Three-dimensional climatology, trends and meteorological drivers of global and regional tropospheric type-dependent aerosols: Insights from 13 years (2007–2019) of CALIOP observations

Abstract. Globally gridded aerosol extinction data from the Cloud–Aerosol Lidar with Orthogonal Polarization (CALIOP) during 2007–2019 are utilized to investigate the three-dimensional (3D) climatological distribution of tropospheric type-dependent aerosols, and to identify the trends in column aerosol optical depth (AOD), partitioned within different altitude regimes, and their meteorological drivers. Using detection samples of layer aerosols, we also yield a 3D distribution of the frequency-of-occurrence (FoO) of aerosol sub-types classified by CALIOP. The results show that the aerosol extinction coefficient (AEC) shows contrasting vertical distribution patterns over land and ocean, with the former possessing significant geographical dependence, while the enhancement of AEC in the latter is mainly located below 1 km. The vertical structures of the type-dependent AECs, however, are strongly dependent on altitude. When the total AOD (TAOD) is partitioned into the planetary boundary layer (PBL) and the free troposphere (FT), results demonstrate that the PBL and FT contribute 61.86 % and 38.13 %, respectively, of the global tropospheric TAOD averaged over daytime and nighttime. Yet, this CALIOP-based partitioning of the different aerosol sub-types in the PBL and FT varies significantly. Among all 12 typical regions of interest analyzed, more than 50 % of TAOD is located in the lower troposphere (0–2 km), while the contribution is less than 2 % above 6 km. In global average terms, we found the aerosol FoO averaged over all layers is 4.45 %, with the largest contribution from ‘clean marine’ (1.79 %) and the smallest from ‘clean continental’ (0.05 %). Overall, the FoO vertical structures of the aerosol layer exhibit a distribution pattern similar to that of AEC. The resulting trend analyses show that CALIOP accurately captures significant regional anomalies in TAOD, as observed in other satellite measurements and aerosol reanalysis. Our correlation analysis between meteorological factors and TAOD suggests the interannual variability of TAOD is related to the variability of precipitation (PPT), volumetric soil moisture (VSM), and wind speed (WS) in the particular regions. For instance, the positive TAOD trend over the equatorial central Pacific is mainly attributable to the increased PPT and decreased WS. In contrast, in dry convective regions dominated by dust and smoke, the interannual variability/trend in TAOD is largely modified by the VSM driven by the PPT. Additionally, we further found these significant regional correlations are more robust within the PBL and significantly weakened or even reversed within the FT. This highlights the superiority of using the TAOD partitioned within the PBL as a proxy variable for the widely applied TAOD to explore the relationships between atmospheric pollution and meteorology.


where AEC( ) z represents the aerosol extinction coefficient for all aerosol types and type-dependent aerosols at altitude z, and z1 and z2 are the minimum and maximum altitude of the specified altitude layer, respectively. The ratio (%) of layer-specific AOD to total column AOD ( total AOD ) in the troposphere was also evaluated, according to the following equation: layer-specific layer-specific total AOD R 100.0% AOD Moreover, in addition to exploring the stratified AOD and its partitioning, this study also pays particular attention to the 165 3D distribution characteristics of FoO with altitude for the CALIOP-derived aerosol subtypes. Consistent with previous studies (Adams et al., 2012;, the FoO (%) of type-dependent aerosols at different altitudes is defined as aerosol total N FoO = N ×100.0% , where aerosol N represents the number of samples detected by CALIOP for a specific aerosol type at the specified altitude layer, and total N is the total number of samples (including all aerosol types and "clean air") at the specified altitude layer. At the same 170 time, the total column (i.e., all layers) FoO (%) for each aerosol subtype at each grid box was further calculated according to the following equation: where i aerosol N represents the number of samples for a specific aerosol type at the specified altitude layer i (maximum number of layers is 208), and i total N is the total number of samples (including all aerosol types and "clean air") at the specified altitude 175 layer i.

Multi-sensor and reanalysis AOD datasets
In order to ensure the accuracy of the AOD trends assessed by CALIOP, five different datasets of AOD at 550 nm, obtained from three satellite retrievals and two aerosol reanalyses, are also used for intercomparison purposes. The three satellite-based AOD retrieval datasets are from MODIS (onboard the Terra and Aqua satellites) and the Multi-angle Imaging 180 Spectro-Radiometer (MISR) (onboard the Terra satellite). In this study, we use the monthly gridded AOD product Collection C6.1 (MYD08_M3 and MOD08_M3 for the Aqua and Terra satellites, respectively), with a 1° × 1° resolution, covering the period 2007-2019, derived from the combined Dark Target and Deep Blue algorithms. MISR monthly AOD data with a 0.5° ×0.5° resolution were obtained from the Level-3 global aerosol product (MIL3DAE), version F15_0032.
The two AOD reanalysis datasets are from the Modern-Era Retrospective Analysis for Research and Applications, version 185 A similar, but much less intense, aerosol belt is apparent in the area north of 40°N latitude. The formation of this aerosol belt is primarily associated with the high-altitude transport of dust and smoke aerosols.
