The lidar system at the MRI in Tsukuba (36.1∘ N, 140.1∘ E) used in this study is an Nd:YAG system operated at a wavelength of 532 nm with capability for both BSR and PDR measurements (Sakai et al., 2016), which has been operated continuously since 2002. We define PDR as S/P, where S and P are the background-subtracted lidar photon counts of the perpendicular (“senkrecht” in German) and parallel components, respectively, with respect to the polarization plane of the emitted laser light. The temporal and height resolutions of the original processed data are 5 min and 7.5 m, respectively. Quality control has been done primarily to flag data points influenced by thick cloud layers. To obtain vertical profiles of BSR and PDR with high signal-to-noise ratios, data were averaged over 150 m and 3 h, with time intervals of 18:00–21:00, 21:00–00:00, 00:00–03:00, and 03:00–06:00 local time (LT) for the use in this paper. BSR data were normalized to unity at 30–33 km of altitude where aerosol backscattering is assumed to be negligible, and PDR values were obtained using the method of Adachi et al. (2001).

The lidar system at Fukuoka (33.55∘ N, 130.36∘ E) used in this study is also an Nd:YAG system operated at a wavelength of 532 nm with PDR measurement capability. This system has been operated manually only during nights under clear-sky and/or non-rainy conditions; during July–September 2018, the system was operated on 11 nights. Vertical profiles were averaged over 900 m and 4 h for each night for the use in this paper. The PDR for Fukuoka is originally defined as S/(P+S), which has been converted to S/P for this paper.

The uncertainties of lidar data discussed here are applicable to both systems. The BSR uncertainties were estimated as follows. The random component was estimated from the photon counts of the backscatter signals at 532 nm after temporal and vertical averaging by assuming Poisson statistics. Other sources of BSR uncertainties (biases) were estimated by assuming the uncertainty of the normalization value of BSR to be 8.5 × 10-3 (Russell et al., 1979, 1982) and that of the extinction-to-backscatter ratio to be 30 sr (Jäger and Hofmann, 1991; Jäger et al., 1995). The total uncertainty of BSR was then estimated to be 2 %–3 %, typically around the tropopause. The PDR uncertainties were estimated from the parallel and perpendicular components of backscatter signals at 532 nm. Other sources of PDR uncertainties (biases) include (1) the uncertainty in calibration of the total depolarization ratio (TDR), due to both particles and air molecules, and (2) the BSR uncertainty. Uncertainty (1) was estimated as follows. In the TDR calibration (Adachi et al., 2001), we subtracted depolarization caused by the lidar system (DEPsys) estimated from the observed TDR and BSR obtained in the altitude region where aerosol backscattering is negligible (i.e. BSR equals unity, and TDR equals the molecular depolarization ratio) or where spherical particles predominate (i.e. in lower-tropospheric water clouds). DEPsys errors result in PDR bias. For example, a DEPsys error of ±0.2 % results in a ±2 % bias in PDR, where BSR = 1.1. Uncertainty (2) arises mainly from our assumption that aerosol backscattering is negligible at 30–33 km of altitude. We also assumed an aerosol extinction-to-backscatter ratio of 50 sr over the whole measurement height range. These assumptions result in errors in BSR and thus PDR. For example, BSR errors of +0.05 and -0.05 result in a bias of -1 % and +3 % in PDR, respectively, where BSR = 1.1 and TDR = 0.7 %. Based on these considerations, we estimate that the total PDR uncertainty (random plus bias errors) is ≤±5 % PDR.

Backward trajectories are calculated with the trajectory model used by Inai (2018) and Inai et al. (2018) as well as the most recent global atmospheric reanalysis dataset by the European Centre for Medium-Range Weather Forecasts (ECMWF), ERA5 (Hersbach et al., 2020), with 37 pressure levels up to 1 hPa and horizontal and temporal resolutions of 0.25∘  × 0.25∘ and 1 h, respectively. ERA5 temperature data in the tropical tropopause layer have been evaluated by Tegtmeier et al. (2020). Lagrangian transport calculations using ERA5 and its predecessor ERA-Interim have been compared by Hoffmann et al. (2019) and Li et al. (2020).

