Optical properties of long-range transported Saharan dust over Barbados as measured by dual-wavelength depolarization Raman lidar measurements

Abstract. Dual-wavelength Raman and depolarization lidar observations were performed during the Saharan Aerosol Long-range Transport and Aerosol-Cloud interaction Experiment in Barbados in June and July 2013 to characterize the optical properties and vertical distribution of long-range transported Saharan dust after transport across the Atlantic Ocean. Four major dust events were studied during the measurements from 15 June to 13 July 2013 with aerosol optical depths at 532 nm of up to 0.6. The vertical aerosol distribution was characterized by a three-layer structure consisting of the boundary layer, the entrainment or mixing layer and the pure Saharan dust layer. The upper boundary of the pure dust layer reached up to 4.5 km in height. The contribution of the pure dust layer was about half of the total aerosol optical depth at 532 nm. The total dust contribution was about 50–70 % of the total aerosol optical depth at 532 nm. The lidar ratio within the pure dust layer was found to be wavelength independent with mean values of 53 ± 5 sr at 355 nm and 56 ± 7 sr at 532 nm. For the particle linear depolarization ratio, wavelength-independent mean values of 0.26 ± 0.03 at 355 nm and 0.27 ± 0.01 at 532 nm have been found.


Introduction
Aerosol particles play a key role in the Earth's climate system and affect the Earth's radiation budget 15 in two different ways; directly by interacting with solar and terrestrial radiation (scattering and absorption) and indirectly by acting as cloud condensation nuclei and therewith influencing the clouds microphysical and optical properties and the clouds lifetime. Up to now the impact of aerosols on the global climate system is not fully understood (Forster and et al., 2007;Penner et al., 2011;Boucher et al., 2013). One main reason is the strong variability of aerosols. The sign and the magnitude of 20 the radiative forcing crucially depends on the vertical distribution of aerosols, their microphysical 1 properties and chemical composition, the reflectance of the underlying surface and the occurrence and amount of clouds (Forster and et al., 2007). However, knowledge of the temporal and vertical aerosol distribution on the global scale is limited (Penner and et al., 2001;IPCC, 2013). Additionally, significant sources of uncertainty result from deficits of satellite-based measurements in the deter-25 mination of global mean AOD (Su et al., 2013), and from the insufficient knowledge of the impact of mixing, aging processes and transport on the aerosol optical and microphysical properties.
Advanced lidar systems like Raman lidar systems (Ansmann et al., 1990(Ansmann et al., , 1992 or high spectral resolution lidar (HSRL) systems (Shipley et al., 1983;Shimizu et al., 1983;Piironen and Eloranta, 1994) with polarization sensitive channels (Sassen et al., 1989;Freudenthaler et al., 2009) (Winker et al., 2009) and has only limited capability to distinguish different types of aerosols (Omar et al., 2009). In contrast, 35 with the Cloud-Aerosol Transport System (CATS) currently flying on board the international space station (ISS) and the future ESA satellite mission EarthCARE polarization sensitive HSRL systems are deployed, having the potential to classify different aerosol types (Burton et al., 2012;Groß et al., 2013Groß et al., , 2015. However, current classification schemes for EarthCARE lidar measurements are mainly based on measurements of pure and fresh aerosol types (Groß et al., 2011a(Groß et al., , 2014Illingworth and 40 et al., 2014). But as the optical properties are related to the microphysical properties like particle size, particle shape and chemical composition (Gasteiger et al., 2011b, a), aerosol aging, mixing and modification during transport can have an impact on the lidar derived optical properties, as well as on their wavelength dependence. For example, for measurements of the lidar ratio over Greece Amiridis et al. (2009) found that both, the value and the wavelength dependence of the lidar ratio 45 of biomass burning aerosols may change with aerosol lifetime. Thus, possible changes of the lidar derived optical properties have to be investigated and considered for proper aerosol classification.
Mineral dust is a major component of the atmospheric aerosol (Haywood and Boucher, 2000;Forster and et al., 2007) with the Saharan desert being the most important source of mineral dust (Goudie and Middleton, 2001;Washington et al., 2003;Shao et al., 2011). Once lifted in the air, 50 mineral dust can be transported over thousands of kilometers (Goudie and Middleton, 2001;Liu et al., 2008) exposed to the effects of aging and mixing. These effects change the optical, microphysical and cloud condensation properties. Coatings on mineral dust particles and mixing with other aerosols change the optical properties (Nousiainen, 2009;Redmond et al., 2010) and thus alter their radiative impact (Bauer et al., 2007). For example, biomass burning aerosols and mineral dust 55 may become internally mixed when aging together (Hand et al., 2010) and thus change their size distribution, optical properties, hygroscopicity and their ability to act as cloud condensation nuclei.
From measurements close to the dust source regions in comparisons to measurements in dust plumes over Cape Verde Weinzierl et al. (2011) found an indication of sedimentation of large particles in Saharan dust plumes during transport although sedimentation of large super-micron dust particles was 60 less pronounced than expected from Stokes gravitational settling. Yang et al. (2013) assume a shapeinduced particle sedimentation from measurements of transported dust with the space-based lidar system onboard the Cloud-aerosol Lidar with Orthogonal Polarization (CaLIOP) satellite mission (Winker et al., 2009). Wiegner et al. (2011) found an increase of the mean particle linear depolarization ratio at 355 nm of an aged Saharan dust plume over Central Europe compared to values 65 measured in fresh Saharan dust plumes (Freudenthaler et al., 2009;Groß et al., 2011b). Up to now the mechanism and magnitude of dust aging is unknown, and whether and how it influences the optical properties of dust.
In this work we present dual-wavelength Raman and depolarization lidar measurements of longrange transported Saharan dust over Barbados. Our study includes a general investigation of aerosol 70 layering and optical depth during our measurement period as well as the characterization of the Saharan dust layer and marine boundary layer by means of the lidar ratio and the particle linear depolarization ratio. These observations are crucial to investigate possible age-induced changes in the intensive lidar optical properties necessary for lidar based aerosol classification schemes. The measurements were performed during the SALTRACE closure experiment. A general description of 75 the SALTRACE campaign, our lidar measurements and data analysis is given is Sect. 2. The results are presented in Sect. 3, and discussed in Sect. 4. Section 5 summarizes this work.