The differences in the optical properties (i.e., absorption and morphology) of different aerosol types allow the CALIOP retrieval algorithm to assign the total aerosol extinction to aerosol subtypes. Fig. 2 presents the 3D distribution of AEC for the three main aerosol types (dust, PD and smoke) classified by CALIOP. These three aerosol species were chosen because of 255 their remarkable climatic effects. In the CALIOP V4.2 classification algorithm, the "dust" type represents uncontaminated pure dust, the "PD" type tends to represent a mixture of dust and PC aerosols or smoke aerosols, and the "smoke" type represents the lifted smoke plumes produced by biomass-burning emissions, which consist mainly of carbonaceous aerosols with strong absorption of black carbon and organic carbon. For dust aerosols, the enhancement of dust AEC (DAEC) is located mainly in arid and semi-arid regions, including the Sahara and Sahel, the Arabian Peninsula, northern India, the Tarim Basin 260 and the Gobi Desert, which are the main contributors of dust globally, and form the well-known dust belt in the Northern Hemisphere. Vertically, dust aerosols can be lifted up to an altitude of 6 km, and with their high values usually located in the near-surface layer (below 3 km) nearby the dust source. In addition, this 3D structure provides direct observational evidence for the widely identified dust trans-Pacific transport phenomenon. That is, mineral dust from desert source areas in East Asia, driven by westerly or northwesterly winds, can be transported all the way along the North Pacific Ocean to near the United 265 States West Coast, with a maximum uplift altitude of 4-6 km (also see Fig. S2b). Similarly, consistent with previous findings (e.g., Adams et al., 2012;Yu et al., 2015), the trans-Atlantic transport phenomenon of dust aerosols is also confirmed. The results show that dust aerosols from the Sahara Desert (SD) can be transported to North and Central America, thus forming a dust belt that can extend to the Caribbean and northern South America, with a maximum uplift altitude of 6-7 km.
Unlike the 3D pattern of dust aerosols, PD is more widely distributed-mostly in regions downstream of the dust source 270 where there are intensive human activities or typical biomass burning, e.g., the eastern United States (EUS), Amazon Zone (AMZ), eastern Africa, southern Asia (SA), and northwestern and northern China. Enhanced PDAEC is usually constrained to altitudes below 3 km near the ground, which can be mainly attributed to the distribution related to anthropogenic pollutant emissions. For smoke aerosols, the distribution of smoke AEC (SAEC) has a distinct regional dependence. Specifically, its high values are mainly distributed in typical biomass burning regions, including central-southern Africa (CSA), the AMZ, 275 southeastern Asia (SEA), and central Siberia, where the maximum lifting altitude of smoke aerosols exceeds 4-6 km. Given their strong absorption properties, the wide distribution range of smoke aerosols has been proven to have a non-negligible impact on weather and climate. For instance, previous studies have shown that smoke plumes generated by forest fire emissions can even be ejected into the stratosphere, where they can remain for up to eight months . In addition, black carbon aerosols-one of the main components of smoke plumes-entering the stratosphere, can also affect the dynamic 280 stability and horizontal circulation of the stratosphere by heating the surrounding air, thus disturbing the radiative balance of the Earth system and ultimately predisposing it to drastic global climate change.
The climatological distributions of the whole-layer integrated AODs for total aerosols and for aerosol sub-types are given in Fig. S3. The CALIOP-based observations show that the global average AOD of total aerosols reaches 0.095, of which dust, https://doi.org/10.5194/acp-2021-467 Preprint. Discussion started: 8 July 2021 c Author(s) 2021. CC BY 4.0 License.
PD and smoke aerosols account for 0.020, 0.011 and 0.008, respectively. In terms of type-dependent AODs as a percentage of 285 TAOD, dust AOD (DAOD), polluted dust AOD (PDAOD) and smoke AOD (SAOD) contribute 15.55%, 10.01% and 7.81% of TAOD, respectively (Fig. S4), while the remaining 66.63% is contributed by other types of aerosol (OTA), including anthropogenic pollution aerosols consisting of sulfate, nitrate and ammonium, and sea-salt aerosols. Spatially, TAOD is similar to the distribution patterns previously obtained based on ground-based observations, satellite retrievals, and model simulation (Hsu et al., 2012;Chin et al., 2014;Che et al., 2015). In contrast, there are significant regional differences in the spatial 290 distribution of type-dependent AODs, which can mainly be attributed to the regional differences in anthropogenic activity intensity, surface type, and climatic conditions. Fig. 3 further presents the distribution characteristics of the layer-specific integrated AODs (at 2-km intervals) with altitude for these aerosol species. The results show that the layer-specific integrated TAOD markedly decreases with altitude, and about 80% or more of the aerosols are located in the lower troposphere (0-2 km altitude range) (Fig. S5). Aerosols over 295 land can still contribute about 10% of the column aerosol loading in the altitude range of 2-4 km, which is indicative of the complexity of aerosol sources over land and their interactions with the topography and meteorological conditions. Compared with aerosols over land, the aerosol loading over the oceans shows a rapid decline with altitude, and the percentage of aerosol loading within 2-4 km drops to less than 5% in all regions except for the ocean areas adjacent to land. When the altitude exceeds 6 km, the high values of TAOD are mainly located in the middle and high latitudes of the Northern Hemisphere, and 300 the main contributors are elevated dust and smoke aerosols. When the TAOD is separated according to aerosol types, differences in the distribution of layer-specific integrated AOD with altitude are apparent for different types of aerosols.
