Ten years of multiwavelength Raman lidar observations of free-tropospheric aerosol layers over central Europe: Geometrical properties and annual cycle

Authors


Abstract

[1] We present geometrical properties and seasonal variations of appearance of aerosol particle pollution in the free troposphere over the central European lidar site at Leipzig, Germany. The data set has been acquired with Raman lidar in the past 10 years in the framework of the German Lidar Network (1997–2000) and since 2000 in the framework of the European Aerosol Research Lidar Network (EARLINET). In summary we analyzed 1028 measurements. Geometrical depth of the pollution layers was ≤1 km in 33% of all cases. Geometrical depths >5 km were found in 10% of all cases. Traces of particle pollution were detected up to the height of the tropopause. Forest-fire burning in North America causes intrusion of particles into the stratosphere. Seven hundred seventeen of all observations were carried out on the basis of a regular measurement schedule which allows us to establish a statistic on the frequency of particle transport in the free troposphere. In 43% of the regular measurements we observed pollution above the continental boundary layer. The lofted particle layers largely result from intercontinental long-range transport. We use backward trajectory analysis to identify the main source regions of the lofted pollution layers. In 19% of all regular measurements, free-tropospheric pollution was advected from North America. Forest-fire smoke from Canada and anthropogenic pollution from urban areas of the United States of America and Canada were the sources of the particle layers. We find a strong seasonal dependence of occurrence of these layers with a peak in June–August of each year. In a few cases we observed forest-fire smoke advected from Siberia and east Asia with winds from westerly directions. Pollution advected from areas north of 70°N presents another transport channel. That pollution consists of Arctic haze or mixtures of haze with anthropogenic pollution. The main occurrence of such particle layers is around springtime of each year. Import of mineral dust from the Sahara represents another transport path. Most of such cases are observed during late springtime and summertime. Free-tropospheric pollution advected from east and southeast Europe and Russia presents one transport channel from within the Euro-Asian continent.

1. Introduction

[2] Despite the importance of free-tropospheric particles on climate and air quality [McKendry et al., 2001; Jacob et al., 1999; Creilson et al., 2003; Collins et al., 2000; Prather et al., 2003] their optical and microphysical properties, their geometrical features, and particle transport mechanisms are still poorly understood. This lack of understanding becomes obvious in the first textbook review on present-day knowledge of long-range transport of atmospheric pollution [Stohl, 2004]. The review mainly focuses on trace gases from manmade pollution. The transport of gases from pollution sources can be described comparably well. Transport mechanisms of gases however cannot readily be applied to aerosol particles as, for instance, particle washout and other cloud processing mechanisms have additional influence on the presence and transport of particle pollution. For example, observations of particles carried out with lidar in the Indian Ocean [Ansmann et al., 2000; Müller et al., 2000; Franke et al., 2003; Léon et al., 2002] and in the framework of the European Aerosol Research Lidar Network (EARLINET) [Bösenberg et al., 2003] are in contradiction to the common assumption that measurements of trace gases can be used to explain the presence as well as the transport of particulate pollution [Lelieveld et al., 2001, 2002].

[3] Andreae et al. [2003] carried out first detailed airborne observations of long-range transport of Euroasian emissions to the remote northeast Pacific troposphere in 1985. The authors observed air masses from east Asia that had traveled in heights around 3–6 km for 4–8 days before observation. Long-range transport of aerosols was observed over the North and South Pacific in the framework of airborne studies in 1990 [Clarke, 1993]. Jaffe et al. [1999] reported on transport of Asian air pollution to North America. Bertschi et al. [2004] and Bertschi and Jaffe [2005] subsequently provided a detailed study of long-range transport to the northeast Pacific on the basis of aircraft flights in 2002. Lofted layers were observed from ground level to 6 km height. Geometrical depth of the layers of 0.2–3 km are reported. The measurements were carried out in the frame of PHOBEA (Photochemical Ozone Budget of the Eastern North Pacific Atmosphere) activities which lasted from 1999 until 2003. Studies on black carbon from biomass burning emission were carried out by Clarke et al. [2001] over the Pacific Ocean and by Clarke et al. [2004] in the outflow region of Southeast Asia.

[4] Aerosol pollution transport in the free troposphere over the North Atlantic was studied with airborne instrumentation in the vicinity of the Azore islands during ASTEX (Atlantic Stratocumulus Transition Experiment) in June 1992 [Clarke et al., 1997]. First detailed airborne studies of optical and microphysical properties of lofted aerosol pollution from North America were reported by Petzold et al. [2002] and Fiebig et al. [2002]. The authors observed an forest-fire smoke plume that had been advected from west Canada to central Europe. These studies were carried out in the framework of LACE (Lindenberg Aerosol Characterization Experiment) [Ansmann et al., 2002]. LACE also provided the first data of such events over central Europe from ground-based Raman lidar observations [Wandinger et al., 2002].

[5] Knowledge on aerosol pollution long-range transport and particularly knowledge on black carbon in such plumes was significantly broadened by Petzold et al. [2008], who carried out detailed airborne studies in summer 2004 in the framework of ICARTT (International Consortium for Atmospheric Research on Transport and Transformation). Clarke et al. [2007] also reported on studies on forest-fire plumes over the North Atlantic in the framework of ICARTT.

[6] Latest results on intercontinental pollution transport are discussed by Ramanathan et al. [2005] and Chin et al. [2007]. The authors focus on results from satellites and transport modeling.

[7] Optical and microphysical properties of boundary-layer aerosols which originate from local and regional emissions of particles and gases, usually are very different from free-tropospheric particles which are often advected over large distances from other continents. Information on the altitude in which particle layers appear is important for calculating aerosol radiative forcing. It depends on the atmosphere's albedo, and it makes a significant difference in the radiative impact whether the aerosol particles are above a cloudy boundary layer or within the boundary layer [Wagner et al., 2001; Keil and Haywood, 2003; Forster et al., 2001].

[8] Particle layers in the free troposphere often are optically very thin with an optical depth of 0.01–0.03 at 500 nm wavelength. Thus their direct effect on climate forcing may be low. However it is widely unknown how these aerosol particle layers influence formation of clouds and precipitation. They increase the free-tropospheric background aerosol load and may therefore have a comparably strong effect on indirect climate forcing, because of their large-scale structures, which may reach continental dimensions [Ansmann et al., 2005] and transport distances on the hemispheric scale [e.g., Damoah et al., 2004; Müller et al., 2003; Wandinger et al., 2002].

[9] Passive remote-sensing instrumentation is limited in documenting and characterizing long-range transport of particles in the free troposphere. For instance first results on long-term satellite observations of aerosol pollution on the global scale were reported by Husar et al. [1997]. Satellite passive sensors however cannot separate between particles in the boundary layer and pollution in the free troposphere. It still is a very challenging task to identify aerosols over land with such instruments. The retrieved parameters rather describe a mixture of particle types. The same also holds true for ground-based sensors like Sun photometers.

[10] Differences between free-tropospheric pollution and boundary layer particles are described by Wandinger et al. [2002]. The authors for the first time documented in detail an event of intercontinental long-range transport on the basis of vertically resolved measurements with multiwavelength Raman lidar. Forest-fire smoke generated over west Canada was transported to central Europe in the course of 6 days [Fiebig et al., 2002], and observed in the framework of the Lindenberg Aerosol Characterization Experiment 1998 (LACE 98) in the summer of 1998 [Ansmann et al., 2002].

