Aerosol Characteristics in the Three Poles of the Earth Observed by CALIPSO

To better understand the aerosol properties over the Arctic, Antarctic, and Tibetan Plateau (TP), the aerosol optical properties were investigated using 13 years CALIPSO L3 data, and the back trajectories for air masses were also simulated using the Hybrid Single Particle Lagrangian Integrated 15 Trajectory (HYSPLIT) model. The results show that the aerosol optical depth (AOD) has obvious spatial and seasonal variation characteristics, and the aerosol loading over Eurasia, Ross Sea, and South Asia is relatively large. The annual average AOD in the Arctic, Antarctic, and TP are 0.046, 0.025, and 0.098, respectively. The Arctic and Antarctic regions have larger AOD values in winter and spring, while the TP in spring and summer. There are no significant temporal trends of AOD anomalies in the 20 three study regions. Clean marine and dust-related aerosols are the dominant types over ocean and land respectively in both the Arctic and Antarctic, while dust-related aerosol types have greater occurrence frequency (OF) over the TP. The OF of dust-related and elevated smoke is large for a broad range of heights, indicating that they are likely transported aerosols, while other types of aerosols mainly occurred at heights below 2 km in the Antarctic and Arctic. The maximum OF of dust-related aerosols 25 mainly occurs at 6 km altitude over the TP. The analysis of back trajectories of the air masses shows large differences among different regions and seasons. The Arctic region is more vulnerable to mid-latitude pollutants than the Antarctic region, especially in winter and spring, while the air masses in the TP are mainly from the Iranian Plateau, Tarim Basin, and South Asia. https://doi.org/10.5194/acp-2020-1159 Preprint. Discussion started: 16 November 2020 c © Author(s) 2020. CC BY 4.0 License.

mean AOD of each grid at different temporal scales was calculated, and the seasonal differences between the northern and southern hemispheres were also considered. In this study, the spring (autumn), 145 summer (winter), autumn (spring), and winter (summer) are defined as March-May, June-August, September-November, and December-February in the north (south) hemisphere, respectively. Note that the aerosol properties in Figures 3, 4, and 7 over the TP region are only for the internal pixels of TP, which is marked by black dots in Figure 1 (c). In addition, the occurrence frequency (OF) of aerosol types was also calculated by counting the number of samples of seven aerosol types in each horizontal 150 grid cell or altitude layer.

HYSPLIT model and GDAS data
The Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model has been widely used in the simulation of atmospheric pollutant transport, dispersion, and deposition (Ashrafi et al., 2014;Jeong et al., 2012;Vernon et al., 2018;Zhao et al., 2009). To fully understand the sources of aerosols, 155 the back trajectories of air masses at ten selected sites over three study regions mentioned above were examined using the latest version (V5.0.0) of the HYSPLIT model (Stein et al., 2015). Simultaneously, the multiple trajectories that are near each other were merged into groups through cluster analysis. In this study, the four Arctic sites are located in Greenland (N1), Northern Europe (N2), Northern Asia (N3), and Northern North American (N4). The four sites in the Antarctic are located on the Antarctic 160 Peninsula (S1), Ross Sea (S2), Dronning Maud Land (S3), and Wilkes Land (S4). The two selected sites in the TP region are located on the northern (TP1) and southern (TP2) edges of the TP region. The locations of these sites are shown in Figure 1, and the detailed information of each site is shown in Table S1. Previous air mass back trajectory simulations in the Polar regions found that it is difficult to simulate the seasonal difference of the air mass with short-term back trajectory simulation, while the 165 long-term back trajectory simulation has great uncertainties in the spatial domain (Hirdman et al., 2010;Sharma et al., 2013), thus a 14-day back trajectory simulation was adopted in this study (Rousseau et al., 2006), and the simulation date was set as the 15th and last day of each month which can help save a lot of computation sources while keeping the simulated back trajectories representative. Asia in the south) is large. In terms of regional differences, the aerosol concentration in the Arctic region is significantly higher than that in the Antarctic region. Meanwhile, the annual average AODs over the Arctic, Antarctic, and the inner region of the TP are 0.046, 0.025, and 0.098, respectively. Antarctic mainly occur in winter and spring. The north wind prevails over the TP region in spring, which makes the dust aerosols originating from the Tarim Basin and the Qaidam Basin have a greater contribution over the TP (Xu et al., 2015). And with the development of the South Asian monsoon and mid-latitude Westerlies, aerosols in South Asia are lifted under the influence of large-scale atmospheric circulation and cross the Himalayas to affect the internal region of the TP (Cong et al., 2015;Lüthi et 195 al., 2015;Xia et al., 2011). The high aerosol concentration in the winter and spring Arctic is known as the Arctic haze phenomenon (Garrett and Zhao, 2006;Mitchell, 1957;Zhao and Garrett, 2015 Heintzenberg, 1989). Among the three study regions, the AOD of the Antarctic is slightly higher than that of the Arctic in the southern hemisphere wintertime, while the AOD of the Antarctic is the lowest in other months. The slightly higher AOD shown in the Antarctic in spring and winter compared to the other two seasons may be due to a similar reason as in the Arctic: stable atmospheric conditions and less precipitation make the aerosols difficult to be removed in spring and winter.