Specifically, dust aerosols are scattered almost equally in the altitude ranges of 0-2 km and 2-4 km, and their enhancement is mainly located in source areas and their downstream. Vertically, the elevated altitude range of dust aerosols can extend from near the ground all the way to more than 8 km, which is mainly the result of the interaction between dust, terrain and wind (Xu 305 et al., 2018).The latest CALIOP aerosol type discrimination algorithm further classifies the mixed dust aerosols over the oceans into PD and DM. Thus, between 0-2 km, PD is almost entirely distributed over land, and the high values of its proportion (> 60%) are mainly distributed in continental regions north of 40°N, northern Africa, coastal regions of South America, northern SA, and eastern China. In contrast, the proportion of PD is higher over the oceans than over land in the altitude range above 4 km. Unlike dust and PD aerosols, smoke aerosols are slightly more abundant in 2−4 km than in 0−2 km, and the vertical impact 310 extent of smoke aerosols can be extended all the way to about 10 km. Overall, the proportion of smoke aerosols on land decreases rapidly with altitude, while the proportion on the ocean does not show a monotonic decreasing trend.

Partitioning of TAOD and type-dependent AODs between the PBL and FT
Accurate partitioning of the loadings in total aerosols and key aerosol types within the PBL and FT is essential for constraining atmospheric models and accurately quantifying the radiative effects of aerosols. Here, the partitioning of the AOD 315 in the PBL and FT was calculated using the monthly CALIOP Level-3 aerosol extinction retrievals and the PBLH data from the MERRA-2 Reanalysis (see section 2). The annually averaged global distribution of PBLH and FT AODs for all aerosol https://doi.org/10.5194/acp-2021-467 Preprint. Discussion started: 8 July 2021 c Author(s) 2021. CC BY 4.0 License. types, dust, PD and smoke over 13 years (2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019) are shown in Fig. 4. Overall, for all aerosol types, the global TAOD is 0.0923 in the troposphere, of which 0.0571 is found in the PBL and 0.0352 in the FT, corresponding to 61.86% and 38.13%, respectively, of the TAOD. The above results are largely consistent in terms of the average partitioning between daytime and 320 nighttime (daytime: 69% and 31% within the PBL and FT, respectively; nighttime: 38% and 62% within the PBL and FT, respectively) obtained by Bourgeois et al. (2018) using Level-2 CALIOP aerosol extinction retrievals during 2007-2015. The slight difference between the two can be mainly attributed to the differences in the PBLH products used and the time-matching scheme. However, if the low bias of the CALIOP aerosol extinction in the lower troposphere (below 2 km) with respect to surface lidar observations noted in prior studies (e.g., Campbell et al., 2012;Papagiannopoulos et al., 2016) is taken into 325 account, the contribution of column TAOD within the PBL determined in this study is expected to be even higher than 62%.
Spatially, high values of TAOD can be seen both in the PBL and FT over west-central Africa, the Arabian Peninsula, India and eastern China. This distribution is expected since these regions correspond to the main continental sources of aerosol mass.
Typical features such as aerosol transport over the Arabian Sea and the Bay of Bengal or long-range transport of dust aerosols from the Sahara over the Atlantic Ocean are also observed in the FT assigned by CALIOP data. 330 Unlike TAOD, which accounts for approximately 1.6 times more in the PBL than in the FT, type-dependent aerosols account for significantly different proportions in the PBL and FT. For example, although the locations of high values of DAOD and PDAOD over land are very different, the AOD partitioning between the PBL and FT is similar for dust (49.48% and 50.51%, respectively) and PD (52.22% and 47.79%, respectively). In contrast, a very different partitioning ratio (23.68% and 76.32% for the PBL and FT, respectively) is apparent for smoke aerosols. For dust aerosols, the high DAOD within the PBL 335 is mainly located in the dust source area, whereas its high values within the FT are located in the downstream region away from the dust source area. The enhanced DAOD within the FT contributes mainly to two regional processes: pyroconvection and orographic lifting, which can transport dust aerosols from the surface to the FT (Yumimoto et al., 2009;Bourgeois et al., 2015). For PD aerosols, the contribution of PDAOD is more or less equal in the PBL and FT over most land areas, except in SA and eastern China where the PDAOD within the PBL is higher than that within the FT. The equivalent partitioning of 340 PDAOD within the PBL and FT is mainly attributable to the adequate mixing of dust aerosols and anthropogenic pollution aerosols within the PBL, as well as to pyroconvection and orographic lifting. For smoke aerosols, the partitioning of SAOD within the FT is almost three times as much as that within the PBL, indicating that light-weight, fine-mode carbonaceous aerosols are more likely to be lifted to higher altitudes by horizontal and vertical motions with a very wide range of effects.