SALTRACE
In June and July 2013 the Saharan Aerosol Long-range Transport and Aerosol-Cloud interaction Ex-80 periment (SALTRACE, http://www.pa.op.dlr.de/saltrace/index.html) took place. SALTRACE was designed as a closure experiment combining ground-based lidar, in-situ and sun photometer instruments, with airborne aerosol and wind lidar measurements of the research aircraft Falcon of the Deutsches Zentrum für Luft-und Raumfahrt (DLR), satellite observations and model simulations.
The main ground-site during SALTRACE was on Barbados where extensive lidar were performed.

85
Barbados is an optimal location to characterize long-range transported dust after transport across the Atlantic Ocean. In addition, the 50 year Barbados dust record (Prospero et al., 1970) provides long-term information on year to year variability of trans-Atlantic dust transport to the Caribbean.
The SALTRACE project continues the work started with the SAMUM-1 and SAMUM-2 (Ansmann et al., 2011) which aimed for characterizing Saharan mineral dust in the source regions and 90 at different stages of dust lifetime. During SALTRACE particular focus was drawn on aerosol aging and mixing, and on aerosol removal processes. Therefore the physical, chemical and optical properties of the long-range transported Saharan dust layers were characterized in-depth to study the impact of long-range transported dust on the Earths radiation budget, clouds and precipitation. During SALTRACE, ground-based measurements at Barbados were performed at two main locations: 95 ground-based in-situ measurements were made at the very eastern edge of the island at Ragged Point, whereas the lidar measurements were carried out at the Caribbean Institute of Meteorology and Hydrology (CIMH) at the south-western side of Barbados (13.14 • N, 59.62 • W). Sun photometer measurements were performed at both measurement sites. For this study we use the AERONET CIMEL (Holben et al., 1998) measurements "Barbados_SALTRACE" performed at CIMH. SALTRACE 100 measurements were carried out between 10 June and 15 July with the main closure experiments taking place between 20 June to 12 July 2013.

