Lidar observations of pyrocumulonimbus smoke plumes in the UTLS over Tomsk (Western Siberia, Russia) from 2000 to 2017

Large volcanic eruptions with the volcanic explosivity index (VEI)  3 are widely known to be the strongest source of long-lived aerosol in the upper troposphere and lower stratosphere (UTLS). However, the latest studies have revealed that massive forest (bush) fires represent another strong source of short-term (but intense) aerosol perturbations in 10 the UTLS if combustion products from the fires reach these altitudes via convective ascent within pyrocumulonimbus clouds (pyroCbs). PyroCbs, generated by boreal wildfires in North America and North-East Asia and injecting smoke plumes into the UTLS, have been intensively studied using both groundand space-based instruments since the beginning of the 21 century. In this paper, we focus on aerosol layers observed in the UTLS over Tomsk (56.48 N, 85.05 E, Western Siberia, Russia) that could be smoke plumes from such pyroCb events occurred in the 2000–2017 period. Using the HYSPLIT 15 trajectory analysis, we have reliably assigned ten aerosol layers to nine out of more than 100 documented pyroCb events, the aftereffects of which could potentially be detected in the UTLS over Tomsk. All of the nine pyroCb events occurred in the USA and Canada: one event per year was in 2000, 2002, 2003, 2015, and 2016, whereas two events per year were in 2013 and 2017. No plumes from pyroCbs originating in the boreal zone of Siberia and the Far East (to the east of Tomsk) were observed in the UTLS over Tomsk between 2000 and 2017. We conclude that the lifetimes of pyroCb plumes to be detected 20 in the UTLS using ground-based lidars are less than about a month, i.e. plumes from pyroCbs generated by wildfires to the east of Tomsk can significantly diffuse before reaching the Tomsk lidar station by the westerly zonal transport of air masses. A comparative analysis of the contributions from pyroCb events and volcanic eruptions with VEI  3 to aerosol loading of the UTLS over Tomsk has also been made. Finally, an aerosol plume from the Aleutian volcano Bogoslof erupted with VEI = 3 on 28 May 2017 was detected at altitudes between 10.8 and 13.5 km over Tomsk on 16 June 2017. 25