The multi-year averaged seasonal variation of AOD
205

The long-term trend of AOD
To study the long-term trend of AOD over the Arctic, Antarctic, and TP, the monthly AODs along with their standard deviations from June 2007 to December 2019 were calculated using valid data in the study areas. In order to remove the clear seasonal variation of AOD as found earlier in the study regions, the deseasonalized trend was carried out by calculating the AOD anomalies. The AOD 210 anomaly here is defined as the difference between the monthly average value of AOD in each month and the average value of AOD for that month in all years. The results of the monthly AOD anomaly over the Arctic, Antarctic, and TP are outlined in Figure 4. The solid line with red color represents the monthly AOD anomaly, the shadow region represents the single standard deviations, and the blue dotted line represents the linear trend based on deseasonalized monthly AOD anomalies from June 215 2006 to December 2019. Figure 4 shows that there are no significant increasing or decreasing trends of AOD anomalies in the Arctic, Antarctic, and TP (slope = -0.00724%~-0. 00219%), although the linear trends show a high confidence level (p > 0.05). It is worth noting that the deseasonalized monthly AOD anomalies over the TP region are relatively high. There are two likely reasons. First, there are anthropogenic emission sources over the TP region. Second, the TP is located in Central Asia 220 surrounded by highly polluted areas, which is easily affected by external aerosol transport.

