Three cloud data sets, each covering four months of observations, were recently recorded with a lidar at Punta Arenas (53°S), Chile, at Stellenbosch (34°S, near Cape Town), South Africa, and aboard the research vessel Polarstern during three north-south cruises. By comparing these observations with an 11–year cloud data set measured with a lidar at Leipzig (51°N), Germany, the occurrence of heterogeneous ice formation (as a function of cloud top temperature) for very different aerosol conditions in the northern and southern hemisphere is investigated. Large differences in the heterogeneous freezing behavior in the mostly layered clouds are found. For example, <20%, 30%–40% and around 70% of the cloud layers with cloud top temperatures from −15°C to −20°C, showed ice formation over Punta Arenas, Stellenbosch, and Leipzig, respectively. The observed strong contrast reflects the differences in the free tropospheric aerosol conditions at northern midlatitudes, that are controlled by anthropogenic pollution, mineral dust, forest fire smoke, terrestrial biological material and high southern midlatitudes with clean marine conditions.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Aerosol particles are required to initiate ice formation in the atmosphere at temperatures down to −37°C. Besides aerosol properties (number concentration, size, aerosol type), ambient meteorological conditions (temperature, relative humidity, vertical motions, turbulent mixing), as well as cloud properties itself (e.g., droplet concentration and size, dynamics) control the nucleation of first ice crystals in water clouds. There is no doubt that aerosol particles serving as cloud condensation nuclei (CCN) or ice nuclei (IN) sensitively influence these cloud processes. One of the basic questions is however, how important this impact is (compared to ambient meteorological influences and cloud dynamics) and whether global models [e.g., Hoose et al., 2010], used to predict future climate change, properly reflect aerosol effects on ice formation and the subsequent impact on the production of rain in different clean to heavily polluted areas around the globe. The lidar study presented here is motivated by the question: Can we provide observational evidence for a strong impact of (continental) aerosols on heterogeneous ice formation?
 A promising way is to contrast cloud observations in the lower and middle troposphere at high northern and southern latitudes. At northern midlatitudes, the aerosol conditions in the free troposphere are widely determined by anthropogenic pollution from North America, Europe, and East Asia, desert dust emissions in Africa, America, and Asia, and release of forest fire and other biomass burning smoke in the latitudinal belt from 35–65°N [Mattis et al., 2008]). In addition biological material (pollen, bacteria, leaf fragments) may contribute to efficient ice nucleation. At high southern latitudes over remote oceanic areas (>50°S), continental aerosol particles are widely absent and even if present the most favorable large IN may already be removed by dry and wet deposition [Bigg, 1973]. Clean marine conditions prevail here [Minikin et al., 2003]. An approach to contrast cloud observations at high northern and southern midlatitudes was performed in the framework of the INCA (Interhemispheric Differences in Cirrus Properties from Anthropogenic Emissions) project that focused on cirrus evolution and properties [Gayet et al., 2004].
 For the investigation of the occurrence of heterogeneous ice formation in low level and midlevel tropospheric clouds, the polarization lidar technique is well suited [Sassen et al., 2003; Sassen, 2005; Ansmann et al., 2005, 2008]. Compared to spaceborne active remote sensing [Choi et al., 2010], ground-based systems have the advantage that cloud virgae can be studied to unambiguously identify heterogeneous ice nucleation. This information is essential, but can hardly be provided when using lidar platforms above the cloud layers because of strong multiple scattering effects and the frequent attenuation of the lidar beam by optically thick liquid-water layers above these virgae.
 We already performed a series of studies regarding the role of desert dust and tropical aerosols in heterogeneous ice formation [Ansmann et al., 2008, 2009; Seifert et al., 2010]. Recently, the impact of volcanic ash was investigated [Seifert et al., 2011]. We already found large differences in the heterogeneous freezing temperatures for Cape Verde and Leipzig [Seifert et al., 2010]. An unambiguous conclusion regarding the role and interaction of meteorological effects and aerosol-related effects could however not be drawn from these measurements alone.
 In this letter, we use the opportunity of two cruises of the research vessel (RV) Polarstern to Antarctica (ANT-XXVI, October 2009 – May 2010, ANT-XXVII, October 2010 – May 2011) to extend our lidar cloud studies toward the southern hemisphere, i.e., to moderately to almost unpolluted areas. We performed lidar observations of clouds aboard the RV Polarstern during the cruise from Germany to Chile in October-November 2009, and back to Germany in April-May 2010, and during another cruise to South Africa in October-November 2010. We deployed the lidar at Punta Arenas (53.2°S, 70.9°W), Chile, in late November 2009 (before the RV Polarstern moved on to Antarctica) and performed continuous observations from 4 December 2009 to 4 April 2010. The same lidar was deployed one year later at Stellenbosch (33.9°S, 18.9°E), 60 km east of Cape Town, South Africa, and performed lidar observations from 2 December 2010 to 13 April 2011. Figure 1 provides an overview of the three cruises of the RV Polarstern and the locations of the lidar sites in South America and South Africa. Cape Verde (14.9°N, 23.5°W) and Leipzig (51.3°N, 12.4°E) are also indicated in Figure 1. As Leipzig, Punta Arenas is located at a high latitude (>50°) and the INCA study [Gayet et al., 2004] already indicated similar meteorological conditions in the free troposphere at these high midlatitudes north and south of the equator. Thus, a possible impact of aerosols on the initiation of ice nucleation should become visible by contrasting the cloud observations at Leipzig and Punta Arenas.
