2.1 Polly and AERONET observations at Punta Arenas, Chile

Lidar observations were performed at the campus of the University of Magallanes (UMAG) at Punta Arenas (53.2^∘ S, 70.9^∘ W; 9 m above sea level, a.s.l.) in the framework of the DACAPO-PESO (Dynamics, Aerosol, Cloud And Precipitation Observations in the Pristine Environment of the Southern Ocean) campaign lasting from November 2018 to November 2021 (Radenz et al., 2021). The main goal of DACAPO-PESO was the investigation of aerosol–cloud interaction processes in rather pristine, unpolluted marine conditions.

A Polly instrument (POrtabLle Lidar sYstem; Engelmann et al., 2016) was used for aerosol profiling. This multiwavelength polarization Raman lidar was continuously running, has 13 channels, and measures elastic backscatter signals at the laser wavelengths of 355, 532, and 1064 nm and respective Raman backscattering at 387 and 607 nm (nitrogen Raman channels) and 407 nm (water vapor Raman channel). These observations permit us to determine the height profiles of the particle backscatter coefficient at the laser wavelengths of 355, 532, and 1064 nm wavelength, particle extinction coefficients at 355 and 532 nm, the particle linear depolarization ratio at 355 and 532 nm, and of the water-vapor-to-dry-air mixing ratio by using the Raman lidar return signals of water vapor and nitrogen. The Polly instrument is designed for the automated continuous profiling of aerosols and clouds and thus was running around the clock. Well-defined breaks were necessary to exchange laser flash lamps, to run different calibration procedures, to check the full setup, and to perform an overall alignment of the Polly instrument.

Details of the stratospheric Polly data analysis, including an uncertainty discussion, are given in the recent publications of Engelmann et al. (2021) and Ohneiser et al. (2020, 2021). The Raman lidar method was exclusively used to determine particle backscatter and extinction profiles up to the top of the Australian smoke layers. The particle backscatter coefficient is obtained from the ratio of the elastic backscatter signal (355 and 532 nm) to the respective Raman signal (387 and 607 nm). The 1064 nm backscatter coefficient is calculated from the ratio of the 1064 nm elastic backscatter signal to the 607 nm nitrogen Raman signal. In the retrieval of the extinction coefficient, a least squares linear regression is applied to the respective Raman signal profiles. To obtain stratospheric background aerosol information at heights above the smoke layer (usually above >25 km in height), the backscatter and extinction properties were determined from the elastic backscatter signal profiles by using the so-called Klett method, which needs a height-independent extinction-to-backscatter ratio (lidar ratio) as input (Fernald, 1984). We used a lidar ratio of 50 sr in the background aerosol retrieval. Calibration heights were generally set into the clean stratosphere (above 30–35 km height). The backscatter signal intensities were always sufficiently above the detection limits at heights of 30–35 km.

Different expressions for the Ångström exponent were used to characterize the spectral dependence of the aerosol optical properties. The Ångström exponent $A_{x,{\mathit{\lambda }}_i,{\mathit{\lambda }}_j}=\ln (x_i/x_j)/\ln ({\mathit{\lambda }}_j/{\mathit{\lambda }}_i)$ describes the wavelength dependence of an optical parameter x (backscatter coefficient β or extinction coefficient σ) in the spectral range from wavelength λ_i to λ_j.

The particle linear depolarization ratio (PLDR) was used to characterize the shape properties of the smoke particles. PLDR is obtained from the cross-to-co-polarized signal ratio after the correction of Rayleigh contributions to light depolarization. The terms “co” and “cross” denote the planes of polarization parallel and orthogonal to the plane of linear polarization of the transmitted laser pulses, respectively. The particle depolarization ratio, defined as the ratio of the cross-to-co-polarized backscatter coefficient, was observed at 355 and 532 nm. Irregularly shaped smoke particles cause considerable light depolarization and PLDR values around 0.2, whereas PLDR < 0.05 in the case of spherical smoke particles with a compact core-shell structure.

Although the Polly data analysis software package automatically delivers calculated aerosol optical properties, the lidar observations presented in this article were manually analyzed. To keep the uncertainties in the derived aerosol quantities at an acceptably low level of around < 10 %, for the backscatter coefficients and depolarization ratios and of the order of 20 %–30 % in the case of particle extinction coefficients and lidar ratios, large vertical signal smoothing and regression window lengths of 500 to 2500 m and signal averaging times from 30 min to several hours had to be applied. More details are given in Ohneiser et al. (2021).

To obtain estimates of the microphysical properties of the smoke particles such as mass, volume, and surface area concentration, a conversion method is available (Ansmann et al., 2021a). The 532 nm particle backscatter coefficients are input in the retrieval process. In the case of the Australian smoke observations, the backscatter coefficients are converted to extinction profiles, in the first step, by assuming a smoke lidar ratio of 85 sr at 532 nm. In the second step, the extinction coefficients were converted to microphysical properties by respective smoke conversion factors. To obtain smoke mass concentrations, we assume a smoke particle density of 1.15 g cm⁻³ (Ansmann et al., 2021a).

Auxiliary data were required in the lidar data analysis in form of temperature and pressure profiles in order to calculate and correct for Rayleigh backscatter and extinction influences on the measured lidar return signal profiles. For auxiliary meteorological observations, we used GDAS1 (Global Data Assimilation System 1) temperature and pressure profiles with 1^∘ horizontal resolution from the National Weather Service's National Centers for Environmental Prediction (GDAS, 2021). We checked the quality and accuracy of the GDAS1 profiles through a comparison with respective temperature and pressure profiles measured with radiosonde at Punta Arenas airport (daily launch at 12:00 UTC). CALIOP, Sentinel-5, and MODIS (Moderate Resolution Imaging Spectroradiometer) data (CALIPSO, 2021b; ERA5, 2021; MODIS, 2021a, b, c) were used to support the complex study of the long-range smoke transport across the Pacific Ocean.

Besides the continuous lidar measurements, an AERONET sun photometer was operated (Holben et al., 1998; AERONET, 2021). Measurement wavelengths and retrieval products of the Punta Arenas sun photometer are given in Ansmann et al. (2021a). Sun photometer aerosol optical depths measured at 380, 500, and 1020 nm were partly combined with column-integrated backscatter coefficients from lidar to estimate stratospheric lidar ratios at 355, 532, and 1064 nm. This approach is described in detail in Sect. 3.3.

To study the impact of the stratospheric wildfire smoke on the stratospheric ozone layer, we made use of the long-term Southern Hemispheric ozone data record (NDACC, 2021). Ozonesonde observations are regularly performed by the Network for the Detection of Atmospheric Composition Change (NDACC). Height profiles of the ozone partial pressure have been measured since the 1990s. Based on these ozone profile observations, clear indications for a smoke-related ozone loss are obtained, as discussed in Sect. 4.

