Interactive comment on “ Extreme levels of Canadian wildfire smoke in the stratosphere over central Europe – Part 2 : Lidar study of depolarization and lidar ratios at 355 , 532 , and 1064 nm and of microphysical properties

Main comment: This manuscript presents the depolarization and microphysical properties of an extraordinary event of Canadian wildfire smoke detected in the stratosphere over central Europe. Being the second part of two papers about this event, the main findings are: 1) the quite complete information provided (lidar ratios and depolarizations at several wavelengths) and 2) the strange high values of the depolarization in


Introduction
In a series of two papers we report on a rather strong Canadian wildfire smoke event.Optically dense aerosol layers crossed central Europe (Leipzig, Germany) on 22 August 2017.Biomass-burning-smoke was detected at almost all heights in the free The measurements are also of importance for the following reasons: (1) The spectrally resolved optical data sets for stratospheric smoke can be regarded as a new and important contribution to the aerosol-typing library used in lidar remote sensing (Omar et al., 2009;Burton et al., 2012;Groß et al., 2013;Illingworth et al., 2015;Baars et al., 2016Baars et al., , 2017)).( 2) The obtained smoke optical properties allow a clear distinction between stratospheric smoke and volcanic aerosols.(3) Our multiwavelength polarization/Raman lidar observations are complementary to the spaceborne lidar observations of the spread of the smoke over the northern hemisphere with CALIOP (Cloud-Aerosol Lidar with Orthogonal Polarization) (Khaykin et al., 2018).The smoke lidar ratios at 532 and 1064 nm can be used to convert the CALIOP smoke backscatter profiles into profiles of the climate-relevant and more interesting smoke extinction coefficients.(4) Spaceborne lidar activities from CALIOP (since 2006, with laser wavelengths of 532 and 1064 nm) to ATLID (Atmospheric Lidar, laser wavelength of 355 nm) of the EarthCARE (Earth Cloud Aerosol and Radiation Explorer) satellite mission (three year mission probably starting after 2020) (Illingworth et al., 2015) will span about 20 years of global aerosol observations.These aerosol observations performed at 355, 532, and 1064 nm need to be harmonized based on multiwavelength lidar observations of all main aerosol types and mixtures.
Our triple-wavelength polarization/Raman lidar of particle extinction, depolarization and lidar ratio can contribute in a rather valuable way to this harmonization of long-term spaceborne aerosol lidar data sets.

Lidars
In the evening and night of 22 August 2017, we run three lidars at the European Aerosol Research Lidar Network (EAR-LINET) station Leipzig (51.3 • N, 12.4 • W, 120 m above sea level).The single-wavelength 532 nm Polly (Portable lidar system) (Engelmann et al., 2016;Baars et al., 2016) measures the total, co-, and cross-polarized elastic backscatter signals at 532 nm, the rotational Raman signals around 532 nm, and the vibrational-rotational Raman signal at 607 nm.Co-and cross-polarized denotes here the plane of polarization with respect to the plane of the linearly polarized laser pulses.The 532 nm Polly allows us to determine height profiles of the particle backscatter coefficient, extinction coefficient, the corresponding extinction-tobackscatter ratio (lidar ratio) and the particle linear depolarization ratio at 532 nm.Specific details to the data analysis are given in Sect.2.2.
The second Leipzig lidar is the dual receiver field-of-view (RFOV) multiwavelength polarization/Raman lidar MARTHA (Schmidt et al., 2013;Jimenez et al., 2017).This lidar is unique because it measures Raman signals at 532 and 607 nm and polarizatiom-sensitive 532 nm backscatter signals at two RFOVs so that besides aerosol profiles, cloud microphysical properties can be retrieved from measured cloud multiple scattering effects.We used the 532 nm particle depolarization ratio measured with the smaller RFOV in the study presented here.Furthermore, the 355, 532, and 1064 nm particle backscatter coefficients, the 355 and 532 nm extinction coefficient profiles and the corresponding lidar ratio profiles are presented in the result section.
The third Leipzig lidar is the triple-wavelength polarization/Raman lidar BERTHA (Haarig et al., 2016(Haarig et al., , 2017)).BERTHA allows us to measure particle linear depolarization ratios and lidar ratios at all three important lidar wavelengths of 355, 532, and 1064 nm.In the present configuration, the 1064 nm depolarization ratio and the 1064 nm lidar ratio can only be measured alternatively (not simultaneously).The 1064 nm depolarization sensitive channel (cross-polarized channel) can be substituted by a 1058 nm rotational Raman channel within 20-30 minutes.This procedure includes adjustment and signal optimizing efforts.On 22 August 2017, we first measured the 1058 nm Raman signal profiles (for 2.5 hours) to obtain the 1064 nm extinction profile, and afterwards the cross-polarized 1064 nm signal component (for 40 minutes), needed in the retrieval of the 1064 nm depolarization ratio.
The laser beams of Polly and BERTHA were tilted to an off-zenith angle of 5 • in different directions, whereas MARTHA was pointing to the zenith, which leads to horizontal distances between the laser beams of the order of 450-750 m at the base of the tropospheric smoke layer at 5 km height and of 1.3-2 km at the base of the stratospheric layer at 15 km height.However, the good agreement of the results as discussed in Sect. 3 indicated that the smoke layers were obviously horizontally homogeneous on scales of 1-2 km.