Overall, the degree of influence of smoke aerosols within the PBL is much weaker than that within the FT. This is evidenced 345 by the fact that biomass-burning aerosols released from the China-Indochina Peninsula, CSA and AMZ produce a broader range of effects within the FT than within the PBL.  Table S1. These ROIs are used to explore the regional differences in the vertical extinction profiles and FoOs for different types of aerosols. Fig. 5 presents the regional-averaged vertical profiles of the multi-year average TAEC over the 12

Regional-averaged vertical distribution of TAEC and type-dependent AECs with altitude
ROIs. The vertical profile of the multi-year averaged TAEC is somewhat representative of the climatological characteristics 355 of the vertical distribution pattern of total aerosol concentration. Comparing the vertical profiles of TAEC for different ROIs shows that the peaks of regional TAEC occur, in descending order, in NC (0.47 km −1 ), ME (0.37 km −1 ), SA (0.32 km −1 ), SC (0.31 km −1 ), SD (0.18 km −1 ), AMZ (0.18 km −1 ), NWC (0.16 km −1 ), CSA (0.14 km −1 ), WEU (0.12 km −1 ), SEA (0.10 km −1 ), NEA (0.08 km −1 ), and EUS (0.07 km −1 ). The vertical profile of TAEC is dominated by a single peak in most regions, and the altitude corresponding to the peak is mostly below 1 km. Among them, TAEC reaches a maximum near the ground in NC and 360 SA, and then shows a rapid decreasing trend with altitude. In CSA and NWC, it is multi-peaked, with the former having a smaller peak around 3 km, mainly attributable to the uplifted smoke aerosols. In addition, the vertical profile of TAEC in NC is accompanied by large interannual fluctuations (corresponding to amplified standard deviations) below 3 km. This large interannual variation in TAEC is mainly attributable to the decreased anthropogenic emissions as a consequence of clean-air actions implemented by the Chinese government in the last decade Gui et al., 2019), resulting in a rapid 365 decline in annual aerosol concentrations from the near-surface to upper layers (see Fig. S6). Similar interannual evolutionary patterns of the vertical profile of TAEC are also apparent for SC, EUS and WEU. However, a similar evolutionary trend (rightward/leftward shift) is not apparent in other ROIs. The enhanced DAEC profile 375 occurs mainly in dust source regions, including the ME, SD and NWC. In the ME, the DAEC reaches a maximum (~0.30 km −1 ) near the ground and then decreases gradually with altitude. In SD, it reaches a maximum (~0.10 km −1 ) at an altitude of 0.5 km.
In contrast, the DAEC peak in NWC occurs at an altitude of 1.5 km. Compared to the PDAEC profile, a more pronounced interannual fluctuation is found in the DAEC from the lower to the middle and upper levels (Fig. S7). In NEA, AMZ, CSA and SEA, the vertical aerosol extinction is dominated by elevated smoke, and the vertical distribution of SAEC shows a 380 bimodal distribution, with the peak locations at 0-1 km and 2-4 km, respectively. Although more prominent interannual differences are apparent in the AMZ and SEA regions than other ROIs, the altitudes corresponding to the peak of the SAEC profile do not change significantly (Fig. S9). It is also noteworthy that the elevated smoke aerosol layer is prevalent over the anthropogenic aerosol-dominated EUS, WEU and SA too, with the peak SAEC corresponding to altitudes of about 2-4 km. and 12 ROIs within a specific layer at 2-km intervals. Meanwhile, the statistics of the relative proportion (%) of the layerspecific AODs to TAODs are also shown in Fig. 7. The results show that, on a multi-year global average, the contribution of the integrated AOD within 0-2 km, 2-4 km, 4-6 km, 6-8 km, 8-10 km, and 10-12 km to TAOD is 80.38%, 15.41%, 3.42%, 0.52%, 0.20% and 0.07%, respectively. In contrast, the contribution of the integrated AOD within these five specific altitude ranges is 70.82% (87.52%), 23.59% (9.41%), 4.82% (2.28%), 0.53% (0.51%), 0.18% (0.21%) and 0.06% (0.07%) over land 390 (ocean), respectively. Ocean-land differences in the contribution of the lower troposphere suggest that aerosols over land are capable of being lifted to higher altitudes by the topography and atmospheric circulation. In all 12 ROIs, more than 50% of the aerosol loading is located in the lower troposphere in the 0-2 km altitude range, while the contribution is less than 2% in altitudes above 6 km. Among these 12 ROIs, the proportion of integrated AOD within 0-2 km is located, in descending order, in WEU (85.12%), SEA (83.54%), NC (81.68%), EUS (80.67%), SC (79.51%), NEA (77.44%), AMZ (77.08%), SA (75.50%), 395 ME (65.98%), SD (58.66%), NWC (54.58%), and CSA (51.91%). The proportion of the layer-specific integrated AOD at 2-4 km to TAOD decreases more than two to three times compared to that at 0-2 km over all ROIs, except NWC and CSA.
Among them, the highest value of regional aerosol loading within 2-4 km, as a percentage of total aerosol loading, occurs in NWC (36.37%), and the lowest occurs in WEU (12.12%). In contrast, the highest and lowest values at 4-6 km occur in the ME (8.15%) and AMZ (1.84%), respectively. The percentage of layer-specific integrated AOD to regional TAOD is less than 400 1%, 0.4% and 0.2% for all ROIs, except NEA at 6-8 km, 8-10 km and 10-12 km, respectively. respectively. This pattern of higher TAOD over land than ocean does not vary across the altitude range (Fig. S10). However, the contribution of dust and smoke to the integrated TAOD within the specified altitude is dominant with increasing altitude, both over land and ocean.
Among the 12 ROIs, the multi-year regional-averaged TAOD values, from largest to smallest, occur in NC (0.427), ME In contrast, dust, PD, smoke and OTA contribute 0.008, 0.021, 0.013 and 0.049, respectively, to the lowest regional-averaged TAOD in the EUS. In terms of regional averages for type-dependent aerosols, we find that the maximum value (0.312) of the regional-averaged DAOD occurs in the ME, while the minimum value (0.007) occurs in CSA.
