The vertically resolved optical properties of stratospheric aerosols have been observed with a 532 nm polarization lidar in Wuhan (30.5° N, 114.4° E) since October 2010 (He et al., 2021, 2022; Yin et al., 2021). A detailed description of the lidar system can be found in previous studies (Kong and Yi, 2015; He and Yi, 2015). Specifications of the polarization lidar system are listed in Table . Benefiting from the installation of a transparent waterproof window in 2017, the lidar system can perform continuously regardless of rainy or snowy conditions (Yi et al., 2021) from then on except for the necessary maintenance.

Raw data are stored with resolutions of 1 min and 30 m. The Fernald method (Fernald, 1984) was used to retrieve the backscatter coefficient βp and aerosol extinction coefficient αp, using a fixed aerosol lidar ratio of 50 sr for the long-term (2010–2021) and 70 sr for smoke events (Canadian wildfire in 2017 and California wildfire in 2020) (Haarig et al., 2018). The volume depolarization ratio δv is calculated as the ratio of perpendicular- to parallel-oriented signals, multiplied by the gain ratio, and then converted into the particle depolarization ratio (PDR) δp (Freudenthaler et al., 2009). The same as Trickl et al. (2024), we use 1 km above the tropopause as the lower limit for sAOD calculation to avoid the influence of tropospheric aerosols and to incorporate the stratospheric aerosols as much as possible. Therefore, the stratospheric aerosol optical depth (sAOD) is calculated by integrating the aerosol extinction coefficient from 1 km above the tropopause to 30 km to minimize disturbances from the troposphere and ensure a sufficient signal-to-noise ratio (SNR). The uncertainties in the derived parameters, as well as the corresponding references, are listed in Table .

In addition, an algorithm developed by Yin et al. (2021) was used to screen out all the cloud-free profiles, utilizing a height resolution of 30 m and a time resolution of 4 h to ensure a sufficient SNR, with a sliding average of 300 m for height. The Rayleigh fit method was used to find the reference altitude between 5 and 20 km, with a width of 1.5 km, which the signal is close to the molecular signal derived from meteorological data provided by the Global Data Assimilation System (GDAS) (Baars et al., 2016). The reference value was set to 0.018 Mm-1sr-1 (corresponding to an extinction coefficient of 0.9 Mm-1). The molecule backscatter coefficient and extinction coefficient are calculated based on the method presented by Bucholtz (1995), with an uncertainty of < 2 %. Fernald forward inversion was then applied up to an altitude of 30 km to calculate the profiles of the backscatter coefficient and extinction coefficient. Due to the weak signal of stratospheric aerosols, different data processing methods can lead to significant variations in the specific values of the results. From October 2010 to September 2021, 775 nighttime cloud-free profiles were selected for long-term statistical analysis.

Absorption by ozone in the stratosphere was also taken into account. The ozone absorption coefficient is given by 1αO3=σO3×nO3, where σO3 = 2.72 × 10-21 cm2 is the ozone absorption cross-section (Gorshelev et al., 2014), and nO3 is the ozone number density (Fig. a). We utilized nO3 profiles from the Copernicus Atmosphere Monitoring Service (CAMS) reanalysis data within latitudes of 28.0–31.8° N and longitudes of 113.0–114.5° E. To correct aerosol extinction by considering ozone absorption, the seasonal average nO3 profiles from 2015 were employed as the representative ozone levels for each season through our measurement period, i.e., March–April–May for spring, June–July–August for summer, September–October–November for autumn, and December–January–February for winter. The relative uncertainty of the backscatter coefficient profile was less than 5 %. As an example, Fig. b shows the backscatter coefficient profile with and without correction for ozone absorption on 5 February 2020. The mean residual between 15 and 25 km is approximately 8 × 10-4 Mm-1sr-1, which cannot be ignored.

The spaceborne lidar CALIOP, carried on the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) satellite, has been widely used to observe the vertically resolved optical and microphysical properties of aerosols and clouds since 2006 (Winker et al., 2007). It is capable of measuring the elastic backscatter at both 532 and 1064 nm, as well as the depolarization ratio at 532 nm near the nadir.

In this study, CALIOP version 4.10 Level-1B data were used to track smoke plumes from September to October 2017 and from October to November 2020, illustrating their vertical distributions and optical properties. In addition, the CALIOP Level-3 stratospheric aerosol profile product (Kar et al., 2019) provides monthly mean aerosol optical properties on a spatial grid of 5° in latitude and 20° in longitude. It should be mentioned that the CALIOP Level-3 product consistently exhibits a bit higher aerosol extinction compared with the other satellite-based datasets (Chouza et al., 2020). Moreover, monthly mean tropopause altitudes from MERRA-2 (Modern-Era Retrospective analysis for Research and Applications, Version 2) reanalysis were provided in the CALIOP Level-3 stratospheric aerosol profile product.

