We summarize our Raman lidar observations which were carried out in Europe, Asia, and Africa during the past 10 years, with focus on particle extinction-to-backscatter ratios (lidar ratios) and Ångström exponents. For the first time, we present statistics on lidar ratios for almost all climatically relevant aerosol types solely based on Raman lidar measurements. Sources of continental particles were in North America and Europe, the Sahara, and south and Southeast and east Asia. The North Atlantic Ocean, and the tropical and South Indian Ocean were the sources of marine particles. The statistics are complemented with lidar ratios describing aged forest fire smoke and pollution from polar regions (Arctic haze) after long-range transport. In addition, we present particle Ångström exponents for the wavelength range from 355 to 532 nm and from 532 to 1064 nm. We compare our data set of lidar ratios to the recently published AERONET (Aerosol Robotic Network) lidar ratio climatology. That climatology is based on aerosol scattering modeling in which AERONET Sun photometer observations serve as input. Raman lidar measurements of extinction-to-backscatter ratios of Saharan dust and urban aerosols differ significantly from the numbers obtained with AERONET Sun photometers. There are also differences for some of the Ångström exponents. Further comparison studies are needed to reveal the reason for the observed differences.
 Spaceborne lidar observations of vertical profiles of the aerosol volume extinction coefficient are needed to improve our knowledge of the influence of aerosols on Earth's climate and to properly account for the geographical, vertical, and temporal variability of aerosols in global atmospheric models. Complementary network observations with ground-based lidars are necessary in addition, because such networks allow for a more frequent aerosol profiling in key areas of anthropogenic pollution, and in marine and desert dust environments. The majority of aerosol lidars, including the recently launched Cloud-Aerosol Lidar and Pathfinder Satellite Observations (CALIPSO) lidar [Winker et al., 2003], are so-called elastic-backscatter lidars. Such lidars allow us to retrieve the particle backscatter coefficient (180° scattering coefficient) only [Ansmann and Müller, 2005]. In contrast, aerosol Raman lidar [Ansmann and Müller, 2005] and High Spectral Resolution Lidar (HSRL) [Eloranta, 2005] permit the direct, unambiguous determination of vertical profiles of the particle backscatter coefficient and particle extinction coefficient, and thus of the extinction-to-backscatter ratio (lidar ratio). A good knowledge of the lidar ratio of key aerosol types such as urban haze, marine and desert dust particles, and biomass-burning smoke is essential to properly convert the retrieved backscatter-coefficient profiles, obtained with elastic backscatter lidars, into extinction profiles. This knowledge is particularly needed in the case of spaceborne lidar observations over coastal and oceanic sites. There, continental aerosol plumes resting on top of the marine boundary layer can frequently be observed. A good knowledge of the lidar ratio of different aerosol types is required to properly resolve the different particle layers in terms of light extinction [Ansmann, 2006].
 We demonstrated in numerous studies [Ansmann et al., 2002; Müller et al., 2002, 2003, 2005; Mattis et al., 2004] that the lidar ratio is a quantity valuable for aerosol characterization. Profiles of the lidar ratio observed with Raman lidar or HSRL help us to separate between the impact of natural and anthropogenic aerosols on climate. The European Space Agency plans to launch two HSRLs, namely ALADIN (atmospheric laser Doppler instrument) in the framework of the Atmospheric Dynamics Mission [European Space Agency (ESA), 1999; Endemann et al., 2004; Stoffelen et al., 2005] and ATLID (atmospheric backscatter lidar) in the framework of the Earth Cloud, Aerosols, and Radiation Explorer mission [ESA, 2004]. These lidars will allow us to map the global aerosol distribution also in terms of the 355-nm lidar ratio. Therefore an aerosol-type-dependent lidar ratio climatology is valuable to better interpret the spaceborne observations also at this wavelength.
 We started our aerosol Raman lidar studies in 1988/89, and we demonstrated that an independent determination of the particle backscatter coefficient and of the particle extinction coefficient is possible from the measured elastic-backscatter and nitrogen Raman signals. These initial investigations were done in cirrus clouds [Ansmann et al., 1992]. It can be concluded from the available literature on aerosols that there is no alternative to the Raman lidar and HSRL techniques if one wants to establish a long-term record of aerosol extinction and lidar ratio profiles.
 Since 1996 we focus our studies on atmospheric aerosols. We initiated the German Lidar Network (GLN) [Bösenberg et al., 2001] that served as the nucleus for the European Aerosol Research Lidar Network (EARLINET) [Bösenberg et al., 2003]. At present more than half of the 24 EARLINET stations operate Raman lidars. A 10-year aerosol data record is now available for the Leipzig site. The data set describes optical properties of central European haze and of lofted, free-tropospheric aerosol layers originating from Africa, North America, and northeast Asia. Furthermore, we participated in several international aerosol field campaigns, which enabled us to characterize the optical properties of marine particles of the North Atlantic Ocean and the tropical Indian Ocean as well as optical properties of continental particles over south Europe, south and east Asia, and North Africa. A detailed overview of our activities is given in the following section 2.
 In this contribution we summarize our observations of the extinction-to-backscatter ratio. We present for the first time a lidar ratio statistic that is solely based on direct measurements of that quantity. After 10 years of observations within two networks and five international field campaigns, a lidar ratio climatology can be presented that covers many of the climatically relevant aerosol types.
 We reanalyzed the entire GLN and EARLINET data record with regard to free-tropospheric aerosol layers that originate from Saharan dust outbreaks and long-range transport of urban haze and forest fire smoke from North America. We include Saharan dust observations we recently carried out in Morocco in the spring of 2006. We also add data from observations in Beijing, China [Tesche et al., 2007]. The measurements were made in January 2005. We present lidar ratios measured at 532 nm wavelength and, where available, also measured at 355 nm wavelength. Simultaneous observations of the lidar ratio at these two wavelengths are important to link the CALIPSO observations at 532 nm to the future observations with ESA's ALADIN and ATLID, and to contribute in this way to establishing a consistent, continuous, long-term aerosol data record as requested by the aerosol community [Diner et al., 2004]. The results are presented in section 3.
