Long-term multi-source data analysis about the characteristics of aerosol optical properties and types over Australia

. The spatiotemporal distributions of aerosol optical properties and major aerosol types, along with the vertical distribution of major aerosol types over Australia, are investigated based on multi-year AERONET observations at nine 10 sites, the Moderate Resolution Imaging Spectroradiometer (MODIS), Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2), Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP), and back-trajectory analysis from the Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT). The annual aerosol optical depth (AOD) at most sites showed increasing trends (0.002-0.028 yr -1 ) except for that at three sites of Canberra, Jabiru, and Lake Argyle, which showed decreasing trends (-0.004 - -0.002 yr -1 ). In contrast, the annual Ångström exponent 15 (AE) showed decreasing tendencies at most sites (-0.044 - -0.005 yr -1 ). The results showed strong seasonal variations in AOD with high values in the austral spring and summer and relatively low values in the austral fall and winter, and weak seasonal variations in AE with the highest mean values in the austral spring at most sites. Spatially, the MODIS AOD showed obvious spatial heterogeneity with higher values appeared over the Australian tropical savanna regions, Lake Eyre Basin, and southeastern regions of Australia, while low values appeared over the arid regions in western Australia. 20 Monthly averaged AOD increases from August to next austral spring peak (typically December-January), and decreases during the March-July. Classification of Australian aerosols revealed that the mixed type of aerosols (biomass burning and dust aerosol) are dominated in all seasons at nine sites, followed by biomass burning aerosol and dust aerosol. The MERRA-2 showed that carbonaceous over northern Australia, dust over central Australia, sulfate over densely populated northwestern and southeastern Australia, and sea salt over Australian coastal regions are the major types of atmospheric 25 aerosols over Australia. The CALIPSO showed that polluted dust is the dominant aerosol type detected at heights 0.5 - 5 km during all seasons. Australian aerosol has similar source characteristics due to intercontinental transport of aerosols a better understanding of the aerosol characteristics and their impacts on climate over Australia. Ground-based observations can provide more accurate data to estimate aerosol properties, while remote sensing technique provides a better understanding of aerosol properties at large scale. In this study, multi-year aerosol optical properties obtained from nine ground-based sun photometers, along with the MODIS AOD product over Australia are 75 and 0.087 at Adelaide Site 7, Fowlers Gap, and Birdsville, respectively. The findings were consistent with the result of MODIS AOD, which also showed higher values during September and February than during other months in the central plains. It was worth mentioning that the bimodal trend weakened at Fowlers Gap and Adelaide Site 7 compared to other sites, especially the first peak during September. This was likely due to the their farther away from the Simpson desert, which 270 was the most active dust source with high aerosol loading during September - February. At Birdsville, coarse mode such as biomass burning, dust, marine and mixtures over Australia. The frequency distributions of AE were skewed towards large AE values at all sites except for Lake Lefroy, Learmonth, and Lucinda, where AE followed the normal probability were transported from the northwestern Indian Ocean and southern deserts, respectively. In western Australia, it was evident that mixed aerosols, biomass burning, and urban/industrial aerosols were dominant during all seasons at Lake Lefory. The mixed aerosols and dust aerosols were abundant during all seasons at Learmonth, while the biomass burning and urban/industrial aerosols were also observed in 315 biomass burning and urban/industrial aerosols during all seasons. In addition, clean marine and dust aerosols were observed during fall and winter. The back trajectories ending at Canberra showed 54.03% of airflow from southwest and 45.97% of airflow from southeast, suggesting a possible transport of 335 biomass burning and clean marine aerosols from forest regions and ocean. Different aerosol types (e.g. clean marine aerosols, dust, mixed aerosols) were observed at Lucinda during all seasons. The back trajectories ending at Lucinda

9 0.002±0.035 yr -1 , -0.004±0.057 yr -1 , respectively. In particular, the highest AOD (0.20) at Canberra was observed in 2003 followed by a decrease to 0.06 in 2004. This is mostly related to the wildfires of southeastern Australia in January 2003, which generated large amounts of smoke aerosols, leading to a maximum in AOD during that observation period (Mitchell et al., 2006). Among the six sites with the increasing trend of AOD, Birdsville is with a more evident AOD increasing 165 trend of 0.028±0.065 yr -1 during the period 2013-2020. It is worth mentioning that significant increasing trends of AOD are observed during the period 2019-2020 at most sites, such as Adelaide Site 7, Birdsville, Fowlers Gap, and Lucinda.
