Statistical aerosol properties associated with fire events from 2002
to 2019 along with a case analysis in 2019 over Australia

Abstract. Wildfires are an important contributor to atmospheric aerosols in Australia and could significantly affect regional and even global climate. This study investigates the impact of fire events on aerosol properties along with the long-range transport of biomass burning aerosols over Australia using multi-year measurements from Aerosol Robotic Network (AERONET) at ten sites over Australia, satellite dataset derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) and Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP), reanalysis data from Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2), and back-trajectories from the Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT). Strong correlation (0.62) was found between fire radiative power (FRP) and aerosol optical depth (AOD) over Australia, suggesting the significant contribution to aerosols from fires. The fire count, FRP, and AOD showed distinct and consistent interannual variations with high values during September–February (Biomass Burning period, BB period) and low values during March–August (non-Biomass Burning period, non-BB period) every year. The annual average contribution of carbonaceous, dust, sulfate and sea salt aerosols to total aerosol were 26.24 %, 23.38 %, 26.36 % and 24.02 %, respectively. The results from AERONET, MODIS, and MERRA-2 showed that AOD values significantly increased with fine mode aerosol dominated during BB period, especially in northern and southeastern Australia. Further, Carbonaceous aerosol was the main contributor to total aerosols during BB period, especially in September–December when carbonaceous aerosol contributed the most (30.08–42.91 %). The great fires during the BB period of 2019/2020 further demonstrated significant impact on aerosol properties, such as the extreme increase in AOD for most southeastern Australia, the dominance of fine particle aerosols, and the significant increase in carbonaceous and dust aerosols in southeastern and central Australia, respectively. Moreover, smoke was found as the dominant aerosol type detected at heights 2.5–12 km in southeastern Australia in December 2019 and at heights roughly from 6.2 to 12 km in January 2020. In contrast, dust was detected more frequently at heights from 2 to 5 km in November 2019, January, and February 2020. A case analysis revealed that significant changes in aerosol properties including aerosol loading, aerosol particle size, aerosol type in central Australia could be caused during the BB period of 2019/2020 due to the long-range transport of biomass burning aerosols from eastern and southern Australia.