In contrast, the maximum value of the regional-averaged PDAOD (0.190) occurs in NC, while the minimum value (0.014) 415 occurs in WEU. For SAOD, the maximum regional-average value (0.117) occurs in CSA, while the minimum value (0.002) occurs in the ME. From different altitude ranges, the distribution modalities of the regional-averaged whole-layer TAOD averaged FoO occurs in the ME (16.43%), while the minimum value occurs in WEU (4.63%). Fig. 10 presents the 3D distribution of FoO for all aerosol types at the specified 208 layers from the altitude range of 0 to 12 km. Also, the 3D distribution of FoO for the seven aerosol sub-types classified from CALIOP are shown in Fig. 11 and Fig.   S11. On a global scale, the 3D distribution pattern of cumulative stratified FoO for all aerosol types with altitude remains highly consistent with that of TAEC (Fig. 1). In land areas other than the desert source, high values of stratified FoO (generally > 455 60%) occur in the lower atmosphere (below 1 km). Vertically, stratified FoO shows a decreasing trend with altitude. In the dust source area, the high values of stratified FoO (70%) can extend from the surface to about 3 km. Over the ocean, near the surface is the altitude where sea salt aerosols are most frequently detected.
The 3D distribution of the FoO for dust aerosols (Fig. 11a) shows that the enhanced stratified FoO at different altitudes is located in the dust belt spanning the Sahara, the ME and NWC, with the highest FoO (~60%) located at altitudes of 1-3 km. 460 It is noteworthy that most regions of the Northern Hemisphere suffer from trans-regional transport of dust aerosols, which is continuously lifted by the topography, enabling the stratified FoO to reach 3%-5% within 4 km in most regions of the Northern Hemisphere except for the dust source. Meanwhile, the maximum value of the stratified FoO (~3%) of dust aerosols, originating from the desert source area in NWC during the prevailing trans-Pacific transport, is located at 3-4 km altitude.
Unlike dust aerosols, high values (~20%-30%) of the stratified FoO for PD aerosols are mainly located in the dust source area, 465 as well as in the downwind region, with a spread altitude of about 0-4 km (Fig. 11b). Conversely, the maximum values of the stratified FoO of PC aerosols are all located at altitudes below 1 km and are mainly controlled by anthropogenic aerosol emissions near the ground (Fig. 11d). In contrast, the maximum value of the stratified FoO for smoke aerosols is located at 3 km altitude in CSA and its adjacent oceans (Fig. 11c). For DM aerosols, the maximum value of FoO (~30%) is located in the altitude range of 0-2 km over the Arabian Sea between Saudi Arabia and SA (Fig. 11e). Also, high FoO (~10%-20%) is 470 observed over the tropical mid-Atlantic (at an altitude of about 0-3 km). Further analysis of the 3D distribution of FoO for both clean-type aerosols (i.e., CC and CC) shows that, in most ocean areas far from the continents, CM aerosols are found to exhibit high FoO values (> 60%) in the lower atmosphere. Instead, in the lower atmosphere over the oceans close to the continents (0-2 km), the values are able to decrease to 10%-30% (Fig. 11f). Unlike CM aerosols, which exhibits high FoO over most of the ocean, the stratified FoO for CC aerosols is low over most land areas (see Fig. S11). 475 peaks of the stratified FoO for CM aerosols are generally located at 0.1-0.2 km altitude in the near-surface layer. DM aerosols mostly occur below 4 km, and the peaks of the stratified FoO are generally located within the near-surface to 0.4 km altitude, with higher frequencies in NC, NEA, SA and the ME, with maxima of 9.2%, 15.6%, 12.4% and 17.0%, respectively. 485

Vertical profiles of regional-averaged FoOs for different types of aerosols
In the ROIs other than SD, NEA, SEA and ME, aerosols are mostly dominated by pollution-type aerosols (i.e., PD, PC and smoke). In addition, pollution-type aerosols tend to have larger interannual fluctuations corresponding to larger standard deviations, influenced by anthropogenic activities and the interannual variability of meteorological conditions. In terms of altitude, the peak values of the stratified FoO for PD and PC aerosols are usually located below 2 km. In contrast, the maxima of regionally stratified FoO for smoke aerosols all occur between 2 km and 4 km. The most dominant role of smoke aerosol is 490 found to appear at about 2 km over CSA. Except for CSA, the FoO with altitude for smoke aerosols shows a distinct singlepeak distribution pattern in AMZ, SA, SEA, SC, NC, NEA, EUS and WEU, corresponding to an altitude of 2.6 km (5.6%), 2.5 km (4.3%), 2.5 km (4.7%), 2.7 km (8.1%), 2.7 km (4.4%), 2.5 km (4.3%), 2.6 km (3.4%) and 2.5 km (2.2%), respectively.
In the three ROIs located in the dust source area, the maximum values of the stratified FoO for dust aerosols are located at 1.1 km (33.4%), 1.5 km (35.3%) and 1.5 km (18.1%) for SD, ME and NWC, respectively. As for the downstream ROIs of the dust 495 aerosols, the maximum stratified FoO is located at 1.7 km (9.8%) and 2.4 km (8.9%) in NC and SA, respectively. Notably, in the SD region, we find low standard deviations for all aerosol types, indicating that the interannual variability of aerosols in this region does not fluctuate significantly.

