The lidar measurements were performed at the German–French AWIPEV research base in Ny-Ålesund, Svalbard, by the “Koldewey Aerosol Raman Lidar” (KARL). This system consists of a Spectra 290/50 Nd:YAG laser, which emits a laser beam at 355, 532 and 1064 nm with 50 Hz and 200 mJ per pulse and color vertically into the atmosphere. The backscattered photons are collected by a 70 cm telescope with a field of view of about 2 mrad in the elastic colors as well as in the Raman-scattered signals at N2 at 387 and 607 nm. Data recording is done via Hamamatsu photomultipliers and Licel transient recorders with 7.5 m resolution. Full overlap is reached at about 700 m altitude. A more detailed description of the instrument is given by . The aerosol backscatter coefficient βaer(λ) [(m sr)-1] and the aerosol extinction coefficient αaer(λ) [m-1] were calculated following the principle of . For better readability and simplicity we use β(λ):=βaer(λ) and α(λ):=αaer(λ) in the following, if not stated otherwise. The lidar ratios, LR=αaer(λ)βaer(λ) [sr], for the other wavelengths were chosen as LR355=50 sr for the troposphere and LR355=70 sr in the stratosphere and for λ=532 nm LR532=36 sr and LR532=42 sr in the troposphere and stratosphere, respectively, if not stated otherwise. High values of LR indicate highly absorbing particles and are therefore mostly related to the refractive index but also to shape and size. The fact that we found LR532<LR355 in the stratosphere was derived by our lidar data via two conditions: same backscatter at different wavelengths for (tropospheric) cirrus clouds and the same gradients for the aerosol backscatter in clear layers of the troposphere. Since we do not expect rapid changes in the physical and chemical composition of aerosols in the lower stratosphere and to reduce the uncertainty due to noise, especially during polar day, we chose a temporal and spacial resolution of 60 min and 150 m for the elastic profiles at 355 and 532 nm. Both signals can always be evaluated up to almost 30 km height.

It is well known from that the solution of the elastic lidar equation is 1P(z)=Const⋅βtot(z)z2⋅exp⁡-2∫z0zαtot(z^)dz^, where P(z) is the signal power at a given height z, Const a constant determined by the instrument itself and its setup, βtot(z)=βaer+βRay the total backscatter signal, and αtot(z)=αaer+αRay+αabs the total extinction, depending on the choice of a lidar ratio and on the choice of a boundary condition. The extensions “aer”, “Ray” and “abs” are the contributions of aerosols and Rayleigh scattering as well as the absorption by trace gases. βaer needs to be prescribed at a certain altitude. pointed out that the solution is stable and less critically dependent on the assumed boundary condition, when the integration is done “backwards”: from large distances back to the lidar.

We set the boundary condition, ϵ, in the altitude interval from 27 to 30 km for March to November. In winter times, polar stratospheric clouds (PSCs) might occur at this altitude , so we lowered the fitting range to 24 to 27 km. No PSCs were found in our data set after this adjustment. In this altitude range we assume that the average total backscatter coefficient in that interval is ϵ532=1.10⋅βRay and ϵ355=1.05⋅βRay for parallel and perpendicular. The signal-to-noise ratio, SNR, is calculated following Eq. (). 2SNR(z)=P(z)Perr(z)=P(z)P(z)+Pbgrd Here Perr is the estimated noise profile and Pbgrd the averaged signal in an altitude range, which is dominated by electronic noise and background light in an altitude range of 52 to 60 km. The lidar signal P(z) is given in the units of “counts” like in the photo-counting channels. This altitude range has a sufficiently small median signal-to-noise ratio. Therefore it is valid to assume there is just noise in the signal. The SNR is shown in Table  at an altitude of 20 km for the different wavelengths and directions of polarization.

With these high values of signal-to-noise ratio we conclude that the data set is sufficiently good for a further evaluation of observations of physical features in the lower stratosphere. Since the depolarization is very low (see Sect. ), the signal is much weaker than for the parallel polarized light, and therefore the SNR is generally smaller. Due to the small SNR at 20 km the signal of λ=1064 nm is neglected for this study.

From the lidar data we derive the following quantities: the lidar ratio LR, aerosol depolarization δaer(λ) and the backscatter color ratio CR were calculated following Eq. (): 3LR(λ)=αaer(λ)βaer(λ),δaer(λ)=β⟂aer(λ)β∥aer(λ),CR(λ1,λ2)=βaer(λ1)βaer(λ2). Here, β⟂aer and β∥aer are the backscatter coefficients with respect to the perpendicular and parallel laser polarization, respectively, which are measured separately in the elastic 355 and 532 nm channels. For βaer and δaer only 532 and 355 nm signals were used. The color ratio is defined with λ1<λ2.

