3.1 Polly^XT

In the framework of ALPACA, the lidar measurements were conducted with the portable multi-wavelength Raman and polarization lidar Polly^XT_IFT (Althausen et al., 2009), as part of the Polly lidar family (Engelmann et al., 2016) and will be referred to Polly^XT in the paper. Technically, the lidar is able to measure the backscattered light at 355, 532 and 1064 nm wavelengths, and Raman-scattered light at 387 and 607 nm to determine profiles of the particle backscatter coefficients at three wavelengths and the extinction coefficients at 355 and 532 nm. In fact, below 1500 m, the laser beam with the receiver field of view of the bistatic system is incomplete; thus, an overlap correction was applied. However, below 400 m, the overlap function is less than 0.5 and a reliable overlap correction is not possible. As a consequence, values of the particle backscatter coefficient were set constant below 400 m height in the lower part of the planetary boundary layer (PBL) under the assumption of well-mixed conditions.

The PBL top height is determined with the wavelet-covariance transformation that supposes a much higher aerosol load in the PBL than in the free troposphere (Brooks, 2003; Baars et al., 2008).

The rather low aerosol content in the area of Punta Arenas caused low signal-to-noise ratios. Thus, the particle extinction coefficient had to be estimated from the 532 and 1064 nm backscatter coefficients by means of appropriate particle lidar ratio (S_P) values.

3.2 Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP)

In April 2006, the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) mission started (Winker et al., 2009). Aboard CALIPSO, the two-wavelength backscatter and polarization lidar, CALIOP, has been operated to achieve a worldwide four-dimensional data set of clouds and aerosols. CALIPSO orbits the Earth at a height of nearly 705 km with a velocity of 7 km s⁻¹ and passes over the same location every 16th day.

The CALIPSO data processing provides profiles of backscatter and extinction coefficients at 532 and 1064 nm within the CALIOP level 2 version 4.10 data. In contrast to version 3 data, the version 4.10 data analysis algorithm aims at distinguishing not six but seven tropospheric aerosol subtypes (Omar et al., 2009, https://www-calipso.larc.nasa.gov/resources/calipso_users_guide/qs/cal_lid_l2_all_v4-10.php, last access: 8 May 2019). In this study, the CALIPSO data are used to determine planetary boundary layer heights and the backscatter-related Ångström exponent, and to retrace the long-range transport of smoke based on total attenuated backscatter profiles at 532 nm (level 1 V4.10) and the aerosol subtype product (level 2 V4.10). Level 3 products of the monthly averaged aerosol optical thickness (AOT) are still based on version 3.10 because version 4.10 is not yet available.

The CALIOP data processing provides several quality flags. Within the scene classification algorithm, the cloud–aerosol discrimination (CAD) score is determined (Liu et al., 2009). Aerosol particles are assigned negative values of −100 to −1 and clouds are assigned values of 1 to 100. The larger or smaller the value, the more confident the discrimination, respectively. Values around zero define an uncertain discrimination.

3.3 Auxiliary data

An Aerosol Robotic Network (AERONET) Sun photometer which measures AOT (column-integrated extinction coefficient) from 340 to 1020 nm at seven channels is located in Rio Gallegos (51.6^∘ S, 69.3^∘ W), Argentina (CEILAP-RG), which is 200 km away from Punta Arenas. Level 2.0 data are used with an AOT uncertainty of 0.01 to 0.02 (Holben et al., 2001).

Two models were used to support the analysis of the air mass transport. The Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) is a model to calculate trajectories of air parcels for simulations of dispersion and deposition at arbitrary locations (Draxler and Hess, 1998; Stein et al., 2015). By means of trajectories, aerosol sources are assigned. FLEXPART is a Lagrangian particle dispersion model, which calculates probabilistic trajectories of a large number of air parcels (Stohl et al., 2005). Thereby, the transport and diffusion of aerosol could be described and a coarse assignment of aerosols to their sources is enabled.