2.2 CORAL observations at Río Grande, Argentina

As part of the SOUTHTRAC-GW (Southern Hemisphere Transport, Dynamics, and Chemistry–Gravity Waves) mission (Rapp et al., 2021), the Compact Rayleigh Autonomous Lidar (CORAL) (Kaifler and Kaifler, 2021) was operated at Estación Astrónomicas Río Grande (53.8^∘ S, 67.7^∘ W; 9 m a.s.l.), which is close to the airport of Río Grande in southern Argentina. The station is 230 km east to southeast of Punta Arenas. The lidar is optimized for density and temperature measurements in the middle atmosphere, approximately between 15 and 90 km. During the stratospheric smoke period, the powerful lidar delivered valuable smoke optical properties at 532 nm for heights from 15 to 30 km. CORAL operates autonomously during clear-sky conditions in darkness and uses three height-cascaded elastic backscatter channels (532 nm wavelength) and one Raman channel (607 nm wavelength), which allows for the independent retrieval of the 532 nm backscatter coefficient, extinction coefficient, and lidar ratio. Meteorological data are used from radiosonde launches at Punta Arenas airport (daily launch at 12:00 UTC; University of Wyoming, 2021).

We applied the Polly data analysis software to analyze the CORAL raw signal counts (at 532 and 607 nm) to avoid any bias caused by different data analysis concepts. We noticed and corrected a small cross-talk effect in the Raman channel (contamination by elastically backscattered laser photons; less than 0.8 %). The signal reference height was set to 30 ± 1 km, with a reference particle backscatter coefficient of 0.001 Mm⁻¹. Vertical signal smoothing and regression windows (extinction and lidar ratio retrieval) was set to 2100 m. Profiles of the particle backscatter coefficient were obtained with smoothing lengths of 500 m.

3.1 The major stratospheric perturbation in January 2020

According to Peterson et al. (2021), the Black Summer fire season of 2019–2020 in southeastern Australia contributed to a rather intense period of fire-induced and smoke-infused thunderstorms. ANYSO resulted in roughly 1.0 Tg of cumulative smoke particle mass being injected into the lower stratosphere. It was characterized as a pyroCb “super outbreak” because of its exceptional scale and magnitude. An unusually large number of 38 distinct convective cloud towers, denoted as pyroCb pulses, developed over a prolonged period of 51 non-consecutive hours. ANYSO occurred in two distinct phases, with the first and largest occurring during 29–31 December, with an overall duration of about 45 h (from 09:30 UTC on 29 December to 06:40 UTC on 31 December 2019). During the first phase, 33 distinct pyroCb pulses injected smoke into the UTLS. Part of the smoke reaching the upper part of the troposphere was further lofted to UTLS heights by strong convective activity in the outflow regime over the Pacific Ocean on the way to New Zealand (Hirsch and Koren, 2021) and by large cyclonic systems that formed further downstream east of New Zealand and facilitated a direct transport of smoke into the lower stratosphere (Magaritz-Ronen and Raveh-Rubin, 2021). The second phase of ANYSO began on 4 January 2020, after 3 d without pyroCb activity. The duration of this pyroCb activity (with five pyroCb pulses) was more typical of previous significant events (such as the Canadian wildfires, PNE, British Columbia, 12–13 August 2017; Peterson et al., 2018), spanning just over 6 h. PNE comprised less than 10 pyroCb pulses; however, it caused the largest smoke-related stratospheric perturbation ever observed over Europe (Ansmann et al., 2018; Baars et al., 2019).

ANYSO’s first phase stands as the largest known stratospheric injection of smoke particles linked to a distinct period of pyroCb activity (0.2–0.8 Tg) (Peterson et al., 2021). ANYSO’s second phase injected an estimated 0.1–0.3 Tg of additional smoke particle mass into the lower stratosphere. The cumulative smoke particle mass injected into the stratosphere by both phases of pyroCb activity was thus 0.3–1.1 Tg and therefore at least 3 times larger than the PNE smoke mass of 0.3 Tg (Peterson et al., 2018). More than half of the 38 observed pyroCb pulses injected smoke particles directly into the stratosphere. Taking maximum updraft velocities of 35–60 m s⁻¹ in these cumulus towers into account (Rodriguez et al., 2020), smoke can be lofted within less than an hour from the surface to the lower stratosphere. The UTLS smoke plumes encircled a large swath of the Southern Hemisphere while continuously rising due to local heating by the smoke absorption of solar radiation. More details to the ascent of the smoke is provided in the following subsection.

Figure 1 provides an overview of the smoke situation over the southern Pacific and Atlantic Oceans in January 2020 (MODIS, 2021a, b, c). Figure 1a shows an optically dense smoke field between Australia (Melbourne) and New Zealand (Wellington) on 5 January 2020. In Fig. 1b, monthly mean January AOT (500 nm) values are visualized, frequently clearly exceeding 0.5, and sometimes 1.0, between Australia and South America. The smoke was mainly transported from Australia across northern New Zealand down to southern South America and Antarctica during the initial phase of the long-lasting stratospheric perturbation. Figure 1c and d magnify the smoke situation close to the source region, where the 10 d AOT mean values were of the order of 1.0 to > 2.0 during the injection phase (1–10 January 2020) and, after long-range transport over 10 000 km southwest of southern South America, still high 10 d mean AOT values of the order of 0.4–0.8 were found.

Figure 1Satellite observations of the stratospheric smoke transport from Australia to South America in January 2020. (a) Brownish smoke layer between Melbourne and Wellington on 5 January 2020 (MODIS, 2021a, b) and fires indicated by red spots in eastern and northern Australia (FIRMS, 2021). (b) The 500 nm AOT in the Southern Hemisphere (MODIS, 2021c); the smoke plumes were traveling mostly eastward. (c) Mean AOT values for the 1–10 January period, close to the source region (MODIS, 2021c). (d) Mean AOT values for the 20–31 January period, close to southern South America (MODIS, 2021c). (e) Time series of daily 500 nm AOT (mean values for the black rectangles are shown in (c) and (d) separately for MODIS Terra and Aqua). Cloud-screened values are shown (MODIS, 2021c). In addition, AERONET sun photometer values of the total (tropospheric + stratospheric) AOT at Punta Arenas, CALIOP-derived AOT values, as shown in Kablick et al. (2020), for the multiday outbreak ANY (Australian New Year), and Polly AOT values for the stratosphere over Punta Arenas, are shown.

Figure 1e provides a time series of daily AOT observed with MODIS (Terra and Aqua satellites), an AERONET photometer, and lidar at Punta Arenas. Initially, the area mean AOT values were close to 2.0 in the black rectangle shown in Fig. 1c and then mostly in the range from 0.2 to 0.5 after 10 January 2020. According to the time series of daily cloud-screened MODIS AOT observations within the black rectangle in Fig. 1d, the rotating smoke-filled vortex close to Punta Arenas caused AOTs of 0.3–0.5 during the last week of January 2020. Details of the rotating smoke fields are given in the next subsection. Note that the Southern Ocean AOT of the clean marine troposphere is of the order of 0.03–0.05 at 500 nm, as indicated by the Punta Arenas AERONET sun photometer observations from 1–6 January 2020.

Nighttime Raman lidar observations of 532 nm extinction profiles are used in Fig. 1e to compute the lidar-derived stratospheric AOTs. First, smoke plumes reached Punta Arenas on 5–7 January 2020. Significantly enhanced stratospheric AOTs were then observed on 8 January 2020 (Ohneiser et al., 2020). AOT values observed with lidar (during nighttime) and a photometer (during daytime) slowly increased towards values around 0.3–0.35 in the last week of January 2020. A few values measured with CALIOP during smoke vortex overflights (close to South America) from 23–31 January 2020 are included in Fig. 1e.