 The methodology applied in this study is described in detail by Ansmann et al.  and Seifert et al. . A portable multiwavelength Raman and polarization lidar system PollyXT was used for cloud observations. The lidar was designed to perform automated and remotely controlled lidar measurements [Althausen et al., 2009]. Laser pulses are transmitted at 355, 532 and 1064 nm with a repetition rate of 20 Hz. The outgoing 355 nm laser pulses are linearly polarized and the cross-polarized and total backscatter signals are measured. From these signals the volume depolarization ratio is computed and used to discriminate cloud layers with dominating liquid drop backscattering from layers with ice crystal backscattering (see Seifert et al.  for more details). The laser beams are tilted (5° off zenith angle) to avoid specular reflection by horizontally aligned planar ice crystals which complicates cloud-phase discrimination. We should emphasize here that we can only identify those clouds as ice-containing clouds which produce a significant amount of ice and thus precipitation (virgae). If ice crystals form but do not show up below the cloud layer we will not be able to identify this cloud as mixed-phase cloud. All these clouds are then counted as liquid-water clouds. However, the good agreement between lidar and aircraft observations regarding statistics of ice-containing clouds [Seifert et al., 2010] indicates that our method provides a trustworthy view on cloud phase statistics.
 At Punta Arenas, the 355 nm depolarization channel worked properly only during the first seven weeks of the four-month long campaign. Clouds, observed during the second half were classified as ice-containing clouds when they produced virgae. It was found during the first half of the campaign that all virgae were ice crystal fall streaks. In the case of cloud layers without virgae, we identified the liquid-water clouds from the combination of observed cloud temporal development (less variable than ice-containing clouds), geometrical depth (usually thin), and cloud extinction coefficient (>1000 Mm−1). Pure liquid-water clouds with virgae consisting of drops only would be misinterpreted as ice-containing clouds which would lead to an overestimation of the ice-containing cloud fraction.
 A cloud layer is defined as one single cloud layer, when it is separated from adjacent ones by 500 m in the vertical or by 5 minutes in time. Long-lasting upper-level cloud layers, temporally not visible because of the presence of optically thick low-level clouds, are counted as one cloud layer. It is assumed in our approach that ice formation is initiated at cloud top, i.e., in the coldest part of the cloud layer [Lebo et al., 2008]. The respective cloud top temperature for each observed cloud was calculated from GDAS1-data that are based on the global data assimilation system GDAS (Global Data Analysis System, http://www.arl.noaa.gov/gdas.php). The analysis of the model-derived temperatures and temperatures from rarely launched nearby radiosonde ascents show deviations of the order of 1.5 K or less in the height range from 2 to 10 km.
 Cloud observations with a lidar are restricted to clouds with optical depths of <2.5 at visible wavelengths. Only for these clouds (transparent for laser light), the top of the cloud can be identified which is a basic requirement in the study of heterogeneous ice formation as a function of temperature. Thus, the following investigation is mainly based on layered clouds in the free troposphere from 2–8 km height.
Figure 2 shows frequency-of-occurrence distributions of cloud depth and cloud top height of all liquid and ice-containing clouds (mixed-phase and cirrus cloud layers) considered in our study. The observations at Leipzig cover the summer months (April to September) of the years from 1997–2008. About 1100 cloud layers (on average 30 min long) are considered in Figure 2. The Punta Arenas observations (about 900 cloud cases on average 50 min long) were performed during the summer months of December 2009 and January 2010, i.e., within the seven-week period during which the depolarization channel was in operation and allowed us to clearly identify ice crystals. By considering also the second part (February–March 2010, no depolarization available) almost the same curve for the ice-containing cloud fraction as shown in Figure 2 is obtained. The differences in the number of cloud cases (and their duration) result from the different measurement strategies. The Leipzig observations were performed in the framework of the European Aerosol Research Lidar Network (EARLINET) project and were preferably conducted during situations without boundary layer clouds or with only few low-level to mid-level cloud layers. EARLINET Lidar measurements are typically taken every third day and last for 30 minutes to two hours. In contrast, around-the-clock lidar observations were performed aboard RV Polarstern, at Punta Arenas, and at Stellenbosch. The lidar was only stopped during situations with precipitation (reaching the rain sensor). This also explains slight differences in the found statistics.
 As can be seen in Figure 2, the depth of liquid clouds was mostly below 1 km at both sites and the top heights of the liquid-water clouds mostly ranged from 1–6 km height. Cloud top heights of the mixed-phase clouds (not presented) show a broad distribution from typically 2–8 km height. Most of mixed-phase clouds were 500–1000 m in depth and less than 3 km thick when considering the virgae layer in the cloud depth calculation. The broad distribution of cloud depths in the case of ice-containing clouds is caused by cirrus and the associated deep virgae layers. All in all, the similarity in the cloud characteristics at both sites point to similar atmospheric conditions during cloud formations at high northern and southern midlatitudes.