 Our study has also been motivated by the recent paper of Cattrall et al. . The authors present a lidar ratio climatology which is based on observations with Sun photometers of the Aerosol Robotic Network (AERONET) [Holben et al., 2001] in 26 locations around the globe. Sun photometers cannot measure the scattering coefficient at 180°, and thus the lidar ratio. The backscatter coefficient has to be retrieved by means of a sophisticated aerosol model from the measured spectrally resolved particle optical depths and the scattering-angle-resolved sky radiances that serve as input. Sun photometers provide column-integrated information only. Optical properties of boundary layer particles that originate from local sources and regional aerosol transport cannot be separated from the optical properties of lofted, free-tropospheric aerosol layers that predominantly originate from long-range transport on regional to intercontinental scales. That circumstance may, for instance, complicate the determination of the Saharan dust lidar ratio over the AERONET site on the Cape Verde Islands, where the boundary layer contains a considerable amount of marine particles even during strong Saharan dust outbreaks from the African continent. In section 4 we compare our results with the data set provided by Cattrall et al. , and we discuss the deviations found. In section 5 we summarize our results.
 The data presented in this study were acquired with three Raman lidars of the Leibniz Institute for Tropospheric Research. In the framework of GLN (1997–2000) [Mattis et al., 2001; Bösenberg et al., 2001] and EARLINET [Mattis et al., 2004; Bösenberg et al., 2003] we carried out routine observations with a three-wavelength elastic-backscatter two-wavelength Raman lidar. We observed central European haze and free-tropospheric lofted particle layers, such as Saharan dust [Mattis et al., 2002], forest fire smoke from Siberia and North America, aged anthropogenic haze plumes from North America [Müller et al., 2005], and haze from polar regions (Arctic haze) [Müller et al., 2004]. The lofted particle layers were transported to Europe during long-range transport events. In the framework of the results discussed in this paper we reanalyzed the entire German Lidar Network and EARLINET data set to improve our Saharan dust statistics and the statistics on long-range transport of forest fire smoke and urban haze from North America.
 We collected data during the Second Aerosol Characterization Experiment (ACE 2) [Russell and Heintzenberg, 2000] in the summer of 1997. We observed pure marine particles which originated from the North Atlantic, and aged urban particles which were advected from central and south Europe. In the framework of the Indian Ocean Experiment (INDOEX) [Ramanathan et al., 2001] we conducted four campaigns in the spring, summer, and autumn of 1999, and in the spring of 2000. The observations were carried out in the Maldives, which are approximately 700 km southwest of India. During this campaign we monitored lofted layers of aged Asian haze which consists of a mixture of particles from fossil fuel consumption and domestic and agricultural biomass burning. The pollution layers were advected from north India, south India, and Southeast Asia [Franke et al., 2001, 2003]. During the summer and autumn seasons air was carried to the field site mainly from the south, i.e., from the tropical and southern Indian Ocean, and we observed pure marine aerosols. On several days, however, we also observed lofted, aged desert dust plumes from East Africa and Saudi Arabia. Pure, fresh Saharan dust plumes were measured at Ouarzazate (30.93°N, 6.9°W) in south Morocco in the spring of 2006 in the framework of the Saharan Mineral Dust Experiment (SAMUM). The field site in Morocco was rather close to the source areas of the dust particles.
 The ACE 2, INDOEX, and SAMUM data were acquired with a mobile aerosol Raman lidar that provides backscatter coefficients at six wavelengths and extinction coefficients at two wavelengths [Althausen et al., 2000]. However, the lidar was equipped with just one aerosol Raman channel at 532 nm during ACE 2. Furthermore, the 355-nm Raman channel worked properly only sporadically during the INDOEX campaigns.
 A small, compact one-wavelength aerosol Raman lidar, which operates at 532 nm wavelength [Althausen et al., 2004], measured during two campaigns in the Pearl River Delta (PRD) in south China in October 2004 [Ansmann et al., 2005] and at Beijing in January 2005 [Tesche, 2006; Tesche et al., 2007]. The PRD is one of the economically fastest growing regions of China. The aerosol at this subtropical site is characterized by a mixture of urban haze and domestic biomass-burning particles. The average optical depth of 0.9 at 532 nm was extremely high compared to the conditions we normally find over central and west Europe. In contrast to the aged south Asian aerosols observed during INDOEX the Chinese aerosol plumes mainly consisted of newly produced and emitted particles. We found very different aerosol conditions during the Beijing campaign. Optical depth was 0.02–0.05 (532 nm) during background aerosol conditions and 0.1–0.6 during situations of moderately polluted air. A case of dust from the Gobi desert could also be observed.
Table 1 presents an overview of the lidar ratio observations which were made with the three lidars. Lidar ratios at 355 nm wavelength (mean value with one standard deviation) are given only for those cases in which we could carry out a systematic analysis of the 355-nm observations. For the rest of the cases, the ratio of the 355–532 nm lidar ratios is only roughly estimated on the basis of several simultaneous observations at 355 and 532 nm. Observations of the lidar ratio at 355 nm wavelength were often limited during INDOEX because of detector problems. In the case of the EARLINET data set, a careful correction of the dependence of the 355-nm elastic backscatter channel on the particle depolarization ratio was necessary [Mattis, 2002]. This time-consuming procedure has not yet been fully applied to the entire set of our EARLINET Saharan dust data.