In the statistical analysis of the aerosol conditions during ALPACA, we used ensemble backward trajectories combined with a land cover classification for a temporally and vertically resolved air mass source attribution. The “software for automated trajectory analysis: trace” is used (Radenz and Seifert, 2019). The land cover is a simplified version of the MODIS land cover (Friedl et al., 2002). At first, a 27-member ensemble of 10 d backward trajectories is calculated using HYSPLIT. Meteorological input data for HYSPLIT are taken from the Global Data and Assimilation Service data set (GDAS1; https://www.ready.noaa.gov/gdas1.php, last access: 8 May 2019) provided by the Air Resources Laboratory (ARL) of the US National Weather Service's National Centers for Environmental Prediction (NCEP). Each ensemble is generated using a small spatial offset in the trajectory endpoint. Whenever a trajectory is below the PBL height provided in the GDAS1 data (“reception height”), the land cover is categorized using custom defined polygons according to land mass boundaries. Hence, an air parcel is assumed to be influenced by the surface if the trajectory is below the PBL height. The residence time for each category is then the total time an air parcel fulfilled this criterion by land cover category. This calculation is repeated in steps of 3 h in time and 500 m in height to provide a continuous estimate on the air mass source and as a first hint on potential aerosol load.

The global aerosol model NAAPS (Navy Aerosol Analysis and Prediction System; https://www.nrlmry.navy.mil/aerosol/, last access: 8 May 2019) and the model results of the Monitoring Atmospheric Composition and Climate (MACC; https://apps.ecmwf.int/datasets/data/macc-reanalysis/levtype=sfc/, last access: 8 May 2019) project are used to obtain the modelled aerosol optical thickness of dust.

4.1 2 March 2010: lofted smoke layers and Patagonian dust

On 2 March 2010, the weather in Punta Arenas was dominated by a northwesterly air flow. Between 09:00 and 12:00 UTC, Polly^XT observed clouds in the range from 4 to 5 km and a PBL height of about 1 km (Fig. 2). An aerosol layer was situated above the PBL up to a height of 2 km. An increased backscatter was also observed at the height range from 4.5 to 5.2 km (06:00 to 12:00 UTC) and from 11.5 to 12.1 km (06:00 to 09:30 UTC). The source apportionment of the three observed aerosol layers by means of FLEXPART and HYSPLIT analyses is shown in Figs. 3, 4 and 5. In Fig. 3a, a FLEXPART analysis for the lower layer (1 to 2 km) reveals that the air masses passed the southern Pacific and southern Patagonia (red colouration in Fig. 3a). HYSPLIT 48 h backward trajectories for heights of 0.5, 1 and 2 km confirm that the ground-level aerosol on 2 March 2010 was locally affected by an extended residence time above Patagonia (Fig. 4). At this time, modelled AOT of dust at 550 nm from the global aerosol model NAAPS regionally peaked at 0.14 and was 0.02 in the area of Punta Arenas (Fig. 4). The AOT of dust at 550 nm from the MACC model reaches values of 0.07 in southern Patagonia.

Figure 2Height–time display of the range-corrected signal observed at 1064 nm with the Polly^XT lidar on 2 March 2010. The analysed period between 06:15 and 09:00 UTC is framed in white. The planetary boundary layer top height is indicated by the black line.

Download

Figure 3FLEXPART simulations for the integrated residence time of the particles that travelled in the whole atmospheric column within the last 10 d until the observation time on 2 March 2010 for the observed heights between 1 and 2 km (a), 4.5 and 5.2 km (b) and 11.5 and 12.1 km (c). The colours represent the logarithm of the integrated residence time (in seconds) in a grid box for 10 d integration time.

Download

Figure 4HYSPLIT 48 h backward trajectories for heights of 0.5, 1 and 2 km for Punta Arenas on 2 March 2010, 09:00 UTC, and modelled dust surface concentrations from the NAAPS aerosol model on 2 March 2010, 06:00 UTC.