35
Biomass burning is a major global source of fine carbonaceous aerosols in the form of organic carbon (OC) and black carbon (BC) (Vermote et al., 2009). Biomass burning aerosols are mainly generated by man-made fires such as agricultural burning, deforestation, and biofuels combustion, and natural fires that include lightning-induced wildfires.
The global major source regions of biomass burning aerosols are the sub-Saharan Africa, South America, Southeast Asia, North Australia, and the boreal forest in north hemisphere (Ito and Penner, 2004;Mitchell et al., 2006). Abundant 40 aerosols emitted from fires could affect the Earth's climate system by both direct and indirect effects (Jacobson, 2014).
Biomass burning aerosols can not only warm the atmosphere and cool the Earth's surface by reducing sunlight through absorption and scattering, but also can modify cloud microphysical properties by acting as cloud condensation nuclei or ice nuclei (Garrett and Zhao, 2006;Ramanathan and Carmichael, 2008;Fujii et al., 2015;Zhao and Garrett, 2015;Grandey et al., 2016;Zhao et al., 2018;Yang et al., 2019). Moreover, biomass burning aerosol can cause environmental 45 pollution and thus affect public health (Crippa et al., 2016;He et al., 2016;Yang et al., 2016;Yang et al., 2018).
The tropical north Australian is dominated by savanna ecosystem with grassland and woodland, while the southeastern Australia is largely vegetated by eucalyptus forest. Meyer et al. (2012) found that approximately 550,000 km 2 of tropical and arid savannahs are burned each year in Australia. Australia contributes about 6-8% of global carbon emissions from biomass burning (van der Werf et al., 2006;Meyer et al., 2008). It is also found that biomass burning 50 aerosol is the dominant aerosol type in northern and southeastern Australia during spring and summer, respectively (Mitchell et al., 2013;Yang et al., 2020b). Northern Australia showed a distinct seasonal cycle in aerosol optical depth (AOD) and Ångström exponent (AE) with the highest aerosol loading occurring in spring (burning season) and the highest AE values occurring in August (Mitchell et al., 2013). Mitchell et al. (2006) investigated the characteristics and radiative impact of the smoke aerosol during the Canberra firestorm of January 2003. They found that the single 55 scattering albedo of the aerosols was~0.96 and the daily mean forcing during the week following the firestorm was a cooling of 50 W m -2 . Additionally, air quality was also significantly affected by smoke from biomass burning over Australia (Meyer et al., 2008;Reisen et al., 2011;He et al., 2016).
During the period September 2019-February 2020, large areas of southeastern Australia were ravaged by deadly wildfire, causing extensive damages to the property and environment. According to the report of Filkov et al. (2020), a 60 total of 5,595,739 and 1,505,004 hectares were burned, 2,475 and 396 houses were destroyed in New South Wales and Victory during the 2019/2020 fire season, which was the most devastating fire season in the states' history. During the wildfire periods, a large amount of aerosols containing smoke, dust, and other burning material were uplifted into the atmosphere, and were even transported around the world, which significantly affected the regional and global atmospheric chemistry, carbon cycle and surface radiation budget. For example, Torres et al. (2020) found that the great 65 fires in the New South Wales of Australia injected large amount of carbonaceous aerosols in the stratosphere. These wildfires significantly affected local air quality and visibility, and can also be transported to the New Zealand, resulting in degraded air quality in some cities. Ohneiser et al. (2020) found that the smoke from extreme wildfires in southeastern Australia crossed the Pacific Ocean and arrived at Punta Arenas in South America.
There are many studies that have investigated the fire events and their association with enhancing aerosol loading 70 and degrading air quality in Australia (Mitchell et al., 2006;Luhar et al., 2008;Meyer et al., 2008;Bouya and Box, 2011;Mitchell et al., 2013;Mallet et al., 2017;Chen et al., 2019). However, most of these studies are carried out based on observations at specific region/site or during short time period. Long-term statistical analysis about the properties of fire-induced aerosols is highly demanded over a large domain. In addition, the northern and southeastern Australia have frequently suffered from wildfires, especially during the fire season of 2019/2020. It is also highly valuable to fully 75 understand the impacts of the huge fire events occurred in 2019/2020 over the southeastern Australia. In this study, multi-year aerosol products from ground-based observations (i.e. Aerosol Robotic Network (AERONET)) and https://doi.org/10.5194/acp-2020-1139 Preprint. Discussion started: 9 November 2020 c Author(s) 2020. CC BY 4.0 License.
satellite-based technique (e.g. Moderate Resolution Imaging Spectroradiometer (MODIS), Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO)) are used to study the impacts of long-term fire events on aerosol properties over Australia. In addition, we further investigated the aerosol properties associated with the huge fire events 80 in 2019/2020 over southeastern Australia, including the impacts on aerosols over central Austria caused by the long-range transport of biomass burning aerosols.
The paper is organized as follows. Section 2 describes the study area, data and method. Section 3 shows the long-term statistical aerosol properties associated with fire events in Australia, aerosol characteristics in Australia during the period in 2019/2020 with huge fire events, along with the aerosol contribution for a case study from long-range 85 transport of biomass burning aerosol. Section 4 summarizes the findings of the study.

Study area
Australia is situated in Southern Hemisphere between the India and South Pacific Oceans. Australian climate varies greatly throughout the eight states and territories and can be divided into six climatic zones (i.e. temperate, 90 grassland, desert, subtropical, tropical and equatorial) based on modified Köppen climate classification system ( Fig.   1(a)). The northern Australia lies in the tropical zone with a wet and dry season. The central of Australia is a desert region with high temperature, evaporation, and low amounts of rain during summer. The southern Australia belongs to the temperate zone with hot dry summer and cold winter. In this study, the observations at ten AERONET sites across Australia were adopted for analysis. Table 1 shows the detailed information of site locations and the available data 95 period at each site. Fig.1(b) shows the Kernel density estimation (KDE) of fire spots over Australia from June 2019 to May 2020, and the Gray shade area is the selected domain for vertical profile analysis in this study.

