We investigate the discrepancies in measurements of light extinction and extinction-to-backsatter ratio (lidar ratio) of desert dust with CALIPSO and ground-based lidar systems. Multiwavelength polarization Raman lidar measurements in the Saharan dust plume performed at Praia, Cape Verde, 15.0°N, 23.5°W, during SAMUM–2 in June 2008 were analyzed and compared to results of nearby CALIPSO overflights. The particle extinction coefficients and thus the optical depth are underestimated in the CALIPSO products by about 30% compared to Raman lidar measurements. A pre-defined lidar ratio of 40 sr at 532 nm is used for mineral dust in the CALIPSO algorithms in agreement with values of 41 ± 6 sr found from constrained retrievals. However, the ground-based lidar observations show much larger values of the order of 55 ± 10 sr. The discrepancies can be explained by the influence of multiple scattering which is ignored in the CALIPSO retrievals. Based on recent observations of the size distribution of dust particles from airborne in-situ observations during SAMUM–1, our model calculations show that the multiple-scattering-related underestimation of the extinction coefficient in the CALIPSO lidar signals ranges from 10%–40%. We propose a method to overcome this underestimation.
 The Cloud–Aerosol Lidar with Orthogonal Polarization (CALIOP) aboard the Cloud–Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) satellite is a conventional backscatter lidar with observation channels at 532 and 1064 nm [Winker et al., 2009]. Retrieval algorithms for Level 2 products require the assumption of an extinction-to-backscatter ratio (lidar ratio) to calculate profiles of particle backscatter and extinction coefficients from the Level 1 product of attenuated backscatter. For aerosol observations, a look-up table with lidar ratios at 532 and 1064 nm for six types of aerosol (dust, smoke, clean marine, polluted continental, clean continental, polluted dust) is applied [Omar et al., 2009]. Aerosol types and thus lidar ratios are assigned according to the geographical location of the measurement and the detected integrated attenuated backscatter and approximate particulate depolarization ratio. The six aerosol types are represented by aerosol models with a prescribed bi-modal size distribution (fine and coarse mode) and a characteristic complex refractive index for each mode and wavelength [see Omar et al., 2009, Table 1]. The models are mainly based on the analysis of Aerosol Robotic Network (AERONET) observations [Omar et al., 2005], with some adjustments to generate lidar ratios that are in agreement with observations [Winker et al., 2009; Omar et al., 2009]. Except for mineral dust, lidar ratios follow directly from the aerosol models by Mie scattering calculations.
 The CALIPSO dust model is based on scattering calculations with the discrete dipole approximation (DDA) technique to account for a realistic mixture of particles with irregular shapes [Omar et al., 2005]. The underlying bi-modal size distribution with mode radii of 0.12 μm (fine mode) and 2.83 μm (coarse mode) shows a considerable amount of fine-mode particles (volume fraction of 22%) leading to an effective radius of the size distribution of 0.4 μm. Lidar ratios of 40 sr at 532 nm and 30 sr at 1064 nm were derived and applied in the CALIPSO Level 2 Version 2 algorithm [Winker et al., 2009]. The lidar ratio at 1064 nm could not be justified experimentally [e.g., Liu et al., 2008] and has been replaced by a value of 55 sr in the Version 3 algorithm [Omar et al., 2009]. Layer-mean lidar ratios can be estimated from CALIPSO measurements directly, when optically thick aerosol layers above clear air or opaque water clouds are observed and thus the layer optical depth can be derived. With the optical depth constraint, Liu et al.  found 532-nm lidar ratios of 41 ± 6 sr, which are in good agreement with the model results, at various locations in the Saharan dust plume off the west coast of Africa.