Introduction
There are many sources of aerosol in the troposphere: bio-and fossil-fuel burning, forest and bush fires, power generation and industrial processes, engines, volcanic eruptions, etc., and conversely, only a few such sources exist in the stratosphere.
Aircraft emissions (combustion products of carbon-containing fuels) (Blake and Kato, 1995;Hendricks et al., 2004;Koehler et al., 2009;Wilkerson et al., 2010;Balkanski et al., 2010) and troposphere-to-stratosphere transport of air (Kremser et al., 30 2016) are responsible for background aerosol loading in the lower stratosphere (LS). Large volcanic eruptions with the volcanic explosivity index (VEI)  3 represent the principal source of strong and long-term stratospheric aerosol perturbations (Robock, 2000;Robock and Oppenheimer, 2003;Kremser et al., 2016), which is confirmed by both spaceborne and ground-based long-term lidar measurements (Vernier et al., 2011;Trickl et al., 2013;Mills et al., 2016;Sakai et al., 2016;Khaykin et al, 2017;Zuev at al., 2017;Friberg et al., 2018). Volcanic plumes persist in the stratosphere for several 5 months to several years, depending on the eruption latitude, VEI, and maximum plume altitude (MPA) after the eruptions (Hofmann et al., 2009). However, studies over the last two decades have revealed that, in addition to volcanic eruptions, there exists another source being able to cause short-term, but locally intense, aerosol perturbations in the LS. This source is massive forest (or bush) fires if combustion products from the fires reach stratospheric altitudes.
Massive forest fires (wildfires), the plumes of which can ascend to the LS, and their aftereffects have been intensively 10 studied since the beginning of the 21 century (Fromm et al., 2000(Fromm et al., , 2005(Fromm et al., , 2006(Fromm et al., , 2008a(Fromm et al., ,b, 2010Fromm and Servranckx, 2003;Jost et al., 2004;Livesey et al., 2004;Damoah et al., 2006;Cammas et al., 2009;Gonzi and Palmer, 2010;Guan et al., 2010;Siddaway and Petelina, 2011;Dahlkötter et al., 2014;Paugam et al., 2016). Smoke plumes of the overwhelming majority of forest fires are located within the planetary boundary layer (Val Martin et al., 2010;Nikonovas et al., 2017;Rémy et al., 2017), and a small number of them (< 5-10 %) can enter the free troposphere (Sofiev et al., 2013;Peterson et al., 2014). 15 Only in exceptional cases aerosol plumes from the fires are able to reach stratospheric altitudes via convective ascent within pyro-cumulonimbus clouds (pyroCb; http://glossary.ametsoc.org/wiki/Pyrocumulonimbus). PyroCbs, injecting aerosol directly into the LS, originate mainly from boreal wildfires in North America (particularly in the Canadian boreal zone) and North-East Asia (Siberia and the Far East) Guan et al., 2010), and bush fires in Australia (Fromm et al., 2006;Siddaway and Petelina, 2011). In particular years, pyroCb events can occur too frequently to be considered as an 20 occasional source of aerosol in the LS. For example, Fromm et al. (2010) identified 17 such pyroCbs in the United States and Canada during the summer of 2002, a part of which reached the LS.
PyroCb stratospheric plumes can spread throughout the hemisphere and are detected by both ground-and space-based lidars for 2 to 4 months after their occurrence (Fromm et al., 2000(Fromm et al., , 2008b. Owing to their potential impact on the climate, a lot of attention is currently paid to monitoring pyroCbs via, e.g., the Geostationary Operational Environmental 25 Satellite (GOES) system (https://www.nasa.gov/content/goes). The data on pyroCb events occurring throughout the world are accumulated on the web page of the Cooperative Institute for Meteorological Satellite Studies (CIMSS): http://pyrocb.ssec.wisc.edu/ and their archives have been available since May 2013.
Ground-based lidar observations of stratospheric aerosol perturbations have been almost continuously performed in Tomsk (56.48 N,85.05 E, Western Siberia, Russia) for more than 30 years (Zuev et al., 1998(Zuev et al., , 2001(Zuev et al., , 2017. In the papers, 30 we mainly discussed and focused on aerosol perturbations in the stratosphere over Tomsk after major volcanic eruptions (with VEI  3), the plumes of which were able to directly enter the stratosphere. To consider the effect of only volcanic eruptions on stratospheric aerosol loading and definitely exclude from consideration any aerosol perturbations in the upper troposphere (UT ) (such as cirrus clouds) and tropopause region, we analyzed the results of lidar measurements at altitudes Atmos. Chem. Phys. Discuss., https://doi.org /10.5194/acp-2018-1153 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 22 November 2018 c Author(s) 2018. CC BY 4.0 License.
3 higher than 13-15 km. It is clear that this altitude limitation could lead to the loss of information on aerosol events like pyroCb plumes in the UTLS over Tomsk.
The possibility to observe stratospheric smoke plumes in Tomsk from massive forest fires occurred in North America was noted in Zuev et al. (2017). In this paper, we analyze all aerosol perturbations in the 11-30 km altitude region over Tomsk that could be caused by massive wildfires in North America and North-East Asia from 2000 to 2017. 5

Lidar instruments and methods
The lidar measurements we consider were made using the aerosol channel of the Siberian Lidar Station (SLS) located in Tomsk. The transmitter of the channel represents a Nd:YAG laser (LS-2132T-LBO model, LOTIS TII Co., the Republic of Belarus) that operates at a wavelength of 532 nm with 100 mJ pulse energy and at a pulse repetition rate of 20 Hz. The channel receiver is a Newtonian telescope with a mirror diameter of 0.3 m and a focal length of 1 m. The backscattered 10 signals are registered by a photomultiplier tube R7206-01 (Hamamatsu Photonics, Japan) operating in the photon counting mode with a vertical resolution of 100 m. Owing to the rearrangement and improvement of the SLS, there were two shutdown periods of the aerosol channel from July 1997 to May 1999 and from February to September 2014. A detailed description of the SLS aerosol channel technical parameters is given in (Zuev, 2000;Burlakov et al., 2010 where the lower limit H 1 = 11 km falls within either the UT or LS due to the variability of the local tropopause altitude and does not allow missing pyroCb plumes in the UTLS, and the upper limit is the calibration altitude H 2 = H 0 = 30 km. 25 When analyzing the perturbed scattering ratio R(H, ) profiles, cirrus clouds are excluded from consideration based on the following two criteria. First, a detected aerosol layer is definitely located in the UT and, second, the layer has a thickness of < 1 km and the value of R(H) > 2.45 for  = 532 nm (see Appendix A). In some cases, however, there is a problem in Atmos. Chem. Phys. Discuss., https://doi.org /10.5194/acp-2018-1153 Manuscript under review for journal Atmos. Chem. Phys.