3.1.1 The unique, self-organized, rotating and ascending smoke-filled vortex

A new atmospheric phenomenon was detected in January–March 2020 (Kablick et al., 2020; Khaykin et al., 2020; Allen et al., 2020; Lestrelin et al., 2021). It coupled smoke occurrence, aerosol–radiation interaction, and dynamical processes. The striking discovery was the observation of several stratospheric smoke plumes that self-organized as quasi-ellipsoidal anticyclonic vortices, persisted over weeks to months, and traveled around the Southern Hemisphere. These plumes transported smoke up to heights above 30 km. The most intense, self-maintained vortex measured about 1000 km in diameter and about 5 km in vertical extent. This highly stable vortex persisted in the stratosphere for over 13 weeks, crossed the Pacific within 2 weeks, and hovered above the tip of South America for more than a week. It then followed a 10-week westbound round-the-world journey that could be tracked over 66 000 km until the beginning of April 2020 (Khaykin et al., 2020). The vortex lofted a confined bubble of smoke and moisture to 35 km altitude (Khaykin et al., 2020), an altitude not reached by tropospheric aerosols since the Pinatubo eruption (McCormick et al., 1995; Stenchikov et al., 2021). Aerosol heating was essential in maintaining the structure and providing the loft. In turn, the vortex created a confinement that preserved the embedded smoke cloud from being rapidly diluted within the environment (Lestrelin et al., 2021). Allen et al. (2020) emphasized that the peculiar motion was related to the steady rise in plume potential temperature of about 8 in January and 6 K d⁻¹ in February, due to local heating by smoke absorption of solar radiation. This heating resulted in an anticyclonic (i.e., counterclockwise) circulation, with winds rotating around the plume at about 15 m s⁻¹ (Kablick et al., 2020). Kablick et al. (2020) highlighted the importance of this detection. This is the first evidence of smoke causing changes to winds in the stratosphere and opens up a whole new vein of scientific research.

The rotating smoke field stayed close to the southernmost tip of South America (see also Fig. 1d) for 10–12 d, until the beginning of February 2020, and ascended by about 500 m d⁻¹. During the steady and monotonic ascent, the smoke reached heights with dominating easterly winds. The center of the vortex was closest to Punta Arenas on 29–30 January 2020.

Figure 2a shows the disk-like structure southwest of the two lidar stations at Punta Arenas and Río Grande. The Sentinel-5 UV aerosol index from 340 and 388 nm is shown for 26 January 2020 (Sentinel-5, 2021). Bright blue colors indicate the presence of absorbing aerosol. The smoke plume was well captured by CALIPSO lidar observation on 25 January 2020, as shown in Fig. 2b (dashed curve in Fig. 2a shows the CALIOP track). The smoke vortex was detected between 19 and 26 km height at latitudes from 52 to 65^∘ S.

Figure 2(a) Sentinel-5 aerosol index (340–380 nm), showing the location of the rotating and ascending smoke vortex southwest of the southernmost tip of South America on 26 January 2020 from 00:00 to 20:00 UTC (Sentinel-5, 2021). Indicated are the two ground-based lidar stations at Punta Arenas and Río Grande (white triangles). The white dashed line indicates the CALIOP overpass. (b) CALIOP observation along the overpass track on 25 January 2020 between 05:50 and 06:02 UTC, showing smoke plumes above 20 km and the smoke-filled vortex at latitudes south of 52.5^∘ S in terms of the 532 nm attenuated backscatter coefficient. The thin vertical line at 52.3^∘  indicates the shortest distance to the two ground-based lidars (triangles in panel a).

Download

According to the Punta Arenas lidar observations in Fig. 3, parts of the rotating smoke field covered southern Chile from 24–31 January 2020. The flat base of the dome-like plume structure (see Fig. 2b) ascended from 19 km on 24 January to 22 km on 30 January and thus with a speed of about 500 m d⁻¹ over the Polly lidar site. This is in agreement with the CALIOP observations shown in Khaykin et al. (2020), who found that the initial ascent rate was at 0.45 km d⁻¹, with an average rate of 0.2 km d⁻¹ for the 3-month period during which this smoke field existed.

Figure 3The rotating and ascending smoke vortex, as observed with ground-based lidar at Punta Arenas from 24–31 January 2020. Shown is the range-corrected signal at 1064 nm wavelength in arbitrary units in a logarithmic color scale. The tropopause is indicated by the white dashed line. Tropospheric clouds and day and night signal background changes caused the many interruptions in the high-quality observations of the stratospheric smoke.

Download

We performed a detailed analysis of the Punta Arenas lidar observations on 29 January 2020 when the vortex center was closest to the Polly lidar site. Table 1 summarizes the essential optical and microphysical properties, as discussed in detail by Ansmann et al. (2021a). The very large 532 nm lidar ratio of 106 sr that is also supported by the AERONET (AERONET, 2021) measurements used in Fig. 8 points to strongly light-absorbing smoke particles. The mass, volume, surface area, and number concentrations indicate a strong perturbation of the usually rather clean stratosphere. At unperturbed conditions, particle extinction coefficients are of the order of 0.1 Mm⁻¹ (Sakai et al., 2016), particle surface area concentrations are in the range of 0.5–1 µm² cm⁻³ (Hofmann and Solomon, 1989; Deshler et al., 2003), and particle number concentrations are around 5–10 cm⁻³. Now we found the particle number concentrations up to 500 cm⁻³, surface area concentrations of >100 µm² cm⁻³​​​​​​​, high mass concentrations (>10 µg cm⁻³), and 532 nm extinction coefficients of 75 Mm⁻¹ in the layer from the 20–26 km height. The smoke volume size distributions for the Australian smoke, presented in Ansmann et al. (2021b) and Ohneiser et al. (2021), showed a pronounced accumulation mode (particles with radius mostly from 150–700 nm) with an effective radius of 280–300 nm. The revealed refractive index values, and the low single scattering albedo of close to 0.8 at all three wavelengths (not listed in the table) again points to strongly light-absorbing particles.

Table 1Layer mean particle optical (532 nm) properties and estimated microphysical properties of the rotating and ascending smoke disk on 29 January 2020, when the center of the vortex was close to Punta Arenas (Ohneiser et al., 2020; Ansmann et al., 2021a). Besides values obtained by the conversion of extinction coefficients into microphysical properties (Ansmann et al., 2021a), the lidar inversion method was applied to Polly multiwavelength backscatter and extinction observations on 26 January (Veselovskii et al., 2002) to add values for the effective radius, refractive index, and single scattering albedo. Uncertainties are of the order of 15 %–30 % in the optical properties and 20 %–50 % in the microphysical properties. We used values of ±0.1 for the real part of the refractive index, within an order of magnitude for the imaginary part, and ±0.05 for the single scattering albedo. The number concentration n₅₀ considers particles with radius >50 nm only.

Download Print Version | Download XLSX

As discussed by Ohneiser et al. (2020), the lidar ratios for Australian smoke at 532 nm were considerably higher than the ones for Canadian smoke observed in August 2017 of 60–80 sr (Haarig et al., 2018). The difference is probably related to the different burning material. Australian's forests mainly consist of eucalyptus trees, whereas western Canadian tree types are predominantly pine, fir, aspen, and cedar.