[11] A detailed summary of our observations of lofted aerosol pollution during the Aerosol Characterization Experiment 2 (ACE 2) in southern Europe in 1997, LACE 98 in central Europe, and four campaigns during INDOEX in south Asia in 1999–2000 is given by Müller et al. [2007a].

[12] In the following years we intensified our observations of free-tropospheric pollution with our stationary Raman lidar in the framework of the German Lidar Network (AFS-Deutsches Lidarnetzwerk) [Bösenberg et al., 2001] and EARLINET [Mattis et al., 2004]. Mattis et al. [2003] reported for the first time height-resolved observations of forest-fire smoke for a complete forest-fire season. On the basis of sophisticated model calculations of air-mass transport it was shown that some part of that smoke circled from its source regions in Siberia and North America around the globe [Damoah et al., 2004]. We determined optical and microphysical properties of smoke layers from Siberia and North America from measurements with our stationary multiwavelength Raman lidar [Müller et al., 2005]. We also observed transport of anthropogenic pollution from North America [Müller et al., 2005], and mixtures of Arctic haze and east European pollution advected from polar regions and east Europe [Müller et al., 2004]. Mattis et al. [2002b] report on optical properties of Sahara dust that was transported from North Africa to central Europe. A detailed case study on optical and microphysical properties of Sahara dust over Leipzig was discussed by Müller et al. [2003]. One large-scale event of transport of Sahara dust was observed at Leipzig and other lidar stations of the European Aerosol Research Lidar Network [Ansmann et al., 2003]. Modification of mineral-dust optical properties could be studied in detail with high vertical resolution.

[13] In this contribution we summarize 10 years of Raman lidar observations at Leipzig. This contribution to our knowledge is the first systematic, height-resolved study on long-range transport of aerosols. Particularly we document for the first time statistical results on aerosol pollution transport from North America to Europe, with information on the height distribution of that pollution. Our study thus fills a gap in our knowledge on hemispheric pollution transport, which primarily deals with trace gases, for example, ozone and carbon monoxide, and individual aerosol species, for example, sulfur and carbonaceous particles, leaving large gaps in a comprehensive assessment of surface particulate matter levels that arise from regional and long-range transport.

[14] We focus in this contribution on geometrical properties, and the annual, seasonal, and monthly variation of appearance of lofted pollution plumes advected on intercontinental distances from North America, polar areas, and North Africa. The results presented in this first of a series of papers can also be obtained by simple backscatter lidars. However, in future contributions we shall extend our results to quantitative descriptions of optical and microphysical particle properties of these pollution plumes, which requires multiwavelength Raman lidar.

[15] In section 2 we briefly discuss the methodology. We present the time series of our observations in section 3. In section 4 we discuss our results. We conclude our contribution with a summary and outlook in section 5.

2. Methodology

2.1. Instrument

[16] The data were collected with the stationary Raman lidar MARTHA (Multiwavelength Atmospheric Raman lidar for Temperature, Humidity, and Aerosol profiling) [Mattis, 2002; Mattis et al., 2002a] at the Leibniz Institute for Tropospheric Research (IfT) in Leipzig, Germany (51.3°N, 12.4°E). Regular measurements were carried out three times each week, i.e., Monday afternoon, and Monday and Thursday after sunset. A statistical overview on the observational results for the timeframe from 2000 to 2003 with the focus on the aerosol in the boundary layer is given by Mattis et al. [2004].

[17] We conducted 717 regular measurements since 1997, which is equivalent to 41% of the maximum possible number of measurements. In about two thirds of the days that are missing, unfavorable weather, such as rain, prevented aerosol observations. We carried out 311 additional observations during special situations as for instance transport of Sahara dust and advection of smoke from North America and east Asia to central Europe.

[18] We observed particle layers in the free troposphere in 478 of our measurements, which is equivalent to 379 measurement days. Of those 379 measurement days, free-tropospheric particle layers were observed on 310 regular measurement schedules (equivalent to 262 measurement days), and on 168 special measurements schedules (equivalent to 117 days).

[19] We covered around 2100 measurement hours in total, of which 1100 hours belong to regular observations. During approximately 480 hours of the regular observations we observed particles in the free troposphere. We carried out 1000 hours of observations during special events, and we observed free-tropospheric particle layers during 540 hours.

[20] MARTHA is located in the upper floor of the main building of the Leibniz Institute for Tropospheric Research (IfT). One Nd:YAG laser generates laser pulses at 355, 532, and 1064 nm wavelength with a repetition rate of 30 Hz. The laser beam is expanded 15-fold and vertically transmitted into the atmosphere. The backscattered radiation is collected by a Cassegrain telescope and transmitted to the signal detection unit. Backscatter signals are detected at the three emitted wavelengths, at 387 and 607 nm resulting from Raman scattering from nitrogen (355 and 532 nm primary wavelength), and at 408 nm resulting from Raman scattering from water vapor (355 nm primary wavelength). The system detects the component of light cross-polarized to the plane of polarization of the outgoing beam at 532 nm. Rotational Raman signals are detected around 532-nm wavelength for measurements of temperature. Those channels were installed in 2000 [Mattis et al., 2002a; Arshinov et al., 2005].

2.2. Planetary Boundary Layer Height

[21] In this work we present statistical information on geometrical properties of the free-tropospheric particle layers. For that purpose we have to determine the bottom height of the lofted layers. That task involves the determination of the top of the planetary boundary layer (PBL, daytime mixing layer height) or of the residual layer height [Stull, 1983]. The residual layer forms in the late afternoon (before sunset) in the upper part of the PBL and is usually present up to the late morning of the next day. In the following discussion we do not distinguish between the PBL top and the residual layer top. Both layer top heights for simplicity are denoted as PBL top height. If we determine the PBL top, we have to consider that sharp gradients in the aerosol concentration may occur in the entrainment zone where cleaner and dryer air from the free atmosphere above is mixed into the aerosol-laden more moist boundary layer [Cohn and Angevine, 2000; Steyn et al., 1999]. Particles in the entrainment zone may thus generate a false signal of a lofted particle layer.

[22] There exists a wealth of literature on how to determine the height of the PBL. The kind of used instruments, of which radiosondes certainly are the most common ones, have crucial influence on the actually determined height. This PBL height always must be seen in the context of the variety of definitions and the measurement site. A critical review is given by Seibert et al. [2000]. Pahlow [2002] presents an extensive comparison on PBL height obtained from lidar backscatter profiles and derived from radiosonde profiles of specific humidity and virtual potential temperature. At times the author finds significant differences.

[23] There has been a long-lasting discussion in the framework of EARLINET workshops on how to define the height of the planetary boundary layer (PBL) on the basis of optical profiles measured with lidar. The chosen method relies on the observation that the intensity of the raw signal from particle and molecular backscattering is a qualitative measure of the presence of particles. Changes of the strength of the range-corrected signal are caused by changes in particle intensive properties, for example, particle size, or extensive properties, for example, particle number concentration.

[24] There are elaborate techniques on how to determine the PBL height from particle backscatter profiles. A fitting method that uses lidar backscatter ratio profiles to determine mixed-layer depth and entrainment zone has been proposed by Steyn et al. [1999]. A sophisticated method which rests upon wavelet analysis [Cohn and Angevine, 2000; Brooks, 2003] has just recently been developed and is now applied to a data set we collected with our Raman lidar POLLY (POrtabLe Lidar SYstem) [Althausen et al., 2004; Baars, 2007]. Recent statistics on PBL top heights determined with lidar are presented by Lammert and Bösenberg [2006], Wiegner et al. [1997], Martucci et al. [2007], and Morille et al. [2007].