H
for the 20 HYSPLIT air mass backward trajectories are calculated above ground level (a.g.l.). Since the SLS is situated at an altitude of 148 m a.s.l., the difference between altitudes H (a.s.l.) and back.
traj.  (Zuev et al., 2017), with the exception of the 13 February 2014 Kelut eruption, the plume of which, however, could not be detected at the SLS due to the 2014 shutdown period (see Sect. 2).
When analyzing aerosol layers observed over Tomsk and pyroCb events documented in the Northern Hemisphere over the 5 period 2000-2017, we have discovered more than 100 pyroCbs (with known and unknown MPAs), the plumes of which could potentially be detected in the UTLS over Tomsk. However, only a few of the detected layers have been reliably attributed to the selected pyroCb events using the HYSPLIT trajectory analysis. To illustrate the correlation between the pyroCbs and corresponding layers over Tomsk, we present only the most successful examples of the HYSPLIT trajectories that passed over or close to the places of origin of the pyroCbs (or near the known pyroCb plume locations when the exact 10 coordinates and time of the pyroCb events are unknown).

Detection of pyroCb smoke plumes in the UTLS
The first aerosol layer we consider was observed in the UTLS over Tomsk at altitudes between 11.4 and 12.  Another aerosol layer potentially associated with a pyroCb event was observed over Tomsk between 10 and 12 km with the maximum R(H) = 1.87 at H = 11.1 km a.s.l. on 29 August 2003 (Fig. 4a). Eleven days earlier, on 18 August, a pyroCb plume was registered in the UTLS over Hudson Bay (61° N, 89° W; Canada) by the TOMS. The pyroCb was previously generated by the "Conibear Lake Fire" in the Wood Buffalo National Park (Alberta/Northwest Territories, Canada) . As seen in Fig. 4b, the HYSPLIT air mass backward trajectories, started from altitudes of 11.75 km a.s.l. over 5 Tomsk at 17:00 UTC on 29 August, passed over the pyroCb plume location at altitudes back. traj.

H
 11.7-12.0 km a.g.l. on 18 August. Based on the behavior of the example trajectories (Fig. 4b) and the tropopause altitudes determined at the three nearest to Tomsk meteorological stations (Fig. 4a), we suppose that the pyroCb plume was spreading in the UT in a given period of time. The next two aerosol layers reliably attributed to pyroCb events were registered at the SLS in Tomsk only 10 years later, 15 in July and September 2013. Namely, the first "weak" layer with the maximum R(H) = 1.27 at H = 11.7 km a.s.l. was observed in the UTLS over Tomsk on 14 July 2013 (Fig. 5a). The HYSPLIT trajectory analysis showed that the layer could represent a smoke plume from a pyroCb generated by large fires that were burning in the Eastmain region of Quebec (52° N, 78° W; Canada) in June-July 2013. The Eastmain pyroCb was discovered using the 1-km resolution GOES-13 0.63 µm visible channel after 21:55 UTC on 4 July (http://pyrocb.ssec.wisc.edu/archives/136). The second "strong" layer was observed over Tomsk at altitudes between 11.2 and 12.8 km with the maximum R(H) = 10 11.4 at H = 11.8 km a.s.l. on 23 September 2013 (Fig. 6a)

H
 10.7-11.7 km a.g.l. on 16 September are shown in Fig. 6b. Despite the high value of the scattering ratio R(H), which is representative of cirrus clouds, the tropopause altitudes determined at the nearest to Tomsk meteorological stations show that the aerosol layer maximum was definitely in the LS (Fig. 6a). This allows us to conclude that the layer was a stratospheric one and could not be a cirrus cloud. H  10.5-11.3 km a.g.l. on 16 May. As seen in Fig. 8a, it is difficult to definitely determine whether the aerosol layer was in the UT or LS over Tomsk. Nevertheless, the fact that the layer was completely higher than 11 km and two out of three 10 tropopause altitudes allows us to conclude that the layer was not a cirrus cloud. The smoke from the pyroCb was also observed in the UTLS over the UK with Raman lidars between 23 and 31 May 2016. (Vaughan et al., 2018).  (Fig. 9a). Three days later, on 29 August, the second ("weak") layer was detected with the maximum R(H) = 1.37 at H = 14.5 km a.s.l. (Fig. 9b).
Finally, the third layer was observed between 14.3 and 16.2 km with the maximum R(H) = 3.1 at H = 15.7 km a.s.l. on 31 10 August (Fig. 9c). In each case considered, the HYSPLIT trajectory analysis showed that the backward ensemble trajectories passed near and/or over the places of origin of both pyroCbs on 12 August (Fig. 10). The initial conditions (times and altitudes over Tomsk) for each trajectory can also be found in Fig. 10. Based upon the end points of the trajectories, the MPAs H MPA for both pyroCbs were definitely in the LS within the range of 13.5-15.0 km a.g.l. We cannot exclude that the layers observed over Tomsk on 26, 29, and 31 August could contain aerosol from the other three pyroCbs detected by the 15 NOAA-18 instruments on the evening of 12 August.