In contrast to the backscatter coefficient, β, the backscatter ratio, R, is dimensionless and does not decrease with increasing height. It is defined as in Eq. (): 4R(λ)=βtot(λ)βRay(λ)=1+βaer(λ)βRay(λ).

An ideal, aerosol-free atmosphere would have R=1, which increases with stronger pollution. The parameter R is therefore a good indication for aerosol existence.

With the ERA5 reanalysis data from the mean height of the tropopause was found at an average height of 9137 m with a standard deviation of 775 m at the study site of Ny-Ålesund during the lidar measurement time and will be further discussed in Sect. . To investigate the properties of stratospheric aerosol, we focus on the altitude range between 10 to 15 km; most of the troposphere is neglected and not shown in the plots.

For this study, we only used automatically flagged data, which fulfilled the following criteria. Each wavelength and profile is tested and eventually removed individually:

elimination of negative or too many (>20) values that are too low (R≤1.003)

A qualitatively good profile has to fulfill all of these criteria at the same time. If one is not satisfied, the profile is eliminated out of the further processed data set. These criteria were chosen by manually looking at the data of different times of the year and atmospheric conditions with the aim of finding a good balance between eliminating as many disturbed profiles as necessary and keeping the good ones.

The color ratio, CR, is a crucial parameter to determine the size of aerosols and is defined according to Eq. (). An overview of the CR in 2021 is shown in Fig. . As a ratio of two small quantities, CR critically depends on noise and numerical assumptions in the lidar evaluation . Hence, some noise is noticeable in the form of rapidly changing colored dots in Fig. .

Typically, CR values around 2 have been found. Remarkable is the fact that the second half of the year becomes more homogeneous with slightly larger values than the first half. From June to October a more homogeneous distribution of a CR∼2 is detected between 320 and 440 K potential temperature. Since CR is a function of particle radius, the particle size decreases towards the end of the year and additionally with increasing height, as was also seen by . According to the generally larger particles (lower CR) in the first half of the year may be an indication of reduced tropospheric advection in the Arctic lower stratosphere in this time of year.

The relationship between β532 and CR is depicted for the previously defined four layers of the lower stratosphere in Fig. . With this dependency a better understanding of the particle size and concentration can be made.

While the spread in β532 decreases with height, CR remains in this same range of about CR∈[1,4]. At altitudes of 330–350 K a small influence of the troposphere might be seen in summer months, since the tropopause can reach these altitudes exceptionally. In the range of CR∈[1,2] all altitude layers have a backscatter coefficient larger at the end of the year than in the beginning. Under pristine conditions, β532≤5×10-8 (m sr)-1, we see always a clear dependence between CR and β: the smaller β (the clearer the atmosphere) becomes, the larger CR becomes because small particles show only a small scattering efficiency. Hence, below β532=5×10-8 (m sr)-1 the stratospheric β is mainly driven by the size of aerosols. This effect levels to above the limit of β532=5×10-8 (m sr)-1. After a certain limit the stratospheric aerosol does not grow apparently. This is in agreement with the annual cycle already shown in Fig. . Since the color ratio does not change much in the four discussed intervals, it also means the aerosol size is very similar, but only the concentration becomes lower the higher up the layer is located.

Another parameter to determine physical properties of aerosol by remote sensing is using the so-called aerosol depolarization, δaer(λ). It is defined according to Eq. (). The laser of KARL is linearly polarized and is always the reference direction of the scattered light. With Mie theory it can be shown that spherical particles do not change the polarization of the laser light, while non-spherical ones do. Due to surface tension liquid droplets are in general spherical .

In general stratospheric aerosols have the property of β⟂aer≪β∥aer. Therefore, it is expected that δaer becomes quite noisy for a standard aerosol measurement at higher altitudes. Due to generally low values of the depolarization we only show and discuss δ532 in this study.

The overview of the aerosol depolarization, δ532, is given in Fig.  for the different layers of the lower stratosphere color-coded for every month. Due to case studies in Sect.  September is highlighted here already. The depolarization is low (< 10 %) for all corresponding values of β532. All four height intervals a common depolarization is between δ532≥1 % and δ532≤4 %. Additionally May is the most diverse month with single cases of δ532→10 %, while January shows the smallest depolarization. In general, summer has a higher depolarization than winter months.

We define a “weak” depolarization with δ532=2 %–5 % (see Table ), a “medium” depolarization with δ532=5 %–10 % (see Table ) and a “strong” depolarization with δ532≥10 %, but this was not found in the lower stratosphere of 2021.