 Although CALIPSO dust retrievals seem to be self-consistent, comparisons with ground-based measurements of Saharan dust show considerable discrepancies, especially with respect to the 532-nm lidar ratio. Lidar observations in pure mineral dust in the Saharan source region during SAMUM–1 (Saharan Mineral Dust Experiment, Southern Morocco, May–June 2006 [Heintzenberg, 2009]) yielded distinct lidar-ratio values of 55 ± 7 sr at 532 nm [Tesche et al., 2009a]. A very similar mean value of 54 ± 10 sr was found in lofted dust layers in the Saharan outflow region over the North Atlantic during SAMUM–2 (Cape Verde, May–June 2008). The variability of lidar ratios reported for long-range dust transport towards Europe and North America is larger and ranges from 30 to 80 sr [e.g., Mattis et al., 2002; Mona et al., 2006; Papayannis et al., 2008; Liu et al., 2008]. One reason might be the mixing of dust with marine airmasses, local pollution and smoke during transport [Papayannis et al., 2008]. Nevertheless, observations of strong and thus mainly undisturbed Saharan dust intrusions over Europe in the framework of EARLINET (European Aerosol Research Lidar Network) on average also revealed lidar ratios of the order of 50–60 sr [e.g., Müller et al., 2003, 2007; Pappalardo et al., 2010]. We attribute the different findings to the influence of multiple scattering on the space-borne lidar observations. As airborne in-situ measurements during SAMUM–1 showed, the size of dust particles is much larger than assumed in the CALIPSO dust model [Weinzierl et al., 2009]. The large particles cause a non-negligible amount of multiple scattering in lidar observations from space and reduce the single-scattering extinction and thus the lidar ratio.
 In this paper, we investigate the multiple-scattering effect quantitatively based on comparisons of quasi-simultaneous measurements of ground-based multiwavelength polarization Raman lidar and CALIOP in pure Saharan dust at the Cape Verde Islands in May–June 2008. Section 2 briefly introduces the instruments and the multiple-scattering model used in this study. The approach to the analysis of multiple-scattering effects is described in Section 3, and case-study results are presented in Section 4. The paper closes with some conclusions for an improvement of the CALIPSO retrieval in Section 5.
2. Lidars and Multiple-Scattering Model
 The six-wavelength aerosol lidar BERTHA (Backscatter, Extinction, lidar-Ratio, Temperature, and Humidity profiling Apparatus) of the Leibniz Institute for Tropospheric Research (IfT) measures elastically backscattered signals at the emitted laser wavelengths of 355, 400, 532, 710, 800, and 1064 nm, and Raman signals of water vapor at 660 nm and of nitrogen at 387 and 607 nm [Althausen et al., 2000; Tesche et al., 2009a]. The aerosol depolarization ratio is determined at 710 nm and can be extrapolated to 532 nm based on a parameterization obtained from collocated observations with a system that measures at 532 nm [Freudenthaler et al., 2009]. The divergence of the co-aligned laser beams is of the order of 100 μrad. Backscattered light is collected with a 0.53–m Cassegrain telescope with a receiver field of view (RFOV) of 800 μrad. Profiles of the volume extinction coefficient of particles at 355 and 532 nm are calculated from the Raman signals. Signals are averaged over 60 to 170 min and vertically smoothed with window lengths of 300 to 1260 m (increasing with height) to reduce the statistical error to values of 10%–25%. Systematic uncertainties caused by the removal of Rayleigh-scattering and air-density effects from the backscatter signals are of the order of 5%–10%. A detailed description of the techniques for the analysis of Raman lidar observations can be found in the work of Ansmann and Müller .
 CALIOP is an elastic-backscatter lidar that orbits the Earth at a height of 705 km and emits light at 532 and 1064 nm. The laser has a beam divergence of 100 μrad corresponding to a spot diameter of 70 m at the Earth's surface. The RFOV of the 1-m telescope is 130 μrad. The system features three measurement channels. Elastically backscattered light at 532 nm is split into a parallel and a cross-polarized signal with respect to the linearly polarized laser light. The total backscattering signal is detected at 1064 nm. The volume depolarization ratio at 532 nm is calculated from the ratio of the cross-polarized and parallel polarized signals. Backscatter-coefficient profiles of aerosol and cloud layers at 532 and 1064 nm are computed from the respective attenuated backscatter profiles (i.e. calibrated range-corrected lidar signals) by applying pre-defined lidar ratios for the attenuation correction (see Sec. 1). Extinction-coefficient profiles are obtained by multiplying the backscatter-coefficient profiles with the pre-defined lidar ratios. Level 2 Version 2 aerosol profiles are provided with a horizontal resolution of 40 km. Winker et al.  gives an overview of the instrument and the retrieval algorithms and refers the reader to the more detailed descriptions in a special collection of articles published in the Journal of Atmospheric and Oceanic Technology.
