuring backscattered radiation in the plane of the laser's polarization and in the cross-polarized plane, the volume depolarization ratio δ can be calculated from the ratio of both signals, provided, that a calibration of these two channels has been performed. This calibration was done by turning the polarization cube 45 degrees in both directions from its normal position. In these positions each detection channel receives half the amount of the signals from both polarization states. From the ratio of these signals, the different sensitivities of the two channels can be determined and accounted for. The volume depolarization ratio can be considered as a first indication of the particle shape. Although δ depends on the relative contribution of particle backscattering to the total backscattering, the volume depolarization ratio allows the distinction between spherical (i.e., biomass burning aerosol) and nonspherical particles (i.e., mineral dust). The volume depolarization ratio is very useful to illustrate the daily temporal development of the boundary dust layer and above the residual layer from the previous day. Since δ is just the ratio of two signals the detection of this layering is possible with a high temporal resolution of 30 s of the raw data (600 shot averages).

[14] As already mentioned, the volume depolarization ratio δ depends on the relative contributions, in terms of backscatter coefficient, of the particles and the air molecules. To get an unambiguous measure of the particle depolarization one has to correct δ for the molecular contribution and calculate the particle depolarization ratio δpar [e.g., Cairo et al., 1999; Murayama et al., 1999]. It can be written as

equation image

[15] The vertical profile of the molecular backscatter coefficient βrayl was determined from the radio sounding launched closest in time at Niamey airport. The particle backscatter coefficient profile βpar was retrieved from lidar measurements as described below. At the wavelength of 355 nm the molecular backscatter is comparably high because of the wavelengths dependence of the Rayleigh scattering coefficient. Thus, the determination of the particle depolarization in the ultra violet is more sensitive to errors than in the visible and near infrared range. The depolarization due to Rayleigh scattering from the air molecules also depends on the width of the interference filter of the detection channels. The bandwidths of the interference filters used with the depolarization channels of POLIS are 1.1 nm and 1.04 nm (FWHM) at 354.7 nm for the cross- and parallel-polarized signal, respectively. Interference filters with a bandwidth of less than 0.5 nm transmit only the central Cabannes line of the Rayleigh scattering spectrum which results in a molecular depolarization of 0.0036 [Behrendt and Nakamura, 2002]. Broader interference filters transmit parts of the rotational Raman bands and the molecular depolarization increases. In the case of POLIS δrayl = 0.0045 was used. As δrayl is very small compared to 1 and small compared to typical values of δ in the dust layer, it is obvious from equation (1) that in case of POLIS the influence of the filter width on δpar is only marginal. The errors of δ and βpar are much more significant, in particular, as βpar depends on the assumed lidar ratio. However, it should be emphasized that in this paper we only aim for a qualitative discrimination of dust particles from BBA and, thus, can tolerate relative errors of 20–30% for a 1 h mean profile.

2.2.2. Extinction Coefficient

[16] The second aerosol optical property that is in the focus of our study is the particle extinction coefficient. Particle extinction profiles are one of the key input parameters for radiative transfer calculations. The conventional retrieval method to derive extinction profiles from elastic backscatter lidar signals [Fernald, 1984; Klett, 1985] requires the knowledge of the ratio of the backscatter and the extinction coefficient, the so called lidar ratio. In contrast to this method the inversion of the inelastic Raman signal allows a direct determination of the aerosol extinction coefficient without a priori assumptions on the lidar ratio [Ansmann et al., 1992]. This is due to the fact that the lidar signal at 387 nm scattered back by nitrogen molecules is only affected by the transmission due to particle extinction and known optical properties of the air molecules. The latter can easily be derived from a radio sounding.

[17] As mentioned above, Raman measurements could only be performed during nighttime. For this reason, the extinction coefficients from the daytime measurements must be calculated using the Fernald method. To get the required lidar ratios we take advantage of our nighttime Raman measurements.