[25] In our study we adopt the definition of PBL height, the way it has been accepted within the EARLINET community: The top of the planetary boundary layer is defined as the steepest gradient of the profile of the particle backscatter signal, i.e., the largest local minimum of the first derivative of the range-corrected signal [Bösenberg et al., 2003]. That definition uses a minimum height of 500 m of the planetary boundary layer as an additional constraint in order to rule out that strong gradients near the surface are defined as boundary layer top height. The method assumes that the aerosol concentration is significantly higher in the boundary layer than in the free troposphere which is usually the case. Exceptions are events with strong Sahara dust plumes, when the bottom of the African-air layer coincides with the PBL top height. We point out that in case of the detection of the PBL top, there is always enough intensity of the backscattered signal and the applicability of the gradient method is not limited. Leipzig is a continental site. Thus the number of particles in the PBL is always high enough to cause backscatter signals which are always significantly above the molecular background (inside the PBL). Furthermore, we use a lidar with high output power, and we collect the signals with relatively high integration times of 1–1.5 hours. Thus it is always possible to obtain high enough signal-to-noise ratios even at the top of highly polluted continental PBLs.

2.3. Geometrical Properties of Free-Tropospheric Layers

[26] We use the range-corrected backscatter signals at 1064 nm to determine the geometrical depth of the particle layers. Because of malfunctioning of the optical setup we use the signals at 532 nm acquired between 1 February 1997 and 19 March 1998.

[27] Figure 1 for illustration shows an example of how we determine bottom and top height, and thus geometrical depth of the lofted particle layers, for the case of a Sahara dust plume. We calculate the first derivative of the signal profile. At the bottom of the dust layer the increase of the particle backscatter intensity causes a local maximum of the derivative at about 1.17 km height. The derivative shows a first local minimum at about 4-km height, which is caused by a strong decrease of the particle concentration. There is a second layer with lower particle content directly on top of this first layer. Thus the bottom height of the second layer coincides with the top height of the first particle layer (at 4-km height). The top height of the second layer is indicated by the second local minimum of the derivative profile at about 5.8 km height.

Figure 1.

(a) Example of how the heights of the bottom and the top of a Sahara dust layer are determined with the gradient method. The measurement shows (left) the range-corrected 1064-nm backscatter signal (no units) of Sahara dust observed over Leipzig from 1645 to 1800 UTC on 30 October 2001, (middle) the corresponding temporally averaged signal profile (red), and radiosonde profiles of potential temperature (green), and relative humidity (blue), and (right) the gradient of the backscatter signal (red) and of potential temperature (green). The radiosonde was launched near Oppin/Halle at 0000 UTC on 30/31 October 2001. Oppin is approximately 30 km to the northwest of Leipzig. (b) Source region of the dust plume. That region is identified on the basis of 10-day backward trajectory analysis with the HYSPLIT trajectory model and the use of meteorological data of the National Centers for Environmental Prediction. The endpoint of the trajectories is at 4, 4.5, and 5 km height above ground level (AGL) at 1700 UTC on 30 October 2001.

[28] We use several criteria to check if the gradient method fails. First, we checked whether the layer boundaries determined by the gradient method coincide with the bottom and the top heights that we recognize in coherent structures of the height time displays of the range-corrected lidar signal. Secondly we test if there is agreement with gradients in potential temperature and humidity. As an example Figure 1 also shows the profiles of relative humidity and potential temperature. The data were acquired with a radiosonde launched by the German Meteorological Service (DWD) at Oppin (51.6°N, 12.1°E) around 4 hours after we started the lidar measurement.

[29] The temperature profile and the profile of the corresponding gradient indicate that the bottom and the top of the first dust layer are at around 1 km and 4 km height, respectively. The top of the second layer is also indicated by a strong positive gradient in the potential-temperature profile.

[30] Stable stratification is observed below 1 km height, which is caused by the warm Sahara dust layer above the comparably cold boundary layer. The potential temperature indicates an almost well-mixed lower part of the dust layer from 1 to 1.8 km height. The relative humidity in the lower layer ranges from 30% to 60%. The second layer is more humid with values between 60% and 80%. The strong decrease of humidity from 80% to 30% around 6 km height indicates the top of the dust plume.

[31] Figure 1 shows that comparably strong changes of particle concentration at the boundaries of the layers are needed in order to identify those changes as a gradient. We used a simple difference ratio calculated from three data successive points. In that way the derivative is calculated as

equation image

where z denotes the height and (z − 1) and (z + 1) the height bin above and below. Here sig is the range-corrected signal. We did not analyze any numerical biases since we are not interested in the absolute values of the derivative profile. We obtain the PBL height and layer boundaries only from the shape of the profile of the signal gradient; that is, only the location of maxima (minima) are of interest, not the maximum (minimum) values.

[32] In few cases of low particle concentration and accordingly small absolute changes of particle concentration with height, particle layers may not be identified with the gradient method at all, because the intensity of the backscatter signals may be too low. In such situations we simply used the height time displays of the range-corrected lidar signal and determined the bottom and the top heights from the coherent structures produced by the lofted aerosol layers.

[33] Indeed this last methodology certainly involves a rather subjective element, if we identify aerosol layers not only by objective methods like thresholds in backscatter coefficient but also with a by-eye identification. We believe, though, that the human eye is the best available tool for pattern recognition. We shall use our experience which we obtained from the analysis of our 10-year data set to develop improved automated layer detection algorithms, based on parameterizations of, for example, vertical and temporal variability of the lidar return signals for the analysis of large data sets.

[34] Finally we would like to make some remarks regarding the separation of clouds from aerosol particles layers. A separation is very important, if we want to avoid strong biases in our statistical analysis. We used several criteria.

[35] 1. We used the gradient method. Signal gradients at cloud bases are much steeper than gradients at the boundaries of aerosol layers.

[36] 2. Water clouds can be detected because of their large optical depth. Usually, all signals get totally attenuated within the first height bins above a water cloud base. Water clouds cause depolarization ratios which are lower than those of the molecular background or of aerosols.

[37] 3. Ice clouds are characterized by high depolarization ratios. Ice clouds can be separated from mineral dust layers, which are also strongly depolarizing, by their different extinction-to-backscatter (lidar) ratios. Typical lidar ratios of cirrus clouds are 5–25 sr whereas Sahara dust is characterized by lidar ratios larger than 50 sr.

[38] 4. Usually, the temporal and vertical variability of lidar signals is lower in aerosol layers than in cloud layers.

[39] 5. All of the above mentioned criteria are used in parallel. There were very few cases in which a doubtless classification regarding cloud or aerosol was not possible. Such cases were excluded from our study.

2.4. Source Regions of Free-Tropospheric Layers

[40] We use backward trajectory analysis with HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory) to identify the main source regions of the free-tropospheric particles. A discussion of the model is given by Draxler and Hess [1997, 1998] and Draxler [2003]. HYSPLIT is available at http://www.arl.noaa.gov/ready/hysplit4.html.

[41] In our analysis we can only determine source regions of continental dimension. We usually calculated trajectories 10 days back in time. The arrival heights of the trajectories were chosen for the altitudes in which we identified particle layers.