Horizontal distribution
In order to examine the spatial and temporal variability of aerosol types, the normalized annual and seasonal averaged OFs of different aerosol types over the (a) Arctic, (b) Antarctic, and (c) TP were 225 https://doi.org/10.5194/acp-2020-1159 Preprint. Discussion started: 16 November 2020 c Author(s) 2020. CC BY 4.0 License.
presented in Figure 5 and Table 1, respectively. The number i~vii represent OF of clean marine, dust, polluted continental/smoke, clean continental, polluted dust, elevated smoke, and dusty marine aerosol, respectively.
In terms of the spatial distribution of aerosol OF among the three study regions, it can be seen from  does not differ significantly between land and sea areas, which can be explained to a certain extent by the fact that the elevated smoke aerosols in the Antarctic and Arctic are mainly transported from the outside.
There is also a significant difference in the OF spatial distribution of different aerosol types in each 245 study region. In the Arctic, dust (ii) and polluted dust (v) aerosol has a higher frequency of occurrence over the Greenland, northeastern Asia, and northern America due to the transport of Asian dust into the atmosphere, which was subsequently transported eastward and reached the high-latitude regions of Northern America (Tomasi et al., 2007;VanCuren et al., 2012). In contrast, polluted continental/smoke (iii) aerosol mainly occurs in Eurasia, which is mainly due to biomass burning (e.g., agricultural 250 burning and wildfires) in the Eurasian region (Soja et al., 2006;Warneke et al., 2010). In the Antarctic, there are obvious spatial differences in aerosol types. Specifically, dust (ii) and polluted dust ( Similarly, polluted continental/smoke and elevated smoke also have a higher frequency in the southern TP. As mentioned above, aerosol types have a distinct seasonal variation. We then investigated the seasonal average OF of different aerosol types. In this study, the number of samples of seven aerosol types in 265 each study region was first counted, and then the normalized OF of different aerosol types was calculated seasonally. Similar to the findings in Figure 5, in general, the dominated aerosol type is clean marine over the Arctic and Antarctic. However, the normalized OF of aerosol types display a substantial seasonal dependence (Table 1). Specifically, the proportion of clean marine aerosols is larger in the Arctic in autumn and winter than that in spring and summer. This may be due to the wind 270 speed in winter half-year in the Arctic region is higher than that in summer half-year, which makes more marine aerosols enter the atmosphere (Erickson et al., 1986). In the summer fire season, the wildfires and agricultural burning occur more frequently over Siberian and North American, which can be transported to the Arctic along with the pollutants, resulting in a high proportion of polluted continental/smoke aerosol and elevated smoke aerosol. This notion is also supported by previous 275 studies (Stohl et al., 2006;Schmeisser et al., 2018;Tomasi et al., 2007). In spring, meanwhile, the proportion of dust and polluted dust increases significantly in the Arctic, which is due to the transported dust from Asian desert sources (Barrie, 1995 that in summer and autumn (Bodhaine, 1995;Weller et al., 2013). Compared with the Antarctic and Arctic regions, the types of aerosols in the TP are relatively simple, which are mainly the dust aerosol 285 and polluted dust aerosol. In spring and summer, the proportion of dust aerosol is relatively high, which is due to the northerly jet over the TP carrying dust aerosols to the internal TP. In autumn and winter, the emission of anthropogenic aerosol increases, and the proportion of mixed anthropogenic aerosol and dust aerosol increases continuously over the TP. In addition, the OF of polluted continental/smoke aerosol in winter is much higher than that in other seasons, which may be due to the increase of 290 biomass combustion.

The vertical extinction coefficient of dominant aerosol type
Knowledge of aerosol extinction coefficient is necessary to enhance our understanding of how atmospheric aerosols impact the weather and climate to a certain extent ( As Figure 6 shows, there is no doubt that the aerosol extinction coefficient profile has a significant 300 regional difference. In general, the aerosol extinction coefficient in the Arctic has a broad vertical distribution at heights ranging from 0 to 12 km, but the vertical distribution of the Antarctic aerosol extinction coefficient is uneven. In the Antarctic, the extinction layer can reach a maximum height at 11 km in winter (k) and spring (b), while it is mainly distributed below 5 km in summer (e) and autumn (h). The vertical distribution of aerosols over the TP is more concentrated, with most aerosols 305 distributed between 2 and 8 km. The vertical distribution of extinction coefficients of different aerosol types also demonstrates large regional differences. The elevated smoke in the Arctic has a larger extinction coefficient when the altitude is greater than 2 km, especially in summer (d) and autumn (g); while in the near-ground area (altitude < 2 km), dust and polluted dust have a larger extinction coefficient, which is in good agreement with previous studies ( The elevated smoke is mainly concentrated at heights about 3 km, while the dust-related aerosol types 315 are more distributed at heights below 2 km. In winter (k), on the contrary, the extinction coefficient of dust and elevated smoke increases significantly above 5 km, and the polluted dust aerosols have large extinction coefficients under 5 km. Unlike the Arctic and Antarctic regions, the extinction coefficients of smoke and dust-related aerosols over the TP region are larger at heights of 4 -9 km and 2 -9 km, respectively. From the perspective of seasonal variation, the vertical distribution of dust-related aerosol 320 extinction coefficient is larger in spring (c) and summer (f) than in autumn (i) and winter (l). These vertical distribution information can help better understand the sources and impacts of aerosols over the three study regions in future. For example, aerosol information below clouds could be particularly important for aerosol-cloud interaction study.