Lidar data analysis: optical properties
Details of the determination of the particle optical properties and the uncertainties in the products can be found in the articles mentioned above.An overview of the retrieval methods is given in Ansmann and Müller (2005); Freudenthaler et al. (2009), andFreudenthaler (2016).The Raman lidar method was exclusively used to determine particle backscatter and extinction profiles.The particle backscatter coefficient is obtained from the measured ratio of the elastic backscatter signal to the respective Raman signal.The extinction coefficients are calculated from the Raman signal profile.In the correction of Rayleigh extinction and backscattering effects, temperature and pressure profiles from the GDAS (Global Data Assimilation System) data base are used (GDAS, 2017).The determination of the particle linear depolarization ratio from the volume depolarization ratio (discussed and shown in part 1) is described in detail by Freudenthaler et al. (2009) and Haarig et al. (2017).
In Sect.3, the lidar results for the time period from 20: 45 -23:15 UTC on 22 August 2017 (shown in Fig. 1 in Sect. 3) are presented and discussed.The lidar signals for the selected time period of 150 minutes were averaged, background-, and overlap-corrected before the optical properties (backscatter and extinction coefficients, depolarization ratios) were computed.This procedure was performed separately and independently for all three lidar data sets.In case of the 1064 nm depolarization ratio observations with BERTHA, we averaged the signals from 23:50 -00:30 UTC.The signal profiles had to be smoothed afterwards to reduce the impact of signal noise to a tolerable level on the final products.In the case of the backscatter coefficients and the particle depolarization ratio (determined from the profile of the ratio of the cross-polarized to co-polarized elastic backscatter signal component), we smoothed the individual signal profiles with vertical gliding averaging window lengths of 50-100 m (backscatter coefficients, troposphere), 100-250 m (backscatter coefficient, stratosphere), and 200-400 m (depolarization ratio, troposphere and stratosphere).
In the retrieval of the extinction coefficient, a least-squares linear regression method was applied to the respective Raman signal profiles.The regression window length was 750 m (532 nm) to 1200 m (355 nm) in the troposphere and 1200 m for both wavelengths in the stratosphere.To obtain the lidar ratios at 355 and 532 nm, the extinction profiles were combined with the respective backscatter profiles.In this procedure, we applied the optimum-effective-resolution concept (Iarlori et al., 2015;Mattis et al., 2016) and used a smoothing window length in the backscatter retrieval which was 0.75 of the regression window length in the extinction retrieval.
In the case of the 1064 nm extinction coefficient, coherent extinction profile structures could not be obtained because the 1058 nm Raman signals were too weak and noisy.The retrieval window lengths are indicated by vertical bars in the figures in the result section (Sect.3).Retrieval window lengths of 750-1500 m in the troposphere and 2500 m in the stratosphere had to be applied to obtain the 1064 nm extinction coefficient values with a reasonably low uncertainty.The retrieval of the 1064 nm lidar ratio by means of these smoothed values is explained in Sect.3.