Download

The FLEXPART analyses for the layers observed between 4.5 and 5.2 and 11.5 and 12.1 km height reveal a clearly defined region of origin. The observed layers were advected to Punta Arenas from Australia and the southern Pacific (Fig. 3b and c). HYSPLIT 13 d backward trajectories were calculated as well (blue and light blue in Fig. 5). Figure 5a also shows active fire spots (red dots) derived by MODIS (Moderate Resolution Imaging Spectroradiometer; Justice et al., 2011), which were detected between 17 and 25 February 2010. On 18 and 19 February 2010, there were eruptions of pyrocumulonimbus (pyroCb) storms in western Australia. On 18 February, two fires, at approximately 32.6^∘ S, 121.0^∘ E and 33.0^∘ S, 122.1^∘ E, spawned pyroCb storms that were active between ∼04:00 and 10:00 UTC. The MODIS (Aqua) 11 µm window brightness temperature minimum in the opaque core of the pyroCb anvil was $\sim -\mathrm{57}$ ^∘C. According to the nearest radiosonde (Esperance, station number 94638; 33.8^∘ S, 121.9^∘ E; 00:00 UTC on 18 February) temperature profile, the brightness-temperature-inferred cloud-top height is ∼12.3 km. A third fire complex near 32.5^∘ S, 122.8^∘ E, produced pyrocumulus convection at that time. Hence, smoke was likely emitted at a range of altitudes during the various stages of pyroconvection from above the PBL up to ∼12 km. On 19 February, another pyroCb was detected that was likely generated from this third fire. The layer between 4.5 and 5.2 km (blue) as well as the layer between 11.5 and 12.1 km (light blue) crossed these regions with pyroconvection in western Australia. In summary, the source apportionment study for 2 March 2010 reveals that the near-surface aerosol layer is likely dominated by Patagonian dust, whereas the two observed lofted layers contain long-range-transported smoke from Australia. The Polly^XT measurement was analysed for the cloud-free period between 06:15 and 09:00 UTC on 2 March 2010. Figure 6 illustrates the profiles of the optical properties with a smoothing length of 150 m. The particle backscatter coefficients reach their maximum values of up to 1.02 Mm⁻¹ sr⁻¹ (532 nm) and 0.63 Mm⁻¹ sr⁻¹ (1064 nm) in the PBL. The observed layers in the height–time display (Fig. 2) are also slightly visible in the profiles of the particle backscatter coefficient in heights of 5 and 12 km. The profile of the particle extinction coefficient (at 532 nm) was reproduced for an assumed constant lidar ratio. According to the origin of the air masses, a lidar ratio of 40 ± 10 sr (Kanitz et al., 2013) was applied to the ground-level layer (up to 2.5 km) and a lidar ratio of 70 ± 10 sr (smoke; Ansmann et al., 2009; Tesche et al., 2011) was used for the lofted layers (Fig. 6c). The AOTs of each single layer result in values of 0.044 ± 0.004 (0 to 2.5 km), 0.004 ± 0.0004 (4.5 to 5.2 km) and 0.002 ± 0.0002 (11.5 to 12.1 km). The backscatter-related Ångström exponent (Fig. 6d) amounts to 0.56 ± 0.21 in the ground-level layer (up to 2 km). Kanitz et al. (2013) determined a Patagonian-dust-related Ångström exponent of 0.4 ± 0.1. In the framework of EARLINET, Saharan-dust-related Ångström exponents of 0.5 ± 0.5 were determined (Müller et al., 2007). In the smoke layer between 4 and 5.5 km, the Ångström exponent was 0.61 ± 0.1. In comparison, biomass-burning-smoke-related Ångström exponents of 1.0 ± 0.4 were determined by lidar observations for long-range-transported smoke from Siberia and Canada (Müller et al., 2007). In Cabo Verde, Ångström exponents of smoke originating from the south of western Africa were 1.06 ± 0.65 (Tesche et al., 2011). Smoke that was transported from Africa to the Amazon rainforest was found to have Ångström exponents of 0.8 (Baars, 2012).