3.2 Case studies from January 2020: optical fingerprints of aged wildfire smoke

To provide an overview of the basic lidar products, we begin with two case studies presented in Figs. 4 and 5. The Polly lidar monitored the smoke layers in terms of particle backscatter coefficients at three wavelengths and the particle extinction coefficient, lidar ratio, and depolarization ratio at two wavelengths. This set of optical properties allows us to characterize the smoke in large detail. However, the Polly instrument is optimized for tropospheric aerosol monitoring up to 10–12 km height, so that the extinction and lidar-ratio profiling was already at its limit for heights above 14 km, especially in the case of the measurement wavelengths of 355 and 387 nm for which the signal attenuation by Rayleigh scattering is strong and signal quality therefore quickly becomes low with height. On the other hand, the powerful CORAL instrument was optimized for stratospheric and lower mesospheric observations up to 90 km height and thus shows good performance at greater heights, as demonstrated in Fig. 5. Unfortunately, the minimum measurement height of CORAL is about 15 km, which makes direct comparisons with the Polly observations problematic.

Figure 4Profiles of smoke optical properties measured with Polly over Punta Arenas on 8 January 2020 between 06:45 and 07:42 UTC. (a) Particle backscatter coefficient at three wavelengths, (b) smoke extinction coefficient, (c) smoke extinction-to-backscatter ratio (lidar ratio), and (d) particle linear depolarization ratio. Error bars show the estimated uncertainties. Profiles of backscatter (in panel a) and depolarization ratio (in panel d) were vertically smoothed with 500–700 m signal smoothing length, and the profiles of extinction coefficient (in panel b) and lidar ratio (in panel c) are obtained with a regression window length of 2–2.5 km. The dotted part of the 355 nm extinction profile indicates that the basic Raman lidar signals were too noisy to allow an accurate determination of the 355 nm extinction coefficient at heights above the layer center.

Download

Figure 5Profiles of smoke optical properties measured with Polly (P; closed circles) over Punta Arenas and with CORAL (C; open olive circles) over Río Grande on 26 January 2020 between 04:27–06:18 (Polly) and 06:14–08:02 UTC (CORAL). (a) Particle backscatter coefficient at three wavelengths, (b) smoke extinction coefficient, (c) smoke extinction-to-backscatter ratio (lidar ratio), and (d) particle linear depolarization ratio (panels b–d at 532 nm). Error bars show the estimated uncertainties. Smoothing and regression window lengths are 500 m (Polly, backscatter, and depolarization) and 2 km (Polly, extinction, and lidar ratio). The 2 km smoothing and regression window lengths are used in the case of all CORAL profiles.

Download

The measurements shown in Fig. 4 were performed when the first strong smoke plumes reached southern South America (Ohneiser et al., 2020). The maximum backscatter and extinction values were found at around 13–14.5 km height, and the layer base and top were at 12.7 and 14.7 km, respectively. The maximum extinction coefficients in the center of the smoke layer were around 70 Mm⁻¹ at 532 nm and 120 Mm⁻¹ at 355 nm. For the applied signal smoothing length of 2000–2500 m, the lidar ratios are fully trustworthy around the layer center only, as shown in Fig. 4. The 532 nm AOT was 0.08. The enhanced particle depolarization ratios of about 0.2 (355 nm) and 0.15 (532 nm) are clear indications for wildfire smoke reaching the stratosphere via fast pyroCb lofting (Haarig et al., 2018; Ohneiser et al., 2020).

In Fig. 5, the Polly and CORAL measurements are compared. The lidar observations on 26 January 2020 were performed close to the edge of the rotating and ascending smoke-filled vortex. The two lidars (230 km apart from each other) saw different parts of the horizontally inhomogeneous smoke field. This explains the differences in the 532 nm backscatter, extinction, and lidar ratio values. Stratospheric AOTs at 532 nm were 0.16 (Punta Arenas) and close to 0.05 (Río Grande).

Based on Figs. 4 and 5, we can summarize the following optical fingerprints of aged wildfire smoke. This aerosol type shows a clear wavelength dependence of the particle backscatter coefficient and a less pronounced spectral dependence of the extinction coefficient. Consequently, the extinction-to-backscatter ratio shows an inverse spectral behavior. The 532 nm lidar ratios are considerably larger than the 355 nm lidar ratios. Furthermore, the 532 nm lidar ratios typically exceed 70 sr and can be even larger than 100 sr. The CORAL observations yielded lidar ratios in the range from 65–85 sr on 5 different days in January and February 2020, but there was also a case with 100 sr on 16 January 2020. As mentioned already, the depolarization ratio of, typically, 0.15–0.2 at 532 nm and 0.2–0.25 at 355 nm are clear indications for smoke that was lofted into the stratosphere via strong pyroCb convection, i.e., within a rather short time period so that the morphological properties of the injected irregularly shaped carbonaceous particles remained widely unchanged when the particles entered the dry stratosphere. Only a small fraction of the smoke serves as cloud condensation nuclei and/or ice-nucleating particles, and the lack of significant precipitation produced in these cloud towers implies that only a small fraction of the smoke is scavenged so that most of it is exhausted through the anvil to the upper troposphere and lower stratosphere (Rosenfeld et al., 2007).

3.3 The decay phase: observation over 2 years (2020–2021)

Figure 6 provides an overview of the stratospheric perturbation from January 2020 to November 2021, as observed with the two ground-based lidars. One set of lidar products per day is considered. Gaps in the data time series are caused by fog and low-cloud events and by instrumental problems (long gap in July 2021). In the case of the CORAL observations, the laser stopped working properly in January 2021. Fortunately, these observations cover the main phase of the major stratospheric perturbation. Because of the designed lidar configuration, CORAL provided data for heights >15 km only.

Figure 6(a) Overview of Polly observations of the UTLS smoke layers (colored bars from layer base to top; one bar per day) from 15 November 2019 to 15 November 2021. Observational gaps between bars are caused by periods with opaque clouds or instrumental problems. The colors in each bar indicate height segments with different 532 nm extinction coefficient levels (see the legend in the panel). Black oval circles highlight periods with frequently detected ascending layers. Panel (b) is the same as panel (a) except for CORAL observations at 532 nm with a minimum measurement height of 14 km. (c) Smoke layer AOT (C denotes CORAL; P denotes Polly) at 355 and 532 nm, calculated from the daily profiles of the particle backscatter coefficient and multiplied by a lidar ratio of 55 sr (355 nm) and 85 sr (532 nm). The Polly instrument measures the total UTLS smoke AOT and the CORAL system measures the AOT between 15 and 28 km. (d) Smoke contribution to total AOT, which includes contributions by the stratospheric background aerosol (dotted lines; Sakai et al., 2016​​​​​​​), by Ulawun volcanic aerosol (Ul), and by Australian smoke reaching the stratosphere from September to December 2019 (smoke). The Ulawun and 2019 smoke contributions decreased with time (dashed curves; see the text for more details). Fluctuations in the smoke fraction around 0 that are visible until May 2020 are caused by measurement noise.