2.2.3. Lidar Ratios

[18] The determination of typical lidar ratios from our measurements is illustrated in Figure 1. It shows the extinction coefficient derived by the Raman method during three nighttime measurements (Figures 1a, 1c, and 1e) and the lidar ratio (Figures 1b, 1d, and 1f) calculated from this extinction profile divided by the corresponding backscatter profile. The lidar ratio profiles clearly show different layers of aerosol. In all three cases the lidar ratio in the boundary layer up to about 2 km varies around 55 sr. Above this altitude the lidar ratio is rising up to values around 75 sr to 80 sr with a peak around 3 km on 28 January (Figures 1a and 1b) and 30 January (Figures 1c and 1d) and at around 2.5 km on 31 January (Figures 1e and 1f). All heights are given in km above sea level (asl). These lidar ratios clearly identify different aerosol types. The large lidar ratio for the elevated layer is due to another aerosol type than observed in the boundary layer. This aerosol type origins from biomass burning as will be shown in the next section. In the boundary layer dust is the dominating aerosol type. As a consequence, the lidar ratio used with the Fernald algorithm for the derivation of the extinction profiles during daytime was set to 75 sr for BBA and to 55 sr for the dust layer and background aerosol. The discrimination of the two aerosol types can be done by means of the particle depolarization. The transition between the values of the lidar ratio in the upper and lower layer was done gradually over a range of 150 m.

Figure 1.

(a) Particle extinction coefficient and (b) lidar ratio profiles for 28 January 2006, (c and d) for 30 January 2006, and (e and f) for 31 January 2006. The Raman extinction coefficient profiles are derived over 300 m up to 1.9 km, over 600 m up to 4.1 km, and over 1200 m above this altitude. Additionally, the calculated lidar ratio profile is smoothed over 1500 m. The error bars indicate an error of 15% for the extinction profiles and of 20% for the lidar ratio.

[19] To obtain an extinction profile up to the tropopause height during daytime an averaging time of about 15 min is sufficient for SOP0. If we restrict our retrieval to the pronounced dust layer and the lower part of the free troposphere, averaging times of 5 min suffice. For the weaker Raman signal averaging over 3 h during nighttime measurements is necessary to achieve an extinction profile in the troposphere up to 5–6 km.

3. Lidar Measurements

[20] Three cases of observed particle profiles over Banizoumbou are discussed in detail in this section. These cases were selected because they show three quite different vertical distributions of mineral dust and biomass burning particles as observed in the dry season in the Sahel. Doing this we want to illustrate the range of aerosol distributions that exist under different meteorological conditions. The first case shown is the case with the highest dust load observed in the lower boundary layer during SOP0. The second case concerns moderate dust load as observed during most of the days. In both cases the dust layer was confined to the lowermost boundary layer, whereas a more or less pronounced layer of biomass burning aerosol was found above but clearly separated from the dust. The third case shows a situation where both aerosol types were mixed.

[21] Three other cases of POLIS extinction profiles (not shown here) were used for a comparison of the extinction profiles measured by an airborne nephelometer during ascends and descends of the British research aircraft FAAM and with the airborne lidar system LAUVA on board the French ultralight airplane. These comparisons are published in a companion paper by Johnson et al. [2008].

3.1. Heavy Dust Load

[22] On 15 January lidar depolarization measurements were taken from morning around 0800 UTC until noon (Figure 2). To visualize the development of the aerosol distribution with very high temporal resolution we selected here the volume depolarization ratio δ rather than the particle depolarization ratio δpar. A pronounced dust layer can clearly be identified between 0.5 and 0.9 km by its high δ values of up to 20% (color coded in red to yellow in Figure 2). The particles in this dust layer are dispersed into the developing boundary layer later in the morning by turbulent mixing. From about 1030 UTC the dust layer is well mixed and extends up to 1.2 km. A less pronounced dust layer with a constant vertical extent of 1.5 km is visible throughout the morning. This is the residual layer from the night.

Figure 2.

Volume depolarization measured on 15 January. In the early hours the nighttime residual dust layer of less than half a kilometer thickness featuring high depolarization values of up to 25% can clearly be identified. Later in the morning the boundary layer was developing and the dust was spread over a 1 km deep layer. The uppermost boundary of the dust layer was at 1.5 km on this day.

[23] The extinction profile achieved by integrating over 1 h from 0800 to 0900 UTC (Figures 3a and 3b) shows very high values of up to 1.6 km−1 in the layer between 0.5 and 0.9 km. Above this layer moderate extinction values around 0.23 km−1 were observed. This layer is reaching up to 1.5 km altitude. The particle depolarization ratio δpar in both layers has high values of up to 29% and 21%, respectively. These values substantiate that the particles have a nonspherical shape and that they certainly consist of mineral dust.

Figure 3.

(a) Particle extinction coefficient and (b) depolarization ratio profiles for 15 January 2006, (c and d) for 19 January 2006, and (e and f) for 23 January 2006. The profiles are vertically smoothed over 375 m. The error bars indicate an error of 15% for the extinction profiles and of 20% for the depolarization ratio.