The Copernicus Atmosphere Monitoring Service (CAMS) atmospheric composition reanalysis dataset produced by the ECMWF (Inness et al., 2019) is used to analyse signatures of the ASM anticyclone and its eastward-shedding vortices, with 25 pressure levels up to 1 hPa and horizontal and temporal resolutions of 0.75∘ × 0.75∘ and 3 h, respectively. Carbon monoxide (CO), temperature (T), and geopotential (Φ) data are primarily analysed in this paper. CO is chosen because it is a good tracer for polluted air of surface origin (e.g. Luo et al., 2018). Although CO and ATAL aerosol particles do not necessarily have the same emission sources, CO is a good chemical tracer to indicate the location of the ASM anticyclone. CO data on pressure levels are projected onto isentropic surfaces using temperature data. In the CAMS, the Measurement of Pollution in the Troposphere (MOPITT) thermal infrared (TIR) satellite total-column CO data are assimilated, but Microwave Limb Sounder (MLS) and Infrared Atmospheric Sounding Interferometer (IASI) CO data are not. CAMS CO data are originally provided as mass mixing ratios (kg kg-1), which are converted to volume mixing ratios (ppbv) for this study. It is noted that a quick comparison (not shown) with MLS version 4.2 level 2 CO data (Santee et al., 2017; Livesey et al., 2020) at a 400 K isentropic surface (in the form of a longitude–time diagram like the one in Sect. 3.2) shows that CAMS CO data are roughly ∼ 10 ppbv greater than MLS CO over Japan during August–September 2018, but it also shows that eastward extension signals coming over Japan agree fairly well qualitatively within the differences in spatio-temporal sampling of the two datasets. The CAMS dataset also includes different types of aerosol particles, but they are not included in this study because relevant variables such as aerosol BSR and NH4NO3 concentration are not included. The Montgomery streamfunction (MSF), defined as MSF =cpT+Φ, where cp is the specific heat of dry air at constant pressure, in isentropic coordinates corresponds to geopotential (height) in isobaric coordinates (e.g. Popovic and Plumb, 2001; Santee et al., 2017; Amemiya and Sato, 2018; Salby, 1996) and is thus a good dynamical indicator of the ASM anticyclone. Potential vorticity (PV) on isentropic surfaces (e.g. at 360–380 K) is often used as a dynamical tracer in studies of the ASM anticyclone (e.g. Popovic and Plumb, 2001; Garny and Randel, 2013; Ploeger et al., 2015; Amemiya and Sato, 2018); however, PV at and above 400 K (the isentropic surface we will focus on in Sect. 3.2) is not very useful to analyse the ASM anticyclone boundary. Thus, we will analyse MSF at the 400 K surface calculated from CAMS data.

MLS version 4.2 level 2 water vapour data (Santee et al., 2017; Livesey et al., 2020) are analysed because water vapour is also a good tracer of the ASM anticyclone. We use MLS data rather than CAMS data for lower-stratospheric water vapour because MLS data have been well validated with e.g. balloon-borne frost-point hygrometers (e.g. Hurst et al., 2016; Fujiwara et al., 2010; Vömel et al., 2007), while reanalysis water vapour data are in general less reliable in the lower stratosphere (e.g. Davis et al., 2017). We found (not shown) that CAMS water vapour volume mixing ratio data (converted from the original specific humidity data) are greater than MLS data at the 400 K isentropic surface over Japan during July–September 2018 (e.g. the differences are roughly ∼ 2 ppmv for the wet signals around the longitudes of Japan in August 2018).

The possible influence of volcanic eruptions and wildfire events is investigated using two satellite aerosol particle datasets, one providing vertical extinction profile data at 675 nm from the Ozone Mapping and Profiler Suite (OMPS) Limb Profiler (LP), level 2 version 1.5 (Chen et al., 2018), and the other attenuated scattering ratio data from the CALIOP on board the Cloud–Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) satellite (Thomason et al., 2007; Winker et al., 2007, 2010). CALIOP level 3 monthly mean stratospheric aerosol data (CAL_LID_L3_Stratospheric_APro-Standard-V1-00) are used in this study; in this data product, clouds and polar stratospheric clouds (PSCs) have been removed based on the information of particulate extinction-to-backscatter (lidar) ratio and the multiple-scattering factor profile (Young and Vaughan, 2009; Kim et al., 2018; https://www-calipso.larc.nasa.gov/resources/calipso_users_guide/data_summaries/l3/lid_l3_stratospheric_apro_v1-00_v01_desc.php, last access: 24 February 2021).