Download

In Fig. 6a and b, layer height, depth, and particle light extinction information at 355 and 532 nm is given. We determined the layer base and top heights by visual inspection. In the case of Polly, the mean height profile of the 1064 nm backscatter coefficient was used, which is rather sensitive to particle backscattering. The base height is defined as the altitude at which the 1064 nm backscatter coefficient started to increase in the middle to upper troposphere after a minimum in the lower to middle free troposphere. The layer top was set to the height where the total-to-Rayleigh backscatter ratio at 1064 nm dropped below a value of 1.1. A similar approach was applied to the elastic backscatter signals, measured with CORAL at 532 nm, to determine the smoke layer top. The vertical bars in Figs. 6a and b are colored to distinguish different levels of the aerosol loading expressed in terms of the particle extinction coefficient. The backscatter coefficients at 532 nm were multiplied by a lidar ratio of 85 sr to obtain the extinction coefficients. A lidar ratio of 85 sr represents the mean extinction-to-backscatter relationship of the entire Australian smoke data set well. The tropopause is given in Fig. 6a to indicate that smoke was also present in the upper troposphere and influenced cirrus formation (Knopf et al., 2018; Ansmann et al., 2021a; Engelmann et al., 2021). After the injection of the first aerosol layers in January 2020, the smoke layers typically extended from 8–10 km height up to 21–25 km for the next 2 years. On average, the layer base and top heights were at 9.8 and 22.5 km, respectively. No long-term trend in the layer geometrical properties is visible. The optically densest part was found between 12 and 18 km height. The light extinction values slowly decreased with time (from red to orange to blue Fig. 6a and b). The periods with isolated, probably rotating and ascending, smoke-filled vortices are highlighted by big oval circles. Isolated smoke features vanished at the beginning of May 2020 (4 months after the ANYSO smoke injection period).

Figure 6c presents the time series of 532 nm AOT measured at Punta Arenas over the 2-year period. The AOT values were obtained by integrating the extinction values, i.e., the backscatter values multiplied with 85 sr, from the layer base to the top. The directly determined extinction coefficients (from the Raman signal profiles) were too noisy, and in many cases even not available for the upper part the smoke layers, and thus did not permit a trustworthy computation of the AOT time series for the 2-year period. The 355 nm AOT values in Fig. 6c were obtained from the 355 nm backscatter coefficients multiplied with a lidar ratio of 55 sr. Most of the variability in the 355 nm AOTs is caused by signal noise. The almost noise-free 532 nm AOT from the CORAL observations between 15 km and smoke layer top is given in Fig. 6c as well. A strong increase in the 532 nm AOT from November–December 2019 to January 2020 and large variability in AOT were found in January 2020 with the two lidars. The AOT reached values up to 1.0 at 532 nm in mid-January over Punta Arenas. The day-by-day AOT variability significantly decreased in February 2020. Since then, a slow and persistent decrease of the stratospheric perturbation has been visible. The AOT values at 532 nm decreased to 0.03–0.06 (February–June 2020), 0.02–0.05​​​​​​​ (July–December 2020), and were about 0.015–0.03 in 2021. The AOT for undisturbed, clean stratospheric background conditions for the height range from the tropopause to 30 km height is about 0.005.

The contribution of Australian smoke to the total AOT is shown in Fig. 6d. The total AOT also includes contributions by Australian smoke reaching the stratosphere from September–December 2019, Ulawun volcanic aerosol, and background particles. In order to determine the smoke fraction caused by the ANYSO smoke injections, we need thus to consider the natural stratospheric background conditions and further contributions by aerosols injected in 2019 (Kloss et al., 2021a, b). In Fig. 6c, different stratospheric background levels for the wavelength of 532 nm are shown as dotted and dashed lines. In November–December 2019, UTLS AOTs around 0.01 (8–30 km height) and stratospheric AOTs around 0.004 (15–30 km height) were observed with the lidars at Punta Arenas and Río Grande, respectively, and in reasonable agreement with results from passive remote sensing for heights >15 km (Kloss et al., 2021b). According to Sakai et al. (2016), who analyzed Japanese lidar observations at Lauder, New Zealand, from 1992 to 2015, the minimum stratospheric background level is reflected in a 532 nm AOT of 0.0025 for the height range from 15–30 km height. The AOT is obtained from the measured minimum column-integrated backscatter coefficient multiplied by a lidar ratio for sulfate aerosol of 50 sr. This minimum background level (for the 15–30 km height range) is indicated by a dark green line in Fig. 6c. From the New Zealand lidar observations, it can be concluded that the UTLS AOT is roughly 0.005 during clean conditions (light green dotted line in Fig. 6c). This estimate was found to be in agreement with the analysis of satellite observations from 1990 to 2010 during volcanic quiescent times (Solomon et al., 2011).

However, further contributions to the AOT measured in November–December 2019 need to be added. The Ulawun volcano in Papua New Guinea (5^∘ S, 151^∘ E) erupted on 26 June 2019 (injection of 0.14 Tg SO₂ into the stratosphere) and on 3–4 August 2019 (injection of 0.3 Tg SO₂; Kloss et al., 2021a) and most probably caused a stratospheric AOT at 532 nm of around 0.005 at the end of September 2019 and of 0.0025 2–3 months later. We estimated the Ulawun-related AOT by comparing SO₂ emissions and the resulting stratospheric sulfate AOT for the Raikoke, Sarychev, and Ulawun eruptions (Haywood et al., 2010; Kloss et al., 2021a, b; Ohneiser et al., 2021). For the height range >15 km, we arbitrarily assumed a volcanic AOT contribution of 0.001 for the November–December 2019 period. To realistically consider the decrease in the Ulawun-related disturbance of the stratospheric aerosol conditions, we assumed a decay time of 180 d.

As a second additional contribution Kloss et al. (2021b) identified Australian smoke emitted from September to December 2019. To match the measured AOT values of 0.005 and 0.01 over Río Grande and Punta Arenas, respectively, this contribution must also have been of the order of 0.0025 (for the UTLS height range) and 0.001 (for the height range >15 km). Here, we assume a decay time of 1 year (365 d). The sum of all three contributions (constant minimum background level plus volcanic AOT plus Australian smoke for the September to December period) is shown as dashed lines in Fig. 6c. By considering these enhanced time-dependent background levels, we can conclude that the ANYSO smoke caused roughly a factor of 4 higher stratospheric AOT values over the midlatitudes in the Southern Hemisphere for more than 1 year. Accordingly, the ANYSO smoke fraction for the UTLS height range, shown in Fig. 6d, was about 80 % during 2020 and 70 %–80 % in 2021. On the contrary, the Ulawun-related AOT fraction was about 25 % in November–December 2019, negligible in January 2020, about 5 % in February–April 2020, and <3 % as of June 2020.

In Fig. 7, the stratospheric aerosol perturbation in 2020 and 2021 is presented as a function of height. Besides the January 2020 mean extinction profile (in red), the February–June 2020 mean profile (orange), and further three half-year mean extinction profiles (blue, green, and black) are shown together with the pre-ANYSO 532 nm extinction profile (November–December 2019 in dark gray). The daily height profiles of the particle backscatter coefficient at 532 nm were multiplied by a representative smoke lidar ratio of 85 sr and then separately averaged for the different time periods. In the case of the November and December 2019 backscatter height profiles (pre-ANYSO profiles), we applied a lidar ratio of 60 sr to consider the contribution of volcanic aerosol (lidar ratio of 45 sr) and stratospheric background aerosol (lidar ratio of 50 sr) in addition to the September–December 2019 smoke from Australia (lidar ratio of 85 sr). The dashed lines in Fig. 7a above approximately 25 km height are calculated by applying the Klett method to the strong elastic backscatter signals (Fernald, 1984), as described in Sect. 2.1. The stratospheric minimum background extinction levels (Sakai et al., 2016) are given as well in Fig. 7 as a thin solid dotted gray line.