Long-term trends in aerosol loading at different altitude regimes
Benefiting from the continuous record of aerosol vertical observations spanning 13 years, this study further investigates 500 the temporal trends of total aerosol and type-dependent aerosol over different altitude regimes, including the whole layer, sublayers, the PBL and FT. This is expected to provide new insights to further our understanding of the long-term changes in the total global aerosol loading over the last decade or so. Fig. 13 shows the long-term trends in the total column AODs and their partitioned AODs within the PBL and FT for total aerosols and the three main aerosol types (i.e., dust, PD and smoke), separately, for the period 2007-2019. These trends were calculated for the annual time series at each 2º × 5º (latitude × 505 longitude) grid box by using Sen's slope method. The results show that TAOD experiences a significant decreasing trend in eastern China, EUS, South America, WEU, and the ME, and a significant increasing trend in SA and central Siberia during the period 2007-2019. These significant regional temporal anomalies, both in terms of spatial patterns and magnitude of variability, are confirmed by satellite observations (i.e., MODIS/Aqua, MODIS/Terra, and MISR) and reanalysis products (i.e., MERRA-2 and CAMS) (Fig. 14). Of course, these regional trends are also consistent with previous studies (e.g., Klingmüller 510 et al., 2016;David et al., 2018;Che et al., 2019;Jin and Pryor, 2020;Gui et al., 2021). Despite the differences in sampling frequency and spatial resolution between different AOD data sources, the spatial consistency of TAOD trends further confirms the accuracy of CALIOP observations, especially the ability to capture long-term variability.
Further analysis of the trends in TAOD partitioned within the PBL and FT shows that the trend patterns of AODs within the PBL (TAOD_PBL) and within the FT (TAOD_FT) are consistent with those of TAOD, but the former have greater magnitude of variation, confirming the dominant driving role of the TAOD_PBL for the total aerosol changes in the whole layer. Also noticeable is a significant reduction in the number of grid points in some regions (e.g., the ME, SA and tropical Pacific) showing significant variability in TAOD_FT compared to TAOD_PBL, which is likely related to the regional differences in vertical transport efficiency (Cui and Carslaw, 2006). Regional differences in the efficiency of the vertical transport of aerosols from the PBL to FT in different regions will determine any regional inconsistencies in the regularity of 520 the vertical distribution of aerosol concentrations. Moreover, when the aerosols within the PBL are lifted to the FT, they can then be transported over long distances, leading to the existence of greater spatial heterogeneity in TAOD_FT than TAOD_PBL. effect of the trans-regional transport of dust aerosols .
Influenced by a combination of anthropogenic emissions and dust aerosols, PDAOD has experienced significant decreases in eastern China and southern South America, and significant increases in the ME and SA. These regional trends are usually dominated by aerosols distributed within the boundary layer, except for eastern China. Aerosols within both the PBL and FT in eastern China are comparable in their contributions to the PDAOD trends. For smoke aerosols, significant decreases 535 in SAOD occur in SC and the EUS, with similar non-significant decreases also occurring in South America and southern Africa. In contrast, SAOD enhancement occurs in SA and the Indochina Peninsula, but the magnitude of the trend is not significant. Different from DAOD and PDAOD, the trend in SAOD is mainly driven by the aerosols distributed within the FT, especially in SC. This phenomenon is mainly due to the fact that smoke aerosols can be lifted to the FT rapidly under the action of a vertical atmosphere. 540 Fig. 15 further shows the global trend distribution of the partitioning of TAOD, DAOD, PDAOD and SAOD within the altitude layer at every 2-km interval. The results reveal that the decline in TAOD in eastern China, the ME and EUS can be mainly attributed to 0-2 km and 2-4 km, while the increase in SA attributes to 0-2 km. Influenced by the continuous reduction in dust aerosol emissions from dust source areas in the last decade, the intensity of the trans-Pacific transport of dust aerosols in East Asia has significantly decreased, which has directly led to a significant weakening of dust aerosol loading within 2-10 545 km altitude over the North Pacific region. In the ME, the reduction in DAOD remains largely consistent across the four altitude layers within 0-8 km. In contrast, the decreasing trend in the layer-specific PDAOD in eastern China occurs mainly within the altitudes of 0-2 km and 2-4 km. In addition, we find that the weakening of TAOD over the North Pacific region is not only attributable to the decreased DAOD, but also contributes significantly to the reduced PD due to the continued diminishing of https://doi.org/10.5194/acp-2021-467 Preprint. Discussion started: 8 July 2021 c Author(s) 2021. CC BY 4.0 License. anthropogenic aerosol emissions in China . Overall, the stratified trend pattern of SAOD at 0-2 km and 2-550 4 km remains largely consistent. However, regionally, the overall decline in SAOD in the EUS (SA) can be mainly attributed to 0-2 km (2-4 km). The decline in smoke aerosols over most of the land directly leads to a significant reduction in SAOD in the upper atmosphere over the ocean.