The increasing trend in AOD is related to the frequent fire activities in Australia from September 2019 to January 2020.
In addition, optical and physical properties of aerosols during the Australia wildfires in 2019 will be discussed in detail in our future study. The annual means of AE were 0.74-1.31 over Australia. The annual AE showed decreasing trends at 170 most sites during the observation period except for Lake Lefory, at which it showed an increasing trend of 0.013±0.216 yr -1 . This result indicates that the size of aerosol increased at Lake Lefory, while the size of the aerosol decreased at other sites. The annual AOD and AE at Jabiru (0.15; 1.20) and Lake Argyle (0.14; 1.27) were higher than those at other sites (0.06-0.10; 0.74-1.13) in Australia, which can be explained by the extensive and frequent wildfires over the tropic North of Australia. Moreover, the annual AOD and AE at Jabiru and Lake Argyle presented significant interannual variations 175 in amplitude. Similar results have also been reported by Radhi et al. (2012) and Mitchell et al. (2013). This is likely a consequence of the relatively high rainfall during the wet season and large smoke emissions during the subsequent burning season. The rainfall likely suppressed the smoke emissions due to the high level of moisture in the air, but it also promoted the growth of vegetation, leading to an increase in smoke emissions during the subsequent dry seasons (Mitchell et al., 2013). 180 Spatial distribution of annual averaged ( Fig. 4 (a-s)) and multi-year averaged ( Fig. 4 (t)) MODIS AODs for the years of 2002-2020 indicates that, overall, the patterns of AOD spatial distributions were similar from 2002 to 2020. However, the magnitude of the spatial AOD distributions varied to some extent. It is worth mentioning that the years of 2002 and https://doi.org/10.5194/acp-2020-921 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License.
high AOD values were observed in Australian Capital Territory (0.152) and Victoria (0.115) in 2003. This was mostly caused by 2003 wildfires in southeastern Australia, which generated large amounts of smoke aerosols (Mitchell et al., 2006). In 2009, there were extreme dust storms in central Australia (Mukkavilli et al., 2019), which lead to a large increase in AOD from 0.038 (2008)  be related to the more frequent wildfire activities during this period in Australia.

Seasonal variations of aerosol optical properties
Seasonal variations of AOD and AE at nine sites in Australia are presented in Fig.6. It is clear that there was a seasonal cycle in AOD at nine sites with high values in spring and summer, and low values in fall and winter. Similar results were reported by Mitchell et al. (2013). Further, the highest seasonal average AOD values were observed in spring 215 at Birdsville (0.09), Jabiru (0.22), Lake Argyle (0.23), and Lucinda (0.13), while they were observed in summer at the other five sites (0.07-0.11). The seasonal variations in AOD observed by AERONET at nine sites were similar to that observed by MODIS at the corresponding sites (Fig.7). Similarly, Mitchell et al. (2013) found that the AOD values at Wagga and Canberra peaked in summer, while AOD values peaked in spring at sites that are located in the arid zone. This is due to the increasingly forested and bushfire-prone characteristics at the more easterly sites (Mitchell et al., 2013). 220 However, the seasonal variation in AE was different from that in AOD. There were no obvious seasonal variations in AE at the nine sites. The maximum seasonal mean AE values (0.92-1.43) were observed in spring at all sites except for Canberra. Further, the seasonal mean AE was greater than 1.0 over all seasons at Canberra, Jabiru, Lake Argyle, Fowlers Gap, and Birdsville, while the mean AE values were less than 1 over all seasons at the Adelaide Site 7 and Luncinda. In addition, at Learmonth and Lake Lefroy, high AE values (0.98-1.07) were observed in spring and fall, and low values 225 (0.56-0.99) were observed in summer and winter.