The Ozone Mapping and Profiler Suite (OMPS) is installed on the joint NASA/NOAA (National Aeronautics and Space Administration and National Oceanic and Atmospheric Administration) Suomi National Polar-orbiting Partnership (Suomi NPP) satellite, launched in October 2011 (Jaross et al., 2014). OMPS comprises three spectrometers: a downward-looking nadir mapper, a nadir profiler, and a limb profiler. The UV Aerosol Index (UVAI) is an effective indicator of UV-absorbing aerosols, discerning between aerosol absorption and Rayleigh scattering, which is provided by the nadir-mapper instrument on the Suomi-NPP satellite at a spatial resolution of 50 km × 50 km. The UVAI has been widely employed in detecting elevated aerosols with significant absorption in the atmosphere, such as smoke and dust (Penning de Vries et al., 2009; Lee et al., 2015; Tao et al., 2022). In this study, the UVAI provided by the OMPS-NPP Level-2 data product was used to show the horizontal spatial distribution of smoke plumes.

Additionally, OMPS-NPP Level-3 data provide aerosol extinction coefficients at a spatial resolution of 5° × 15° lat–long grid measured by a limb profiler sensor. The monthly mean sAOD at 510 nm at 32.5° N, 112.5° E, provided by OMPS-NPP Level-3 data product, was used to estimate the evolution of Canadian smoke aerosols over Asia from August to November 2017. Furthermore, the monthly mean tropopause altitudes from August 2020 to September 2021 were also presented.

The NOAA/ARL (National Oceanic and Atmospheric Administration/Air Resources Laboratory) Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model (Draxler and Rolph., 2003; Stein et al., 2015) can simulate the forward and backward trajectories of an air mass by giving a starting time and an initial altitude and geographical location. The meteorological field from the GDAS archive (Kanamitsu, 1989) was used to drive the model in the calculation. In this study, the simulated backward trajectories were used to track the transport pathway and to confirm the source of smoke plumes.

Figure  shows the long-term evolution of vertically resolved aerosol backscatter coefficient and sAOD over Wuhan from 2010 to 2021. The monthly mean tropopause altitudes are provided by the CALIOP (before August 2020) and OMPS (after August 2020) Level-3 data products (represented by the white curve). In Fig. a, the white pixels and stripes are due to a lack of observations or insufficient SNRs. A stratospheric background aerosol layer, known as the “Junge Layer”, consisting primarily of sulfate from the oxidation of tropospheric SO2 or OCS (carbonyl sulfide) (Junge, 1960; Junge and Manson, 1961; Kremser, et al., 2016), persistently appeared during that period at altitudes of 19.5–23.0 km, with an average extinction coefficient of 0.36 Mm-1. Three significant episodes with more abundant tropospheric aerosols injected are evident: volcanic aerosols from the 2011 Nabro eruption and the 2019 Raikoke eruption, as well as smoke aerosols from the 2017 Canadian wildfire event. Moreover, a weak aerosol plume from the 2020 California wildfire event was also observed.

The stratosphere-quiescent period from January 2013 to August 2017 can be defined as reflecting the background level of stratospheric aerosols over Wuhan. The Volcanic Explosivity Index (VEI) is a general indicator of the explosive character of a volcanic eruption (Newhall and Self, 1982). It compositely estimates Walker's magnitude, intensity, destructiveness, dispersive power, and energy release rate and is assigned a value from 1 to 8. Two volcanoes with a VEI ≥ 4 erupted during this period: Kelud in Indonesia (7.9° S) in February 2014 and Wolf in the Galápagos Islands (0.0° N) in May 2015. Kelud volcanic aerosols were reported to be detected over high latitudes in Tomsk (56.5° N) (Zuev et al., 2017) and the Observatoire de Haute-Provence (OHP) (43.9° N) (Khaykin et al. 2017) in January 2015 due to the meridional aerosol transport. However, no aerosol plumes were observed by our lidar over Wuhan in the first quarter of 2015, and the increase in sAOD during this period cannot be attributed to a significant influence of Kelud. Several reasons are considered. First, it cannot be completely ruled out that a few Kelud aerosol plumes passed over Wuhan but were not observed due to weather conditions or hardware maintenance. Second, aerosol meridional transport from tropical into extratropical (middle) latitudes generally intensifies during the cold half of the year (October to March of the following year) (Niwano et al., 2009), causing an increase in aerosol optical properties in winter (Sakai et al., 2016; Zuev et al., 2017). An increase in sAOD at the turn of the year might be largely attributed to the seasonal variation of tropopause height (Fig. b). A detailed discussion of seasonal characteristics will be presented in Sect. . Therefore, it cannot be determined whether the increase in sAOD at the beginning of 2015 was due to the influence of Kelud or the seasonal meridional transport of tropical aerosols.