Figure 7January 2020 mean (red), February–June 2020 mean (orange), and half-year mean profiles of smoke extinction coefficient at 532 nm. (a) Polly observation (1–30 km). (b) CORAL (15-30 km). The particle extinction coefficients were retrieved by multiplying the backscatter coefficient profiles by 85 sr. All solid profiles are average profiles calculated with the Raman lidar method. The bars indicate 1 standard deviation caused by atmospheric variability. Dotted parts indicate bars ending at values <0.03 Mm⁻¹, i.e., outside the frame. The dashed lines in panel (a) were calculated with the Fernald method (as explained in Sect. 2.1). The thin, solid dotted gray lines show stratospheric background values (Sakai et al., 2016). The November–December 2019 mean extinction profile (before the ANYSO event) is given as dark gray curve, here a lidar ratio of 60 sr is used to convert backscatter into extinction values.

Download

Figure 8Smoke layer mean optical properties (daily values) observed from 1 January 2020 to 10 January 2021. (a) Backscatter-related Ångström exponent A_β. (b) Extinction-to-backscatter ratio (lidar ratio) from Polly (P) and CORAL (C) measurements in January–February 2020 and from combined lidar–AERONET observations (since March 2020) by using lidar ratios of 30 (A30), 40 (A and A40), and 50 sr (A50) in the estimation of the tropospheric 532 nm AOT from lidar observations (see the text for a further explanation). Stratospheric lidar ratios at 355 nm (circles) and 1064 nm (squares) are shown for a lidar ratio of 40 sr in the tropospheric AOT retrieval. (c) Particle linear depolarization ratio (PLDR). The scatter in the data is mainly caused by signal noise and to a minor degree by atmospheric variability.

Download

The extinction profile in November–December 2019 indicates already significantly enhanced aerosol pollution levels that have been caused by the Australian fires since September 2019 (Kloss et al., 2021b) and volcanic sulfate aerosol (Ulawun eruption; Kloss et al., 2021a). Then, in January 2020, the aerosol extinction coefficient increased by 1–2 orders of magnitude compared to the November–December 2019 extinction levels in the height range from 8–25 km and more than 2 orders of magnitude with respect to the minimum perturbation (Sakai et al., 2016). The structure of the January 2020 profile is dominated by the arrival of several single, very dense smoke layers, such as the rotating and ascending smoke-filled vortex in the last January week. Monthly mean extinction values of 10–30 Mm⁻¹ were found in January 2020. The following half-year mean extinction profiles showed much lower values. However, these profiles indicate quite a slow decay of the perturbation from February 2020 to November 2021. In the second half of 2021, the extinction values were still a factor of 5–10 higher than the extinction levels representing a minimum stratospheric aerosol load (Sakai et al., 2016). Note also that the extinction levels were also clearly enhanced between 26 and 30 km height in 2020–2021.

3.4 Smoke intensive parameters

Figure 8 shows time series of smoke-intensive properties for the year 2020. The backscatter-related Ångström exponents (defined in Sect. 2.1), the extinction-to-backscatter ratio, and the particle linear depolarization ratio enable a good optical characterization of aged wildfire smoke. The strong variability in the data, especially in the case of the backscatter-related Ångström exponent for the short wavelength range (355–532 nm), is caused by signal noise rather than changes in the microphysical and chemical properties of the smoke particles. Mean values of the Ångström exponents and lidar ratios shown in Fig. 8 are given in Table 2. The statistics consider 326 values for each Ångström exponent, 7–10 values for each of the lidar ratios, except for the Polly–AERONET 532 nm lidar ratio (19 values). The PLDR statistics are based on 14 (355 nm) and 17 values (532 nm).

Table 2Mean values and standard deviations of the Ångström exponents (backscatter related, A_β, lidar ratio related, A_LR, and extinction related, A_σ) and lidar ratios (LR) computed from all stratospheric values shown in Fig 8. Lidar ratio index includes the following: A – AERONET–Polly retrieval; P – Polly observation; C – CORAL observation; all – mean values by considering all determined and retrieved LR values. The standard deviations include atmospheric variability and dominating retrieval uncertainties. A further explanation is given in the text.

Download Print Version | Download XLSX

As can be seen in Fig. 8a, the Ångström exponents for the 355–1064 nm and the 532–1064 nm wavelength ranges accumulated between 1 and 2. Such values were already observed in the stratosphere for aged Canadian smoke observed over central Europe in August 2017 (Haarig et al., 2018) and for aged Siberian smoke in the High Arctic (Ohneiser et al., 2021) and points to a pronounced accumulation-mode-dominated particle size distribution and the absence of a particle coarse mode. The less noisy Ångström exponent values for the long wavelength range (532–1064 nm) accumulate at around 1.5 in January 2020 and are around 1.8 in the second half of 2020. This increase reflects a shift in the size distribution towards smaller particles. Larger particles may have been removed by sedimentation processes.

Lidar ratios at 355 and 532 nm could be measured directly with the two lidars in January and February 2020 only (see Fig. 8b). The lidar ratios measured with Polly in January 2020 (blue and light green) are also given in Table 1 in Ohneiser et al. (2020). The values range from 50–100 sr (355 nm) and 75–115 sr (532 nm). The CORAL values for 532 nm in Fig. 8b accumulate around 80 sr. Since March 2020, the Raman signal profiles have been too noisy and the smoke extinction coefficients too low and have no longer allowed a lidar ratio retrieval. In order to obtain some snapshot-like estimates for the smoke lidar ratio at 355 and 532 nm, and even for the IR wavelength of 1064 nm, we combined stratospheric AOT estimates from AERONET sun photometer observations at 380, 500, and 1020 nm with respective lidar-derived stratospheric-column-integrated backscatter values at 355, 532, and 1064 nm. In this approach, we estimated the stratospheric AOT from the total (tropospheric + stratospheric) AOT measured the AERONET photometer at Punta Arenas during favorable, stable, and temporally constant aerosol conditions. We subtracted the tropospheric AOT contribution from these AERONET AOT values. The tropospheric AOT was estimated by using the tropospheric backscatter and extinction profiles (from the surface to the tropopause) obtained with the Fernald method that was applied to the elastic backscatter signal profiles (Fernald, 1984). We assumed reasonable tropospheric lidar ratios of 30, 40, and 50 sr for a mixture of continental and marine aerosols in the Fernald data analysis to obtain the tropospheric AOTs at 355, 532, and 1064 nm. We combined the lidar observations with the near-range and far-range telescope (receiver units) to obtain the full extinction profile from close to the ground up to the tropopause. In order to transfer the total (tropospheric + stratospheric) AERONET AOT values at 380, 500, and 1020 nm to the ones for the lidar wavelengths at 355, 532, and 1064 nm, we used climatologically mean tropospheric Ångström values measured over Punta Arenas during undisturbed stratospheric aerosol conditions from January to July 2019. The ratio of the stratospheric AOT (from the combined AERONET and Polly observations) to the respective stratospheric column backscatter coefficient is shown in Fig. 8b. The label 532-A40 indicates, e.g., the use of AERONET observations (index A) and the use of a lidar ratio of 40 sr (index 40) in the retrieval of the tropospheric AOT at 532 nm. In Fig. 8b, we only show smoke lidar ratio solutions for 355 and 1064 nm for the tropospheric lidar ratio assumption of 40 sr to avoid overloading the figure with too many results.