325
Aerosol types not only have significant spatial and temporal variations, but also vary with height.
CALIPSO data provides the vertical distribution of aerosol types at 208 levels, ranging from surface to 12 km. We here investigate the vertical distribution of seven aerosol types, as shown in Figure

Back trajectory
In order to better understand the origins of the air masses arriving in the study regions, the latest version (V5.0.0) of the HYSPLIT model was used in this study to simulate the back trajectories of air 345 masses. Ten sites listed in Table S1 were selected in this study, and the 14-day back trajectories for the Arctic, Antarctic, and TP sites were simulated. A total of 3,120 (10×2×12×13) back trajectories were computed at a height of 500 m above the surface at all ten sites. The seasonal climatologies (January 2007~December 2019) of air mass trajectories were created and the cluster analysis was implemented to examine the long-range transport pathways of air masses. The cluster analysis determines the final 350 number of clusters based on the total spatial variance (Draxler and Hess, 1998). Figure 8 reveals the seasonal climatological characteristics of the back trajectories after cluster analysis. It can be seen that the back trajectories over different study regions have distinctive characteristics, especially in the TP region. Compared with the Antarctic, the air mass trajectory in the Arctic region has a shorter transport distance. This is most likely due to the fact that the temperature in the Arctic is higher than that in the 355 Antarctic, which decreases the pressure gradient and reduces the wind speed. In the Arctic, the difference of back trajectories between summer and winter half-year is obvious, with more proportion of air masses from the Eurasian in winter and spring. At the same time, Asian dust storms prevail in spring, resulting in a greater proportion of dust and polluted dust in spring. In contrast, the influence of external transport of aerosols is relatively small in autumn, and the larger wind speed allows more 360 marine aerosols enter the atmosphere, which together make the contribution of clean marine aerosols in autumn relatively large in the Arctic.
In the Antarctic region, the seasonal difference in air mass trajectories is relatively small compared with the Arctic region, and the air mass trajectories were mainly controlled by circumpolar westerly winds (Ravi et al., 2011). While it is not clearly shown by the air mass back trajectory simulation 365 results, dust and polluted dust over East Antarctica was likely caused by the transport from South America and Africa, and the polluted dust over West Antarctica was more likely affected by the aerosol transport from South America and Australia. Generally speaking, under the influence of steady and https://doi.org/10.5194/acp-2020-1159 Preprint. Discussion started: 16 November 2020 c Author(s) 2020. CC BY 4.0 License. strong westerly winds, dust and carbonaceous aerosols in South America, Australia, and Africa have a certain impact on Antarctic pollution (Li et al., 2018;McConnell et al., 2007;Zou et al., 2018).

370
Different from the Arctic and Antarctic, the back trajectories of air masses over the TP have significant seasonal variation. In spring and summer, the air masses located on the northern slope of TP mainly come from the northern desert area. In autumn, the air masses from the north begin to weaken, while the air masses from Iranian Plateau begin to increase and reach the maximum in winter (93.15 %). For the station on the southern slope of the TP, the air masses mainly come from the Iranian Plateau in 375 spring and winter, while in summer they mainly come from South Asia, which causes more biomass burning aerosols to enter the TP.

Summary and Conclusions
Aerosols play a crucial role in the radiative budget of the Earth-atmosphere system, but due to insufficient understanding of aerosol properties, at least partly, the uncertainty of the total radiative 380 forcing by aerosols in the climate mode is still the largest. Understanding the properties of aerosols is highly demanded. The satellite active remote sensing can make up for the insufficiency of ground-based remote sensing to obtain long-term and large-scale aerosol properties. In this study, the spatial and temporal distribution of the aerosol optical depth (AOD) and aerosol type over the Arctic, Antarctic, and Tibetan Plateau (TP) regions were investigated. In addition, ten typical sites were 385 selected and the back trajectories of air masses were simulated using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model. The main findings are as follows.
The distribution of AOD over the three study regions shows distinctive spatial and seasonal differences.
In general, the AOD over the Arctic and Antarctic decreases with the increasing latitude. In the Arctic, the AOD over land is greater than that over the ocean, while the opposite is true for the Antarctic.