Table 1. Mean Values and One Standard Deviations of the Particle Lidar Ratios S355 and S532 at 355 and 532 nm, Respectively; the Ratio of the Lidar Ratios Measured at 355 and 532 nm; Ångström Exponents Derived in the Wavelength Range From 355 to 532 nm and From 532 to 1064 nm; and Particle Depolarization Ratio δpar
Planetary boundary layer (PBL) indicates local and regional aerosol, and free troposphere (FT) indicates aged particles after long-range transport.
The Ångström exponents are derived from measurements of the particle extinction coefficients at the two indicated wavelengths. The data for China (last two rows) were derived from Sun photometer measurements of optical depth at 381 and 502 nm.
The Ångström exponents are derived from measurements of the particle backscatter coefficient at 532 and 800 nm, or 532 and 1064 nm. The data for China were derived from Sun photometer measurements of optical depth at 502 and 1046 nm.
The particle depolarization ratio has only been analyzed in detail in the case of mineral dust observed at Leipzig and during SAMUM. In the case of the INDOEX data the channel used for depolarization measurements did not operate properly. The lidar that was used for the observations in the PRD and in Beijing did not possess a channel for depolarization measurements. In all other cases we roughly estimated the particle depolarization ratio to be less than 5%.
Measurements of the depolarization ratio were done at 532 nm during ACE 2 and EARLINET. The channel at 710 nm was used during INDOEX and SAMUM. That channel was not operational during dust advection from Saudi Arabia.
The ratio of the lidar ratio has been roughly estimated from a few lidar observations. In these cases no uncertainty values are given.
 Particle depolarization ratios were determined from the mineral dust observations at Leipzig (EARLINET data set), and during SAMUM. Details on the used technique are given by Cairo et al.  and Murayama et al. . In the case of biomass-burning aerosols observed at Leipzig and anthropogenic pollution observed at Leipzig and during ACE 2 we merely estimated an upper limit of the particle depolarization ratio. In the case of the INDOEX data no reasonable results could be derived as the detection channels did not work properly during that time. The lidar instrument that was used during the field campaigns in China did not possess a channel for measurements of the depolarization ratio.
 Lidar ratios for the planetary boundary layer (PBL) in Table 1 represent only the upper PBL above approximately 500–800 m height. Overlap correction [Wandinger and Ansmann, 2002] uncertainties prohibit a trustworthy analysis in the near field. Only data collected since 2000 in the framework of EARLINET observations are considered in this analysis. The results for lofted Saharan dust and North American forest fire smoke and urban haze, however, are based on GLN and EARLINET observations carried out since 1997.
 Our values of the lidar ratios represent the average of all atmospheric situations. We did not differentiate between cases of high and low relative humidity. We assume a bias of approximately 10% of the mean lidar ratios, if we go from dry to wet humidity conditions.
 We estimated the Ångström exponents [Ångström, 1964] for the 355–532 nm and 532–1064 nm wavelength ranges from measurements of the particle extinction and backscatter coefficients. Details on the methodology that is used to derive the Ångström exponents are given by Ansmann et al.  and Franke et al. .
 The results, which are shown in Table 1, characterize the observed aerosol types in more detail. Ångström exponents describing the short-wavelength range are sensitive to particles of the fine-mode fraction of the size distribution (particle diameters <1 μm). Particles in the fine mode dominate the size distribution of urban haze. Coarse-mode particles (with diameters typically >1 μm) may govern the optical properties of the particle size distribution, as may be the case for desert dust and marine particles. Ångström exponents in the short-wavelength and the long-wavelength range are low in such situations.
 In the case of INDOEX, most Ångström values were estimated from the backscatter coefficients at 400, 532, and 800 nm [Franke et al., 2003]. Most EARLINET Ångström values were derived from backscatter coefficients at 355, 532, and 1064 nm. The Ångström exponents of ACE 2 marine cases, and the results of the PRD and Beijing (urban haze) result from daytime Sun photometer observations performed at the lidar site [Tesche, 2006].
3.1. Marine Particles
Table 1 shows that marine particles cause the lowest lidar ratios. Large sea spray particles in the PBL lead to values of 20–26 sr. That result is in agreement with findings of Tomasi et al.  (southern Italy, 20–25 sr at 351 nm) and Masonis et al.  (Hawaii, 25 ± 4 sr at 532 nm). Our values of the lidar ratios of marine particles are also in agreement with the results of Ackermann .
 Lidar ratios range from 25–35 sr at 355 and 532 nm between about 1000 and 2500 m height in the free troposphere over the tropical Indian Ocean. Ten-day backward trajectories clearly indicated marine air that was never close to any landmass.
 These larger values may be explained by two mechanisms. First, number concentration of large marine particles decreases with increasing height. Second, the influence of particles, which are formed from precursor gases (secondary aerosol production) over the oceans, increases with height and leads to the observed larger lidar ratios.
 In general, we find small Ångström exponents for marine particles. During INDOEX the marine boundary layer over the Maldives was frequently polluted during the northeast monsoon season. The lidar ratios for polluted marine air vary between approximately 30 sr and 40 sr at 532 nm. These values are not included in Table 1.
3.2. Urban Haze/Industrial Aerosol
Table 1 shows that local and regional anthropogenic haze over the central European site at Leipzig causes lidar ratios that typically range from 45–60 sr with a mean value of 53 sr at 532 nm. The values are 10% larger at 355 nm. Lofted layers observed during ACE 2 were caused by pollution outflows from the European continent. Most of the particles traveled more than 4 days before arriving at the field site. Aging and mixing with dust and anthropogenic particles over southern Europe caused average lidar ratios of 45 ± 10 sr. The size of the particles is comparably small. Accordingly, the particle depolarization ratio is rather low.