One should emphasize that these retrieved stratospheric lidar ratios must be regarded as rough estimates. In this combined AERONET–Polly approach, we ignore an underestimation of the stratospheric smoke AOT from AERONET sun photometer observations up to 20 %–30 %, as discussed by Ansmann et al. (2018). Furthermore, measurements with lidar and the photometer did not exactly cover the same time periods and thus may have been conducted at slightly different boundary layer, free tropospheric, and stratospheric aerosol conditions. Thus, the shown lidar ratios provide only a very general view on the spectral behavior of the smoke lidar ratios. Nevertheless, the findings are in line with previous observations (Haarig et al., 2018). When averaging all lidar ratios obtained with the different methods, the resulting lidar ratios are 69, 91, and 120 sr for 355, 532, and 1064 nm, respectively.

Table 2 summarizes the observations in Fig. 8a and b. With the mean lidar ratios of 69 and 91 sr, the lidar-ratio-related Ångström exponent A_LR was calculated and also the extinction-related Ångström exponent A_σ with $A_{\mathit{\sigma }}=A_{\mathrm{LR}}+A_{\mathit{\beta }}=\mathrm{0.64}$. The low values of A_σ and the negative value of A_LR are clear optical fingerprints of wildfire smoke (Ohneiser et al., 2020; Haarig et al., 2018).

Figure 8c shows the evolution of the particle linear depolarization ratio (PLDR) at 355 and 532 nm. High-quality depolarization ratio observations were frequently only possible in the lower half of the smoke layer. The shown values may thus not be at all representative for the entire layer (from base to top). Initially, the PLDR showed values close to 20 % for fresh smoke in the stratosphere. In the first 2 months, the smoke PLDR was very variable, with values from almost 0 % to 20 %. After 2 months, the PLDR decreases to values below 5 %. After 3 months, only a few useful measurements were possible, and the PLDR values were below 5 % for both wavelengths. The smoke particles obviously became rather compact and spherical in shape within a short time of about 4–6 weeks. The depolarization ratios are further discussed in the next section.

Figure 9Map showing the ground-based lidar stations at Punta Arenas (ANYSO smoke observations in 2020/2021), in the High Arctic (Polarstern, SILBE smoke observations in 2019/2020), in Europe (PNE smoke observations in 2017/2018). In addition, the ozonesonde stations at Lauder, New Zealand, at the Neumayer Station, and at the South Pole are indicated.

3.5 Comparison of three major stratospheric smoke events: ANYSO (2020) vs. PNE (2017) vs. SILBE (2019)

In this section, the decay behavior of the stratospheric perturbation caused by the Australian fire smoke is discussed in more detail. We use the opportunity to compare the ANYSO event with respective observations of two major smoke events occurring in the Northern Hemisphere in 2017 and 2019–2020 (see Fig. 9). The record-breaking Canadian fire storm (PNE, 2017)​​​​​​​ was already mentioned in Sect. 3.1. Strong smoke injection into the UTLS regime occurred during a cluster of about five pyroCb pulses over British Columbia, Canada, on 12–13 August 2017 (Peterson et al., 2018, 2021). Lidar network observations of aged smoke after long-range transport over weeks to months were conducted over Europe (indicated by a black circle in Fig. 9) from August 2017 to January 2018 (Baars et al., 2019). We further include lidar observations of the decay phase of a dense stratospheric smoke layer observed over the High Arctic with a Polly lidar aboard the ice breaker Polarstern. These observations, from October 2019 to May 2020, were performed in the framework of the MOSAiC (Multidisciplinary drifting Observatory for the Study of Arctic Climate) expedition (Engelmann et al., 2021; Ohneiser et al., 2021). Extremely strong fires occurred over central and eastern Siberia, north and northeast of Lake Baikal (SIberian Lake Baikal Event, SILBE), from mid-July to mid-August 2019. The fire smoke reached the UTLS height range, most probably by so-called self-lofting processes, within 2–7 d after injection into the lower troposphere (Ohneiser et al., 2021). The pyroCb activity was absent over the huge Siberian fire region in July and August 2019. Since there was an injection of Siberian smoke into the UTLS height range about 2500–3500 km south of the High Arctic, the plumes had to travel a short distance only with the prevailing westerly to southerly winds.

In Fig. 10, the three major stratospheric smoke events are compared regarding their geometrical and optical properties. As can be seen, the vertical extent of the Canadian smoke layers was of the order of 1–4 km, and these smoke layers were clearly observable over Europe with lidar up to 150 d after injection only. The Siberian smoke formed a 7–10 km thick layer and was monitored from day 80 up to about 300 d after injection. From December 2019 to May 2020, the smoke was trapped around the North Pole under the influence of a rather strong polar vortex. The stratospheric perturbation could be observed until the strong polar vortex collapsed in May 2020 so that the smoke became dispersed over a larger region towards the midlatitudes in the late spring and summer of 2020. The Australian fires were by far the strongest and produced a 10–15 km thick layer. These smoke layers were clearly detectable even 2 years after injection. Unfortunately, we had to stop the DACAPO-PESO campaign and thus our Punta Arenas smoke observation at the end of November 2021. The CORAL system, on the other hand, has been measuring again since December 2021.

Figure 10Comparison of geometrical and optical properties of three stratospheric fire smoke events measured with different Polly instruments. Day 0 is the 12 August 2017 (PNE), 23 July 2019 (SILBE), and 31 December 2019 (ANYSO). Canadian (PNE) and Siberian (SILBE) fire smoke data are taken from Baars et al. (2019) and Ohneiser et al. (2021), respectively. (a) Overview of Polly observations of UTLS smoke layers (colored bars from layer base to top) in the days after smoke injection into the UTLS height regime. (b) The 532 nm AOT times series for the different wildfire events; straight lines indicate the decay of the perturbations according to the given e-folding decay times in months (M). In the case of PNE and ANYSO observations, AOT for heights >13 km is used. (c) Particle depolarization ratios (layer mean values at 532 nm) for the different fire events.

Download

The layer base height of the ANYSO and SILBE smoke layers was always close to or slightly below the tropopause (mostly at 8–10 km height). In contrast, the base height of the PNE smoke layer was typically several kilometers above the tropopause. The top height was mostly around 20 (PNE), 18 (SILBE), and 23–24 km (ANYSO). The different top heights had a sensitive impact on the decay of the stratospheric perturbation described in terms of e-folding decay times.

Table 3Comparison of the three wildfire smoke events (ANYSO – Australian fires; PNE – Canadian fires; SILBE – Siberian fires) in terms of mean values and SD of smoke layer geometrical and optical properties and perturbation decay time. The observations were performed over Europe (PNE), Punta Arenas (ANYSO), and the North Pole region (SILBE). For these sites and regions (see Fig. 9), the mean tropopause height and SD and, for Europe, the typical tropopause height range are given as well.

Download Print Version | Download XLSX

In Fig. 10b, the AOT data sets of the three wildfire events are compared, and the decay behavior of the different stratospheric perturbations are presented. As expected, the PNE AOTs were, on average, 3 to 4 times lower than the ANYSO AOTs (see the discussion in Sect. 3.1). In order to see a clear reduction in the AOT with time, the shown PNE and ANYSO AOTs were computed from the smoke extinction coefficients between 13 and 30 km height and thus for a height range several kilometers above the tropopause. The selection of the 13 km level as base height was necessary in the case of the ANYSO AOTs to avoid the impact of new aerosol plumes in 2020 and 2021 that regularly appeared in the tropopause region on the decay time computations. The decay behavior is indicated by straight lines in Fig. 10b. The respective e-folding decay times are given as numbers in Fig. 10b and were computed by using the AOT time series from day 40 to 160 (PNE) and from day 60 to 560 (ANYSO). The first 40–60 d (after smoke injection) were excluded because they showed a rather strong AOT variability and no clear trend.