Detection of the Bogoslof volcanic plume in 2017
The  (Fig. 11a). The aerosol layer was initially considered as a smoke plume from a pyroCb that was generated by a wildfire started to burn at Air mass backward trajectories started from altitude of 12.7-12.8 km a.s.l. over Tomsk at 18:00 UTC on 16 June 2017.

PyroCb events in 2004-2012 15
Several strong pyroCbs, the plumes of which reached UTLS altitudes with H MPA  12 km a.s.l. and could potentially be detected in the UTLS over Tomsk, were documented in the Northern Hemisphere between 2004 and 2011 (Table 1).
However, no aerosol layers associated with these pyroCb events were observed at the SLS during the period. This was due to unfavorable weather conditions or pyroCb plumes could have diffused or passed by the SLS and, therefore, might not be detected. Note also that twelve explosive eruptions with VEI = 3-4 of both tropical and northern extratropical volcanoes, the 20 Atmos. Chem. Phys. Discuss., https://doi.org /10.5194/acp-2018-1153 Manuscript under review for journal Atmos. Chem. Phys. 14 aftereffects of which were reliably registered in the stratosphere over Tomsk, occurred in the 2004-2011 period (Zuev at al., 2017). We do not exclude that pyroCb plumes could hardly be discerned against the background of the volcanic plumes in the UTLS over Tomsk in this period. There were no significant events (volcanic eruptions and pyroCbs) to be recorded at the SLS in 2012.

Time series of the integrated aerosol backscatter coefficient (2001-2017)
To estimate the contribution of the pyroCb events discussed above to aerosol loading of the UTLS over Tomsk, we have two volcanic eruptions that could perturb the UTLS over Tomsk occurred for a given period of time (Table 3). Six pyroCb events injected smoke into the UTLS in 2013 and 2015-2017 (Table 2)  The trends in Fig. 12 definitely show that for Tomsk region the aftereffects of tropical and northern volcanic eruptions with VEI  3 are stronger and longer-lasting than those of pyroCb events occurred mainly due to wildfires in North America.
Indeed, volumes and lifetimes of primary (volcanic ejecta) and secondary (sulfuric acid) aerosols in the UTLS from explosive volcanic eruptions are known to be higher (Hofmann et al., 2009) compared to those of aerosols from pyroCb plumes . Hence, twelve volcanic eruptions occurred in time interval (b) naturally led to an increase in 10 aerosol loading of the UTLS over Tomsk and, therefore, to a positive trend in the annual average a  B values. PyroCbs generated by wildfires from 2004 to 2011 (including documented ones listed in Table 1) also had to perturb the UTLS over