 For comparison, Bösenberg et al.  report an average lidar ratio of 55 sr at 351 nm for the lowermost 2–3 km over Hamburg, Germany. That value is based on all lidar observations carried out during the EARLINET period. The highest lidar ratios with values close to 70–80 sr were observed at Hamburg [Matthias and Bösenberg, 2002] and Leipzig in cases in which masses were advected from east Europe. Such high values indicate the presence of a pronounced fine mode and a rather low impact of maritime particles.
 Compared to the more northern EARLINET stations, mean lidar ratios were typically below 50 sr within the PBL at the Italian and Greek stations [Bösenberg et al., 2003]. Most of the southern stations are located very close to the Mediterranean Sea, which means that marine particles always have a strong influence on the aerosol optical properties.
Amiridis et al.  analyzed 4 years of EARLINET observations (2000–2004). The authors find 355-nm lidar ratios of 40 ± 21 sr during conditions of rapid transport of air from westerly directions (Atlantic, France, Italy, Adriatic Sea). Values are 29 ± 11 sr during stagnant conditions with slow transport of air from southerly directions (Mediterranean, southern Greece). Mean lidar ratios are 40–47 sr (±16–25 sr) when air was advected from continental eastern Europe (Bulgaria, Romania, the Ukraine, Russia).
Tomasi et al.  found annual mean 351-nm lidar ratios of 42–48 sr (±25 sr) below 2 km height at Lecce, south Italy. Pisani  found seasonal mean 351-nm lidar ratios of 55 ± 4 sr (spring), 51 ± 3 sr (summer), 52 ± 4 sr (autumn), and 72 ± 5 sr (winter) for the height range from 1 to 2 km over Naples, south Italy.
3.3. Mixture of Urban Pollution and Arctic Haze
 A special situation occurs when pollution is advected from polar regions. Although there are only few anthropogenic pollution sources north of 70°N, lidar observations at Leipzig sometimes show particles in the free troposphere. This pollution, referred to as Arctic haze [Shaw, 1984, 1995], originates from precursor material advected from industrialized areas of the Northern Hemisphere to the polar regions [Rahn and Heidam, 1981; Barrie and Hoff, 1985].
 Up to now we have analyzed only a few such cases. Lidar ratios are around 60 sr at both wavelengths, and thus well within the variability found for European anthropogenic pollution. One notable difference however are the comparably large Ångström exponents. As has been shown by Müller et al.  the mean size of such particles is smaller than what is usually found for free-tropospheric anthropogenic pollution [Müller et al., 2002, 2005].
 Particle properties within the boundary layer do not differ too much from the free-tropospheric conditions in cases in which Arctic-haze-like material passes over strongly urbanized areas in eastern Europe on the way from the polar regions to central Europe. It has been speculated that the optical properties of east European particles and Arctic haze must be rather similar [Heintzenberg et al., 2003; Müller et al., 2004].
3.4. Aged North American Urban Haze and Forest Fire Smoke After Long-Range Transport
Table 1 also lists lidar ratios of North American urban haze. The prevailing westerly winds frequently cause intercontinental aerosol transport in the free troposphere toward Europe during the summer season. On average we observe ten of such events within each summer half year, which translates to 20% of all EARLINET observations carried out each year. These aerosols need on the order of seven to ten days to travel across the Atlantic. Mean lidar ratios of 45 sr and 38 sr are found at 355 and 532 nm, respectively. Standard deviations are comparably large for these rather aged particles. Urban haze can be unambiguously distinguished from forest fire smoke by the spectral slope of the particle lidar ratio at 355 and 532 nm (see Table 1). The ratio of the 355-nm to the 532-nm lidar ratio is found to be, on average, approximately 1.4.
 Several statistics of lidar ratio observations in North America are available for comparison. Anderson et al.  used a nephelometer-based setup for ground-based in situ observations of the lidar ratio at 532 nm at Bondville, Illinois. On average, the lidar ratio was 47 sr in the late summer of 1999. The authors obtained 64 ± 4 sr during stagnant atmospheric conditions. Rapid air mass transport resulted in lidar ratios of 40 ± 9 sr.
 The mean lidar ratio was 68 ± 12 sr at 355 nm at the CART site, Oklahoma, during the period from April 1998 to April 1999 [Ferrare et al., 2001]. AERONET photometer measurements carried out at the CART site were used to retrieve size distribution and refractive index information. A mean lidar ratio of 53 sr was calculated at 532 nm wavelength on the basis of that information.
 Long-range transport of anthropogenic pollution to Europe seems to decrease the lidar ratio. Lower lidar ratios may be linked to an increase of particle size and/or a reduction of the light absorption capability of the particles. Possible reasons may be mixing with marine particles over the North Atlantic, particle growth due to uptake of water or precursor gases, photochemical reactions, and particle growth because of coagulation.
 A completely different wavelength dependence of the lidar ratio was observed for free-tropospheric aerosols originating from forest fires. In that case we find, on average, a value of 0.8 for the ratio of the 355-nm to the 532-nm lidar ratio. Similar values as those reported in Table 1 and a similar wavelength dependence of the lidar ratio were observed for Siberian forest fire smoke over Japan [Murayama et al., 2004]. According to our recent publication [Müller et al., 2007] forest fire smoke particles grow during long-range transport. According to the same paper, the Ångström exponent that describes particle light extinction in the wavelength range from 355 to 532 nm is near zero, whereas the backscatter-related Ångström exponent is about one for aged forest fire smoke after long-range transport. This remarkable behavior of the optical properties of forest fire particles was first shown by Wandinger et al.  and Mattis et al. .
 Despite the large size of particles from forest fire smoke in North America, which expresses itself in terms of rather low Ångström exponents, we find particle lidar ratios that are similar to lidar ratios of urban haze from North America and Europe. Another noteworthy fact is that relative humidity of these smoke plumes in many cases shows low values below 50% [Müller et al., 2005].