[42] Figure 1 shows the example of 10-day backward trajectories calculated for different arrival heights of the particle plume. According to the model calculations the origin of the particles was in western parts of North Africa (Algeria, Lybia, Mali, Mauretania, Morocco). As indicated by the traces, the air mass that contained the dust was below 4 km height above ground level throughout most of the time. It is very likely that dust was picked up in that area. For example, we carried out lidar observations in Morocco in May/June 2006 in the framework of the Sahara Mineral Dust Experiment (SAMUM), and we found that dust layers reached top heights of 5 km above ground in most of our observations [Tesche et al., 2008].

[43] It is rather obvious that such an analysis may be affected by strong uncertainties. We cannot rule out that particles were injected over other regions that were along the track of the air masses. For instance in the example of Figure 1 anthropogenic pollution may have been injected into the plume when it traveled over Europe. Measurements of the particle depolarization ratio, however, show that there was a significant amount of mineral dust in that plume. We do not intend to determine the contribution of different particle types to the particle concentration in the lofted particle layers. For that reason we identify North Africa as source region of the particle plume.

[44] Furthermore we selected several arrival heights within one particle layer, if its geometrical depth was larger than approximately 1000 m. We also calculated trajectories with arrival heights below a particle layer and trajectories with arrival heights above a particle layer. Thus we could better constrain the main source region.

[45] In recent years we also analyzed in more detail the origin of lofted pollution layers for different cases of particle types and source areas [Müller et al., 2004, 2005; Müller, 2007]. We used the Lagrangian particle dispersion model FLEXPART [Stohl et al., 1998; Stohl and Thomson, 1999; Damoah et al., 2004]. The case studies show how a combination of backward trajectories, simulations with FLEXPART, and an optical and microphysical characterization of particles can be used for a rather trustworthy identification of the source regions of the investigated particle plumes.

3. Free-Tropospheric Particle Layers

3.1. Time Series

[46] Figure 2 shows our complete data set that we acquired from 1 February 1997 until December 2006. The total number of observations is 1028. Out of this number 717 observations belong to regular observations within the German Lidar Network and EARLINET. No measurements were possible from 6 October 2004 until 2 June 2005 because of a severe laser failure.

Figure 2.

Geometrical depth of lofted particle layers (thick vertical bars) advected to Leipzig in the years 1997–2006. Each short, thin vertical line along the time axis denotes one measurement case. Regular EARLINET measurement times were on Monday noon, and Monday and Thursday after sunset. Additional measurements were carried out during special events, for example, dust import to central Europe from the Sahara, and advection of forest-fire smoke from North America and Siberia. No lidar measurement could be carried out between 6 October 2004 and 2 June 2005 because of a major laser failure. The numbers below each year denote the number of regular measurements in which we observed free-tropospheric pollution, and the percentage of such events in relation to the total number of regular measurements in that specific year, respectively.

[47] Black vertical lines indicate the observed free-tropospheric layers from bottom to top. If a haze plume consisted of several sublayers, which at times showed a sandwich-like structure [Müller et al., 2007b], the bottom height was calculated for the lowest layer, and the top height was defined by the top of the uppermost lofted layer. We are currently analyzing these layered structures in more detail. Results will be presented in another publication.

3.2. Discussion

[48] We observed free-tropospheric pollution events on 310 out of our 717 regular observations. The number of pollution events in the free troposphere varies considerably among the different years. We find free-tropospheric particle layers in at least 30% of regular observations carried out per year. This value rises to 72% in 2006. On the 10-year average we observe free-tropospheric particle pollution in 43% of our observations.

[49] The geometrical depth of the plumes strongly varies from several hundred meters up to 10 km. We did not explicitly define a minimum geometrical depth of the aerosol layers. According to the range resolution of the raw signals that we acquire with our lidar and some atmospheric variations the thinnest aerosol layers that we can detect are 180 m thick, which is equivalent to three height bins of the data acquisition system.

[50] Furthermore those very thin layers were only counted as lofted layers, if they are clearly separated from the PBL by an aerosol free region of more than approximately 500 m geometrical depth. Thus we can be sure that such very thin layers are indeed lofted layers, and not caused by entrainment effects at the top of the PBL.

[51] The accuracy of our estimate of geometrical depth depends on the correct determination of the PBL height. We use radiosonde profiles of potential temperature in order to estimate that uncertainty.

[52] Figure 3 shows the comparison of PBL height derived from lidar signals with the gradient method and from radiosonde data acquired at the station Oppin/Halle. For that comparison we use 110 lidar observations. We picked the timeframe from July 2001 to October 2004, because the radiosonde station became operational in 2001, and our laser failed in October 2004.

Figure 3.

(a) Comparison of the height of the planetary boundary layer determined from the potential-temperature profiles from radiosonde and derived with the gradient method from lidar profiles. The radiosonde launches were made at Oppin at 1200 UTC and 0000 UTC from 2001 to 2004. The linear regression line y = bx was fit to the data. We obtain b = 0.97 ± 0.01. The correlation coefficient R = 0.97. (b) Frequency distribution of the height of the planetary boundary layer determined from range-corrected lidar signals on the basis of the gradient method. One hundred ten lidar measurements in the timeframe from July 2001 until October 2004 were used in that analysis.

[53] The analyzed 110 cases are all cases with free-tropospheric layers for which nearby radiosonde observations are available. “Nearby” means a spatial distance of 30 km (for the Oppin/Halle radiosonde station), and a mean temporal distance of 167 ± 98 min between start of the lidar measurement and the radiosonde launch. In that regard the chosen cases do not represent special atmospheric conditions. We analyzed both nighttime and daytime measurements in the same way. However, we had more nighttime than daytime measurements at hand for our analysis. The ratio is approximately 4 to 3.

[54] The German Weather Service uses Vaisala RS80 and RS90 radiosondes at Oppin, Lindenberg, Meiningen, and Dresden. The Bergen station uses Graw-sondes. Sondes are launched every 6 hours at Lindenberg, and every 12 hours at the other four stations.

[55] Balloon rise rate and sampling rates are not reported in the public data set which we used. Usually the raw vertical resolution of the DWD sondes is very high. From those vertically highly resolved profiles “characteristic” points are extracted which are reported to the public. Characteristic points are points where changes in the gradients of temperature and/or pressure and/or humidity exceed certain threshold values. The procedure of extracting characteristic points follows the common rules of the World Meteorological Organization (WMO). In that respect we emphasize that the uncertainty of the radiosonde observations can be neglected concerning the atmospheric variability.

[56] We find a correlation coefficient R = 0.97 for the PBL heights determined with the two instruments. In the worst case we find a deviation of 600 m of PBL height determined from the lidar and radiosonde signals, respectively. In 64% of all cases the deviation is ≤120 m.

[57] For completeness Figure 3 also shows the frequency distribution of the PBL height of the 110 lidar observations. The frequency distribution peaks at 2–2.5 km. The distribution is mainly determined by the conditions in the three summer seasons (2002, 2003, 2004) in which we carried out many more lidar measurements than during the winter seasons (2001/2002, 2002/2003, and 2003/2004).