POLIS lidar system
In this work we present measurements of the small portable Raman and depolarization lidar sys-  . Thus profiles of the particle extinction coefficient α p and backscatter coefficient β p , of the lidar ratio S p , and of the volume and particle linear depolarization ratio δ v and δ p at 355 and 532 nm can be retrieved. The full overlap of POLIS is at 110 about 200 to 250 m depending on system settings. The range resolution of the raw data is 3.75 m; the temporal resolution is 5-10 s depending on atmospheric conditions. The repetition rate of the frequency doubled and tripled Nd:YAG laser is 10 Hz with a pulse energy of 50 mJ at 355 nm and 27 mJ at 532 nm.

Data analysis 115
The particle extinction coefficient α p is retrieved from the Raman signals at 387 and 607 nm (Ansmann et al., 1990), the particle backscatter coefficient β p is derived from combined Raman and elastically backscattered lidar returns at 355/387 and 532/607 nm (Ansmann et al., 1992). The height dependent lidar ratio S p = α p /β p can be derived from the ratio of both properties. Due to the low signal-to-noise ratio of the Raman channels during daytime, these measurements were restricted to 120 night-time only. Furthermore, typical temporal averaging of one hour is necessary for analyzing α p and S p to achieve a sufficient signal-to-noise ratio. The temporal stability of the atmosphere within this time period has been validated by assessing the temporal evolution of the range corrected signal Pr 2 over the whole smoothing period. A typical vertical smoothing of ≈ 940 m (250 rangebins) is applied to further increase the signal-to-noise ratio. The errors of the retrieved optical properties are 125 calculated according to Groß et al. (2011c).
From the co-and cross-polarized elastically backscattered signals the volume linear depolarization ratio δ v and the particle linear depolarization ratio δ p (Biele et al., 2000;Freudenthaler et al., 2009) 4 are derived. The relative calibration factor of both polarization channels was determined with the ±45 • calibration method (Freudenthaler et al., 2009) by manually rotating the receiver optics behind 130 the telescope. Although the signal-to-noise ratio of the elastic channels is much better than for the Raman channels the same temporal average was used for the analysis of the nighttime Raman and depolarization measurements to get comparable results. The vertical average of the elastic signals is typically ≈ 550 m (150 rangebins), otherwise the vertical smoothing length is specified in the text.
Details of the depolarization calibration and system performance can be found by Freudenthaler 135 et al. (2009,2015). The error calculation of δ v and δ p was done analogue to Freudenthaler et al. (2009).
To determine the dust contribution within the boundary layer and the intermediate layer, we determined the profile of the dust backscatter coefficient applying a procedure based on the work of Shimizu et al. (2004) and described by Tesche et al. (2009a) and Groß et al. (2011a) assuming a two-140 type mixture of dust and marine aerosols (based on coordinated in-situ measurements). The linear depolarization ratios used as input for the aerosol type separation are set to δ d = 0.30 at 532 nm for dust and δ nd = 0.02 for marine aerosols (relative humidity ≥40%) according to the findings for pure Saharan dust and marine aerosols (Freudenthaler et al., 2009;Groß et al., 2011b). The dust extinc- is taken from Tesche et al. (2009b) and is in good agreement with the mean S p values we find for long-range transported Saharan dust during SALTRACE.

Dust source regions and transport
To identify dust source regions and transport way and time, we use a combination of backtrajectorie calculation and satellite observations. The trajectories were calculated with the Hybrid Single Parti-150 cle Lagrangian Integrated Trajectory (HYSPLIT) model (Draxler and Rolph, 2012)  at Barbados are identified from infra-red (IR) dust index images calculated from Meteosat Second Generation (MSG) Spinning Enhanced Visible and Infra-Red Imager (SEVIRI) observations. As described in detail in Schepanski et al. (2007), active dust sources are identified and recorded on a 1 • × 1 • map covering Africa north of 10 • N.