Figure 5(a) Map of the HYSPLIT 13 d backward trajectories for heights of 4.5 to 5.2 km (blue) and 11.5 to 12.1 km (light blue) arriving in Punta Arenas (brown star) on 2 March 2010, 09:00 UTC: MODIS fire counts between 17 to 25 February 2010 (red dots) and CALIPSO tracks (green). (b) Height of the according trajectories.

Download

Figure 6Vertical profiles of (a) particle backscatter coefficients at 532 and 1064 nm, with error bars indicating 10 % uncertainty, (b) particle extinction coefficient at 532 nm, with error bars resulting from a lidar ratio uncertainty of ±10 sr, (c) particle lidar ratio, with error bars of ±10 sr and (d) backscatter-related Ångström exponent, with propagated error bars, derived by Polly^XT on 2 March 2010, 06:15 to 09:00 UTC. The planetary boundary layer top height is indicated by the dashed black line.

Download

In a next step, CALIPSO lidar observations were used for the characterization of the lofted aerosol layers during long-range transport. Six CALIPSO overpasses were found for the observation of the two lofted layers (see green lines in Fig. 5a), which provide intersections of the intercontinental transport of the smoke plume. On 21 February 2010, CALIPSO passed over the region of origin of the smoke in southern Australia. Figure 7a and b show the height–time display of the attenuated backscatter coefficient and the determined aerosol subtypes (with CAD score $<-\mathrm{80}$) on 21 February 2010. Between heights of 4 and 7 km, a section of increased backscatter can be identified. According to the CALIOP data algorithms, the lofted aerosol layer is a mixture of smoke (black) and continental aerosol (green), which confirms the analysis of their origin discussed above. The AOT of the layer determined by CALIOP amounts to 0.1 (at 532 nm). AERONET Sun photometer measurements show a mean AOT of 0.165 (at 500 nm) in Canberra, southern Australia, on 21 February 2010. In contrast, a biomass-burning-related AOT of 0.55 (at 532 nm) was determined in the Amazon rainforest during the dry season (Baars, 2012). Additionally, an aerosol layer between 10 and 12 km is identified as clean continental aerosol.

Figure 7Height–time display of the total attenuated backscatter at 532 nm (left panels) and corresponding aerosol subtype (right panels) derived by CALIOP from six CALIPSO overpasses on (a, b) 21, (c, d) 22, (e, f) 24, (g, h) 27 February, (i, j) 2 and (k, l) 3 March 2010. The aerosol subtypes are only illustrated if the CAD score is below −80 excluding uncertain classifications.

Download

On 22 February 2010, the smoke layer is also visible between heights of 4 and 8 km (Fig. 7c and d). On the following days, CALIOP was not able to determine unambiguously the smoke layer (Fig. 7e, f, g and h). A possible reason might be the decreasing smoke concentration along its transport route caused by dispersion and deposition (Bigg, 1973). The simultaneous occurrence of clouds and aerosol (for 24 February at 8 km height, see Fig. 7e, f; for 27 February at 5 to 11 km height, see Fig. 7g, h), which constrains the detection of the optically thin smoke plume, might be another reason. However, the CAD score of the detected smoke layers is smaller than −80, which indicates high discrimination accuracy (Liu et al., 2009). In the region of Punta Arenas (on 2 March 2010 at a height of 4 km), CALIOP is again not able to doubtlessly determine the smoke layer, because all determined aerosol layers are in the vicinity of clouds. On 3 March 2010, the smoke layer was confidently detected by CALIOP (Fig. 7l).