Lidar data analysis: microphysical properties
An overview of the theoretical background of the lidar inversion method applied to obtain microphysical particle properties such as the particle effective radius, volume and surface area concentrations and refractive index characteristics from the measured optical properties, i.e., from particle backscatter coefficients at 355, 532, and 1064 nm and extinction coefficients at 355 and 532 nm, is given in Ansmann and Müller (2005).In the present smoke data analysis, we use the method developed by Veselovskii et al. (2002Veselovskii et al. ( , 2010)).The data analysis assumes spherical smoke particles in the tropospheric layer.In the retrieval of the microphysical properties of stratospheric smoke, spherical as well as spheroidal particles are assumed.The single scattering albedo (SSA) of the smoke particles, presented in the result section (Sect.3) as well, is computed from the retrieved particle size distribution and the most appropriate refractive index characteristics (real and imaginary parts) used as input in the lidar inversion procedure.

Overview
An introduction to the record-breaking Canadian wildfire smoke event on 22 August was given in part 1 (Ansmann et al., 2018).
It was shown that the total (tropospheric+stratopsheric) smoke-related AOT at 532 nm reached values close to 1.0 during the noon hours.Smoke was present at all heights in the free troposphere as well as in the lower stratosphere up to 16 km height.
An optically dense stratospheric layer extended from 14-16 km height and showed a 532 nm AOT of 0.6.
Figure 1 shows the aerosol layering over Leipzig in the night of 22 August 2017, about 10 hours after the maximum smoke burden occurred over Leipzig.Tropospheric aerosol layers were present from the surface to about 6.5-7 km height in the night.

Smoke profiling with three lidars
In Figs. 2 and 3, the results of the observations with our three polarization/Raman lidars are presented.Mean height profiles of the optical properties for the time period from 20:45-23:15 UTC on 22 August 2017 are shown, except for the 1064 nm depolarization ratio (23:50-0:30 UTC, see Fig. 1, and explanations in Sect.2.2). Figure 2 shows the smoke optical properties in the tropospheric layer.According to the backward trajectory analysis presented in part 1, the wildfire smoke traveled about 10 days from the fire sources in western Canada to central Europe.Figure 3 contains the respective findings for the stratospheric smoke layer.This aerosol was probably directly lifted into the stratosphere within deep cumulus towers.
As can be seen in Figs. 2 and 3, a good agreement between the observations with BERTHA, MARTHA, and Polly is given for all parameters.Highlight of our lidar measurements are the lidar ratio and depolarization profiles in Figs.2c, 2e, 3c, and 3e.However, a high impact of signal noise on the retrieved profiles is visible as well.The MARTHA 355 nm extinction profile could be measured up to about 15.3 km only.The high signal noise is due to the fact that we avoided overloading of the photomultipliers (operated in the photon counting mode) so that even the strong near-range signals in the lowest part of the troposphere were properly measured.As a consequence, the signals were comparably weak in the middle and upper troposphere and lower stratosphere and therefore the influence of signal noise likewise high.This measurement strategy was selected to obtain reliable backscatter and extinction profiles almost from the ground to the top of the stratospheric smoke layer so that the full extinction profiles (as well as the integral) is available for comparison with AERONET sun photometer observations.
In the case of the 1064 nm extinction coefficient, we only can show a few values in Figs. 2 and 3.The retrieval window lengths are indicated by vertical bars.The shown 1064 nm extinction coefficients (and the respective lidar ratio) can be interpreted as mean values for these vertical regression-fit intervals.In the case of stratospheric smoke, a regression window length of 2500 m was required to obtain the 1064 nm extinction coefficient as discussed above.However, the vertical extent of the layer was 1250 m only.Because the densest part of the smoke layer was between 15 and 16 km (as indicated by the backscatter coefficient profiles in Fig. 3), we multiplied the 1064 nm extinction coefficient derived for the 2500 m layer (from 14.4-16.9km) by a factor of 2 to obtain a trustworthy estimate for the main layer from 15-16 km height.This multiplication yields the correct value if extinction by smoke outside the 15-16 km height range is zero.This assumption is reasonably valid as all backscatter coefficient profiles indicate.To obtain finally the 1064 nm lidar ratio, we combined the extinction value for the 1250 m layer with the respective backscatter coefficient computed from signal profiles smoothed with a window length of 937.5m vertical window length according to the effective resolution concept (Iarlori et al., 2015;Mattis et al., 2016).