Figure 3 shows the fraction of ice-containing clouds related to all observed cloud layers for eight classes (5 K intervals) of cloud top temperature between 0°C and −40°C. Curves are shown for Leipzig and Punta Arenas, and for comparison, also for Stellenbosch, the RV Polarstern cruises, and Cape Verde (January-February 2008 observations [Ansmann et al., 2009]). The strongest increase in the ice-containing cloud fraction with decreasing temperature, expressing most efficient heterogeneous ice formation, is observed over Leipzig. The weakest increase is found for Punta Arenas. The increase in the ice-containing cloud fraction with decreasing temperature is even weaker for Punta Arenas when all clouds of the four-month period are considered. Less than 10%, <20%, about 30%–40% and around 70% of the cloud layers with cloud top temperatures from −15°C to −20°C, showed ice formation over Cape Verde, Punta Arenas, Stellenbosch/Polarstern, and Leipzig, respectively. More than 90% of the cloud layers contained ice over Leipzig and Stellenbosch at temperatures from −25°C to −30°C, whereas this 90% level was not reached at Punta Arenas before cloud top temperatures were below −35°C. Half of the observed loud layers were ice-containing at −13°C at Leipzig and at −30°C at Punta Arenas. At Cape Verde, a westerly air flow from remote tropical oceanic regions prevails with an apparently low IN number concentration so that heterogeneous ice formation is weak at least for temperatures down to −25°C.
 The RV Polarstern data set includes all clouds observed during the three journeys from Bremerhaven (53°N) to Punta Arenas and back, and to South Africa, and thus represents almost mean conditions of heterogeneous ice formation for tropical and midlatitudes. The Stellenbosch curve is rather similar to the Polarstern curve. Aerosol conditions are controlled by local African sources of particles and long-range transport of dust, smoke, and anthropogenic pollution from South America.
 The strong contrast found in the cloud observations over Leipzig and Punta Arenas is obviously related to the rather different aerosol conditions at the polluted site in the northern hemisphere and the rather clean maritime site south of 50°S. The large deserts in the northern hemisphere permanently inject favorable coarse mode particles into the atmosphere. Global observations by the spaceborne Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) suggest also a link between the locations of deserts and enhanced occurrence of glaciated clouds [Choi et al., 2010].
 A further impact by anthropogenic particles can not be excluded. Ebert et al.  recently reviewed in situ observations of ice particle residuals and presented new results obtained in mixed-phase clouds occurring at the high alpine research station Jungfraujoch, Switzerland, at 3580 m height above sea level, and about 600 km southwest of Leipzig. The authors concluded that admixtures of anthropogenic aerosol components are responsible for the increased IN efficiency within the mixed-phase clouds. Especially lead-bearing particles (most likely originating from combustion of leaded aviation fuel used by helicopters and small aircraft) and complex mixtures with silicates or metal oxides were found to be favorable ice nuclei. Most ice nuclei identified were internal mixtures.
 In contrast, the aerosol concentration in the free troposphere at high southern midlatitudes can be regarded to be rather low, and the main particle source is the ocean. Punta Arenas is situated in the Antarctic low pressure belt that causes westerly air streams during most of the year. Thus maritime air masses dominate aerosol conditions throughout the troposphere. Sources for favorable coarse mode IN are absent. This conclusion is in agreement with the shipborne observations of Bigg  in and around Australia from 20–80°S, and also with the study of Ebert et al. . They found that particle groups containing sulfates and sea salt were not enriched in the ice particle residuals.
 With decreasing temperature, the immersion freezing mode becomes more and more efficient. However, if soluble sea salt and other marine, sulfate containing particles are inside the droplets, ice formation may be significantly suppressed [Zobrist et al., 2008]. That may explain the likewise slow increase of the ice-containing cloud fraction with decreasing temperature at Punta Arenas.
 By contrasting liquid-water and ice-containing cloud data sets observed with polarization lidar at several places at northern and southern midlatitudes and aboard the RV Polarstern a clear north-to-south decrease in the efficiency of heterogeneous ice formation was found. This decrease is well correlated with the decreasing contribution of anthropogenic aerosols and mineral dust particles to the overall aerosol load in the free troposphere when going from northern to southern hemispheric sites in or close to the Atlantic Ocean. The study documents the significant role of aerosol particles in the glaciation of shallow cloud layers in the northern hemisphere.
 Funding by the Wilhelm Gottfried Leibniz Association (OCEANET project in the framework of PAKT) is appreciated. We thank the RV Polarstern team, the Alfred Wegener Institute for Polar and Marine Research (AWI) and the project leader of OCEANET Andreas Macke for their permanent support and interest in this work during the cruises ANT-XXVI and ANT-XXVII.
 The Editor thanks an anonymous reviewer for assistance in evaluating this paper.