2.3. Multiple-Scattering Model
 For the multiple-scattering calculations we used the model described by Hogan and Battaglia . The model is fast and well suited to simulate space-borne lidar observations, since it considers high-order and wide-angle scattering as well as temporal pulse stretching. Input parameters are the characteristics of the lidar system in terms of altitude of the platform, laser wavelength, beam divergence, and RFOV, and the properties of the scattering medium, in particular the effective particle radius reff and the extinction profile.
 For our study we chose measurements taken during the second field phase of SAMUM–2 in May–June 2008. Comparisons for SAMUM–1 in the Saharan dust source region were not possible, because CALIOP started its operation shortly after the end of the campaign in June 2006. SAMUM–2 was dedicated to observations in the outflow regime of Saharan dust over the North Atlantic. The major field phase took place in winter (January–February 2008), when dust is transported westward at low altitudes and mixed with marine aerosol and smoke from the African continent most of the time. In contrast, during summer undisturbed dust plumes travel at larger heights from the Sahara towards the Caribbean and South America. Therefore, only the observations during the summer campaign of SAMUM–2 provide the opportunity to compare CALIOP results to reliable Raman lidar measurements within an environment that is dominated by pure mineral dust.
 The measurements with BERTHA were performed between 24 May and 16 June 2008 at the airport of Praia, Cape Verde, 15.0°N, 23.5°W. Observations typically showed 4–5 km deep dust layers above the marine boundary layer. CALIPSO passed the ground station 13 times at distances <500 km. We used HYSPLIT backward trajectory analysis (available at http://ready.arl.noaa.gov/HYSPLIT.php) to investigate the flow pattern and to make sure that the chosen segment of the CALIPSO track and the temporal segment of the ground-based measurement provide reasonable comparisons. For the case study, we chose the observations on 3, 11, and 15 June 2008 which are characterized by high dust load, sufficient signal-to-noise ratio of the CALIOP measurements, and low influence of clouds.
 No airborne observations and thus no information on in-situ aerosol size distributions are available for the SAMUM–2 summer campaign. However, the analysis of the 25-day lidar time series yielded similar results in terms of lidar ratio and particle depolarization ratio as were found for fresh dust during SAMUM–1 [Tesche et al., 2009a; Freudenthaler et al., 2009]. CALIPSO measurements also reveal that there is little change in the optical properties of dust during the first 3–5 days of transport over the Atlantic in summer [Liu et al., 2008]. Therefore, we regard the airborne in-situ measurements in pure dust taken over Southern Morocco in the same season during SAMUM-1 as representative for the Saharan dust plume in general. Effective radii between 1.2 and 6.8 μm with a mean value of 3.2 ± 1.2 μm and a negligible contribution of fine-mode particles were obtained [Weinzierl et al., 2009].
 For the multiple-scattering calculations we chose effective radii of 3 and 6 μm as representative input values. The calculations showed that the multiple-scattering effect can be neglected for ground-based dust measurements with BERTHA. Therefore, we regarded BERTHA extinction-coefficient profiles to be the truth and combined them with the CALIOP observing geometry to compute profiles of the effective extinction coefficient that would be seen by CALIOP provided that dust particles of the pre-defined effective radius had caused the true extinction profile.
4. Case Studies
Figure 1a shows the comparison of backscatter-coefficient profiles at 532 nm taken with BERTHA (light green) in the night of 10–11 June, 2134–0001 UTC, and with CALIOP during the overpass at 0309 UTC at a distance of 127 km. The measurement shows the presence of a dust layer up to 5.5 km height. BERTHA profiles were cloud-screened before averaging. CALIOP profiles are cut a the top of the marine boundary layer because of clouds as well. CALIPSO Level 2 (Version 2) 40-km aerosol product profiles number 112–115 (thin grey lines) were averaged to yield a mean profile (light blue). The ground-based and space-borne profiles are in very good agreement. Integrated backscatter coefficients for the dust layer from 0.8 to 5.2 km (dashed horizontal lines) agree within 5%.
 Lidar-ratio and depolarization-ratio profiles are presented in Figure 2. Lidar ratios measured with BERTHA in the center of the dust plume are about 30% larger than the prescribed lidar ratios taken in the CALIPSO retrieval. The 532-nm BERTHA particle depolarization ratio clearly indicates the presence of mineral dust with values around 0.3 which are similar to the ones observed by Freudenthaler et al.  during SAMUM–1. The CALIPSO volume depolarization ratio profile was obtained by averaging the appropriate Level 1 (Version 2) products. Volume depolarization ratios derived from CALIOP and BERTHA measurements showed reasonable agreement for all comparisons we performed.