Main findings
The main results can be summarized as follows: The backscatter and extinction profiles and respective Ångström exponents show typical smoke features.For aged Canadian fire smoke a clear and strong backscatter wavelength dependence is usually observed, whereas the wavelength dependence is weak, absent, or even negative in the case of the extinction coefficient in the 355-532 nm spectral range.Consequently, the 355 nm smoke lidar ratio is smaller than the 532 nm smoke lidar ratio.The reason for the different backscatter and extinction wavelength dependencies is probably related to the fact that the particle backscatter coefficient is a complex function of particle shape, size distribution, and composition (aerosol mixture) whereas extinction of light is mainly a function of size distribution and chemical composition (and corresponding absorption and scattering features), and depends only weakly on particle shape properties.The lidar data inversion analysis, discussed in detail below, revealed that a high imaginary part of 0.04 of the refractive index was needed to reproduce the strong backscatter and low extinction wavelength dependence.This means that high absorption by the smoke particles was most probably responsible for the observed different extinction and backscatter-related Ångström exponents.Our findings are in good agreement with laboratory and field studies of smoke optical properties (Renard et al., 2001(Renard et al., , 2002(Renard et al., , 2005;;Zhang et al., 2008;Lewis et al., 2009;Adachi et al., 2010;China et al., 2015).
The most surprising finding is the strong difference between the depolarization spectrum in the tropospheric and stratospheric smoke layers as shown in Figs.2e and 3e.The depolarization ratios were at all below 3% for tropospheric smoke, a clear indication that the particles were spherical and/or small.In strong contrast, high depolarization ratios of 22% and 18% were observed at 355 and 532 nm, respectively, in the stratosphere.The depolarization ratios was again low (4%) at 1064 nm.Strong depolarization of the transmitted linearly polarized laser radiation points to irregularly shaped particles.
Table 1 provides an overview of the found optical properties of the tropospheric and stratospheric smoke layers.The presented values can be interpreted as layer-mean values.The microphysical properties in Table 2 are obtained by applying the lidar inversion method described in Sect.2.3 to the extinction coefficients at 355 and 532 nm in Table 1 and the corresponding backscatter coefficients at 355, 532, and 1064 nm computed from the extinction coefficients, lidar ratios, and respective Ångström exponents in Table 1.The particle mass concentrations were computed from the volume concentrations assuming a smoke particle density of 1.35 g cm −3 (Reid and Hobbs, 1998).Mass concentrations were 5-6 µg m −3 in the tropospheric layer and much larger with values close to 40 µg m −3 in the stratospheric layer at the nighttime hours.A clear indication for the presence of highly absorbing stratospheric particles is the low SSA of 0.80-0.85 at 532 nm and 1064 nm.
Figure 4 shows the lidar-derived particle mass size distributions which belongs to the results in Table 2 and is also a product of the lidar data inversion analysis.The size distribution for the particle mass concentration is obtained by multiplying the derived volume size distribution with the smoke particle density of 1.35 g cm −3 .The respective particle mass size distribution derived from the AERONET observation at Lindenberg in the morning of 23 August 2017 is shown for comparison (Holben et al., 1998;AERONET, 2018).The AERONET observation describes the aerosol in the entire vertical column from the surface to the top of the stratopsheric layer.To convert the AERONET column values to stratospheric volume and mass concentrations so that we can compare sun-photometer-derived and lidar-derived stratospheric volume and mass concentrations, we assumed The lidar-derived and AERONET-derived mass size distributions in Fig. 4 provide a consistent picture of the smoke-related tropospheric and stratospheric size distributions.The pronounced accumulation mode in the AERONET column observation is clearly caused by stratospheric smoke particles.The coarse-mode of the AERONET size distribution is most probably the result of light extinction by boundary-layer aerosol particles (surface soil dust, road dust).By comparing the tropospheric and stratospheric size distributions we see that the particles were comparably small in the tropospheric layer.The size distribution in the stratosphere in Fig. 4 is in good agreement with airborne in situ smoke observations (Fiebig et al., 2002;Petzold et al., 2007;Dahlkötter et al., 2014).
The lidar inversion results in Table 2 and the size distribution in Fig. 4 do not change much when assuming spheroidal instead of spherical particles in the lidar inversion procedure.We hypothesize that the reason for the low impact of particle shape on the retrieval products is the absence of a particle coarse mode in the stratospheric smoke layer so that the particles were at all likewise small.At these conditions shape aspects have a low impact on the lidar inversion products.A very different spectral behavior was found in the case of the particle linear depolarization ratio in the tropospheric and the stratospheric layer (Fig. 5b), whereas the lidar ratios showed quite similar values and a similar wavelength dependence in both layers (Fig. 5a).As mentioned, the particle depolarization ratio was low at all three wavelengths in the tropospheric layer.These low depolarization values are indicative for spherical or almost spherical particles.The particles must have been compact in shape and many of them may have been composed of a solid soot core with liquid sulfate shell (Dahlkötter et al., 2014).The influence of the shape variability from spherical to rather irregular shaped smoke particles on the optical and radiative properties are discussed by Zhang et al. (2008); Adachi et al. (2010); China et al. (2015).The slightly enhanced tropospheric depolarization values of 2-3% at 355 and 532 nm may have been partly caused by traces of soil dust injected into the atmosphere by the hot fires and associated strong and turbulent winds at the ground (Nisantzi et al., 2014).