4.2 17 February 2010: Patagonian dust

This subsection discusses the observation of a low-level aerosol layer that was observed with Polly^XT on 17 February 2010 (see Fig. 8). A cyclone situated in the northwest of Punta Arenas caused northerly air flows on this and the previous days. The calculated 96 h HYSPLIT backward trajectories for 0.5, 1, 1.5 and 2 km reveal the South Pacific and southern Patagonia as the origin of the air masses (Fig. 9). The corresponding modelled NAAPS and MACC AOT (at 550 nm) exceeded 0.2 and 0.08 in southern Patagonia during this time, respectively (Fig. 9).

Figure 8Height–time display of the range-corrected signal observed at 1064 nm with the Polly^XT lidar on 17 February 2010. The analysed period between 02:00 and 06:00 UTC is framed in white. The planetary boundary layer top height is indicated by the black line.

Download

Figure 9HYSPLIT 96 h backward trajectories for heights of 0.5, 1, 1.5 and 2 km for Punta Arenas on 17 February 2010, 03:00 UTC, and modelled dust surface concentrations from the NAAPS aerosol model on 17 February 2010, 06:00 UTC.

Download

In the lidar measurement between 00:00 and 06:00 UTC on 17 February 2010, an enhanced aerosol load can be deduced by an increased backscatter coefficient up to 2.5 km height. The surface layer below 1.2 km height seems to contain two sublayers, one below and one above 0.6 km height, respectively.

Figure 10Vertical profiles of (a) particle backscatter coefficient at 532 and 1064 nm, with error bars indicating 10 % uncertainty, (b) particle extinction coefficient at 532 nm, with error bars resulting from a lidar ratio uncertainty of ±10 sr, (c) particle lidar ratio with error bars of ±10 sr and (d) backscatter-related Ångström exponent, with propagated error bars, derived by Polly^XT on 17 February 2010, 02:00 to 06:00 UTC. The planetary boundary layer top height is indicated by the dashed black line.

Download

Vertical profiles of the particle backscatter coefficient of this measurement were generated for the time period from 02:00 to 06:00 UTC along with a vertical smoothing length of 150 m (framed at the bottom of Fig. 8). It is illustrated up to 5 km because the atmosphere was free of clouds and aerosol above this height. Figure 10a shows the particle backscatter coefficients at 532 nm (green) and 1064 nm (red), as derived with the Raman method. Both reach their maximum values of 2.5 Mm⁻¹ sr⁻¹ (532 nm) and 1.8 Mm⁻¹ sr⁻¹ (1064 nm) in the PBL. In the profiles of the particle backscatter coefficient, an aerosol layer up to 3 km height is visible. At heights above 3 km, the particle backscatter approaches zero. The profile of the particle extinction coefficient (532 nm) was reproduced for a Patagonian-dust-related lidar ratio of 40 ± 10 sr (Kanitz et al., 2013) (Fig. 10b). The particle extinction coefficient reaches maximum values of 100 Mm⁻¹. The vertical integration of the particle extinction coefficients below 3 km results in an AOT of 0.09 ± 0.01. That AOT value is well between values which would be derived for lidar ratios of other species of dust. For instance, assuming Saudi Arabian dust (lidar ratio of 38 sr; Müller et al., 2007), Indian dust (43.8 sr; Schuster et al., 2012) or Saharan dust (53 ± 7; Preißler et al., 2011) would yield AOTs of 0.08, 0.1 and 0.12, respectively. Spaceborne measurements with MODIS between 16 and 18 February 2010 however indicate an increased AOT >0.2 over the southern Patagonian Desert. In the region of Punta Arenas, the AOT amounts to 0.07 up to 0.1. Whether the increased AOT close to Punta Arenas is caused by dust load could not be clearly identified. Comparable AOTs were also determined by MODIS in the southwest of Patagonia.

The mean Ångström exponent was 0.48 ± 0.06 (Fig. 10d), which is within the range of the values for Patagonian dust described in Sect. 4.1 and which is also confirmed by Müller et al. (2007) and Kanitz et al. (2013).