The percentage trends (% decade −1 ) in TAOD, DAOD, PDAOD and SAOD over the globe, land, ocean, and 12 ROIs for the different altitude regimes are summarized in Fig. 16. These altitude-related trends were calculated from regionally averaged 555 annual time series. Overall, we can see a 4.4% decade −1 decrease (P < 0.05) in TAOD during 2007-2019. This significant reduction in global TAOD contributes to the weakening of TAOD, to varying degrees, over both land (−6.1% decade −1 , P < 0.1) and ocean (−3.3% decade −1 , P < 0.05). In terms of altitude, the reduction in column TAOD is attributable to both TAOD_PBLH and TAOD_FT globally, over land and over ocean, but the former is the more dominant driver. In addition, this reduction in TAOD is also found simultaneously within all partitioned altitude layers. Among the 12 ROIs, column TAOD 560 shows a significant decrease (P < 0.05) in NEA, NC, SC, ME, WEU, EUS and AMZ, with trend values reaching −19.9%, −31.3%, −37.3%, −21.0%, −15.5%, −29.6% and −16.9% decade −1 , respectively. These significant regional trends are present in both the PBL and FT, in addition to the FT in WEU. In NEA, NC, SC, EUS and AMZ, the decreasing sign of TAOD can extend all the way from 0-2 km to 8-10 km. However, we find completely reversed trend signs in the altitude range of 0-4 km and above 4 km in WEU, SEA and NWC. Considering that aerosols play different climatic roles at different altitudes, these 565 findings highlight the importance of exploring the stratified AOD trends. The regional differences in trends within the lower and upper atmosphere may be related to the high-altitude transport of aerosols.
For dust aerosols, which are mainly driven by meteorology, a non-significant decline in column DAOD is found globally, over land, and over ocean, and this decline can spread to different altitude regimes. Regionally, the negative DAOD trends in different altitude regimes are seen simultaneously in NEA, NC, SC and ME. For PD aerosols, PDAOD is found to decrease 570 by 5.3% (P < 0.05), 4.4% and 7.2% (P < 0.05) decade −1 for the globe, land and ocean, respectively. We also see consistent negative trend signs in all altitude regimes, albeit of different magnitudes. Regionally, column PDAOD shows a consistent negative trend in NEA (−38.5 % decade −1 , P < 0.05), NC (−38.1 % decade −1 , P < 0.05), SC (−47.7 % decade −1 , P < 0.05), while in SA and ME the trend values are completely reversed. These regional trends do not change within the PBL and FT. Also, the same consistent regional trend sign is seen for the altitude stratification within 0-6 km over the above ROIs. For 575 smoke aerosols, we find that SAOD experiences a 13.3% decade −1 (P < 0.05) decline, which is the result of a simultaneous decline in SAOD over land (−14.7% decade −1 , P < 0.1) and ocean (−10.9% decade −1 , P < 0.05). We do not find a shift in the sign of these regional trends in the different altitude regimes. Consistent with DAOD and PDAOD, the trend in SAOD at different altitude regimes is also negative in sign over the NEA, NC and SC regions. However, a positive SAOD trend sign does extend through different altitude regimes in the SA and ME regions, except for stratification above 6 km.

Effects of meteorological conditions
To explain the trends in total aerosol loading and different types of aerosol loading derived from CALIOP, we examine their relationship with the interannual variability of three key meteorological factors (PPT, VSM and WS). These meteorological factors have been demonstrated to be closely linked to the processes of emission, dispersion, transport and deposition of AAs. The correlation coefficients (R) were thus calculated between CALIOP-retrieved TAOD and these 585 meteorological factors from 2007 to 2019. Also, the relationships between meteorology and DAOD and SAOD are examined.
As shown in Fig. 17, the most prominent region where TAOD is highly correlated with PPT and WS is Indonesia and its extension into the equatorial central Pacific (ECP). In the ECP region, TAOD is positively correlated with WS and negatively correlated with PPT, indicating that WS and PPT regulate the interannual variability of TAOD mainly by affecting the process of emission and wet deposition of sea-salt aerosols. This implies that the significant decrease in TAOD observed in the ECP 590 region during 2007-2019 (see Fig. 13a) is mainly attributable to the significant increase in PPT (~30%-60% decade −1 ) and significant decrease in WS (~40%-80% decade −1 ) in the region (Fig. 18). In addition, TAOD is positively correlated with WS over the seas of the Southern Hemisphere, implying a dominant driving role of WS for marine aerosols in this region. In contrast, the interannual variability of land aerosols, especially dust and smoke aerosols, is closely related to the variability of PPT and VSM, while the role of WS is not prominent. This can be explained by the fact that PPT mainly modifies surface 595 conditions (e.g., roughness and vegetation cover) in arid or semi-arid areas by affecting VSM, which in turn affects the intensity of dust emissions. For instance, in the ME, a significant increase in PPT (30%-50% decade −1 ) and a slight increase in VSM are responsible for the decreased TAOD over the region (Fig. 18 and Fig. 13a), especially its dust component (i.e., DAOD) ( Fig. 13b).