 In all investigated cases we found particle depolarization ratios <5%; that is, despite their large size and the low relative humidity these smoke particles seem to have spherical shape. The only notable exception from that result is given by a measurement during the Lindenberg Aerosol Characterization Experiment (LACE) 98 [Wandinger et al., 2002]. We found depolarization ratios of up to 10% at 532 nm wavelength. Relative humidity in that particle layer varied between 40% and 60%. A possible reason for the larger depolarization ratio might have been dust particles injected into the smoke plume through the intense heat at the source of the fire in Canada. In situ observations with airborne instruments showed that the particles contained hematite [Fiebig et al., 2002], which is indicative of mineral dust.
3.5. South and Southeast Asian Aerosol
 As outlined by Cattrall et al.  and corroborated by our INDOEX [Franke et al., 2001, 2003], PRD [Ansmann et al., 2005] and Beijing observations [Tesche, 2006; Tesche et al., 2007], Southeast Asian aerosols are distinctly different from European pollution. In addition to the fine mode (with particle diameters typically <500 nm), a pronounced coarse mode is present. These large particles originate mainly from coal and dried plants which are used for domestic heating and cooking. Low combustion temperatures as well as less strict environmental regulations favor the emission of large particles. In addition, road dust is omnipresent in south (India) and east Asia (China). The effective (i.e., cross-section weighted mean) particle radius was found to be about 0.15 ± 0.06 μm, 0.2 ± 0.07 μm, 0.24 ± 0.06 μm, and 0.26 ± 0.01 μm for European [Müller et al., 2002; Wandinger et al., 2002], north Indian [Müller et al., 2003], PRD [Müller et al., 2006], and south Indian aerosols [Müller et al., 2003], respectively. As a consequence the lidar ratios and the Ångström exponents for the wavelength pair 355/532 nm for Asian aerosols listed in Table 1 (PRD, Beijing, south India) are smaller than the ones for central European aerosols.
 Exceptionally large lidar ratios were observed when air masses were advected from north India. The 532-nm lidar ratios accumulated at 50–80 sr for the highly light-absorbing particles [Franke et al., 2003; Müller et al., 2003]. Intense domestic wood burning occurs in north India. Wood-burning aerosol is highly light-absorbing. The lidar ratio distribution describing aged north Indian aerosols is shifted by about 20 sr toward larger values, compared to the values for aged European (ACE 2) particles. These values are consistent with an aerosol black carbon content of about 20% for north Indian and 5% for aged European aerosol [Novakov et al., 2000; Franke et al., 2003].
 When comparing the lidar ratios of different regions in Asia the influence of long-range transport (particle aging) must be kept in mind. The INDOEX aerosols from north India and Southeast Asia (mainly Thailand, Malaysia, Indonesia) traveled more than six days from the source region to the lidar field site at the Maldives International Airport. In contrast, observations were taken directly in the pollution regions in south and north China during the PRD and Beijing campaigns, respectively. We assume that locally produced pollution (young particles) had a comparably strong impact on the observed aerosol optical properties in the planetary boundary layer.
 Lidar ratio seems to increase during long-range transport from the Asian continent. That finding is consistent with the assumption that the influence of the coarse-mode fraction of Asian particles decreases with increasing distance from the source region. Accordingly the Ångström exponents change. Coarse-mode particles may have been partly removed during transport by sedimentation or their sizes became reduced by cloud processes (cloud formation and successive evaporation etc.). In this context it remains unclear in which way differences of aging of particles from biomass burning and fossil fuel burning influence the lidar ratios and the Ångström exponents.
 Only a few lidar ratio observations of Asian pollution are available for comparison. In agreement with our PRD observations, Anderson et al.  found a mean lidar ratio of 45 ± 10 sr at 532 nm as mean value during the ACE-Asia campaign in the spring of 2000. That value describes east Asian haze over the Pacific Ocean close to China. The data were obtained from airborne in situ observations of particle scattering measured at 550 nm wavelength with nephelometer and of particle light-absorption measured with particle/soot absorption photometer at the same wavelength. In cases dominated by the fine-mode fraction of the particle size distribution (pollution cases), i.e., >60% of the scattering was caused by particles with radius <500 nm, the mean lidar ratio was 50 ± 10 sr. In contrast, events dominated by the coarse-mode fraction of the particle size distribution (dust cases), i.e., <30% of the scattering was caused by particles with radius <500 nm, the mean lidar ratio at 550 nm was 47 ± 6 sr. Pollution in the fine mode was generally accompanied by a substantial component in the coarse mode, whereas coarse-mode dust was frequently observed in almost pure form. The mean lidar ratio was 42 ± 9 sr for mixed cases, i.e., 30%–60% of the scattering was caused by particles with radius <500 nm. Murayama et al.  analyzed combined ACE-Asia lidar and aircraft measurements over Japan. Mean lidar ratios in the polluted lower troposphere were 35–45 sr, and Ångström exponents were 1.5–2 in the wavelength range from 450 to 700 nm.
 In general we found low particle depolarization ratios <5% during our INDOEX activities. As outlined before no depolarization measurements were possible in China. The particle depolarization ratios are in the range of numbers that describe anthropogenic pollution from Europe and North America. However, the depolarization channels did not work properly during our INDOEX campaigns, and there remains uncertainty regarding the low values. Because of that uncertainty we did not attempt to separate between the different advection channels that are described by Franke et al. .
3.6. Desert Dust
 Lidar ratios of desert dust can vary strongly as a function of size distribution, particle shape, which may be different for the fine-mode and the coarse-mode fractions, and particle chemical composition, i.e., content of light-absorbing hematite. Barnaba and Gobbi  computed 532-nm lidar ratios and found values between 35 sr and 50 sr under the assumption of spheroid-like particle shapes. Ackermann  obtained lidar ratios around 20 sr by assuming spherical shape of the dust particles. Liu et al.  found 532-nm dust lidar ratios between 20 sr and 40 sr and between 45 sr and 60 sr for spheroids. In their calculations the authors increased the aspect ratio from 1.4 to 1.7. The aspect ratio describes the ratio of the radius of a particle along its longest axis to the radius along its shortest axis.