[24] Above 1.5 km altitudes the extinction coefficient is around 0.05 km−1 and δpar is close to zero. Between 2.5 km and 4 km altitude, the extinction coefficient is rising slightly to values of 0.08 km−1 indicating another aerosol layer. The particle depolarization in this layer is quite low but slightly higher than zero. However, a quantitative assessment of δpar is impossible because of the large retrieval errors so that a discrimination of the different aerosol types is not realistic. Above this layer the particle extinction is below the detection limit.

[25] To support the particle classification, auxiliary data were used (see Figure 4). On the one hand, the origin of the air masses arriving at Banizoumbou was determined from HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory) [Draxler, 1988] back trajectories that have been calculated for several altitudes (graphs below the each of the maps in Figure 4). On the other hand, we used information on active fires that can be considered as primary source of BBA. Corresponding sites can be identified by the active fire detection algorithm using brightness temperatures derived from the 4 and 11 μm channels of MODIS [Giglio et al., 2003]. Maps can be downloaded from the web (http://rapidfire.sci.gsfc.nasa.gov). The fires are marked with red and yellow dots.

Figure 4.

HYSPLIT back trajectories ending at Banizoumbou on (a) 15 January 2006, (b) 19 January 2006, (c) 23 January 2006, and (d) 30 January 2006 and at different height levels. The underlying image is a MODIS 10-day composite from the land temperature and fire product. These fire maps accumulate the locations of the fires detected by MODIS on board the Terra and Aqua satellites over a 10-day period. Each colored dot indicates a location where MODIS detected at least one fire during the composition period. Color ranges from red where the fire count is low to yellow where number of fires is large.

[26] It was found that the trajectories arriving in the pronounced aerosol layer at altitudes of 0.5 and 1 km (see Figure 4a) came from the northeast, i.e., directly from the desert region. This confirms our conclusion from the high depolarization ratio, that this layer is dominated by mineral dust particles from the Sahara.

[27] In the layer above, the back trajectory arriving at 3 km altitude indicates that the air masses origin from areas southeast of Niger. These are the regions where biomass burning took place as indicated by the underlying MODIS analysis. Thus, this layer certainly contains a significant fraction of biomass burning aerosol. Similar layers of biomass burning aerosol were observed on most days of lidar measurements during SOP0.

3.2. Moderate Dust Load

[28] The observations of 23 January are chosen as representative for the vertical aerosol distribution of most days during the dry season: moderate dust load in the boundary layer and an elevated layer of biomass burning aerosol above (Figures 3e and 3f). The dust layer of this day was only a few hundred meter thick with a maximum of the extinction coefficient of 0.18 km−1 at around 1.1 km altitude. Its rather low extinction compared to the case of 15 January can be attributed to the natural variability, different source strength, or sedimentation during transport.

[29] At higher altitudes between 2.5 and 4.8 km, a pronounced aerosol layer with extinction coefficients between 0.12 to 0.16 km−1 was detected. This example demonstrates that the extinction coefficients in the dust layer and the BBA layer can be of the same order of magnitude. The particle depolarization ratio δpar in this layer was relatively low with values between 5 and 8%. The significant differences in δpar in the elevated layer and the lowermost layer suggest different microphysical properties of those particles.

[30] The discrimination of aerosol types was again obtained from the depolarization ratio. Former studies showed that biomass burning aerosol has δpar between 3% and 11% [Ferrare et al., 2001; Fiebig et al., 2002; Murayama et al., 2004]. We assume that the elevated layer consists of BBA, while the lower layer with higher δpar is expected to be mineral dust. However, the maximum value of δpar was only 20% which is lower than on 15 January. This suggests either a mixture of dust particles with other aerosol types, possibly of urban origin, or changes in the shape and size distribution of the dust particles.

[31] As before, HYSPLIT back trajectories and MODIS images were used to substantiate our conclusions. Back trajectories shown in Figure 4c indicate that the air masses observed at 3 km altitude came from areas south of Niger where forest fires were observed. The air masses at altitudes at 1 km came from the Saharan desert in the northwest, whereas the air masses arriving at 2 km came also from the desert but remained close to Banizoumbou for about 36 h. Thus, the layer close to the ground can clearly be identified as dust. A mixture with urban aerosol is at least plausible as the back trajectory arriving at 1 km indeed indicates airflow from the northwest where populated places exist. On 15 January (see section 3.1) they came from the northeast and north, directly from the desert region, so that pure dust is more probable.