Table  lists the lidar-derived sAOD at different sites worldwide. Note that the sAOD values at Tsukuba, Lauder, and Garmisch-Partenkirchen are obtained by multiplying the integrated backscatter coefficients (IBCs) by a lidar ratio of 50 or 46 sr. The mean background sAOD (1 km above the tropopause to 30 km) over Wuhan was 0.0044 (± 0.0019), as obtained from January 2013 to August 2017. Due to the weak signal of stratospheric aerosols, different data processing methods can lead to variations in the exact sAOD values. Furthermore, it is noteworthy that the sAOD during the stratosphere-quiescent period over OHP increased from 0.0024 during 1997–2003 to 0.0028 during 2013–2014, suggesting the contribution of volcanic aerosols in the first decade of the 21st century. It is interesting to note that before 2006, volcanic eruptions with VEI ≥ 4 mainly occurred in the tropics (Chouza et al., 2020) and did not cause a noticeable enhancement of sAOD (or IBC) at mid-latitude sites in the Northern Hemisphere, revealing that tropical volcanic aerosols emitted during these events were not significantly transported northward.

The volcanic aerosol layers from the Nabro and Raikoke eruptions can be observed above Wuhan at altitudes 15–25 km during the second half of 2011 and 2019, respectively (Zhuang and Yi, 2016; Jing et al., 2023). An enhanced sAOD of 0.013 was observed during both the Nabro and Raikoke events, which was 2.9 times the background sAOD (0.0044). Another enhancement of the backscatter coefficient appeared around 20 km after September 2017, caused by the smoke aerosol injection from the Canadian wildfire event in August 2017. The intense pyrocumulonimbus (PyroCb) released approximately 0.1–0.3 Tg aerosols into the low stratosphere, comparable in quantity to those emitted from a moderate volcanic eruption (Peterson et al., 2018). The smoke plumes over Wuhan will be analyzed further in Sect. .

In addition, stratospheric aerosols are contributed by other tropospheric sources. Periodic increases in the backscatter coefficient were observed below an altitude of 18 km during the summer. The majority of these aerosol layers existed below the tropopause and are known as the ATAL (Vernier et al., 2015; Yu et al., 2015). The AMA emerges in response to persistent deep convection over India and southeast Asia during the boreal summer (Garny and Randel, 2016), leading to efficient vertical transport from the surface to the UTLS (Baker et al., 2011). Due to extensive human activities and the influence of the AMA, the ATAL forms and exists in the UTLS at altitudes of 13–18 km over megacities, consisting mainly of sulfate and carbonaceous aerosols (Vernier et al., 2015). The observation of the ATAL confirmed that anthropogenic aerosols and/or their gas-phase precursors can be transported to the UTLS, although they are generally removed effectively through precipitation, according to previous understanding (Mari et al., 2010).

Understanding the seasonal variations in stratospheric aerosol patterns is crucial for gaining deeper insights into their feedback on weather and climate. Figure a shows the profiles of the seasonal mean backscatter coefficient βp over Wuhan for each season during the stratosphere-quiescent period from January 2013 to August 2017 as well as during the whole measurement period. Here, the four seasons are defined as follows: spring (March–April–May), summer (June–July–August), autumn (September–October–November), and winter (December–January–February). The seasonal mean tropopause heights were 13.7 km in spring, 16.1 km in summer, 15.0 km in autumn, and 13.2 km in winter. At altitudes of 19.5–23.0 km, an enhancement of aerosol extinction was observed in all seasons, with a mean βp of 7.6 × 10-3 Mm-1sr-1. This non-seasonal background is interpreted as the Junge layer, a global-wide stratospheric aerosol layer at around 20 km altitude (Junge, 1960; Junge and Manson, 1961).

In summer, another distinct aerosol layer was observed at 13–18 km, with a mean βp of 0.023 Mm-1sr-1, approximately 1.4–1.6 times larger than the βp values in other seasons. This layer was contributed by ATAL. The AOD of ATAL at 13–18 km over Wuhan was 0.0057, consistent with an increasing AOD of 0.002–0.006 at 13–18 km in the entire ATAL region (15–45° N, 5–105° E) during 1995–2013 (Vernier et al., 2015). The AMA facilitates efficient vertical transport from the surface to the UTLS, serving as a primary source of young air in the lower stratosphere (Randel et al., 2010) and bringing anthropogenic aerosols and/or their gas-phase precursors from the lower troposphere to the UTLS. Furthermore, injections of aerosols from volcanic eruptions and wildfires caused the larger βp of 0.019 Mm-1sr-1 at 16–22 km in summer and autumn (compared with 0.016 Mm-1sr-1 in winter and 0.012 Mm-1sr-1 in spring) because Wuhan was generally affected by those events during the summer and autumn months by the summer monsoon circulation, which promotes effective mixing between the extratropics and tropics and influences the tropical seasonal cycle of different atmospheric components (Abalos et al., 2013).