Spectral smoke depolarization and lidar ratios
In contrast to the tropospheric depolarization ratio spectrum, high depolarization ratios were observed at 355 nm and 532 nm and a strong wavelength dependence were found in the stratospheric layer.The particles were clearly non-spherical.The strong differences between the tropospheric and stratospheric smoke were confirmed by the triple-wavelength polarization lidar observation at Lille, northern France, performed during the same smoke period in August 2017 (Hu et al., 2018).They measured values around 23% (355 nm), 20% (532 nm), and 5% (1064 nm) on 24 August 2017.Furthermore, Burton et al. (2015) observed a rather similar wavelength dependence in a well-defined layer of wildfire smoke advected from the Pacific Northwest of the United States to the Boulder-Denver region at 8 km height.They found depolarization ratios of 21% (355 nm), 9% (532 nm), and 1% (1064 nm) in 8 km height.In contrast to the strongly different behavior of the depolarization ratio, the lidar ratio spectrum was found to be rather similar in the tropospheric and stratospheric smoke layers.The same origin of the aerosol and thus similar aerosol composition resulting in similar basic scattering and absorption properties may be the reason for the less variable lidar ratios in the tropospheric and stratospheric layers.However, it remains an open question why the difference in the shape characteristics does not have a noticable influence on the lidar ratio spectrum.
The observed mono-modal smoke particle size distributions, i.e., the absence of a coarse mode, is probably the key to understand the found similarities and differences in the tropospheric and stratospheric lidar ratio and depolarization ratio spectra.We hypothesize that the smoke particles were too small to have an impact on the extinction-to-backscatter ratio via different shape properties.A significant impact of the particle shape on the lidar ratio is for example given in the case of desert dust when coarse-mode particles (particles with diameter >1 µm) control the optical effects.Spherical coarse dust particles would cause a lidar ratio around 20 sr, the irregular shape of the particles however leads to lidar ratios of about 50 sr because of a strongly reduced backscatter efficiency of the irregularly shaped dust particles (Mattis et al., 2002).
On the other hand, the aged accumulation-mode smoke was at least able to significantly depolarize laser light by about 20%.This aspect was already discussed in part 1 (Ansmann et al., 2018).The found smoke depolarization-ratio wavelength spectrum is very similar to the one for fine-mode mineral dust.According to Mamouri and Ansmann (2017), which based their discussions on laboratory studies of Järvinen et al. (2016), fine-mode dust causes depolarization ratios of 20-22% (355 nm), 14-17% (532 nm) and <10% (1064 nm).The spectral behavior as shown in Fig. 5b is no longer visible when coarse-mode dust particles dominate the depolarization features (Haarig et al., 2017).
The irregular shape of the smoke particles in the stratosphere is partly the result of the very low RH values (of 1% according to the Lindenberg radiosonde profiles between 13 and 17 km height) so that the smoke particles dried out, lost most of the liquid aerosol components, and thus the potential to form a compact particle with spherical shape.Pictures of black carbon aerosol particles taken at 18-21 km height corroborate the irregular shape of stratospheric soot particles (Strawa et al., 1999).
At the end it should be mentioned that also in the case of the stratospheric smoke we can not exclude that soil dust reached stratospheric heights together with the smoke and thus were partly responsible for the enhanced depolarization ratios (Nisantzi et al., 2014).However, the single scattering albedo values derived from the lidar observation point to values around 0.8 at 532 nm (see Table 2) which clearly indicates the dominance of soot particles.