 Simultaneous observations of the lidar ratio of Saharan dust at 355 nm and 532 nm clearly showed the impact of dust shape on the particle backscatter coefficient [Mattis et al., 2002]. It is now well accepted that the nonspherical shape introduces a significant reduction of the 180° scattering efficiency, whereas light extinction of dust particles remains nearly unchanged. The latest state-of-the-art computations by Dubovik et al.  show lidar ratios of about 70 sr (for a mean particle volume radius rv = 5 μm), 55 sr (rv = 2 μm), and 45 sr (rv = 1 μm) for typical dust size distributions, spheroid-like particle shapes, and Ångström exponents 0.5 in the wavelength range from 440 to 870 nm.
 Our SAMUM observations at Ouarzazate, Morocco, show lidar ratios mostly between 50 sr and 60 sr at both wavelengths. Preliminary backward trajectory analysis shows that the main portion of the dust plumes originated from the north-central parts of the Sahara desert. We did not find any wavelength dependence of the lidar ratio. The dust layer typically extended from the surface to 4–6 km height above sea level; the lidar station was at 1133 m above sea level. The mean lidar ratio was found to be, on average, still high with values around 60 sr at 532 nm for aged Saharan dust after long-range transport to central Europe. The lidar ratio at 355 nm on average was higher by about 10–20%. We always observed that the dust plumes over central Europe were separated from the boundary layer aerosol. The top heights of the dust plumes were usually detected between 5 km and 7 km height above sea level. We assume that contamination with anthropogenic pollution and marine particles was very weak.
 Our findings are in accordance with EARLINET Raman lidar measurements in south Italy [Tomasi et al., 2003]. Layer mean lidar ratios of 51 sr, 53 sr, and 57 sr at 351 nm wavelength, respectively, were determined during three heavy outbreaks of dust from the west Sahara region. Pisani  analyzed the Naples, Italy, EARLINET data set and found a mean Saharan dust lidar ratio of 46 sr (2500–3500 m height range), 58 sr (3500–5000 m height range), and 77 sr (5000–8000 m height range). Standard deviations were below 10 sr in each case. Amiridis et al.  presented statistics of the 4-year EARLINET Raman lidar data record taken at Thessaloniki, Greece. The authors find an average lidar ratio of 57 ± 29 sr at 355 nm for the backward trajectory cluster that describes transport of dust from the Sahara.
 A long-term study of Saharan dust outbreaks at the EARLINET station Potenza in south Italy however revealed a mean lidar ratio of 38 ± 15 sr. The Ångström exponents, derived from the 355 and 532-nm backscatter coefficients, range mostly from 0.5 to 1. These values point at the mixing of dust with nondust particles, which may have caused the comparably large Ångström values. The Potenza observations agree with our observations of dust that was advected from the Arabian Peninsula to the Maldives during INDOEX. We find lidar ratios around 38 sr at both wavelengths (see Table 1) in a lofted dust layer in 3.5–5 km height. The Ångström exponents of that dust were significantly higher than our Ångström values of Saharan dust.
Balis et al.  find an anticorrelation between the layer mean Ångström exponent, derived from backscatter coefficients measured at 355 and 532 nm, and the 355-nm lidar ratio of lofted plumes that originated from the Sahara. Ångström values of <0, 0.5–1.5, >1.5 were linked to lidar ratios of >60 sr, 40–60 sr, and <40 sr, respectively. Especially the highly variable Ångström values of the dust layers suggest that a (nondust) fine-mode component at least in part dominated the Saharan dust layer.
 As mentioned before strong shape-related variations of the backscatter coefficient and the lidar ratio can never be excluded. Shape effects were most probably responsible for the rather low and varying dust lidar ratios during a dust event we observed in Beijing in January 2005 (see Table 1). The 532-nm lidar ratio was 40 sr and remained almost constant throughout the dust layer which extended from 800 to 2500 m height during the period of highest dust-related optical depth (around 0.25 at 532 nm). The dust layer was well separated from the wintertime urban aerosol layer over Beijing. In the beginning and at the end of the dust storm the lidar ratio was considerably lower. Values varied between 30 and 35 sr. Note that Beijing is only about 70 km away from arid areas.
 Typical lidar ratios of Asian dust vary between 40 sr and 60 sr after long-range transport. Mean values of 46 ± 5 sr were found for a pure dust layer over Japan. The dust was identified by a high depolarization ratio [Sakai et al., 2002]. Liu et al.  found mean dust layer lidar ratios of 42–55 sr for several dust events. Furthermore, Sakai et al.  pointed out that the Ångström exponent of Asian dust is about zero or even negative in the 532–1064 nm wavelength range. The authors present a case with increasing Ångström exponent (from 0 to 1) and relative humidity (from 20% to 60%) within a lofted layer which was classified as dust layer according to backward trajectories. It was hypothesized that such cases again indicate a mixture of dust with sulfate particles.
Murayama et al.  show a case with complex layering of different aerosols over Japan. Below 2 km height, the Ångström exponent was 1.5–2, the particle depolarization ratio ranged from 5% to 10%, and the lidar ratio at 532 nm varied from 35 sr to 45 sr. In the particle layer in 2–3.6 km height, the Ångström exponent decreased to 0.3–1, the depolarization ratio increased to 15%–25%, and the lidar ratio increased to 45–50 sr. Only the uppermost layer above 3.6 km showed a clear dust signature with Ångström values of 0–0.5 and particle depolarization ratios of 30%–35%. In this layer the 532 nm lidar ratio varied between 45 sr and 60 sr.