3.3. Dust and BBA Mixing

[32] An unusual and pronounced aerosol layer was observed on 19 January between 1.5 km and 3 km (Figures 3c and 3d). This layer has high extinction coefficients of up to 0.5 km−1 and particle depolarization ratios of 10% to 15%. Whereas such high extinction values are typical for dust layers from the Saharan desert, the depolarization ratios is comparably low and just slightly larger than expected for biomass burning layers. Thus, the values of δpar indicate a mixture of dust and BBA. From airborne in situ sampling during SOP0 [Formenti et al., 2008] show that mineral dust particles were found also in the BBA layers. The mean relative contribution of dust mass ranged from 93 (±7)% in dust layers to 72 (±16)% in the elevated BBA layers. Therefore, a significant part of this mixed aerosol can be assumed to be dust and the extinction profile can be calculated using a lidar ratio of 55 sr. This choice is confirmed by the values of the integrated extinction from the lidar profile to the AOD measured by the Sun photometer. The lidar integrated extinction is 0.84 and the AOD from AERONET was 0.83. More details on the AOD comparison are described in section 4.1.

[33] Below 1.5 km several distinct layers with extinction coefficients around 0.2 km−1 were observed. The particle depolarization ratio in these lower layers is between 15% and 20%, and is thus larger than in the upper layer. This is an indication of a strong dust component of the aerosol.

[34] Regarding the back trajectories arriving at different heights (see Figure 4b), the lowermost back trajectory at 1 km is approaching from the north and stays close to the ground during the last 48 h. This indicates dust particles from the desert and particles from the local sources at these altitudes. The back trajectories arriving at 2 and 3 km height indicate that the air masses originally came from the central Saharan desert, but were then transported westward over the Sahel region with high fire activity. During most of the transport time, the air masses were descending, so that 1 and 2 days before their arrival at Banizoumbou they were very close to the ground; only during the last 24 h a strong updraft occurred. This transport path supports our statement that dust was lifted up to higher altitudes and was mixed into the biomass burning aerosol layer.

4. Time Series

[35] Three cases of quite different aerosol profiles were discussed in the previous chapter. In two cases, on 15 and 23 January, the dust load was different but the dust was concentrated in the planetary boundary layer as indicated by the high depolarization ratio. Separated from that layer, a pronounced elevated BBA layer was observed on both days. In the third case, on 19 January, the distribution was completely different as mixing of both aerosol types was observed building an optically thick but only moderately depolarizing layer of particles in those altitude ranges where otherwise only biomass burning aerosol was observed. In conclusion it seems that two height ranges can be distinguished above and below an altitude of approximately 1.5–2.0 km. We want to briefly discuss the time series of the optical depth, the mean extinction coefficient and the depolarization ratio for these two height ranges. Though the time period of our lidar measurements was limited to SOP0, we feel that this information gives a deeper insight in the “climatological” conditions during the dry season in the Sahel region. Data from the AERONET are included for comparison.

4.1. Aerosol Optical Thickness

[36] The time series from 14 January to 2 February 2006 of the aerosol optical depth as derived from the vertical integration of the lidar profiles is plotted in Figure 5a. The lidar measurements were compared to the aerosol optical depth (AOD) measured by the AERONET [Holben et al., 2001] Sun photometer at Banizoumbou. The AOD from the lidar is achieved by integrating the extinction profiles up to 6 km. From the surface to the lowermost height of the retrieved lidar profile, which was 225 m above ground, a constant extinction coefficient is assumed, because no ground based in situ measurements of the extinction coefficient was available. If we assume an error of 100% in the lowermost 225 m, the error for the whole profile due to this extrapolation is only 4%. Above 6 km aerosol extinction is assumed to be negligible. The values are plotted against the daily mean level 2.0 AOD at 440 nm, the spectral channel of the AERONET Sun photometer closest to the POLIS wavelength. POLIS AOD is calculated from morning mean extinction profiles or in some cases, especially when operating in the Raman mode, from the nighttime profiles.

Figure 5.

(a) AOD derived from POLIS extinctions profiles compared to AERONET Sun photometer AOD at 440 nm measured at Banizoumbou station, (b) POLIS AOD, (c) maximum extinction, and (d) maximum volume depolarization values below and above 2 km altitude.