PPT-driven changes in VSM can also alter atmospheric humidity conditions by affecting surface evaporation, which 600 further modifies the frequency of biomass-burning events when induced by high temperatures. This is further confirmed by the dominant negative correlation between TAOD or SAOD and PPT and VSM in typical biomass-burning areas (e.g., the AMZ, central Siberia, SEA, and western United States). Therefore, the observed decrease in TAOD in central Siberia may be related to the decrease in VSM. Reduced VSM can increase the fire risk by changing vegetation conditions as well as atmospheric moisture conditions. However, the changes in these meteorological factors are not really sufficient to explain the 605 decrease in TAOD in eastern China, as we did not detect significant trends in WS and PPT throughout eastern China, except for the observed increase in VSM to varying degrees in north-central China and SC. The increase in VSM (unfavorable to fire occurrence) in SC may be partially responsible for the decrease in regional SAOD. In SA, which is also a region dominated by anthropogenic aerosols, TAOD shows a negative spatial correlation with PPT. Nevertheless, we see a non-significant decreasing trend in PPT only. The above results suggest that a slight decrease in PPT (by affecting wet deposition) may also 610 further enhance regional aerosol pollution levels in the context of increasing anthropogenic aerosol emissions in SA.
CALIOP's unique vertical observation advantage gives us the opportunity to explore the effect of the interannual variability of meteorological factors on TAOD_PBL and TAOD_FT trends. Next, benefitting from this, we focus on a key https://doi.org/10.5194/acp-2021-467 Preprint. Discussion started: 8 July 2021 c Author(s) 2021. CC BY 4.0 License. question: do the interannual variations in TAOD_PBL and TAOD_FT obey the same relationship between TAOD and meteorological factors? For this purpose, we further distributed TAOD within the PBL and FT and performed the same spatial 615 correlation analysis, as shown in Figs. S12 and S13. The results show that the spatial patterns of the R between the meteorological factors and TAOD_PBL remain basically the same as those of TAOD. In terms of the magnitude of the correlation, the R between the meteorological factors and TAOD_PBL becomes more robust compared to the column TAOD ( Fig. S12). In contrast, the distribution patterns of the R between meteorological factors and TAOD_FT are quite different from those of TAOD_PBL, along with a significant weakening of the strength of the correlations, and the signs of R are even 620 reversed in most regions (Fig. S13). The above results indicate that surface meteorological elements affect the interannual variation in column TAOD mainly by influencing the TAOD_PBL. However, most previous studies on the formation mechanisms of regional air pollution were based on the relationship between column TAOD (as a key proxy for atmospheric pollution) and meteorological factors. Therefore, our study suggests that future aerosol-related research should use the variable TAOD_PBL, which is closely related to near-surface pollutant concentrations, to more realistically elucidate the mechanisms 625 underlying the influence of meteorological factors on the interannual variability of regional air pollution.

Conclusions and implications
The vertical distribution of different types of tropospheric aerosols can influence the Earth's climate system through varying radiative effects. Based on CALIOP Version 4.2 monthly gridded aerosol extinction profiles, averaged over both daytime and nighttime from 2007 to 2019, this paper comprehensively examines the 3D climatological distribution of AECs 630 for total aerosol and its sub-types (dust, PD and smoke) in terms of different altitude regimes, including the whole layer, stratified layers, PBL, and FT. The contribution of TAOD through its partitioning within these different altitude regimes to column TAOD is also quantified, globally and over 12 ROIs. Then, the FoOs of the seven aerosol sub-types are quantified using the detection samples of layer aerosols included in the CALIOP Level-3 products to examine the differences in the vertical distribution of type-dependent FoOs over the globe, land, ocean, and 12 ROIs. On this basis, this study further evaluates 635 the long-term trends in TAOD and its sub-types, and focuses on elucidating the effects and contributions of long-term changes in TAOD_PBL and TAOD_FT on TAOD trends. Finally, the effects of three key meteorological drivers on the interannual variability of TAOD, DAOD, and SAOD partitioned within the whole layer, PBL, and FT are preliminarily explored by performing spatial correlation analysis.
Over the ocean, enhanced TAEC (0.1 km −1 ) is mostly distributed below 1 km, while the vertical distribution of TAEC 640 over land varies significantly at regional scales. The hotspots of land TAEC are generally located in the lower levels of dust source areas and areas dominated by anthropogenic emissions, which show a decreasing trend with altitude. For dust aerosols, the enhancement of DAEC occurs mainly in the lower atmosphere (below 3 km) inside the dust source and its adjacent region.
In contrast, dust aerosols emitted from the source are mixed with downstream anthropogenic aerosols driven by the atmospheric circulation, directly contributing several PDAEC hotspots (e.g., SA and eastern China) over areas of intense 645 anthropogenic activity along the dust transport path. However, the enhanced PDAEC is often limited to below 3 km as it is https://doi.org/10.5194/acp-2021-467 Preprint. Discussion started: 8 July 2021 c Author(s) 2021. CC BY 4.0 License. occur in SC and the EUS. Different from DAOD and PDAOD, the deceased SAOD trend is mainly driven by the aerosols 680 distributed within the FT, especially in SC.
We attempt to interpret these significant regional trends by linking the interannual variability/trends in aerosol loading to changes in meteorological drivers. Our correlation analysis shows that the interannual variability of TAOD, DAOD and SAOD can be related to variations in PPT, VSM and WS in the particular regions. The positive TAOD trend over the ECP is mainly attributable to the significant increase in PPT and significant decrease in WS, which emphasizes that WS and PPT regulate the      Table S1 for more information).

Figure 7.
Multi-year regional-average layer-specific total AODs (integrated at 2-km intervals between 0 and 12 km) for the globe, land, ocean, and 12 ROIs. The colored numbers above each bar represent the relative proportions (%) of the layer-specific AODs to the total AODs.