The MOSAiC observations started about 80 d after the injection of the smoke into the UTLS regime so that a removal of strongly varying AOT values in the decay time calculations was not necessary in the case of the Siberian smoke. Furthermore, the entire smoke layer, from base to top, was considered in the AOT calculations. The impact of tropospheric aerosol sources on the smoke AOT evolution over the High Arctic was found to be negligible. In terms of the total AOT (from base to top layer), the ANYSO and SILBE smoke events were similar regarding the injected smoke amount and corresponding stratospheric AOTs.

Strong differences in the decay behavior of the stratospheric perturbation in the Northern and Southern hemispheres were found. A very large value of the e-folding decay time of 20 ± 2 months was obtained from the ANYSO AOT data set and a quite short decay time of 8 ± 1 months in the case of PNE (2017)​​​​​​​. If we remove the impact of the stratospheric background aerosol and Ulawun volcanic sulfate aerosol (shown in Fig. 6c), the ANYSO AOT decay time decreases to 19 ± 1 months. Several processes influence the decay time. Besides the sedimentation of particles and self-lofting of the absorbing smoke particle, horizontal (meridional) dispersion, and even the growth of the smoke particles by water uptake and condensation of gases on the particles have to be taken into account. Larger particles produce larger optical effects and thus lead to an increase in AOT.

The top height of the smoke layer plays an important role as well. The ANYSO smoke was distributed over the entire lower stratosphere up to heights around 24 km. And this comparably large top height in combination with particle sedimentation was responsible for the e-folding decay time of 19–20 months, as we will further discuss below. The PNE smoke was mostly confined between 14 and 20 km height and thus could be removed quite efficiently (from the stratosphere above 13 km). Furthermore, complex horizontal dispersion features in the Northern Hemisphere have been reported in the case of the Canadian smoke event (Kloss et al., 2019; Pumphrey et al., 2021). The smoke became efficiently distributed towards the tropics and towards the North Pole within a few weeks. On the other hand, a clear detection of the Canadian smoke was only possible until January–February 2018 (up to 5–6 months after injection), so that an accurate decay time could not be computed from such a short time series with quite strongly varying AOT values.

The short decay time for the High Arctic smoke (SILBE, 2019–2020) of 5 months is related to the specific polar meteorological conditions (impact of the strong polar vortex) and specific vertical air mass exchange mechanisms (related to the Brewer–Dobson circulation) and also to the fact that we considered the AOT from the layer base in the upper troposphere to the top at around 18 km height in the computation. Thus, even removal processes in the upper troposphere (e.g., by cirrus evolution and removal of smoke particles by falling ice crystals) had an impact on the decay behavior.

Table 3 summarized the key features for the three events and allows further comparison. It is worthwhile to also compare the ANYSO decay time with decay times of volcanic perturbations found after the Pinatubo eruption in 1991 at midlatitude lidar stations in the Northern and Southern hemispheres. The Pinatubo aerosol layer also reached about 25 km height. In the case of volcanic aerosol, the dominating removal process is sedimentation. Self-lofting and particle growth aspects can be ignored in the analysis of the decay of a volcanic perturbation 1–5 years after the eruption. For the Northern Hemisphere, our long-term lidar observations of the Pinatubo sulfate aerosol at Geesthacht (close to Hamburg) at 53.4^∘ N in northern Germany, yielded decay times of 14–15 months (445 d; Ansmann et al., 1997). Nagai et al. (2010) reported e-folding decay times of 14 and 16.5 months obtained from lidar observations at Tsukuba (near Tokyo, 36.1^∘ N), Japan, and Lauder (45^∘ S), New Zealand, respectively. According to Sekiya et al. (2016), the Pinatubo-related decay time was 12–13-months on a global scale. Sekiya et al. (2016) performed a global-scale modeling constraint to the lidar observations at Tsukuba and Lauder. The shorter Pinatubo decay times are related to the fact that the volcanic sulfate particles showed a characteristic (effective) radius of 0.4–0.5 µm (Ansmann et al., 1997), whereas the Australian smoke particles were much smaller, with effective radii of 0.2–0.3 µm (Ansmann et al., 2021a), so that the sedimentation speed was much smaller. Self-lofting processes and particle growth effects had no impact on the decay behavior a few months after injection when the AOT was mostly <0.05.

In Fig. 10c, we finally compare the observed particle depolarization ratios of the three fire events. As can be seen, in the case of Canadian smoke over Europe (PNE, 2017), the depolarization ratio decreased slowly with time. This points to a slow smoke aging process in the dry stratosphere at low relative humidity conditions and low levels of condensable gases. It took several months before the initially irregularly shaped particles became very compact and, at the end, spherical so that the depolarization ratio approached values <0.05. The aging process seemed to be much faster in the Southern Hemisphere (ANYSO, 2020). A relatively large amount of water vapor and condensable gases was available in the stratosphere so that the smoke particles became spherical in shape within 4–6 weeks after injection. Khaykin et al. (2020) reported a remarkable increase in the stratospheric abundance of the gaseous combustion products which were injected together with the smoke aerosol. Furthermore, the injected mass of water was estimated to be very high (27 ± 10 Tg; about 3 % of the total mass of stratospheric overworld water vapor in the southern extratropics). Unfortunately, it was difficult to determine trustworthy particle depolarization ratios over Punta Arenas from February 2020 onwards. However, as shown in Fig. 8c, the few 355 nm depolarization ratios from March to June 2020 indicate the comparably fast decrease in the depolarization ratio with time and thus fast particle aging. All depolarization values in Fig. 8c were ≤10 % since February 2020.

In strong contrast to these pyroCb-related smoke events, the Siberian smoke (SILBE, 2019) showed a rather different behavior. Rather low depolarization ratio values were found at all sites after the beginning of the MOSAiC campaign in October 2019 (about 80 d after injection into the UTLS height range over Siberia). Lidar measurements of stratospheric smoke layers over Leipzig, Germany, in August 2019 (originating from the SILBE fires, according to backward trajectory analysis) and with CALIOP also revealed low smoke depolarization ratios (Ansmann et al., 2021b).

This strong contrast to the ANYSO and PNE depolarization features again points to the fact that different lofting processes took place. The pyroCb convection was responsible for the fast lofting of smoke towards stratospheric heights (within less than a few hours) in the case of PNE and ANYSO, so that emitted particles retained their irregular shape. On the other hand, a slow ascent over days caused by self-lofting prevailed in the case of SILBE smoke. The aging process could be completed within these few days so that the spherical shapes of the core shell particles dominated. Additional effects can influence the optical properties, especially the lidar ratio, such as fire type, fuel material, and meteorological conditions, as mentioned above and in Ohneiser et al. (2020) and Ansmann et al. (2021b). A significant fraction of the slowly lofted smoke particles may be spherical tarballs forming from the emitted gases at low heights about 3–6 h after injection (China et al., 2013; Sedlacek et al., 2018; Adachi et al., 2019; Yuan et al., 2021).