Anderson et al.  report on average lidar ratios of 46 sr (at 550 nm) observed during ACE-Asia. That number describes pollution events that were dominated by the coarse-mode fraction of the particle size distribution.
 Regarding particle depolarization ratios we find values between 30% and 35% at 710 nm wavelength for the SAMUM data set. For comparison we find values between 10% and 25% for dust over Leipzig. The measurement wavelength is 532 nm. It is difficult to decide in how far this difference in range of values of the depolarization is caused by different properties of dust near the sources in north Africa (SAMUM) and in the far-field of the dust sources (EARLINET). Another reason may be the wavelength dependence of the particle depolarization ratio. In that respect measurements of the particle depolarization have also been carried out at 355, 532, and 1064 nm during SAMUM with three more lidar instruments operated by the University of Munich and the German Aerospace Center (DLR). These measurements may help to answer that question.
4. Comparison to AERONET Results
Cattrall et al.  recently published a lidar ratio climatology derived from observations at 26 AERONET Sun photometer stations around the globe. The climatology covers very different aerosol types from pure marine to heavily polluted areas. The authors present a list of uncertainties that can significantly influence the retrieval products. For this reason, we compare our findings with the AERONET climatology.
 First, the most important source of uncertainty arises from the fact that a Sun photometer cannot measure the particle backscatter coefficient. The lidar parameters are computed from particle size distributions and complex refractive indices which are obtained from inversions of spectral measurements of sky radiance and atmospheric transmittance. Especially in the case of nonspherical dust particles, care has to be taken in the interpretation of the retrieved lidar ratios. Cattrall et al.  point out that the true shape of nonspherical aerosol particles such as mineral dust is very complex. At present, only light-scattering properties of spheroids [Mishchenko et al., 1997] can be modeled with reasonable computational efficiency across the required range of particle size and complex refractive index. The best method how to deal with the nonspherical shape of particles in remote-sensing application remains an open debate.
 Second, a Sun photometer delivers column values of particle extinction rather than profile information. Thus the impact of different aerosols in the PBL and in the free troposphere on the measured optical properties cannot be separated. For that reason it is difficult to determine the lidar ratio of desert dust which is located mainly in the free troposphere and above the marine boundary layer at ocean sites and islands. Similarly, lofted continental aerosol layers may complicate the determination of lidar ratios of the marine boundary layer.
 Third, the AERONET climatology is based on monthly mean values for selected months with dominating impact of just one aerosol type. A less uncertain approach would be to average selected cases throughout the year for a given aerosol type on the basis of backward trajectory analysis of the observed air masses. Thus it cannot be excluded that lidar ratios of different aerosol types were averaged in the study by Cattrall et al. .
 Fourth, if we compare data products of the boundary layer aerosol derived from ground-based lidar and AERONET Sun photometer we must always keep in mind that a lidar is unable to resolve accurately the lowest several hundred meters of the atmospheric column because of the large uncertainties that are involved in the correction of the incomplete laser-beam receiver-field-of-view overlap [Wandinger and Ansmann, 2002]. Thus lidar data at continental sites do not include the impact of freshly produced tiny aerosol particles at street level. It is usually expected that these small particles cause comparably large Ångström exponents and large lidar ratios.
 Finally, Raman lidar observations are mainly carried out at nighttime whereas the Sun photometer data describe daytime conditions. It may be speculated that different humidity conditions, as well as the vertical variation of relative humidity particularly in the planetary boundary layer may have some effect and cause deviations between the results obtained with Raman lidar and Sun photometer. Our measurements at Leipzig, however, do not show significant variations of relative humidity distribution between day and nighttime observations. It should be pointed out that a similar result, i.e., rather invariable relative humidity conditions between daytime and nighttime, was also found from long-term routine observations of water vapor with Raman lidar at the Southern Great Plains Cloud and Radiation Testbed in Oklahoma, USA [Turner et al., 2000].
Figure 1 presents results obtained from observations with AERONET Sun photometers and our Raman lidars for five well-defined aerosol types. In the case of marine aerosol particles the results for AERONET are obtained by averaging the observations of all five marine stations listed by Cattrall et al. [2005, Table 1]. In the case of Saharan dust, only the stations of Banizoumbou (Nigeria) and Capo Verde (Spain) are included in the computations. The result for European haze only considers the French AERONET stations given by Cattrall et al. [2005, Table 1]. The values for Southeast Asia describe the observations at all four stations in South Korea, Taiwan, and China as given by Cattrall et al. [2005, Table 1]. For better comparison, the results shown in Figure 1 only include the lidar observations at the Chinese field sites (PRD, Beijing). For the comparison of European haze lidar ratios we only use the EARLINET observations in central Europe. For better comparison of the dust properties, we only selected the cases of pure Saharan dust which we observed with Raman lidar in Morocco (SAMUM). The results of AERONET were computed from the numbers given by Cattrall et al. [2005, Tables 1 and 2].
Figure 1a compares the lidar ratio at 532 nm from lidar to the one at 550 nm from Sun photometer. Figure 1b for illustration shows the color ratio of the lidar ratio at 355/532 nm obtained with lidar. In the case of the Sun photometer observations only the ratio at 550/1020 nm is available.
 The backscatter-related Ångström exponent is shown in Figure 1c. This parameter has been determined with our Raman lidar for the wavelength pair at 532/800 nm and 532/1064 nm, respectively, and the assumption of lidar ratios at the long measurement wavelengths (800 and 1064 nm). Details are discussed in the published literature, see section 2. The Sun photometer data describe the backscatter-related Ångström exponent at 550/1020 nm. The numbers were calculated from the backscatter ratios listed by Cattrall et al. [2005, Tables 1 and 2].