[37] In general, the two time series agree well. The agreement of the lidar and the photometer derived optical depths confirms that the wavelength dependence of the aerosol must have been small. This is consistent with the dominating or at least significant contribution of desert dust. The overall variability of the AOD during SOP0 did not change with the move of the lidar from Banizoumbou to Niamey on 30 January. On 10 days the AOD was below 0.5, on 7 days the AOD was between 0.5 and 1 and only on 3 days the AOD was larger than 1. The maximum optical depth was observed on 18 January, when unfortunately no lidar measurements were available. As even the maximum AOD was not larger than 1.2, no day of the SOP0 showed spectacularly high dust load. Inspection of the AERONET time series demonstrates that the highest aerosol load measured during the year 2006 in Banizoumbou was on 8 March with an AOD of 4.1. If we assume that dust storms are associated with AOD larger than 1, the AERONET database shows, that dust storms occur only between 5 and 10 days each year in the region of Banizoumbou, mostly in spring. Insofar our lidar observations during January 2006 give a representative overview over the different dust situations during the dry season in the Sahel.

4.2. Extinction Coefficient and Depolarization Ratio

[38] Figures 5b5d show the AOD (Figure 5b), the maximum extinction coefficient (Figure 5c), and the maximum volume depolarization (Figure 5d) each separated for the atmosphere below and above 2 km height, as derived from our lidar measurements. This allows the distinction of the contributions of dust particles and BBA to the total AOD and the highlighting of differences in the particle properties of the planetary boundary layer and the layer above. The height of 2 km was chosen for all days because the planetary boundary layer was below 2 km during the whole measuring period, except on 19 January. This particular day will be discussed below. The contribution to the AOD from the aerosol content between the real boundary layer height on each day and 2 km is only 0.025 to 0.075 and does not change our conclusions.

[39] Figures 5b5d show that the particles were predominantly present in the lower layer. The AOD, the maximum extinction coefficient and the maximum volume depolarization are generally higher below 2 km. Only in a few exceptions the AOD and the maximum extinction coefficient above 2 km was comparable or even larger. On these days, 23 and 24 January, the dust load in the boundary layer was very low (see also Figures 3e and 3f). On these days the back trajectories arriving in the lower boundary layer came from the west (see Figure 4c for 23 January) and the influence from the desert was relatively low. The larger depolarization in the lower layer suggests that there are always more dust particles in the boundary layer than in the free troposphere, except during a 3-day period starting on 18 January (AOD = 1.2 observed by AERONET), measured by POLIS on 19 and 20 January and decaying on 21 January. Here a minor dust storm transported the dust from the boundary layer up into the free troposphere and mixed it with the biomass burning aerosol layer. As indicated in section 3.3, on 19 January significant amounts of aerosols were transported up to 3.5 km altitude. Therefore, the maximum extinction and depolarization values in Figures 5c and 5d have the same value on this day. But the integrated extinction (Figure 5b) up to 2 km shows that even on this particular day most of the particles are below 2 km.

[40] The general picture of the particle distribution in the Sahel region during the dry season, that can be concluded from the lidar profiles, is that a varying but persistent dust layer is present in the planetary boundary layer and a lofted layer of biomass burning aerosol is constantly present in the free troposphere. Mixing of these particle layers can occur during dust storm episodes, when the dust particles are transported into the free troposphere and get mixed with particles origin from biomass burning.

5. Discussion and Conclusions

[41] During the special observation period SOP0 of the AMMA field campaign, the first time ground-based lidar measurements were performed to characterize the aerosol in the dry season of the annual weather cycle in the Sahel region. The typical aerosol distribution observed in the dry season was dust in the lowermost boundary layer reaching up to 2 km altitude and an elevated layer of biomass burning aerosol between 3 km and 5 km altitude. The dust was transported to Niger from the Saharan desert at low levels by the Harmattan winds. The biomass burning aerosol originated from active forest fires south of Niger and rose to higher altitudes by convection. These aerosol types were clearly identified by the optical property profiles measured by the lidar POLIS. High particles depolarization ratios identify the nonspherical particles in the dust layer while the particles origin from biomass burning only moderately depolarize the laser light.