Smoke depolarization and lidar ratios: an updated literature review
Numerous articles on biomass-burning smoke are available in the literature.Multiwavelength lidar studies of fire smoke are presented by, e.g., Wandinger et al. (2002); Murayama et al. (2004); Mattis et al. (2003); Müller et al. (2005) Veselovskii et al. (2015Veselovskii et al. ( , 2017)); Giannakaki et al. (2015); Ortiz-Amezcua et al. (2017), andHu et al. (2018).Our literature review is summarized in Table 3 and considers multiwavelength lidar observations only.It was already noticed more than 10 years ago that the 355 nm lidar ratio for aged smoke after many days of long-range transport is considerably lower than the 532 nm lidar ratio (Müller et al., 2005).This is consistent with the discussion and the findings presented above.As can be seen  3, the difference between the 355 and 532 nm lidar ratios can be as large as 15-25 sr.For fresh smoke, advected from fire sources to the lidar stations within less than 2-3 days, the lidar ratios at 355 and 532 nm are similar or the 355 nm values are larger.
Only a few observations of the particle depolarization ratio in aged and fresh smoke are available as can be seen in Table 3.
The low values are indicative for small and spherical smoke particles and moderately increased depolarization ratio may be related to the presence of some soil dust in the smoke plumes or that the smoke particles were partly irregularly in shape.The high depolarization ratios at 355 and 532 nm and the strong wavelength dependence of the depolarization ratio as observed in the August 2017 stratospheric smoke layers are a relatively new features and were first observed in an elevated aged smoke layer in the upper troposphere by Burton et al. (2015).