General overview
During SALTRACE we observed a sequence of dust events with Saharan air masses being transported with easterly winds over the Atlantic to Barbados. The dust episodes often lasted several for days and provided, apart from shallow cumulus clouds, optimal measurement conditions in the dry and aerosol rich air. The dust events were interrupted by wet periods with strong convective activity 165 and precipitation. Here we focus on the analysis of four major dust events. CIMEL sun-photometer measurements, (c) the fraction of pure dust optical depth (orange) and total dust optical depth (brown stars) to the total AOD, and (d) aerosol optical depth (AOD) at 500 nm derived from CIMEL sunphotometer measurements (dark green) and at 532 nm derived from POLIS lidar measurements (light green) and optical depth of the pure dust layer (orange) and of the whole dust contribution in the atmospheric column (brown stars) at 532 nm derived from POLIS lidar measurement.
The aerosol optical depth (AOD) at 500 nm and 532 nm during these major dust events reached values of up to 0.55. The corresponding Angström Exponent (AE) between 440 nm and 870 nm showed very low values of 0.2 and lower. The overall aerosol situation was characterized by a three layer structure (Fig. 1). The optical properties of the boundary layer (up to 0.5 to 1.0 nm) were 170 6 mostly dominated by marine aerosols, except during the first and last measurement days. At heights from about 1.0 to 2.0 km the aerosol layer was composed of a mixture of predominantly dust and marine aerosols. This layer showed high variability with respect to aerosol load and mixture. During SALTRACE almost all cloud processes in the lower troposphere took place within this layer. Above this intermediate layer a Saharan dust layer was present almost permanently during our measurement 175 period, except on 8th and 9th July when Tropical Storm "Chantal" dominated the weather situation.
During the main Saharan dust events this uppermost dust layer showed AOD values of about 0.2 at 532 nm, in some cases the AOD at 532 nm even reached values of more than 0.3. The contribution of this pure Saharan dust layer to the total AOD at 532 nm usually ranged between 30 and 60 %, in some cases up to 80 %. The total contribution of Saharan dust to the total AOD at 532 nm was 50-180 80 %, except during Tropical Storm Chantal, when the Saharan dust contribution to the total AOD at 532 nm was only 20 %. An overview over the vertical layering and the AOD is given in Fig. 1.

Case Studies
We present four case studies, which are representative for the four dust events that occurred during the core period of SALTRACE. The date and time of the chosen cases studies are 20 June (23:00- For the four case studies (20 June, 27 June, 1 July, 11 July 2013), different dust sources regions are found to be active (Fig. 3). A brief overview on the dust contributing source regions and the meteorological regime resulting into dust uplift will be given in the following. HYSPLIT backtrajectories are analyzed to identify the dates on which the Saharan air mass observed over Barbados 195 where likely over dust source regions over North Africa. For the first two cases (20 and 27 June), dust source activation over North Africa was dominated by the Harmattan flow (a dry and dusty trade wind over West Africa). The latter two July-cases (1 and 11 July) show an increase in deep convective activity and Haboobs (heavy dust storm) become a more frequent dust uplift mechanism compared to the June-cases.   Layer (SAL), the relative humidity is low (notice that lidar measurements and radiosonde measurements have an offset of about 2 h in this case study) and the air masses were transported from mainly north-easterly directions. In the lowermost height level the relative humidity shows values between 60-80 %. For the analysis of the vertical distribution of the extinction coefficient, the lidar ratio and the particle linear depolarization ratio (Fig. 6)   in the SAL is low with values < 40 % and the temperature profile shows a weak inversion at the 240 lower edge of the SAL. The dusty air masses arrived from mainly easterly directions. Within the SAL α p was about 0.08 km −1 and decreased at heights above ≈ 3.5 km. S p in the SAL shows mean values of 56 ± 5 sr at 532 nm and 55 ± 7 sr at 355 nm, the corresponding δ p shows wavelength independent values of 0.26 ± 0.01 at 532 nm and 0.26 ± 0.03 at 355 nm (Fig. 7). Within the boundary   The top of the SAL observed over Barbados was at about 4.8 km height. Figure 10 gives an overview of the night measurements from 10 to 11 July (00:00-01:00 UTC). High intensity of the range corrected signal was found in the boundary layer below about 0.5 km. The lowest signal inten- In the boundary layer we find wavelength independent mean S p values of 35 sr with corresponding δ p of 0.1±0.01 at 532 nm and 0.14±0.01 at 355 nm, indicating a certain amount of dust mixed into the boundary layer.