[58] One reason for the deviation of PBL height may be the distance between the lidar and radiosonde station, and the different measurement times of the two instruments. Another reason may be the height resolution with which the signals were acquired with the two instruments. In the case of the Raman lidar we use 60 m vertical resolution of the signals. Given the fact that a minimum of three data points is needed to identify a gradient in the raw signal we may easily obtain an uncertainty of about 100 m for the PBL height. We applied a height resolution of 60 m in our analysis of the temperature profiles from radiosonde. That resolution may also cause an uncertainty of 100 m for the height in which the gradient is identified. In the worst case we have to add the uncertainties from both methods, and thus we may end up with an overall uncertainty of 200 m.

[59] The radiosonde data from Oppin/Halle were available for the years 2001–2006, only, and the station was shut down in August 2006. The next nearest radiosonde stations are in Lindenberg (51.6°N, 12.1°E), which is approximately 150 km to the northeast of Leipzig, Meiningen (50.6°N, 10.4°E), which is approximately 170 km to the southwest of Leipzig, and Dresden (51.1°N, 13.8°E) which is approximately 100 km to the southeast of Leipzig.

[60] Despite that distance we also analyzed the radiosonde data from those stations. In summary we analyzed 498 temperature profiles acquired at those three stations from 1997–2006. In the most extreme case we find a deviation of 780 m between the PBL heights from lidar (gradient method) and from radiosonde (potential temperature profiles), respectively. The large distances between our Raman lidar and the radiosonde stations certainly is one cause for such large differences of PBL height. Despite such deviations we find a correlation coefficient of R = 0.94 on the basis of a linear regression of PBL height from lidar and radiosonde data.

[61] The radiosonde data were used to decide whether we picked the correct local minimum of our gradient in the lidar profiles, but we chose, of course, the height from lidar as the exact height. There certainly remains some uncertainty as we did not determine the distance of the location of the radiosonde from the lidar station at the time when the sonde was passing through the altitude regime subsequently identified as the top of the PBL. However in view of the rather high number of analyzed measurement cases we believe that this comparison on PBL height as determined from lidar and radiosonde shows that the gradient method basically provides us with reasonable values.

[62] Another source of uncertainty that affects the correct choice of the bottom height of free tropospheric particle layers are particles that may rest on top of the PBL. Figure 4 shows that in approximately 36% of the regular measurements there was no geometrical separation between the lofted layers and the particles in the planetary boundary layer. That number clearly points at the crucial problem of separating lofted particle layers from particles in the PBL.

Figure 4.

Percentage of measurements in which we observed a separation (in kilometers) between the lowest free-tropospheric layer and the planetary boundary layer particles. The grey-shaded columns are based on the 717 regular measurements carried out between 1997 and 2006. The white columns show the results of all 1028 measurements.

[63] We used radiosonde data to exclude that residual layers were falsely counted as free-tropospheric particle layers. If the profile of potential temperature indicated that the lofted layer was possibly part of the residual layer, we excluded that particle layer from our statistics. We took into account only particle layers above the residual layer.

[64] Approximately 48% of the lofted layers were separated by up to 1 km from the boundary layer aerosols. In about 15% of the cases the free-tropospheric layers were separated by more than 1 km from the PBL. Boundary-layer particles and particles in lofted layers were geometrically separated in 64% of all our measurements. We do not find significant differences, if we use all 1028 measurement cases in our analysis.

[65] Figure 5 shows the number of cases of free-tropospheric layers in each month of the year. Most events occur in late spring and throughout the summer months. The lowest number of events is observed during December.

Figure 5.

Frequency distribution of free-tropospheric particle events over Leipzig in the time period from 1997 to 2006. (a) The percentage of such events observed during the regular measurement times relative to the number of regular observations carried out per month. The statistics are based on all 717 regular lidar observations. The number on top of each column denotes the absolute number of events in that month, respectively. The number in brackets denotes the sum of all events of free-tropospheric pollution. (b) The frequency distribution in relation to all 1028 measurements we carried out from 1997 to 2006. (c) For comparison, the frequency distribution of particle pollution transport from North America, which is based on the regular measurements.

[66] Figure 5 shows the frequency distribution on the basis of the regular observations, as well as on the basis of all measurements. Trustworthy, unbiased statistics on the properties of free-tropospheric pollution events require measurements on a regular schedule. However, we see that there are not really significant differences between the statistics obtained from two data sets. One reason may be that these extra measurements were mainly carried out during the summer months in which we obtain a high percentage of free-tropospheric pollution anyway.

[67] Figure 5 also shows for comparison the frequency distribution of particle pollution transport from North America; see section 4. As will be shown, pollution from North America is the most important source of particle layers in the free troposphere over Leipzig.

[68] Geometrical features of the lofted particle plumes are shown in Figure 6. The bottom of the lowest free-tropospheric layer is at or below 2-km height in approximately 54% of all regular measurement cases. If we consider all 1028 observations, the number for bottom height is similar; that is, we find ≤2 km in 50% of all cases.

Figure 6.

Frequency distribution of (a) bottom height, (b) top height, (c) center height, and (d) geometrical depth of free-tropospheric layers in the timeframe from 1997 to 2006. Also shown is the cumulative distribution function (CDF) (black trace). The data are based on our 717 regular measurements.

[69] The top height of the free-tropospheric layers is at or above 2-km height in approximately 96% of all regular measurements. The same percentage is found, if we take account of all 1028 observations. In the most extreme case the top of the layer is around 12 km height. The geometrical depth of the aerosol layers is ≤2 km in approximately 48% of the 717 regular cases.

[70] In that analysis we took specific care that the top height was not determined by the effect that the backscatter signal became totally attenuated owing to water or ice clouds. As we operate a lidar with high radiation output power we are also not confronted with the problem that optically dense dust or smoke plumes cause total attenuation of the laser beam. However, there were cases where we observed a cloud on top of the particle layers. In these cases the cloud base was used as layer top, even though there might have been some aerosol particles also above the cloud, which were then masked by the attenuating cloud.

[71] The center height of the lofted layers follows in a straightforward manner from the frequency distributions of bottom and top height. We find that in 50% of all cases the center heights are between 2.5 and 4 km height.

[72] We also determined the percentage of cases in which aerosol layers were above 3, 4, and 5 km height. The probability that these layers are free-tropospheric layers is very high because the diurnal cycle of the PBL seldom reaches heights >3 km at the continental station at 51°N. We find that the top of the particle layers is above 3, 4, and 5 km in 69%, 46%, and 29% of all regular measurement cases, respectively.

[73] The peak of the frequency distribution of geometrical depth of the 717 regular measurements is between 1 and 1.5 km. The median value is at 2 km height. In the most extreme case we find a geometrical depth of nearly 10 km. If we use all measurements we find the same values for the peak of the frequency distribution and the median value.

[74] In this work we focus on particle layers in the troposphere. For that reason we report only on top heights of approximately 12 km of the lofted particle layers, i.e., particle layers below the tropopause.

[75] The tropopause is quite well defined and much less variable than the height of the PBL. We determined the tropopause for each individual measurement using the temperature and pressure profiles from the corresponding radiosonde observations. The tropopause height was determined according to the definition by WMO (http://www.wmo.ch/pages/prog/www/WMOCodes/Manual/WMO306_Vol-I-1-PartA.pdf). (1) The upper limit of the troposphere, by convention, the first tropopause, is defined as the lowest level at which the lapse rate decreases to 2 K km−1 or less, provided also the average lapse rate between this level and all higher levels within 2 km does not exceed 2°C km−1. (2) If, above the first tropopause, the average lapse rate between any level and all higher levels within 1 km exceeds 3°C km−1, then a second tropopause is defined by the same criterion as under item 1. This second tropopause may be either within or above the 1 km layer. In this paper we always use the “first tropopause.”