In Fig. , we present the differences in the mean sAOD between the cold half-year (October–next March) and warm half-year (April–September). The mean sAOD in the cold half-year was 0.0054, approximately 69 % larger than that of 0.0032 in the warm half-year because the lower tropopause height in winter leads to a wider integration range of sAOD compared with summer. In addition, aerosols transported meridionally from the tropics to middle and high latitudes generally intensified the cold half-year in the Northern Hemisphere (Niwano et al., 2009). This process provides additional aerosol mass from the stratospheric tropical aerosol reservoir. The winter backscatter above 24 km was also larger than that of summer (Fig. ). Similarly, Zuev et al. (2017) found that the integrated backscatter coefficient was larger over Tomsk, Russia, during the cold half-year from 2000 to 2016. This pattern was also observed in Tsukuba, Japan, where the stratospheric aerosol backscatter ratio was 40 % larger in winter than in summer at 22–23 km from 1997 to 2004 (Sakai et al., 2016).

Stratospheric aerosols modify radiative fluxes by scattering and absorbing solar radiation. Hence amounts of aerosols emitted from those great volcanic eruptions, such as the 1815 Tambora eruption and 1991 Pinatubo eruption, can significantly cool the global climate (Solomon et al., 2011). Figure  shows the estimated radiative forcing (RF) induced by stratospheric aerosols and sAOD over Wuhan from 2010 to 2021. Radiative forcing was calculated by multiplying the sAOD by a conversion factor of -25 Wm-2 (black points) (Hansen et al., 2005; Solomon et al., 2011).

The radiative forcing of smoke is very complicated, depending on the surface albedo as well as the composition (Heinold et al., 2022). Black carbon (BC) in wildfire smoke generally exhibits strong absorption of solar radiation, unlike sulfate aerosols that reflect solar energy back to space, leading to different climate responses. Therefore, the conversion factor from sAOD to RF should be re-estimated to obtain a more reliable RF. The organic carbon (OC) and BC emissions from biomass burning are generally proportional. According to Koch (2001), the organic matter (OM) to BC mass ratio (OM/BC) is set to be 7.9 and the OC to OM mass ratio (OC/OM) is assumed to be 1.3. As estimated by Hansen et al. (2005), the conversion factors from AOD to RF are -13 Wm-2 for OC and 60 Wm-2 for BC. As a result, the contribution of sAOD to RF can be divided into three parts: background sAOD (sAODbackground), OC sAOD (sAODOC), and BC sAOD (sAODBC). The RF during the smoke injection period can be calculated as follows: 2RFsmoke=sAODbackground×(-25)+sAODOC×(-13)+sAODBC×60, where the sAODbackground is 0.0044, as given in Sect. . The corrected RF during the smoke intrusion period is presented (red points) in Fig. . After the correction, smoke RF becomes slightly weaker.

The RF values mainly range from -0.03 to -0.31 Wm-2, with a mean of -0.14 Wm-2. The 2011 Nabro and 2019 Raikoke eruptions caused remarkable cooling effects as the RF of -0.31 Wm-2. The stratosphere-quiescent period shows a stable trend of RF around -0.11 Wm-2, representing the stratospheric background aerosol level over Wuhan. On the contrary, Solomon et al. (2011) found that global stratospheric aerosols have been increasing by ∼ 7 %yr-1 based on satellite observations in the previous decade from 2000 to 2010. The abundant volcanic activities caused a significant increase in stratospheric aerosols. There were 12 eruptions with VEI ≥ 4 in the tropics and Northern Hemisphere before 2010 as reviewed by Chouza et al. (2020). By contrast, only three volcanos erupted with VEI ≥ 4 after 2010, i.e., the Nabro volcano in Eritrea in 2011, the Kelud volcano in Indonesia in 2014, and Wolf in the Galápagos Islands in 2015.

The mean RF during smoke observation periods was -0.06 Wm-2, consistent with the conclusion drawn by Hansen et al. (2005) that biomass burning (BC + OC) shows a negative forcing. This value was close to the background level of stratospheric aerosols and resulted in weaker negative forcing compared with sulfate aerosols. This is due to the absorption of solar radiation by BC, which offsets a portion of solar radiation reflected by OC and background aerosols.

For comparison, Chouza et al. (2020) measured a close value of -0.11 Wm-2 over Mauna Loa Observatory, Hawaii, during a volcanic-eruption-quiescent period in 2013–2019. Khaykin et al. (2017) measured the background value of -0.06 Wm-2 over OHP, France, during 1997–2003. Meanwhile, different data processing methods can lead to significant variations in the specific values of the results due to the weakness of stratospheric aerosols.