A big variability between the months can be observed in Table  for the depolarization of δ532=2 %–5 %. While the depolarization is in general lower in the winter months, it reaches its maximum in August and has a clear annual cycle. The depolarization never reached values > 2 %. This indicates that the stratospheric aerosol observed above Ny-Ålesund is not constant throughout the year but changes its chemical composition and/or its shape . Furthermore, a more frequent depolarization is found at lower altitudes than at higher ones. Our lidar observations fit very well with the physical properties of aerosols in the Asian tropopause aerosol layer (ATAL), which consist of organics, nitrate, sulfate and ammonium, especially in combination with Fig. , where it can be seen that the aerosol is also slightly non-spherical . The altitude at which this aerosol layer in the Arctic was found also matches very well with the altitude of the ATAL.

Table  shows on the other hand the relative occurrence of δ532=5 %–10 % for the entire layer of 330–410 K. Only in a few months was a “medium” depolarization observed. Even though the depolarization has an annual cycle, all values are very small compared to lower latitudes . No young extreme pollution events, like wildfire aerosol, dust intrusion or volcanic ash with typical values of δ532≥16 % , were found in the lidar measurements over Ny-Ålesund in 2021. Therefore we conclude that our data set indeed mainly represents stratospheric background conditions even with the contribution of the ATAL. Therefore we conclude that our data set represents stratospheric conditions free from the influence of local events; however there is evidence of a seasonal impact of ATAL particles on the lower stratosphere during summer.

Figure  shows the two months with the lowest (February) and highest median depolarization (August) for comparison for the lower stratosphere and for further investigating the difference between the relative occurrence of depolarization.

While February has a more or less constant depolarization at δ532=1.35 % in the atmosphere between 320 and 480 K, August on the other hand has in general higher depolarization values, especially in the lower parts of the atmosphere (until about 420 K). The grey shaded area highlights the difference between both months. Especially in the troposphere and lower stratosphere a significant deviation can be found and shows the impact of the ATAL. Additionally, the summer month also has a maximum around 315 K, where it reaches δ532=2.5 %, followed by a second, local maximum at around 340 K with δ532=2.0 %. From that point on it decreases to about δ532=1.25 % and converges to the same value as in February. Due to solar radiation in August, the data become noisy quicker than during polar night. At an altitude of 400 K the depolarization in August is still significantly higher than in February. The median tropopause height according to radiosonde data is 307 K in February, while it is 324 K in August. This plot also shows that even if no distinct biomass burning layer has been seen in our lidar data during summer, the slightly less spherical shape of the particles indicates a non-sulfuric component that is an evidence for an impact of ATAL particles on the Arctic background aerosol up to ∼ 450 K in August 2021, even though aged biomass burning aerosol may still contribute to the stratospheric aerosol background, which was even found at altitudes of up to 27 km .

Our measurements indicate evidence for the impact of ATAL particles on the Arctic stratosphere during summer; therefore we now present the annual cycle of β355 and β532 of the background aerosol in Figs.  and , respectively, for two different heights of 360 and 380 K. Data from ±10 K around the chosen level are taken into account for the following study.

The annual cycle of β355 at 360 K altitude (Fig. a) shows a small seasonal dependency: while the first half of the year is characterized by lower values of the backscatter coefficient, the mean values of August, September and November are >10-7 (m sr)-1. While winter (DJF) is the clearest season, May, July and September are the months with the broadest distribution. Overall the median and mean values are for all months very similar, indicating that β355 follows almost a Gaussian distribution.

Likewise there is a similar annual cycle in the data at the 380 K level (Fig. b) with larger values from June to November. Also here DJF is the clearest season. While the absolute values of β355 are consistently lower than at 360 K height, the relative spreading of β355 is also smaller. Typical values for 360 K altitude are β355∈[7,11]×10-8 (m sr)-1, and they are smaller for the 380 K level, β355∈[6,10]×10-8 (m sr)-1. In summary, a clear annual cycle in the stratospheric backscatter coefficient has been found.

While the values of β532 are by about a factor of 1.5 to 2 smaller than for β355, a similar pattern is shown in Fig. a compared to β355 with some differences: the variability in the first half of the year is more pronounced, and the increased backscatter coefficient peaks in August. On the other hand β532 is less homogeneously distributed than β355 throughout the entire year. The 380 K level (Fig. b) shows the same trend throughout the year with a higher month-to-month variability in April and May. Median and mean values of β532 spread more than for β355, indicating a more asymmetrical distribution and more variability within a month. Typical values for 360 K altitude are β532∈[5,7]×10-8 (m sr)-1 and for the 380 K level β532∈[1,6]×10-8 (m sr)-1.

The contributions of young air masses from South Asia and the tropical Pacific over the course of the year 2021 to the air over Ny-Ålesund is inferred for the heights (360±5) K (Fig. a) and (380±5) K (Fig. b), respectively. Mean values are calculated for the surface origin tracers for each day.