Conclusions
An extreme fire smoke event with aerosol layers in the troposphere and stratosphere permitted us to characterize Canadian wildfire smoke after long-range transport in large detail.We used this unprecedented event also to demonstrate the unique potential of our triple-wavelength polarization/Raman lidar to contribute to atmospheric aerosol research.As described in part 1 (Ansmann et al., 2018), the stratospheric particle extinction coefficients reached 500 Mm −1 and were thus a factor of 20 higher than the maximum light-extinction coefficients measured over Germany after the major Pinatubo eruption in 1991.The smoke-related AOT was 0.9 at 532 nm on 22 August 2017 as observed with lidar and corroborated by AERONET and MODIS observations.The peak mass concentration of the smoke particles were estimated to be 70-100 µg m −3 in the stratosphere.
In part 2, we focused on the optical and microphysical properties of the fire smoke in two pronounced layers in the troposphere and stratosphere.Three lidars were involved in the studies.Particle backscatter and extinction coefficients, respective lidar ratios, and linear depolarization ratios at all three lidar wavelengths were measured and allowed us to derive microphysical, morphological, and composition-related information about the smoke layers.Very different smoke properties were observed in the tropospheric and stratospheric smoke layers.Very low depolarization ratios (<3% at all three wavelengths) were found in the troposphere (reflecting the predominance of spherical smoke particles), whereas the particles in the stratosphere lead to high depolarization ratios of 22% at 355 nm and 18% at 532 nm and a comparably low value of 4% at 1064 nm in the stratosphere.It was concluded that the strong wavelength dependence is attributed to the narrow size distribution (accumulation mode) of irregularly shaped soot particles and the absence of a smoke particle coarse mode.The layer mean particle lidar ratios, on the other hand, were 40-45 sr (355 nm), and 65-80 sr (532 nm), and 80-95 sr (1064 nm) in both layers which may be an indication for a similar chemical composition of the smoke in the troposphere and stratosphere and thus similar scattering and absorption properties.The single scattering albedo was estimated to be 0.8 at 532 nm in the stratosphere, a clear indication for the presence of soot.The smoke particles were rather small (effective radius of 0.17 µm) in the tropospheric layer and much larger (effective radius of 0.32 µm) in the stratosphere.
The spectrally resolved optical data sets for stratospheric smoke can be regarded as an important new contribution to the aerosol-typing library used in lidar remote sensing (Omar et al., 2009;Burton et al., 2012;Groß et al., 2013;Illingworth et 2015;Baars et al., 2016Baars et al., , 2017)).The presented optical properties of stratospheric smoke enable a clear and unambiguous discrimination of biomass burning smoke and volcanic aerosol in the stratosphere, and thus to identify and separate these major contributors to stratospheric aerosol perturbations.
The presented triple-wavelength lidar observations are of great value for spaceborne lidar data analysis (harmonization of long-term observations, aerosol trend analysis).NASA's CALIOP (in space since 2006) is operated at 532 and 1064 nm, whereas ATLID of the EarthCARE mission will measure aerosol optical properties at 355 nm.The data harmonization efforts require 1064-nm-to-532-nm, 1064-nm-to-355-nm, 532-nm-to-355-nm conversion factors for backscatter, extinction, and particle depolarization ratio for all important aerosol types and frequently occurring aerosol mixtures.Lidar ratios for 355, 532, and 1064 nm are needed as well.Our smoke lidar-ratio observations at 532 and 1064 nm can already be used in the present CALIOP data analysis of the 2017 smoke event (Khaykin et al., 2018) to estimate smoke extinction coefficients from the measured backscatter profiles.
The unique Canadian wildfire season 2017 can be regarded as a rather favorable opportunity to test atmospheric transport models which consider biomass burning smoke.Especially transport, removal, upward motion of soot particles, and the impact of stratospheric smoke on ice cloud formation, radiative fluxes and chemical processes must be properly modeled to permit state-of-the-art future-climate-change studies.Never before, so many smoke observations with ground-based lidars, organized in several well-organized networks (including EARLINET) were available for intensive model-observation comparisons.Together with the spaceborne CALIOP observations the spread of the smoke over the northern hemisphere is well documented for the year 2017.
As an outlook, in the next step of our studies of the 2017 stratospheric smoke layers over Europe, all PollyNET and EAR-LINET lidar observations since the summer of 2017 will be analyzed.Smoke layers were observed over whole Europe for several months.Aging of the smoke particles and associated changes in the optical and microphysical properties will be investigated (Baars et al., 2018b).We will also carefully analyze the lidar observations regarding the noticed apparent ascent of the smoke layers from heights of 15-17 km to heights of 23-27 km (Baars et al., 2018a) which is most probably caused by gravito-photophoresis forces (Renard et al., 2008).

Data availability
The lidar data are available at TROPOS upon request (info@tropos.de).AERONET sun photometer data are downloaded from the AERONET web page (AERONET, 2018).Table 1.Optical properties of smoke aerosol in the tropospheric layer (5-6.5 km height) and stratospheric smoke layer (15-16 km height).
Layer mean values of the particle extinction coefficient σ, lidar ratio S, linear depolarization ratio δ, and backscatter-related and extinctionrelated Ångström exponent a σ,λ 1 /λ 2 and a β,λ 1 /λ 2 for the wavelength range from λ1 to λ2 are given.The mean values are based on all lidar observations taken in the night of 22 August 2017.