[76] We observed increased values of particle backscattering and extinction in the free troposphere over Leipzig in spring/summer 2003 [Mattis et al., 2003], and sometimes enhanced values were also detected in the lower stratosphere, which indicates that forest-fire smoke may have been injected into the lower stratosphere by “pyro-cumulonimbus” events [Fromm et al., 2000; Fromm and Servranckx, 2003; Immler et al., 2005]. We reanalyzed our data with respect to intrusion of particles into the stratosphere and we find almost a dozen cases in which there was smoke above the tropopause in spring/summer 2003. These results will be discussed in another publication.

4. Source Regions

[77] Figures 7 and 8present an overview on the main source regions from where lofted particle layers were transported to our site. Because of our crude approach of using backward trajectories for identifying the origin of the particle layers, we restrict our analysis to large, continental-scale source areas. In that way we come up with five areas, i.e., North America, polar areas north of 70°N, the Sahara, Europe, and Siberia.

Figure 7.

Frequency of free-tropospheric haze layers transported to Leipzig in the years 1997–2006. The statistics are based on all 717 regular measurements. Shown are the percentages of events (columns) relative to the total number of regular observations in the respective month. The results are presented according to the different source areas, i.e., (a) North America, (b) areas north of 70°N, (c) Siberia, (d) the Sahara, (e) Europe, and (f) the sector labeled as unknown. We include the results of areas north of 70°N also in the source region North America and Europe. An explanation for that is given in the main body of the text. The number on top of each column denotes the total number of events observed per month. With respect to the class Europe we also show results for the area east/southeast Europe (crosses, and numbers in italic). The sum of these events in each transport channel is given as first number in each bracket. The second number in each bracket denotes the number of events we observed on the basis of all 1028 observations.

Figure 8.

Seasonal occurrence of free-tropospheric layers over Leipzig. Shown is the percentage of regular events (grey columns) in relation to (a) the number of regular observations carried out in that respective season, i.e., 138 in winter (December, January, February), 183 in spring (March, April, May), 216 in summer (June, July, August), and 180 in fall (September, October, November), and split up according to source regions, i.e., (b) North America, (c) areas north of 70°N, (d) Sahara, (e) Europe, and (f) of unknown origin. Also shown is the seasonal frequency of events (at the regular observation times) in relation to the total number of free-tropospheric pollution events (at the regular observations times) of each transport class (squares). The total number of events of all seasons (columns in Figure 8a) is not exactly the sum of the respective numbers in Figures 8b–8f. An explanation is given in the main body of the text.

[78] In section 3, we discussed the frequency of the pollution events, disregarding the layering structures and the source regions of the different detected layers. In section 4, this layering is now taken into account. A pollution event may be characterized with one, two, or more pollution layers from different source regions. For that reason we attribute pollution events to different source regions, if backward trajectories indicate multiple source regions. Of course there were also cases with more than one particle layer which all originated from the same main source region. Those events were counted as one single layer in Figure 7.

[79] For example if we look at Figure 5 we observed in total 50 lofted layers during 48 measurements in the month of April. Out of these 48 measurements there were 34 regular observations, see Figure 5a, in which we observed 35 lofted layers. The remaining 14 observations belong to the so-called special events (see Figure 5b) in which we observed 15 additional lofted layers.

[80] On 7 April 1997, which belongs to the cases of regular observations, we observed two layers. One layer is classified as of unknown origin, whereas the other layer belongs to the class North America. On 30 April 2004 we observed two layers. One layer originated from the Sahara, and the second layer originated from North America.

[81] To put the above examples in more general words: on the one hand there were events with two particle layers of different origin, as for instance North America and the Sahara. On the other hand there can be particle layers with two different possible source regions, as for instance Arctic haze and North America, or for instance Arctic haze and Europe.

[82] Above all it has to be kept in mind that the number of the regular events of free-tropospheric pollution is less than the number of all events in which we observed particulate pollution in the free troposphere.

4.1. Pollution From North America

[83] Figure 7 shows the annual cycle of the frequency distribution of particle transport from North America to central Europe (Leipzig) on a monthly basis. North America is the most important source of pollution transport to central Europe. Forest-fire smoke [Wandinger et al., 2002; Mattis et al., 2003; Müller et al., 2005] and anthropogenic pollution [Müller et al., 2005] are the main contributors to particles in the free troposphere. Intercontinental transport peaks around June. At that time the forest-fire season reaches its maximum.

[84] Anthropogenic pollution layers typically have lidar ratios of 53 ± 10 sr at 355 nm and 39 ± 10 sr at 532 nm. The 355-nm lidar ratio is about 40% larger than the 532-nm lidar ratio; that is, the ratio of the lidar ratios is around 1.4. The Ångström exponent is around 1.7 [Müller et al., 2007a].

[85] In contrast to anthropogenic pollution, forest-fire smoke is characterized by an inverse ratio of the two lidar ratios. The mean 355-nm lidar ratio of 46 ± 13 sr is significantly smaller than the 532-nm value of 53 ± 11 sr. The Ångström exponent of forest-fire smoke typically is around 1 [Müller et al., 2007a].

[86] The source regions of the layers were determined by backward trajectory analysis. As mentioned before, such an analysis may be affected with large uncertainties. For instance we cannot rule out that some of the lofted layers were generated over Europe.

[87] Another difficulty of this simple trajectory analysis is that we cannot separate North American anthropogenic pollution from forest-fire smoke. Up to now, we analyzed several measurements with a combination of trajectory analysis, simulations with FLEXPART, and optical and microphysical particle characterization, and we showed that such a combined data analysis may provide sufficient information to separate the two aerosol types [Müller et al., 2005; Müller, 2007].

[88] Stohl [2001] presents a 1-year Lagrangian climatology of warm conveyor belts (WCB) [Wernli and Davies, 1997; Cooper et al., 2002] in the Northern Hemisphere troposphere and lowermost stratosphere. According to three-dimensional trajectory calculations with FLEXTRA the author finds that WCB inflow to western and central Europe occurs mainly in summer and fall. The author shows that Europe is within the main corridor of the WCB flow from North America, and that many air masses entering the WCBs traverse the high-emission regions [Benkovitz et al., 1996].

[89] The annual cycle as shown in Figure 7 is also in good agreement with a recent publication of Ramanathan et al. [2005]. Moderate Resolution Imaging Spectroradiometer (MODIS) and Aerosol Robotic Network (AERONET) observations of the particle optical depth at 500 nm were assimilated into a regional chemical transport model. The computations yield a maximum of the transport of aerosol pollution to Europe in April to June and a clear minimum in winter. According to the study, the seasonal-mean, anthropogenic aerosol optical depth over the North Atlantic reached values of 0.05–0.1 in April–June. In winter (December–March) the anthropogenic particle optical depth is less than 0.05 over the North Atlantic between North America and Europe. Note that the particle optical depth results from light-extinction contributions from the PBL and the free-tropospheric aerosol.

4.2. Haze From Polar Areas

[90] Observations show particles in the free troposphere north of 70°N [Shaw, 1984, 1995; Quinn et al., 2007]. This pollution, referred to as Arctic haze, originates from precursor material advected from industrialized areas of the Northern Hemisphere. This pollution is transported back to North America and the Euro-Asian continent. Only few vertically resolved observations of Arctic-haze-like material have been made outside polar areas so far [Müller et al., 2004; Müller, 2007].