[42] The lidar ratio of this biomass burning aerosol was found to be 75 sr ± 15 sr which is at the upper boundary of the range of lidar ratios measured for biomass burning aerosol in previous experiments. Lidar ratios of BBA have been measured at quite different sites including fresh and aged particles. For aged aerosol from Canadian forest fires observed over Europe [Wandinger et al., 2002] a range between 40 and 80 sr was found. During a Saharan dust outbreak toward Europe lidar ratios between 50 and 80 sr at 532 nm have been observed by EARLINET and AERONET stations [Müller et al., 2003]. During a biomass burning episode in Greece in 2001, Balis et al. [2003] measured a mean lidar ratio of 62 sr at 355 nm for fresh aerosol. Most recently, A. Papayannis (personal communication, 2007) measured lidar ratios at 355 nm between 60 and 80 sr for very fresh biomass burning aerosol during the forest fires over Athens at the end of June 2007. Trajectory analyses showed that the biomass burning aerosol observed over Banizoumbou came from central Africa and was thus from freshly burned material. During SOP0 BBA and dust mineralogical composition and optical properties were also measured on board FAAM over West Africa by, e.g., Chou et al. [2008] and Formenti et al. [2008]. Their results show that BBA mainly consist of small particles and that the BBA layers, also the elevated layers, always includes a small contribution of dust particles. The latter is confirmed by the lidar particle depolarization ratios, that were close to 10% in the BBA layer, for example on 23 January (see Figures 3e and 3f).

[43] The lidar ratio observed in the dust layer at Banizoumbou by POLIS was 55 sr ± 5 sr. From a preliminary analysis of SAMUM measurements the lidar ratio of dust particles close to the dust source was also determined as 55 sr ± 6 sr [Müller et al., 2007]. From an episode, when Saharan dust particles were transported toward Europe [Ansmann et al., 2003] the lidar ratio was 59 sr ± 11 sr. Thus, the lidar ratio observed during AMMA indicates that we can assume mainly pure dust in the lower boundary layer at Banizoumbou. Only small amounts of local urban aerosol may be present as well. The composition of the aerosol close to the ground has been analyzed in detail from in situ measurements performed at Banizoumbou. An overview over all in situ measurements at Banizoumbou during SOP0 is given by Rajot et al. [2008].

[44] This dust layer can be regarded as typical for the Sahel region during the dry season before the monsoon onset. The mean extinction coefficients for dust were around 0.2 km−1. For days with direct import of dust loaded air masses from the Saharan desert, the extinction coefficient could exceed 1.5 km−1. However, the vertical extent of such layers is only a few 100 m. We also found layers mixed of dust and biomass burning aerosols. The extinction coefficients can be of the same order as the pure dust layers, but the particle depolarization ratio is lower.

[45] Our measurements have demonstrated that a lidar system such as POLIS is a good candidate for routine observations in severe environments. In particular, the depolarization mode is of benefit for the identification of dust particles. The only limitation is that in its present state, POLIS cannot be operated unattended. Therefore, the representativeness of our measurements, though covering about 120 h, cannot be assessed definitely. Comparisons with long-term AERONET data, however, suggest, that the period of January 2006 did not differ significantly from other situations, at least with respect to the aerosol optical depth.

[46] In May and June 2006 a comprehensive characterization of the Saharan dust plume close to the source region and their radiative effects was performed in the framework of the German Saharan Mineral Dust Experiment SAMUM. The measurements took place in Morocco, and POLIS was one out of four deployed lidar systems. In contrast to the observations in Niger the vertical extent of the dust layer was typically 5 km, and no BBA was observed. The detailed results of SAMUM will be published in a special issue of Tellus. Aerosol distributions, more comparable to the conditions during AMMA, have been observed during the second phase of the SAMUM project at the Cape Verde Islands in January and February 2008. Preliminary results show distinct layers of mineral dust particles in the lower troposphere not higher than up to 2 km asl. BBA was observed frequently at higher altitudes up to 5 km asl.

[47] The assessment of vertically resolved extinction coefficient is one of the key parameters for the calculation of components of the radiation budget. The lidar ratio and the particle depolarization ratio can be used to provide an estimate of the single scattering albedo. Single scattering albedo and extinction coefficients can be used to calculate heating rates and thus have direct influence on climate modeling. As the assessment of aerosol properties from spaceborne radiometers still have problems, in particular over land surfaces, lidar measurements can be used for improving present retrieval algorithms by providing data for validation.

Acknowledgments

[48] We would like to thank the anonymous reviewers for their fruitful comments and suggestions. We thank J.-L. Rajot for establishing and maintaining the Banizoumbou station and D. Tanré for the implementation of the Banizoumbou AERONET site. This project was funded by the European Union under grant 4089. On the basis of a French initiative, AMMA was built by an international scientific group and is currently funded by a large number of agencies, especially from France, UK, US, and Africa. It has been the beneficiary of a major financial contribution from the European Community's Sixth Framework Research Programme. Detailed information on scientific coordination and funding is available on the AMMA International web site http://www.amma-international.org.

Ancillary