Parameter
Troposphere  Table 3. Literature overview of multiwavelength lidar observations of smoke lidar ratios and particle linear depolarization ratios of fresh and aged biomass-burning smoke in the troposphere and stratosphere.For better comparison, the tropospheric triple-wavelength depolarization ratio observation of Burton et al. (2015) performed in aged northwestern American smoke is listed in the third line, i.e., before the tropospheric section.The lidar and depolarization ratios of Hu et al. (2018)   For more details, see Sect.2.2.
Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-358Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 6 April 2018 c Author(s) 2018.CC BY 4.0 License.The top of the planetary boundary layer (PBL) was at 1.8 km height.Between 8 and 13-14 km height, the atmosphere was almost free of smoke.A strong smoke layer is visible between 15 and 16 km, traces of smoke occurred also between 14 and 15 km height.The stratospheric layer was about 3-4 km above the tropopause.The 532 nm AOT of the stratospheric layer had decreased from 0.6 around noon to 0.2-0.25 in the night of 22 August 2017.
Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-358Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 6 April 2018 c Author(s) 2018.CC BY 4.0 License.that (a) the stratospheric smoke contributed 60% to the total AOT (as observed with lidar) and thus to the column volume concentration, and (b) that these 60% can be assigned to the 1 km thick stratospheric layer from 15 and 16 km height.With this information, the AERONET column volume values for each size bin were converted into volume and mass concentrations as shown in Fig. 4 and interpreted as the stratospheric contribution to the total column mass size distribution.

Figure 5
Figure 5 highlights the most important result.The shown depolarization and lidar ratio values are taken from Figs. 2 and 3.

Figure 1 .Figure 2
Figure 1.Canadian wildfire smoke layers in the troposphere (mostly between boundary-layer top at 1.8 and 6.5 km height) and in the stratosphere (15-16 km height) observed with lidar at Leipzig on 22-23 August 2017, 20:45-00:30 UTC.Shown is the range-corrected cross-polarized 532 nm backscatter signal measured with temporal and vertical resolution of 10 s and 7.5 m, respectively.The indicated tropopause height (GDAS, 2017) is in agreement with nearby radiosonde profiles.

Figure 3 .
Figure3.Same as Fig.2, except for the stratospheric aerosol layer and for different signal smoothing lengths (as explained in Sect.2.2).In the case of the 1064 nm extinction coefficient (red solid diamond) , a retrieval window length (least-squares method) of 2500 m had to be applied (indicated by the long vertical bar).We estimated the layer mean 1064 nm extinction coefficient (red open diamond) for the 1250 m thick layer from 15-16.25 km height by multiplying the obtain value for 2500 window length by a factor of 2, assuming that the extinction below and above the 1250 m thick layer was close to zero.In the subsequently calculation of the 1064 nm lidar ratio we used this 1250 m layer mean extinction value (open diamond) together with an appropriately smoothed backscatter coefficient (see text for more details).

Figure 4 .
Figure 4. Particle mass size distribution derived from column (tropospheric + stratospheric) AERONET observations at Lindenberg, 180 km northeast of the lidar site, in the morning of 23 August 2017 (green) and obtained from the inversion of lidar-derived optical properties in the tropospheric layer (red) and stratospheric layer (black, magenta).Small particles prevailed in the tropospheric layer and comparably large accumulation-mode particles dominated in the stratospheric layer.The black and magenta curves are obtained by assuming spherical and spheroidal particle shapes in the lidar data inversion, respectively.

Figure 5 .
Figure5.Comparison of the spectral dependence of the tropospheric (5-6 km height) and stratospheric (15-16 km height) particle lidar ratio (a) and particle linear depolarization ratio (b).A strongly contrasting spectral behavior is found in the case of the depolarization ratio and an almost similar wavelength dependence (in the troposphere and stratosphere) is found for the lidar ratio.Only BERTHA values are considered.