Figure 1d presents the Ångström exponent derived from our extinction coefficients measured at 355 and 532 nm (lidar) and from our optical depths measured at 550 and 1046 nm (Sun photometer measurements in China). The AERONET Ångström exponents were calculated for the wavelength pair 440/870 nm. We extended our results of the Ångström exponent on the basis of our particle extinction measurements at 355 and 532 nm, and information on the backscatter-related Ångström exponent at the wavelength pair at 532/1064 nm (see Table 1 of this paper).
 Some of the differences between the two data sets may be caused by the use of different measurement wavelengths, which is particularly true for the color ratio of the lidar ratio. Different measurement locations may be another reason. Nevertheless, there are deviations, which in some cases are quite significant.
 The agreement is reasonable in the case of marine particles. However, mean lidar ratios of four of the five marine AERONET stations included in the statistics were 32–34 sr. The Ångström exponents were 0.7–1. These values indicate mixing of marine with anthropogenic pollution, which may explain the somewhat larger Ångström exponents compared to the values obtained with lidar. Similar lidar ratios from lidar were found in the polluted marine PBL over the Maldives during INDOEX [Franke et al., 2003]. Because the marine PBL mainly contributes to the column value, the marine lidar ratio that is retrieved from Sun photometer data should be of the order of 20–25 sr as discussed before.
 In the case of desert dust, deviations have to be expected because of the unknown shape characteristics of dust particles. It was stated above that especially the backscatter coefficient is rather sensitive to particle shape.
 The AERONET retrievals result in a mean lidar ratio of 35 sr for Banizoumbou and 38 sr for Capo Verde. The Ångström exponents are close to zero and clearly indicate the dominant impact of dust. Provided that the Raman lidar measures the true Saharan dust lidar ratio, the backscatter coefficient of dust is overestimated by roughly 10%–20% in the case of the Sun photometer data. The error in the calculation of the backscatter coefficients can also be seen for the backscatter-related Ångström exponents. We find deviations to the values measured with lidar [Cattrall et al., 2005, Table 2].
 In contrast to the lidar, where we only used data that describe dust above 2000 m height, the Sun photometer observed the entire column. Thus a weak impact of other continental particles in the case of Banizoumbou and of marine particles at Capo Verde cannot be excluded. Especially marine particles can lower the lidar ratio significantly.
 Another source of natural aerosols is biomass burning. We find similar values of the lidar ratio. However the AERONET observations result in significantly larger extinction-related Ångström exponents. The main reason for the strong difference to the lidar-derived values may lie in the location of the fire sources. The AERONET data describe fires in tropical areas of Africa and South America [see Cattrall et al., 2005, Table 1] whereas the lidar-derived values describe aged smoke from boreal areas of the Northern Hemisphere (Siberia and Canada). As discussed by Müller et al.  there may be a significant increase of particle size during long-range transport. Much smaller particles are observed near the fire sources in Africa and South America [Reid et al., 2005]. In contrast to the difference of the extinction-related Ångström exponents we find similar values of the backscatter-related Ångström exponents.
 We find differences for continental haze (Europe, Southeast Asia), which are particularly large for the lidar ratio. The lidar ratios determined from the Sun photometer data are considerably higher (about 20 sr) than the lidar ratios measured with Raman lidar. For example we obtain mean values around 50–55 sr at 532 nm for central European haze, whereas mean lidar ratios of about 75 sr were derived from the Sun photometer observations at 550 nm at the French AERONET sites. The extinction-related Ångström exponents of the two data sets agree within their ranges of variability. The backscatter-related Ångström exponents differ from one another. The very large lidar ratios may be the result of an underestimation of the backscatter coefficient in the AERONET data analysis, caused by, e.g., an incorrect complex refractive index which is used in the calculation of the backscatter coefficients.
 Also in the case of Asian pollution, the lidar ratios derived from the AERONET data appear to be too large. The mean photometer-derived lidar ratio is close to 60 sr, whereas our Raman lidar observations and other observations mentioned before clearly indicate a mean value of approximately 40–45 sr. Main reason for the lower lidar ratios may be the strong impact of particles in the coarse-mode fraction that is always present in Southeast Asia and which significantly lowers the lidar ratio. However, the Ångström exponents derived from lidar and Sun photometer agree very well. Obviously both lidar and photometer observed similar aerosol conditions regarding fine-mode and coarse-mode fractions. The low Ångstöm exponents clearly indicate the impact of coarse-mode particles on the observed optical effects. Again, only an underestimation of the backscatter coefficient in the case of the AERONET data analysis remains to explain the discrepancy between the lidar ratios.
 For the first time aerosol-type-dependent lidar ratio statistics solely based on Raman lidar measurements are presented. The statistics cover the most important aerosol types such as marine particles, desert dust particles, aged biomass-burning smoke, and urban/industrial haze representing conditions in developed (Europe) and less developed regions (south and east Asia) of the world.
 Such statistics are needed for elastic-backscatter lidars, such as the recently launched lidar aboard the CALIPSO satellite. That lidar cannot measure particle extinction without reasonable assumptions on the lidar ratio.
 The European Space Agency plans to launch two high-spectral-resolution lidars (ALADIN and ATLID) in the coming years, which also calls for such a lidar ratio climatology. These two systems will deliver the lidar ratio at 355 nm, and it will be necessary to link the observations of these instruments to the aerosol climatology of CALIPSO.
 The comparison with a photometer-based lidar ratio climatology that strongly relies on modeling of the 180°-scattering reveals significant differences for some of the investigated aerosol types. The reasons for the discrepancies could not be unambiguously determined. Simultaneous observations with Raman lidar and AERONET photometer, as done, e.g., during the recent SAMUM campaign, are thus highly recommended to identify the causes for the different results.
 EARLINET has been funded by the European Commission under grant EVR1-CT-1999-40003. The SAMUM research group is funded by the Deutsche Forschungsgemeinschaft (DFG) under grant FOR 539.