Seasonal distributions of MODIS AOD are shown in Fig. 7. The spatial distribution of MODIS AOD in each season https://doi.org/10.5194/acp-2020-921 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License. was similar to the annual-averaged spatial distribution pattern. High AODs were observed in Spring (~0.048) and Summer (~0.058), while low AODs were observed in fall (~0.031) and winter (~0.029). The main contributors to the high AODs in spring and summer were smoke emissions from biomass burning, dust storms, and marine biogenic emissions (Rotstayn 230 et al., 2010). The occurrence frequency and intensity of dust storm activities and wildfires decreased during fall and winter, resulting in low AOD values over Australia. One consistent feature among the spatial distributions in AOD observed in each season were the high AODs (with different magnitude) in the northern, central, southwestern, and southeastern Australia. wind speeds during spring were observed in northern Australia (north of 18°S), which may lead to the increasing AOD from biomass burning and long-range transport marine biogenic emissions. During summer, the decreasing trend in AOD was significant over northwestern regions, consistent with the large increase in precipitation and decrease in wind speeds.
However, the increase in AOD in eastern and southeastern regions during summer could be associated with the increase of biomass burning and sea salt aerosols that were transported from the Pacific Ocean. 240 July. In addition, significant differences in monthly AOD variations were found among regions of Australia. Therefore, 245

Monthly variations of aerosol optical properties
we classified the nine AERONET sites as four categories based on their locations, which are (1) Jabiru and Lake Argyle in northern Australia, (2) Learmonth and Lake Lefory in western Australia, (3) Adelaide Site 7, Birdsville, and Fowlers Gap in central Australia, and (4) Canberra and Lucinda in eastern Australia. In order to avoid bisect the aerosol peak, the timeline was adopted from July to June (rather than January to December) (Mitchell et al., 2017). In northern Australia, https://doi.org/10.5194/acp-2020-921 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License.
spring. The trajectories ending at Lake Lefory showed that biomass burning and urban/industrial aerosols could be transported by the western and southwestern airflows. The back trajectories ending at Learmonth showed 22.76% of airflow from the eastern deserts and 32.47% of airflow from southern inland. These results indicated that dust in eastern deserts (e.g. Gibson Desert Great Victoria Desert and Lake Eyre Basin) and urban/industrial aerosols from southern cities (e.g. Perth) could be transported to the Learmonth. There were 44.77% of airflow from the Indian Ocean. However, clean 320 marine aerosol was seldom found at Learmonth due to the fact that it is situated on the North-Western dust pathway (Strong et al., 2011). In central Australia, the mixed aerosols, biomass burning, and urban/industrial aerosols are found during all seasons. Dust aerosols with high AOD (>0.15) and low AE (<0.5) were observed at Birdsville and Fowlers Gap during spring and summer, while clean marine aerosols were observed during fall and spring. Different aerosol types can be found under relatively clean atmospheric conditions with AOD <0.2 and AE < 1.5 at Adelaide Site 7. The back 325 trajectories ending at Birdsville showed that the southern (25.42%) and southwestern (19.28%) airflows may bring the clean marine aerosols and dust aerosols from the Indian Ocean and Lake Eyre Basin to Birdsville, while the eastern (25.42%) and southeastern (33.31%) airflows may bring the biomass burning aerosols to Birdsville. The back trajectories ending at Fowlers Gap and Adelaide Site 7 showed that more than 67% of airflow were originated mainly from the Indian Ocean, which could transport clean marine aerosols to the two sites. Further, there were 32.71% and 29.53% of airflows 330 at Fowlers Gap and Adelaide Site 7 from the southeastern Australia, which implies a possible transport of biomass burning aerosols. In eastern Australia, the Canberra site exhibited a wide range of AOD values and high AE values ranging from ∼1.5 to∼2.5, which indicated the existence of biomass burning and urban/industrial aerosols during all seasons. In addition, clean marine and dust aerosols were observed during fall and winter. The back trajectories ending at Canberra showed 54.03% of airflow from southwest and 45.97% of airflow from southeast, suggesting a possible transport of 335 biomass burning and clean marine aerosols from forest regions and ocean. Different aerosol types (e.g. clean marine aerosols, dust, mixed aerosols) were observed at Lucinda during all seasons. The back trajectories ending at Lucinda https://doi.org/10.5194/acp-2020-921 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License. illustrated that the airflows from southwest and northwest may bring dust and biomass burning to Lucinda, respectively.