Figure 1b presents profiles of the extinction coefficients and respective integrated values (optical depth) measured with BERTHA (light green) and CALIOP (light blue) together with profiles resulting from multiple-scattering calculations (ms). The BERTHA profiles are vertically smoothed and thus show somewhat less structure than the backscatter profiles. The CALIPSO integrated extinction coefficient (optical depth) is 28% lower than the one measured with BERTHA. An extinction reduction by 20% and 30% with respect to the BERTHA profile is found from multiple-scattering calculations for the CALIOP geometry with effective radii of 3 (red) and 6 μm (magenta), respectively, and can thus explain the deviation. For comparison, the multiple-scattering effect is of the order of 4% for an effective radius of 0.4 μm (dark blue) as assumed in the CALIPSO dust model. If the CALIPSO extinction profile is converted with a dust lidar ratio of 55 sr (dark green) or 60 sr (not shown), the optical depth is in agreement with the BERTHA observation within ±5%.
Table 1 summarizes the findings of all three comparisons on 3, 11, and 15 June 2008, when CALIPSO passed the lidar station in distances of 171, 127, and 478 km, respectively. Integrated backscatter coefficients of BERTHA and CALIPSO deviated by ≤15% in all cases. The differences are assumed to be determined by the atmospheric variability rather than by systematic retrieval errors. Taking the backscatter deviation into account, the underestimation of the optical depth by CALIPSO is between 25% and 35% and can be explained with the multiple-scattering influence caused by particles with an effective radius between 4.5 and 6 μm. Using a lidar ratio of 55 sr instead of 40 sr to convert CALIPSO backscatter into extinction coefficients leads to a reasonable agreement of ground-based and space-borne values of optical depth in all cases.
Table 1. Height Integrals of Backscatter (in sr−1) and Extinction for Three Comparison Days in June 2008a
The height range of integration is stated in the header. The spatial distance between ground station and CALIPSO ground track is 171, 127, and 478 km, respectively. Times of observation are shifted by 13, 4, and 17 hours, respectively. Percentage values are related to the BERTHA measurements (100%, see also Figure 1).
Height range of integration
CALIPSO, Level 2
Integrated Extinction (Optical Depth)
CALIPSO, Level 2
CALIPSO, backscatter × 55 sr
BERTHA, ms, reff = 0.4 μm
BERTHA, ms, reff = 3.0 μm
BERTHA, ms, reff = 6.0 μm
 Recent findings on the size distribution of Saharan dust imply that multiple scattering is not negligible in space-borne lidar observations of mineral dust. By taking the multiple-scattering effect into account, we can explain the discrepancies found between ground-based measurements of lidar ratios in pure Saharan dust, which are about 55 sr on average, and the value of 40 sr applied in the CALIPSO retrievals. The latter value is an effective lidar ratio which accounts for the increased atmospheric transmission caused by multiple scattering and gives reasonable backscatter coefficients that compare well with ground-based observations. However, if the same value of 40 sr is applied to convert backscatter into extinction coefficients, a systematic underestimation of extinction and optical depth by 25%–35% is introduced. This artifact can easily be overcome by applying two different look-up values for the lidar ratio of mineral dust in the CALIPSO retrieval algorithm, i.e., an effective value of 40 sr for the backscatter retrieval and a single-scattering value of 55 sr, which is typical for pure dust, for the backscatter-to-extinction conversion. Dust aerosol is usually well identified in the CALIPSO aerosol mask based on threshold values of the approximate particulate depolarization ratio [Omar et al., 2009]. Particle depolarization ratios >0.25 clearly indicate the presence of pure mineral dust composed of large non-spherical particles [Freudenthaler et al., 2009], for which the multiple-scattering effect becomes important and a second look-up value of the lidar ratio should be used. Following the approach of Tesche et al. [2009b] for the separation of dust in mixed aerosols, further refinements of the look-up tables with respect to mixed dust aerosols should be investigated.
 The SAMUM research group is funded by the Deutsche Forschungsgemeinschaft (DFG) under grant FOR 539. We thank Robin Hogan for providing the multiple-scattering code.