Time–height distributions of BSR and PDR observed in the UTLS at Tsukuba are shown in Fig. 1, and the corresponding vertical profiles are shown in Fig. 2. Because the PDR has more missing data points, the TDR time–height distribution is also shown in Fig. A1. Days with missing data (white regions; Fig. 1) are due to thick summertime rain clouds in the lower to middle troposphere, which prevented the laser light from reaching the middle stratosphere. However, some events with enhanced particle signals are evident just above the tropopause at 15.5–18 km and last for a few days, mainly in August but with some in September. In particular, the event peaking at around 21 August and spanning 18–26 August was the strongest one among the events that the lidars successfully measured during the 3-month period. We also observe another strong event around 9 August at 15–17 km, although missing observations before and after this date prevent characterization of the temporal scale of the event; furthermore, the tropopause height was highly variable at the time and was located at 17 km on that date, situating the aerosol-enhanced layer temporarily in the troposphere. Figure 2 shows that enhanced particle signals at 15.5–18 km were often observed in August and sometimes in September but not in July. Typical BSR and PDR values of enhanced signals are ∼ 1.10 (1.07–1.18) and ∼ 5 % (3 %–10 %), respectively (Figs. 1 and 2). Below the tropopause, strong signals were sometimes recorded with BSR values of > 1.25 and with PDR values ≫ 10 %. In general, the PDR values are 0 % for spherical particles (i.e. water clouds in the troposphere and liquid H2SO4 particles in the stratosphere) and > 25 %–30 % for ice cirrus particles (e.g. Sakai et al., 2003; Fujiwara et al., 2009). Strong signals in the upper troposphere are thus due to ice cirrus clouds. Enhanced signals in the lower stratosphere (15.5–18 km) may be due to solid particles, as indicated by PDR values of ∼ 5 % (3 %–10 %). Taking PDR uncertainties (Sect. 2.1) into account, these values can be considered small but non-zero. The PDR values of these signals, together with the region being above the local tropopause in most cases, strongly suggest that they are not ice cirrus particles. However, the possibility of a mixture of spherical H2SO4 particles (i.e. background stratospheric sulfate particles) and highly non-spherical particles, such as ice, volcanic ash (Prata et al., 2017), and wildfire smoke (Haarig et al., 2018), cannot be precluded only with our lidar data. We will come back to this issue in Sect. 3.3 after investigating other data. Before looking at the Fukuoka results, it is noted that for Tsukuba we do not plot the data with a “relative” uncertainty of PDR larger than 30 %; this treatment resulted in removing data points with BSR values lower than ∼ 1.05 at which background spherical sulfate particles (with PDR values of < 2 %) were presumably predominant.

Vertical profiles of BSR and PDR observed at Fukuoka for 11 clear-sky and/or non-rainy nights during July–September 2018 are shown in Fig. 3. Again, enhanced particle signals were observed mainly in August above the tropopause at 15.5–18 km. The BSR values were in the range 1.09–1.14, with PDR values of 1 %–3 %, which are smaller than those observed at Tsukuba. It should be noted that the dates of lidar operation at Fukuoka did not overlap those at Tsukuba when strong enhancement was observed above the tropopause (e.g. 9, 18–26 August and 9 September), perhaps partly explaining the differences between Figs. 2 and 3.

The 10 d kinematic backward trajectories (using vertical wind) from Tsukuba and Fukuoka are shown in Figs. 4 and 5, respectively, with contrasting cases with or without aerosol particle enhancement in the 390–410 K potential temperature range (around 16.5–17.5 km at these stations). A potential temperature of 400 K corresponds to altitudes of ∼ 17.1 km at Tsukuba and ∼ 17.3 km at Fukuoka on average during July–September 2018 (based on twice-daily radiosonde data at each site, taken from http://weather.uwyo.edu/upperair/sounding.html, last access: 24 February 2021), i.e. near the centre of the lower-stratospheric BSR enhancement. By comparing the results from Santee et al. (2017) with our own analysis, the 65 ppbv contours of monthly mean CAMS CO data at 400 K potential temperature are chosen as an index of the boundaries of the ASM anticyclone (i.e. within the anticyclone, CO concentration is > 65 ppbv). These trajectories indicate that air masses over both stations come mainly from the west, sometimes via the north of Japan (indicative of the existence of vortices), and originate from the ASM anticyclone well within 10 d. They also indicate that air masses with enhanced aerosol particles at this height tend to originate in regions within the ASM anticyclone at altitudes of 16.5–18 km, i.e. around or just below the tropopause, whereas those without enhanced aerosol particles tend to originate from edge regions surrounding the anticyclone. Note that there is a trajectory that originates in the Pacific south of Japan as low as 4 km (Fig. 4b, a small-scale spiral in purple); this is associated with upward transport in the typhoon Soulik.