285
The main findings of the four case studies are summarized in Table 1.

General findings
The mean values of S p and δ p within the SAL and within the boundary layer are shown in Fig. 13 and summarized in Table 2 Europe (Wiegner et al., 2011) furthermore enable to investigate whether not only the transport time but also the transport path has an effect of particle aging. An overview of δ p measurements at 355 and 532 nm is given in Fig. 14. For each measurement campaign a number of days (4, 8, 3, 13) is included in this study. The main findings are: δ p and its 320 wavelength dependence does not change for measurements of fresh Saharan dust at Morocco, close to the source regions, and for dust mid-range transported at Cape Verde. The wavelength dependent overall mean values are 0.31 ± 0.01 (532 nm) and 0.25 ± 0.07 (355 nm) for fresh Saharan dust (Freudenthaler et al., 2009) and 0.3 ± 0.01 (532 nm) and 0.25 ± 0.03 (355 nm) for mid-range transported Saharan dust . For long-range transported Saharan dust towards Central 325 Europe we found slightly higher values of 0.34 ± 0.02 and 0.30 ± 0.05 at 532 and 355 nm, respectively (Wiegner et al., 2011). At Barbados we find a slightly lower mean δ p value of 0.27 ± 0.01 at 532 nm and a rather constant mean value of 0.26 ± 0.03 at 355 nm. The observed differences be- also the transport path, and the conditions during transport may be of importance when investigating the effects of aging and transport. This assumptions will be subject of further studies. A similar analysis was performed for S p measurements during SAMUM-1 (mean value), SAMUM-2 (5 cases), the Munich event in 2008 (3 cases) and during the SALTRACE campaign  (Tesche et al., 2009b), for mid-range transported Saharan dust on 5 days during SAMUM-2 at Cape Verde in 2008 (SAMUM08) , for long-range transported dust over Central Europe on 3 days at Munich, Germany in 2008 (EARLINET08) (Wiegner et al., 2011), and for long-range transported Saharan dust over the Caribbean on 10 days during SALTRACE at Barbados in 2013 (SALTRACE13  Fig. 15). For the fresh Saharan dust in Ouazazate wavelength-independent mean val-340 ues of S p = 55 ± 7 sr at 355 nm and S p = 56 ± 5 sr at 532 nm were found (Tesche et al., 2009b).
Slightly but not significantly higher lidar ratios were found during the Munich dust event and the SAMUM-2 campaign for mid-range transported dust with wavelength-independent mean values of about 60 sr (Wiegner et al., 2011;Groß et al., 2011b). These slightly higher values are in good agreement with the study of Schuster et al. (2012) investigating lidar ratios of mineral dust for different 345 regions over Northern Africa and finding the highest mean values for Saharan dust at Cape Verde.
During SALTRACE the lidar ratios range between 47 and 63 sr with mean values of S p = 56 ± 7 sr at 532 nm and S p = 53 ± 5 sr at 355 nm. Altogether no significant changes in the lidar ratio can be found within the error bars for the fresh and the transported Saharan dust cases.
The mean values and mean uncertainties of S p and δ p for the different dust measurements are 350 summarized in Table 2.