Concluding remarks
The increasing number and intensity of boreal forest fires in North America and North-East Asia due to climate warming 5 for the last decades (Wotton et al., 2010;Sofiev et al., 2013;Rémy et al., 2017) lead to an increasing number of pyroCbs, the plumes of which are able to reach the UTLS Guan et al., 2010). Boreal wildfires are usually active during the warm half year (April to September) and spread in the UTLS for long distances mainly due to the westerly zonal transport of air masses in the Northern Hemisphere. Therefore, the plumes of pyroCbs occurred in North America are frequently detected in the UTLS over Europe, and more rarely over Siberia, and the Far East by both ground-and space-10 based lidars.
In this study, we have considered and analyzed aerosol layers in the UTLS (11-30 km) over Tomsk that could represent smoke plumes from pyroCbs generated by massive wildfires in North America and North-East Asia between 2000 and 2017.
Using the HYSPLIT trajectory analysis, we have reliably assigned ten such layers to nine out of more than 100 documented pyroCb events, the aftereffects of which could potentially be detected at the SLS. All of the nine pyroCb events occurred in 15 North America: one event per year was in 2000, 2002, 2003, 2015, and 2016, whereas (Table 1) against the background of more powerful plumes from twelve volcanic eruptions observed during this period (Table 3).
Massive forest fires generating pyroCbs are also known to occur in North-East Asia (pyrocb.ssec.wisc.edu). However, no plumes in the UTLS over Tomsk from pyroCbs occurred in the boreal zone of Siberia and the Far East (to the east of Tomsk) 5 were detected at the SLS between 2000 and 2017. We can assume that the lifetimes of pyroCb plumes to be detected in the UTLS using ground-based lidars are less than about a month. In other words, plumes from pyroCbs generated by wildfires to the east of Tomsk can significantly diffuse before reaching the SLS due to the westerly zonal transport. This probably explains a comparatively "low" contribution from pyroCbs to aerosol loading of the UTLS over Tomsk and, therefore, the negative trends in the annual average a  B values in the absence of, and at a low, volcanic activity in time intervals (a) and 10 (c), respectively (Fig. 12).
Based on the results of lidar observations at the SLS between 2000 and 2017, we can conclude the following. During a short-term period (up to three weeks) after pyroCb events have occurred in North America, their aftereffects in the UTLS over Tomsk are comparable to those of volcanic eruptions with VEI  3. The 10-day average a  B value after the events can be even higher than that after volcanic eruptions. For example, the 10-day average a  B for the 20-30 September period 15 reached the maximum value of 3 3.20 10   sr -1 after pyroCb event 5 (Table 2) of 16 September 2013 (Fig. 12). Moreover, smoke plumes reached the UTLS over Tomsk from two or more pyroCbs in a single year can lead to a marked increase in aerosol loading compared to that in the previous year. For example, the annual average a  B value increased by 27.1% in 2013 and 14.8% in 2017 (together with the 2017 Bogoslof eruption). Nevertheless, the contribution from pyroCbs (generated by wildfires in North America and injecting smoke into the UTLS) to the annual average a  B value integrated over the 11-30 20 km altitude range is noticeably lower (for Tomsk region) than the contribution from both tropical and northern volcanic eruptions with VEI  3 (due to, among other things, secondary sulfuric acid aerosol).

Data availability
The NOAA's HYSPLIT model used to calculate all air mass backward trajectories is available at http://ready.arl.noaa.gov/HYSPLIT.php. The volcanic eruption data we used can be found at http://volcano.si.edu/ and the 25 data on pyroCb events occurred after May 2013 are located at http://pyrocb.ssec.wisc.edu/. The radiosonde data for the Kolpashevo, Emeljanovo, and Novosibirsk meteorological stations are on the web page http://weather.uwyo.edu/.
Author contributions. VVZ and VVG performed main analysis of all data and wrote the paper. AVN made measurements at the SLS and processed lidar data. ESS performed the HYSPLIT trajectory analysis. VVG and ESS retrieved data on pyroCbs, the plumes of which could potentially be detected in the UTLS over Tomsk.
Competing interests. The authors declare that they have no conflict of interest.
Acknowledgements. We thank Michael Fromm (the US Naval Research Laboratory) for the information on the aerosol cloud 5 coming to Tomsk from several strong pyroCbs generated by wildfires in British Columbia (Canada) in August 2017.

Appendix A: Scattering ratio R(H, ) values for cirrus clouds
Aerosol layers detected in the UT with ground-based lidars are identified as cirrus clouds if the scattering ratio R(H) > 10 for a laser wavelength  1 = 532 nm (Tao et al., 2008;Samokhvalov et al., 2013). However, according to Sassen et al. (1989), the minimum value of R(H) can be 5.2 in the case of invisible to the naked eye co-called "subvisual" cirrus clouds (for a laser 10 wavelength  2 = 694.3 nm) with a thickness of < 1 km. To calculate the minimum R(H) value for  1 = 532 nm, one can use the fact that the aerosol backscatter coefficient a ( , )   H is considered to be independent of the scattered light wavelength if aerosol particles are much greater than the wavelength (Measures, 1984). Since cirrus cloud particles (25 m, Sassen et al., 1989) Substituting  1 = 532 nm,  2 = 694.3 nm, and R( 2 ) = 5.2 into Eq. (A2), we finally obtain R( 1 ) = 2.45.