Horizontal distributions of CO and MSF at the 400 K potential temperature surface during 18–23 August 2018 from the CAMS reanalysis data are shown in Fig. 6. Again, note that a potential temperature of 400 K corresponds to ∼ 17.1 km at Tsukuba and ∼ 17.3 km at Fukuoka during July–September 2018. The distribution of MSF indicates the location of the ASM anticyclone from the dynamical viewpoint. The strongest particle signals during the 3 months were observed on 21 August in the lower stratosphere over Tsukuba. The air mass with high CO concentrations was transported eastward from the ASM anticyclone centred over the Tibetan Plateau (Fig. 6), with an anticyclonic vortex of ∼ 20∘ longitude scale reaching the Japanese archipelago on 21 August, providing a clear signature of eastward-shedding vortices from the ASM anticyclone (e.g. Luo et al., 2018). Daily averaged longitude–time CO distributions over 30–40∘ N are shown in Fig. 7, with that latitude band chosen here because it includes the two lidar sites. The ASM anticyclone spans roughly 15–40∘ N, whereas the eastward-shedding vortices are often located slightly to the north at around 25–45∘ N, as indicated in Figs. 4–6; the latitude band must therefore be chosen carefully, depending on the research focus. In Fig. 7, the 60 ppbv CO contour may be a good indicator of eastward-shedding vortices. In July 2018, the eastward extension was weak, but in August there were three events that directly affected the two lidar sites on 3–15, 20–24 (Fig. 6), and 28–31 August. In September, there were three events on 3–8, 14–17, and 28–29 September. A comparison with Fig. 1 indicates that aerosol-particle-enhanced events correspond relatively well to CO-enhanced events, although missing lidar data points (due to low-level clouds) result in only the 20–24 August event being relatively well observed, with the 3–15 August event being captured only on 9 August. The ASM anticyclone is also characterized as an air mass hydrated by active convection from below (e.g. Santee et al., 2017). The longitude–time distribution of MLS water vapour at 400 K, averaged over 30–40∘ N with 8∘ longitude and 3 d bins, is shown in Fig. 8. The water-vapour-enhanced events over Japan correspond well with the CO-enhanced events over the same region shown in Fig. 7.

Lidar is sensitive to various types of volcanic aerosol (e.g. Yasui et al., 1996; Sakai et al., 2016; Khaykin et al., 2017). The lower stratosphere is continuously influenced by volcanic eruptions (GVP, 2013), which inject various types of particles and gases into the atmosphere (e.g. Robock, 2000). Among them, solid ash particles may remain in the stratosphere for up to a few months, while liquid H2SO4 particles resulting from the reaction of volcanic SO2 and H2S gases with OH and H2O may remain for a year or more. Aerosol particles are also emitted from biomass burning and forest fires, and although these particles rarely reach the stratosphere, extensive fire events can influence the stratospheric aerosol loading (e.g. Khaykin et al., 2018, 2020; Peterson et al., 2018; Kablick et al., 2020). In this section, the global lower-stratospheric aerosol loading during the summer of 2018 is investigated through an analysis of satellite aerosol data.

The time–latitude distribution of zonal-mean lower-stratospheric aerosol optical depth (AOD) at 675 nm from the OMPS LP satellite instrument is shown in Fig. 9. At high NH latitudes, the lower-stratospheric AOD increased in the summer of 2017 due to extensive wildfires in Canada (Khaykin et al., 2018; Peterson et al., 2018), but their influence became negligible by early 2018. In July 2018, the eruption of Ambae (or Aoba; 15.389∘ S, 167.835∘ E; GVP, 2019), Vanuatu, in the tropical western Pacific, caused increasing stratospheric AOD in the tropics. We also observed very weak signals around the same latitude from the beginning of April 2018, possibly due to the eruption of Ambae again during March–April 2018 (GVP, 2018). However, the lower-stratospheric AOD at 25–40∘ N was relatively low during July–September 2018, at least on a zonal-mean scale. The monthly mean CALIOP attenuated scattering ratio distribution due to aerosol particles at 17 km in July and August 2018 is shown in Fig. 10 where the ATAL is evident, with enhanced aerosol signals over the ASM region. In August there was also a hint of eastward extension of the ATAL to Japan, with a slight increase in the scattering ratio. By August, effects of the Ambae eruption had extended to about half of the tropics but had not reached Japan directly, at least not in a monthly mean view (see also the 10 d backward trajectories; Figs. 4 and 5).

Finally, Chouza et al. (2020) showed that lidar measurements at Mauna Loa, Hawaii, indicated no signals from volcanic eruptions during the summer of 2018. Also, at the OHP lidar site in France (43.9∘ N, 5.7∘ E), no enhancement in the lower-stratospheric aerosol abundance was observed during the summer of 2018. In summary, enhanced aerosol particle signals observed at Tsukuba (36.1∘ N) and Fukuoka (33.55∘ N) were unlikely to be due to volcanic eruptions or northern wildfires.