Impact on aerosol classification
It has been shown that the lidar ratio and the particle linear depolarization ratio are quite different for different types of aerosol (Cattrall et al., 2005;Müller et al., 2007;Sakai et al., 2010;Burton et al., 2012). Therefore aerosol classification schemes both at 355 and 532 nm have been developed based 355 on these intensive lidar optical properties Burton et al., 2012;Groß et al., 2013;Illingworth and et al., 2014). Up to now those classification schemes do not sufficiently account for the effect of aerosol aging on the thresholds for the discrimination of the different aerosol types. With our measurements during the SALTRACE campaign, in combination with the findings of former measurements of fresh and mid-range transported dust during the SAMUM project and long-range 360 transported dust measurements over Central Europe, we are now able to investigate the effect of transport and aging on the lidar optical properties of Saharan dust.  for EarthCARE (Illingworth and et al., 2014;Groß et al., 2015). Additionally we plotted our results 365 found during the SALTRACE campaign for pure Saharan dust and for the boundary layer. Within the boundary layer our results fit quite well with former results found for marine aerosols or marine aerosol mixtures, indicating that the boundary layer was dominated by marine aerosols with various amount of dust mixed into the boundary layer on specific days. Regarding the Saharan dust layer one can see that the δ p -S p -space at 355 nm shows a good agreement to former dust measurements 370 during SAMUM. The specified threshold of 0.23 ≤ δ p ≤ 0.33 to identify pure Saharan dust from combined δ p -S p measurements  is still valuable for long-range transported dust.
At 532 nm the SALTRACE results for long-range transported Saharan dust show slightly lower δ p -values compared to the threshold (δ p ≥ 0.28) used for the identification for Saharan dust . Thus this threshold has to be adapted to a slightly lower δ p value of 0.26 to consider 375 long-range transported Saharan dust.

Conclusions
We presented optical properties of Saharan dust long-range transported across the Atlantic Ocean to Barbados. For this purpose we analyzed measurements with the lidar system POLIS at 355 and 532 nm, in particular we calculated the extinction coefficient α p , the lidar ratio S p , and the particle 380 linear depolarization ratio δ p . While the first properties gives us information about the aerosol load, the latter two properties are intensive lidar properties and thus only dependent on the aerosol type and not on its amount. Therefore, these properties are used for aerosol classification schemes based on lidar measurements (Burton et al., 2012;Groß et al., 2013;Illingworth and et al., 2014;Groß et al., 2015). The measurements and results of this work follow up former measurements performed 385 during the SAMUM-1 (Freudenthaler et al., 2009;Tesche et al., 2009b) and SAMUM-2 Tesche et al., 2011) campaigns and during a strong Saharan dust event over Central Europe observed in the framework of EARLINET (Wiegner et al., 2011). Thus, we are able to study possible changes of the optical properties of Saharan dust caused by long-range transport.
For the long-range transported Saharan dust over Barbados we found typical values of δ p between 390 0.26 and 0.3 at 532 nm and between 0.24 and 0.29 at 355 nm. The mean systematic errors are 0.01 and 0.03 at 532 nm an 355 nm, respectively. Compared to δ p measurements at 532 nm during the SAMUM campaigns we see slightly lower values for long-range transported Saharan dust to Barbados, while over Central Europe slightly higher values have been found. This leads to the assumption that not only transport time but also the transport path, and the transport conditions have an influence 395 of possible changes of the optical properties of Saharan dust. At 355 nm we do not see significant changes in the δ p values although the overall mean values are slightly higher for long-range transported Saharan dust to Barbados as well as to Central Europe. For long-range transported Saharan dust we do not see a significant wavelength dependence anymore.
Mean values of the lidar ratio of the long-range transported Saharan dust over Barbados are 56 ± 400 7 sr at 532 nm and 53 ± 5 sr and thus agree well with the values found for fresh Saharan dust over Morocco (Tesche et al., 2009b). Although these values are slightly lower than the values found for long-range transported Saharan dust over Central Europe (Wiegner et al., 2011) and of mid-range transported Saharan dust over Cape Verde  they agree for the measurement uncertainties. Thus we do not see a significant change in this optical properties during transport.

405
Though the presented measurements are a good test bed to study the optical properties of longrange transported Saharan dust, there are a number of questions remaining unsolved, e.g. the impact of transport condition on the changes of optical and microphysical properties. Thus further studies will combine lidar measurements with information of the transport conditions and path, e.g. from 20 model calculations. Furthermore lidar measurements will be combined to in-situ measurements to 410 get more inside the relationship between optical and microphysical properties, e.g. the cloud condensation properties, and about a possible vertical sorting within the dust layer as recently suggested by Yang et al. (2013).
Acknowledgements. This work has been partly funded by the Deutsche Forschungsgemeinschaft (