5.1 General aerosol conditions

The general aerosol conditions in Punta Arenas are presented in Fig. 11a in terms of monthly averaged AOT at 532 nm wavelength as determined with CALIOP for the grid cell of Punta Arenas (blue curve) from 1 January 2009 to 31 December 2010. The averaged AOT in Punta Arenas from 2009 to 2010 was 0.02 ± 0.01. The annual course of the monthly averaged AOT indicates the absence of a pronounced seasonal cycle. The large standard deviation may have been caused by the low number of observations but may also be the result of the CALIOP aerosol typing limitations in coastal regions (Kanitz et al., 2014). For comparison, the AOT as obtained at the AERONET station of Rio Gallegos is shown, too. In Rio Gallegos, the mean AOT between 2009 and 2010 (0.02 ± 0.02) is in the same range as that in Punta Arenas, although Rio Gallegos is situated closer to the Patagonian Desert and 400 km east of the west coast of Latin America. AERONET AOT measurements in Rio Gallegos confirm the very low mean AOT values (2009: 0.02 ± 0.01, 2010: 0.02 ± 0.01) of CALIOP and indicate clean marine conditions in Punta Arenas and Rio Gallegos (Fig. 11b). Such low AOT values were also found in other coastal and remote oceanic areas. During three meridional transatlantic cruises from 50^∘ N to 50^∘ S, shipborne lidar measurements revealed AOTs (at 532 nm) of the marine boundary layer of below 0.05 in 78 % of the cases (Kanitz et al., 2013). Wilson and Forgan (2002) determined a mean AOT (at 500 nm) of less than 0.04 in Cape Grim, Tasmania, from 1986 to 1999. AERONET Sun photometer measurements in marine areas show AOTs (at 500 nm) of 0.085 ± 0.01 over the Pacific and 0.06 ± 0.02 over the Southern Ocean (Smirnov et al., 2009).

Figure 11(a) Monthly average of CALIOP level 3 AOT (at 532 nm) and their standard deviation in Punta Arenas (blue) and Rio Gallegos (red) in 2009 and 2010. The ALPACA campaign is indicated by the shaded area. (b) Daily average of AERONET Sun photometer AOT (at 500 nm) in Rio Gallegos for the ALPACA campaign. (c) Height–time display of the range-corrected signal (at 1064 nm) measured by Polly^XT. Panel (d) shows the accumulated residence time of backward trajectories separated by different regions of origin.

Download

Figure 11c shows the height–time display of the range-corrected signal at 1064 nm wavelength measured by Polly^XT for the entire measurement period. As expected, most of the aerosol load is contained within the PBL up to around 1200 m (reddish colours). The free troposphere is characterized by a very low aerosol load but a frequent occurrence of clouds (grey and white colours).

A comprehensive analysis of aerosol source regions based on ensemble HYSPLIT trajectory calculations shows (Fig. 11d) that the influence of the ocean on air parcels (given in accumulated residence time of the backward trajectories within the PBL over the respective surface type) reaching Punta Arenas is a factor of 100 larger in contrast to the continents Africa, Australia and even South America.

5.2 Vertical aerosol distribution

After introducing the general aerosol conditions in Punta Arenas, in this section, the vertically resolved measurements with Polly^XT are investigated to determine the vertical aerosol distribution. First of all, the ALPACA lidar measurements are applied to ascertain the height of the PBL. Figure 12a presents the monthly means of the PBL heights. During the period with the strongest warming in the southern hemispheric summer (December and January), the maximum averaged top heights were reached (1230 ± 331 m and 1177 ± 365 m, respectively). The PBL heights decrease to 1106 ± 317 and 984 ± 347 m, respectively, in February and March, due to decreasing solar irradiation. However, the trend to low values is within the range of the standard deviation, which is indicated as error bars. Conducting stationary and continuous lidar measurements with Polly^XT, the determination of the temporal development of the PBL heights is possible (Cohn and Angevine, 2000; Baars et al., 2008). Figure 12b shows the averaged diurnal variation in the PBL height with a time resolution of 3 h. However, no explicit diurnal variation is identifiable within the limits of the error bars. The values vary between 850 ± 290 and 1280 ± 370 m. In contrast, for a northern hemispheric site with comparable latitude, PBL heights larger than 2 km were observed in 59 % of all considered cases in Leipzig (51^∘ N, 12^∘ E; Mattis et al., 2008). In Granada, Spain, which has a similar distance to the coast, measurements showed an annual mean of the PBL height around 1.7 ± 0.5 km (Granados-Muñoz et al., 2012).