Artificial tracers released in the model simulations since 1 December 2020 in Fig.  do not reach Ny-Ålesund at levels of potential temperature of 360 and 380 K before May 2021. During summer and autumn, the highest fractions of tropospheric air over Ny-Ålesund are from South Asia and the tropical Pacific region. Especially in September and October almost 20 % of the air is from South Asia. These model results support the evidence that the Arctic lower stratosphere is impacted by aerosol particles from the ATAL . All other regions combined are of minor importance. With few exceptions the tropical Pacific region contributes constantly more, and sudden changes in its impact are rare.

The accumulation of young air masses in the lower stratosphere over Ny-Ålesund is lower than 100 % during the year 2021. The remaining fraction consists of aged air that is older than 1 December 2020 when the CLaMS simulation was initialized.

Figure  gives for two exemplary days in September (13 and 28 September) the distribution of the surface origin tracer for South Asia (a marker for air from the Asian monsoon region) over northern Europe at 360 and 380 K potential temperature. A high fraction from South Asia is found over the Arabian Peninsula marking the western outflow of the Asian monsoon anticyclone e.g.,. Air masses contributing to the ATAL can be transported eastwards along the subtropical jet and enter the lower stratosphere by quasi-horizontal transport. As a result, thin filaments with enhanced signatures of air mass tracers originating from South Asia were found over the northern Atlantic Ocean and Europe e.g.,. On 28 September a filament located over Ny-Ålesund with a contribution from South Asia up to 40 % is simulated. This filament was not found on 13 September.

Figure  shows the impact of Southeast Asia on the stratosphere for the two different heights of 360 and 380 K. On 13 September the contribution is quite low at 360 and 380 K (Fig. a and b). As can be seen in both figures, the contribution is constant over the whole of Europe but increases with increasing height and changes on timescales of less than a day. As can be seen at both altitudes, the impact is already very low over Svalbard (≈ 10 %). On 28 September a filament was seen faint at 360 K altitude (Fig. c) but very pronounced at 380 K (Fig. d). The presented values of 20 % above Ny-Ålesund are large for the site but small compared to locations at lower latitudes.

When comparing Figs.  and with each other the contributions of the two main sources of tracers can be identified. The region “tropical Pacific” is responsible more for the background amount, which can be observed in Fig. . Contrary to this, “Asia” has a smaller contribution but shows also a very large day-to-day variability. These spikes can be nicely seen as filaments in the three-dimensional simulations of tracers of Fig. .

The following preliminary conclusion can be made: there is a large day-to-day variability in the distribution of young air masses from Asia contributing to the Arctic lower stratosphere. This variability strongly depends on the dynamics of the Asian monsoon anticyclone itself and the transport of air masses from Asia to the extratropical lower stratosphere.

Backward trajectories are well suited to analyze the detailed transport pathway of an air parcel and therefore complement the three-dimensional CLaMS simulations e.g.,. Within this study, we calculated an ensemble of backward trajectories for the two exemplary days (13 and 28 September 2021). Within ±2° in latitudinal and longitudinal direction around Ny-Ålesund and in a height range of 360 to 380 K, 891 trajectories have been calculated in total for each day. Markers in Fig.  show the location where the trajectory first reaches the boundary layer. The color indicates the transport time.

As can be seen in Fig. a, on 13 September in total 41 trajectories (4.6 %) reached the boundary layer in Southeast Asia after traveling for about 3 to 4 months, and 6 other trajectories (0.7 %) are outside of the monsoon region. A total of 850 trajectories (89.6 %) did not touch the boundary layer at all within the model run time.

The situation is a bit different for 28 September 2021 (Fig. b), where the fraction of contacts outside of the monsoon region is larger (6.1 %) but also the fraction of endpoints in Southeast Asia (17.8 %​​​​​​​). On this day 76.1 % of the 891 trajectories did not reach the model boundary layer within the computational time. It took the air parcel on average 87 d to reach Ny-Ålesund on 28 September. As can be seen on both days, back trajectories also end in different parts of the world, like the equatorial Pacific Ocean but also Central America. This result agrees very well with the forward trajectories of aerosol tracers of Sect. . The average transportation time from the Asian summer region to Ny-Ålesund is about 1 month longer on 13 September (111 d) than on 28 September (87 d).

The presented numbers are also given in Table  for a better overview, with the absolute and relative numbers of back trajectories touching the boundary layer inside and outside of the monsoon region in South Asia.

Additionally one trajectory ensemble is given for each day in Fig. . As can be seen in both cases, the artificial tracer circulates from Ny-Ålesund first around the pole before it moves to lower latitudes until it reaches South Asia, where it also circles until it reaches the boundary layer. Since both trajectories have similar pathways, we conclude that the intake of aerosols into the lower stratosphere depends on the local conditions in South Asia and the effectivity for aerosols being lifted into the ATAL.