As indicated earlier, the relationship between AOD and AE can be used to distinguish aerosol types. However, it is difficult to discriminate biomass burning from urban/industrial using this relationship alone (Mishra and Shibata, 2012). 340 Further, a better analysis can be made through correlations between SSA and AE.  2017) also found that the aerosol sources over Australia were driven by mechanisms that do not vary greatly either on regional or continental scales by 345 analyzing the monthly mean data at 22 sites. They pointed out this is associated with the intercontinental transport of biomass burning aerosol. In northern Australia, biomass burning aerosols accounted for a relatively large proportion at Jabiru and Lake Argyle in spring, autumn and winter, which was related to the biomass combustion during the dry season.
In addition, dust aerosols were also observed during summer, autumn and winter. Fire events were also a main provider of dust aerosols in the atmosphere because the pyro-convection could accelerate the dust entrainment during the fire 350 events. The results indicated that the aerosols at the two sites were affected by the fire-related dust emissions and dust transported from the southeastern deserts. In western Australia, the urban/industrial aerosols were observed during all seasons, making fine mode aerosols dominated throughout the year (Section 3.1.3) at Lake Lefory. The urban/industrial aerosols over Lake Lefory may originate from Perth, which were transported to Lake Lefory by the western airflow. The Learmonth was found with heavy loadings of dust aerosols, which is consistent with our earlier result. In eastern Australia, 355 dust aerosol was another dominant aerosol type at Lucinda. The classification results may be inaccurate due to the lack of SSA data at Canberra. However, many previous studies showed that urban/industrial aerosols and biomass burning aerosol were the main components of aerosols at Canberra (Mitchell et al., 2006;Provenç al et al., 2017). https://doi.org/10.5194/acp-2020-921 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License.

Spatio-temporal characteristics of different aerosol types
MERRA-2 data was used to determine the contribution of different kinds of aerosols to the AOD over Australia. 360 Considering the similar emission sources of Organic Carbon and Black Carbon aerosols, we combined the two as carbonaceous aerosol for analysis. Carbonaceous aerosol and sulfate aerosol are produced from biomass burning, fossil fuel combustion, biofuel consumption, while dust and sea salt aerosols mainly originate from natural emissions (Randles et al., 2017). The spatial distributions of carbonaceous, dust, sulfate, sea salt AOD over Australian continent are shown in Fig. 16. Carbonaceous aerosols were mainly distributed in northern and southeastern Australia. Carbonaceous aerosols 365 in these two regions could be highly related to the fires in the grasslands, forests, and croplands during the dry seasons.
Dust aerosols were mainly distributed in the central plains of Australia. The dust aerosols over the central plains primarily originated from the Lake Eyre Basin, one of the southern hemisphere's most significant dust sources. Mukkavilli et al.
(2019) found that the spatial distributions of dust aerosols across Australia demonstrated concentrated values in the Lake Eyre Basin by using the ECMWF Monitoring Atmospheric Composition and Climate (MACC) reanalysis product. Sulfate 370 aerosols were mainly observed in the northwestern (such as Darwin) and southeastern (such as Melbourne, Canberra, and Sydney) Australian coastal regions, where human activities were highly frequent. The near coastal region generally had higher sea salt aerosol loadings than the continental interior region due to the land-sea breeze effects. Prijith et al. (2014) found that the higher wind speed would lead to more sea salt aerosol formation, and the corresponding shorter transport time would lead to weaker loss. Sea salt aerosols had a relatively large impact on the northern coastal regions of Australia, 375 which were mostly due to the higher wind speeds in the northern coast (Fig.8). Overall, carbonaceous over northern Australia, dust over central Australia, sulfate over densely populated northwestern and southeastern Australia, and sea salt over Australian coastal regions were the major types of atmospheric aerosols in Australia.
To determine the temporal distributions of different kinds of aerosols in Australia, seasonal variation analyses of aerosols are performed. Figure 17 shows the seasonal variations of carbonaceous, dust, sulfate, and sea salt aerosols in 380 https://doi.org/10.5194/acp-2020-921 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License.