Figure 12Height of the PBL determined by Polly^XT between 4 December 2009 and 31 March 2010. (a) Monthly averaged PBL top heights including the standard deviation as error bars and (b) averaged PBL top height as function of time of day (averaged over 3 h) including standard deviation as error bars. The red numbers indicate the number of measurements of the according bar. Local time is UTC−4:00.

Download

Figure 13a illustrates the frequency distribution of the obtained PBL heights as obtained from Polly^XT, CALIOP, radiosonde and GDAS1. For the ground-based lidar observations, in 74 % of all cases, the PBL extends up to heights between 750 and 1500 m, and the mean PBL top height amounts to 1151 ± 347 m. CALIPSO passed over Punta Arenas 31 times in the period from 1 May 2009 to 30 April 2010. The CALIPSO aerosol subtype data set is used for the classification of the PBL height. From these, the upper limit of the lowest aerosol layer, if present at all, was assumed to be the PBL height. In total, PBL heights could be estimated for a total of 21 overpasses. The mean PBL height identified with CALIOP is 1169 ± 532 m (Fig. 13b). Overall, 52 % of the PBL heights determined by CALIOP are between 750 and 1500 m. Furthermore, PBL heights were derived from radiosondes (each at 12:00 UTC) at the airport of Punta Arenas (37 m a.s.l., 15 km away) between 4 December 2009 and 31 March 2010 (Fig. 13). The mean PBL height is 1019 ± 376 m, and 64 % of all values are between 750 and 1500 m. In total, 73 % of all radiosonde ascents show a further lower layer with a mean height of 190 ± 13 m, which is indicated by a strong temperature gradient. The PBL thus seems to consist of two thermodynamically distinct layers. The reason for that may be explained by the terrain in the area of Punta Arenas which is directly located at the Strait of Magallanes. The southern Andes mountains and the Pacific are situated in the west (Fig. 1). The lower aerosol layer might be considered as the inflow of the marine boundary layer into the continental boundary layer. However, this layer is rarely detected by Polly^XT due to the effect of incomplete overlap below 400 m height. For comparison, only the upper layer is considered. The PBL heights provided by the GDAS1 data set (see Sect. 3.3) are calculated on the basis of the gradient Richardson number (Stull, 1988) and their mean is about 1024 ± 463 m (Fig. 13). Although the radiosonde data are assimilated into GDAS1 data set, the peak in low PBL heights derived from the radiosondes does not appear in the GDAS1 data. The horizontal resolution of GDAS1 data is $\mathrm{1}{}^{\circ }\times \mathrm{1}{}^{\circ }$ and the vertical one is approximately 250 m in the lowermost 1000 m. Locations between these grid points are interpolated, which might explain the observed differences in the PBL height distributions. In 6 % of all cases, the GDAS1 PBL heights solely fall below 250 m. Generally, the PBL heights determined by the decrease in aerosol backscatter using Polly^XT and CALIOP agree with those determined by the potential temperature profile using GDAS1 and radiosondes. That means that the top height of the aerosol layer coincides with the temperature inflection; hence, the aerosol accumulates within the PBL in the region of Punta Arenas.

Figure 13Frequency distribution of the PBL heights determined by (a) Polly^XT between 4 December 2009 and 31 March 2010, (b) CALIOP between 1 May 2009 and 30 April 2010, (c) radiosonde (12:00 UTC) and (d) GDAS1 data (12:00 UTC) from 4 December 2009 to 31 March 2010 with an increment of 250 m, respectively. N gives the number of samples. The horizontal error bar indicates the standard deviation.