[91] Müller et al. [2007a] discuss that lidar ratios are around 60 sr at 355 and 532 nm wavelength, respectively, which is well within the variability found for European anthropogenic pollution. One notable difference however are the comparably large Ångström exponents of 1.9. As has been shown by Müller et al. [2004] the mean size of such particles is smaller than what is usually found for free-tropospheric anthropogenic pollution [Müller et al., 2002, 2005].

[92] Figure 7 shows how often and at what times in the years 1997–2006 air masses were advected from latitudes north of 70°N. The maximum number of events occurs in springtime. Most of such transport events occur in April.

[93] One main difficulty that arises in using backward trajectory analysis again is the lack of separating the impact of different pollution sources on the air masses that move to our lidar site [Müller et al., 2004]. For that reason we include our results also in the source region North America, if trajectories indicate that the path of the investigated air parcels crossed the North American continent before arriving over Europe. In summary we identified 57% of such cases. In these cases we additionally considered that the height above ground of the air parcels had to be less than 3 km, when they crossed the North American continent. If backward trajectories indicate that air moved across Europe directly from polar areas we label such events as European pollution. We find 27% of such cases.

[94] Thus we have only 16% of cases for which we can assume that there was no mixing between Arctic haze and urban and/or industrial pollution over North America or Europe. Tracer modeling will be used in a reanalysis of our measurements for a more accurate assessment of these Arctic haze events.

4.3. Sahara Dust

[95] Most studies on Sahara dust transport over Europe and the Mediterranean area rely on satellite observations [Dulac et al., 1992; Moulin et al., 1997a, 1997b; Prospero et al., 2002; Kaufman et al., 2002; Moulin and Chiapello, 2004; Kaufman et al., 2005] and measurements with AERONET Sun photometers [Holben et al., 2001]. In recent years, several studies on the basis of lidar observations were published [Hamonou et al., 1999; Gobbi et al., 2000; Sarra et al., 2001; Gobbi et al., 2002; Dulac and Chazette, 2003; Barnaba and Gobbi, 2004; Tomasi et al., 2003; Papayannis et al., 2005; Amiridis et al., 2005; Mona et al., 2006; Müller et al., 2003]. Ansmann et al. [2003] describes for the first time an optically dense desert dust plume over Europe on the basis of EARLINET lidar observations which were carried out coherently with high vertical resolution on the continental scale. Just recently, Papayannis et al. [2008] analyzed more than 90 EARLINET observations of Sahara dust transport to Europe in the timeframe 2000–2002.

[96] Optical properties of Sahara dust depend not only on particle size and composition, but also on particle shape. Literature reports a large variety of dust optical properties. In the free troposphere over Leipzig we typically observe 532-nm lidar ratios around 60 sr, low Ångström exponents of 0.5 ± 0.5 and high linear particle depolarization ratios of 10%–25%.

[97] Figure 7 shows a summary of the total number of Sahara dust events observed over Leipzig during the years 1997–2006. The number of dust events is low compared to, for example, long-range transport of pollution from North America. Most events occur during springtime. As in the case of Arctic haze, some Sahara dust events also show transport of pollution from North America. For that reason we attributed such events also to the source region North America.

4.4. Pollution From Inside Europe

[98] Particles that originate from inside Europe present the second main source for free-tropospheric haze layers over Leipzig. Figure 7 does not show a clear seasonal dependence of such pollution transport.

[99] The 532-nm lidar ratio of anthropogenic pollution is largest close to the sources within the European PBL (53 ± 11 sr) [Mattis et al., 2004]. Lower values have been measured in the outflow plume from Europe toward the Atlantic (45 ± 9 sr). Typical Ångström exponents of 1.4 have been observed in the central European PBL as well as in lofted layers in SW Europe.

[100] We try to specify the possible source regions a bit more on the basis of backward trajectory analysis. In many cases of European pollution transport the origin of the particle layers rather likely was in east and southeast Europe. Such events often show a rather pronounced sandwich-like layering of pollution layers, which is also observed if air is transported from North America. We found pollution events, for which some particle layers may have been affected with Sahara dust. Thus, we attributed such pollution events not only to the class Europe, but also to the class Sahara dust.

[101] In summary we have to state, however, that backward trajectory analysis is insufficient for a more detailed analysis. Simulations with FLEXPART will be needed for a better source characterization.

4.5. Smoke From Siberia

[102] Smoke from forest fires in Siberia may be carried to central Europe, too. For example exceptionally intense forest fires occurred in Siberia in 2003, which gave rise to transport of smoke plumes around the globe [Damoah et al., 2004]. Twelve such events of transport of forest-fire smoke from Siberia could be identified so far. Ten of those events were observed during our regular measurement schedule [Müller et al., 2005]. However, because of the long transport times and distances backward trajectory analysis in general is insufficient to identify such cases with confidence. We have to postpone an analysis of our time series regarding such events. From a few case studies we found that optical properties of smoke from Siberia are similar to optical properties of smoke from North America [Müller et al., 2005].

4.6. Unknown Source Regions

[103] Figure 7 finally shows results of measurements labeled with “unknown.” In 22% of our regular measurements we could not attribute the haze layers to any of the source regions listed above.

[104] Our conclusions regarding source regions of the observed free-tropospheric particle layers mainly rest upon backward trajectory analysis. In that context we already point out here that we shall undertake another, much more sophisticated source identification, which is currently in progress. We shall analyze our data toward quantitative profiles of particle optical and microphysical properties. We shall add results from chemical trace modeling on the basis of FLEXPART simulations [Stohl, 2001; Stohl and Thomson, 1999; Stohl et al., 2003; Forster et al., 2001; Heintzenberg et al., 2003; Huntrieser and Schlager, 2004; Huntrieser et al., 2005; Müller et al., 2005].

[105] Particularly with respect to the source regions labeled unknown we expect further clarification regarding the origin of the observed particle layers. In some cases we could not attribute the haze layers to any of the main source region defined in this paper. In most of these cases the trajectories indicate an air mass transport from midlevel altitudes (2–8 km) above the Atlantic Ocean during the last 10 days before the lidar measurement. There is no known aerosol source in this geographical location and height level.

[106] It might be that the uncertainty of the vertical trajectory position is that large that the observed aerosol layers did not originated in midlevel altitudes, as indicated by the trajectories, but originated within the marine boundary layer. Further it might be that the vertical position of the trajectories is correct but the observed particles are older than 10 days and came from North America or Africa. We did not calculate trajectories more than 10 days back in time because the uncertainty of trajectories increases rapidly with transport distance and transport time. One may also speculate that the aerosol came from the tropopause region, from stratospheric intrusions, or from airplane emissions.

[107] From the available information we cannot decide which of the above mentioned options is the most appropriate one, and thus we label the origin of aerosol layers as unknown if their corresponding trajectories do not enter any of the aforementioned aerosol source areas.

4.7. Seasonal Variation

[108] Figure 8 summarizes how often we observed lofted layers above the Leipzig lidar site in each season of the year. Figure 8b shows the results disregarding the source region. Figures 8b8e summarize the results according to the main source areas of lofted particle layers. The numbers are based on our 717 regular measurements.