Download

In the following, statistics of the vertical aerosol distribution are presented. During the ALPACA campaign, 59 measurement periods were analysed. The low amount of determined aerosol profiles is caused by the frequent occurrence of low and mid-level clouds (about 83 % of the measurement period; Kanitz et al., 2011) hampering an evaluation of the aerosol lidar data. Furthermore, the analysis is limited due to the frequent presence of marginal aerosol concentrations in the atmosphere and a corresponding low signal-to-noise ratio measured in the lidar signals. Figure 14 shows the analysed clear-sky profiles of the particle backscatter coefficient at 532 nm (Fig. 14a) and 1064 nm (Fig. 14b). The general vertical smoothing length was 330 m. Well-mixed homogeneous aerosol conditions are assumed to be present in the PBL, such that the particle backscatter coefficient is set constant in the overlap region below 400 m height. The profiles of the particle backscatter coefficient which were derived by either the Raman or Klett method (Fig. 14) indicate that the particle backscatter coefficient is close to 0 Mm⁻¹ sr⁻¹ above a height of 2 km. Hence, the majority of the aerosol is located in the PBL, as also indicated in the good correlation between radiosonde- and Polly^XT-derived PBL heights, discussed above. The particle backscatter coefficients reach mean values of 0.74 ± 0.56 Mm⁻¹ sr⁻¹ (532 nm) and 0.43 ± 0.32 Mm⁻¹ sr⁻¹ (1064 nm). However, aerosol layers were almost never observed in the free troposphere. Therefore, the free troposphere may be considered as a region representing pristine background aerosol conditions, making it an ideal region of low-aerosol reference for studies of aerosol–cloud interactions (Kanitz et al., 2011).

Figure 14Single (coloured lines) and averaged (black lines) height profiles of the particle backscatter coefficients at 532 nm (a) and at 1064 nm (b) as derived from Polly^XT observations. The error bars indicate the standard deviations.

Download

Lidar measurements as well as AERONET measurements provide spectrally resolved information. The spectral behaviour is expressed by means of the Ångström exponent. Figure 15 displays the frequency distribution of the vertical averaged backscatter-related Ångström exponent (at 532 and 1064 nm) for the time periods from 4 December 2009 to 4 April 2010 using Polly^XT (Fig. 15a) and 1 May 2009 and 30 April 2010 using CALIOP (Fig. 15b). For comparison, Fig. 15c illustrates the AOT-related Ångström exponent of the AERONET station in Rio Gallegos for the ALPACA period. In all three cases, the maximum values of the distribution are between 0 and 1. The mean values are 0.8 ± 0.3, 0.3 ± 0.7 and 0.5 ± 0.5 for the measurements with Polly^XT, CALIOP and AERONET, respectively. By means of these values, the size of the particles can be derived qualitatively. Very low or negative Ångström exponents indicate very large aerosol particles (sea salt or dust) (Moulin et al., 1997; Müller et al., 2007). In turn, large Ångström exponents (>2) point towards small aerosol particles (such as fresh smoke; Baars et al., 2012). The fraction of low Ångstöm exponents is largest for Rio Gallegos. This is caused by the occurrence of more dust events in the outflow of Patagonia. The values derived by CALIOP are too noisy to allow a reasonable interpretation. The Ångstöm exponents determined by Polly^XT are representative of continental aerosol as the lowermost heights of the atmosphere below 400 m height containing most likely a marine contribution could not be analysed due to instrumental limitations as discussed above.

Figure 15Frequency distribution of the vertical integrated particle-backscatter-related Ångström exponent (532 and 1064 nm) of (a) Polly^XT and (b) CALIOP with an increment of 0.2. Additionally, the frequency distribution of the AOT-related Ångström exponent determined by Sun photometer measurements in Rio Gallegos is illustrated. N gives the number of measurements. The horizontal error bar indicates the standard deviation.

Download