[109] Most transport events occur in spring and summer. The largest number of events are due to pollution transport from North America. The forest-fire season starts in the boreal areas of the Northern Hemisphere in spring, and reaches its maximum intensity in summer. As has been mentioned before, warm conveyor belts may be the main reason for lifting pollution into great heights over the east coast of North America from where it is carried to Europe.

[110] The squares in Figure 8a show the seasonal distribution of pollution transport from North America in relation to the total number of transport events (regular measurement schedule only) from North America. The maximum number of events occurs in summer, if we consider all measurements of the years 1997–2006. Lower optical depths and or lower particle number concentrations over North America, low probability of forest fires, low PBL top heights, and less convective activity and weaker lifting processes in winter, and stronger washout processes in winter and spring may be responsible for a significant reduction of particle layers in the free troposphere in the winter half year (from October to February; see Figure 7).

[111] Transport of pollution from or across areas north of 70°N mainly occurs in spring. If the frequency of such transport events is calculated in relation to the total number of such cases (1997–2006, only regular measurements) we find that 42% of all events occur in spring, see squares in Figure 8c.

[112] Dust outbreaks from the Sahara contribute comparably little. Approximately 12% of all cases are related to particle transport from North Africa. The squares in Figure 8a show that the maximum of events occurs in spring. For comparison, dust was observed on 61 out of 270 regular EARLINET observations at the lidar station in Naples in Italy from May 2000 until August 2003; that is, almost 23% of all measurements were affected by mineral dust. Forty percent of these events occurred in springtime, 25% of the dust cases were observed in summer, 20% in fall, and 15% in winter [Pisani, 2006].

4.8. Height Variation

[113] Figure 9 shows the frequency distributions of bottom and top and center height, and geometrical depth of the lofted layers in dependence of our transport channels. Again we omit the cases of pollution transport from Siberia.

Figure 9.

Frequency distribution of bottom and top and center height, and geometrical depth of free-tropospheric layers in the timeframe from 1997 to 2006 and for the different source classes defined in Figure 8, i.e., North America, areas north of 70°N, the Sahara, Europe, southeast Europe (crosses and thin lines) and the sector labeled as unknown. Also shown is the cumulative distribution function (CDF) for each transport channel (black trace).

[114] In all cases the maximum frequency of the bottom height of the lofted layers is between 1 and 2.5 km. The height range in which 50% of all cases are found varies between 1.2 (class Sahara) and 2.9 km (classes Europe, North America) for the different classes. The shape of the frequency distributions of the bottom heights is rather similar for all seven transport classes.

[115] With respect to the top height of the lofted layers we find a wide range among the transport classes. We note a rather strong limitation of top heights of the lofted layers around 4 km. The height range in which 50% of all top heights are found, varies between 3.5 and 6 km in the case of the classes North America, Arctic haze, and Sahara. For Europe, we obtain top heights between 3 and 5 km.

[116] The center height of the particle layers at which 50% of all regular cases are found is at 3.4 km (classes North America and areas north of 70°), 3.2 km (class Sahara), and 2.8 km (class Europe). We find that in 90% of all cases of lofted layers the center height is ≤5 km for the North America class, ≤5.5 km for the Sahara class, and ≤3.8 km for the European class. The geometrical depth of the lofted layers is around 2.5 km in 50% of all cases of each class, respectively. In 90% of all cases and each class the geometrical depth does not exceed 5 km height.

5. Conclusion

[117] We summarize 10 years of Raman lidar observations of pollution layers in the free troposphere over the central European lidar site at Leipzig, Germany. Our work for the first time provides statistical information on the occurrence of lofted particle layers (annual, seasonal, monthly) and geometrical properties. We analyzed 1028 lidar observations.

[118] We find that in 47% of all 717 regular observations lofted pollution layers are present. Most of these events occur during spring and summer of each year. We use backward trajectory analysis to track back the particle layers to their source regions. Because of this approximation method we can only decide on source regions on the continental scale. We identify North America, Siberia, polar regions north of 70°N, the Sahara, and Europe as important source regions for free-tropospheric aerosol layers over central Europe.

[119] Advection of pollution from North America occurs throughout the year. We find a rather clear maximum of events in summer time, which may be caused by the peak of the forest-fire season during that time of year. At times we observe forest-fire smoke from Siberia, which travels sometimes more than 2 weeks with the westerly winds before it arrives at our lidar site. Mixtures of Arctic haze and anthropogenic pollution are primarily advected during springtime of each year. Transport of Sahara dust mainly occurs during springtime and early summer. Pollution transport within Europe is also an important transport class. We find that free-tropospheric haze layers often originate in east and southeast Europe. A clear seasonal variation is not found.

[120] The lofted layers often rest on top of the planetary boundary layer, which makes it difficult to clearly separate particles in the planetary boundary layer from free-tropospheric pollution. The geometrical depth of the lofted layers varies between a few hundred of meters to several kilometers. The top of the particle layers sometimes is in the tropopause region or even extends to the lower stratosphere. In this first analysis of geometrical properties of the free-tropospheric pollution layers we did not take account of the fine structure or filament-like structure of the particle layers. Many pollution events are characterized by two or more particle layers. Investigations on the filament-like structures have to involve a discussion on mechanisms that may lead to such filament structures. We shall carry out such a study in future work. Such an analysis however requires algorithms for the automated and objective detection and separation of particle layers. Such work is in progress.

[121] One main difficulty we face in our analysis is the fact that simple backward trajectory analysis at times does not give a clear indication of the source region. Furthermore we find that several source regions may be responsible for a single free-tropospheric pollution event. Therefore we attributed several cases of pollution events to several source regions, respectively. An improved identification of source regions will be made in a future reanalysis of our data set on the basis of backward trajectory analysis and chemical tracer modeling with, for instance, FLEXPART.

[122] We shall analyze our 10-year data set with respect to optical and microphysical particle properties. Free-tropospheric particle layers may contribute up to 90% of optical depth [Müller et al., 2003]. The mean share of optical depths of all lofted layers for a given particle type (advection channel) may easily reach 25% [Müller, 2007]. Microphysical properties such as effective radius of free-tropospheric particles may differ considerably from the particle size in the planetary boundary layer [Müller, 2007].

[123] In the coming years satellite-borne lidar instruments like NASA's CALIOP (Cloud-Aerosol LIdar with Orthogonal Polarization) system aboard CALIPSO (Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations) [Winker et al., 2003, 2007] and the GLAS (Geoscience Laser Altimeter System) lidar aboard ICESat (Ice, Cloud and land Elevation Satellite) [Spinhirne et al., 2005] will significantly improve our knowledge on the four-dimensional distribution of aerosols. The potential of such systems in monitoring aerosol fields from space has been impressively shown in first publications [Hart et al., 2005; Hlavka et al., 2005; Hoff et al., 2005; Hu et al., 2007]. However, ground-based long-term monitoring with sophisticated multiwavelength Raman lidars, which should preferably be operated in networks like the European Aerosol Research Lidar Network will remain an indispensable tool for a comprehensive particle monitoring.

Acknowledgments

[124] The German lidar network was supported by the Federal Minister for Research and Education under grant UKF07AF108. The EARLINET activities at Leipzig were funded by the European Commission under grant EVR1-CT1999-40003 from 2000–2003. EARLINET is currently funded by the European Commission under grant RICA-025991 in the framework of EARLINET-ASOS. The trajectory model HYSPLIT is available at http://www.arl.noaa.gov/ready/hysplit4.html.

Ancillary

Advertisement