100
The AERONET is a global aerosol monitoring network based on ground-based sun-sky photometers. It autonomously measures direct-sun irradiance and directional sky radiance and provides long-term and high-quality datasets of aerosol optical, microphysical, and radiative properties (Holben et al., 1998). The uncertainty of AERONET direct measurement AOD is ±0.01 at visible wavelengths and ±0.02 at near-UV wavelengths (Dubovik et al., 2000). In this study, level 2.0 quality controlled and cloud screened data of the aerosol spectral deconvolution algorithm (SDA) 105 AOD, AE, volume size distributions from AERONET Version 3 were used. However, Level 1.5 data were used from January 2018 to May 2020 due to the lack of level 2.0 data at a few sites during the period. In addition, the Canberra site in southeastern Australia ceased to provide data in August 2017, while the neighboring Tumbarumba site began providing observations from July 2019. Due to the similarity of aerosol sources at the two sites, we combined their data to analyze aerosol properties in southeastern Australia.

Satellite remote sensing data
The MODIS is a 36-band imaging radiometer onboard the NASA EOS (Earth Observation System) Terra and Aqua platforms and provides globally long-term atmospheric data to monitor the characteristics and dynamics of aerosols (Remer et al., 2005). Wei et al. (2019) reported that the MODIS DB algorithm over Australian continent can retrieve 1.7-1.9 times more AODs than the DT algorithm with more than 82% of AOD retrievals falling within the expected 115 error envelopes and with small root mean squared error (RMSE) (0.052). Therefore, in this study, the record named "Deep_Blue_Aerosol_Optical_Depth_550_Land_Best_Estimate" from Aqua MODIS C61 level2 aerosol product (MYD04_l2) with a spatial resolution of 10 km during the period July 2002-May 2020 was used to investigate the AOD spatio-temporal variations. The MODIS active fire product (MCD14ML) data refer to the active fire hotspots https://doi.org/10.5194/acp-2020-1139 Preprint. Discussion started: 9 November 2020 c Author(s) 2020. CC BY 4.0 License. determined using the thermal anomalies at 1 km pixel resolution (Giglio et al., 2016). The fire hotspot count and fire 120 radiative power (FRP) from the MCD14ML data with confidence level greater than 50% from January 2002 to May 2020 were used to explore their relationship with AOD. The CALIPSO was launched in April 2006 equipped with CALIOP (Cloud-Aerosol Lidar with Orthogonal Polarization), and observes global aerosol-cloud vertical distributions at 532 and 1064 nm during both day and night, which provides a possibility to study the impact of clouds and atmospheric aerosols on the Earth's weather, climate, and air quality (Winker et al., 2003). Omar et al. (2013) found that 125 when cloud cleared and extinction quality controlled CALIOP data was compared with AERONET data with AOD less than 1.0, the mean relative difference between the two measurements was 25% of AERONET AOD. Additionally, the CALIPSO AOD showed a good agreement (R=0.65) with AERONET AOD at Lake Argyle in northern Australia. In this study, the Vertical Feature Mask (VFM) data from level 2 profile product (V4.20) were used to obtain the information of aerosol types (e.g., smoke, dust, polluted dust, etc.) over Australia.

Reanalysis data
The Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2) is the NASA's latest global atmospheric reanalysis product, which is produced by the NASA Global Modeling and Assimilation Office (GMAO) (Gelaro et al., 2017). MERRA-2 provides long-term aerosol assimilation with the horizontal resolution of 0.5°× 0.625°on 72 sigma-pressure hybrid layers between the surface and 0.01 hPa from 1980 to the present (Randles et al., 135 2017). In this study, the monthly MERRA-2 products were used to analyze the spatial and temporal variations of aerosols over Australia from January 2002 to May 2020. ECMWF (European Centre for Medium-Range Weather Forecasts) Reanalysis v5 (ERA-5) (Hersbach and Dee, 2016) is the latest climate reanalysis product developed by the European Centre for Medium-Range Weather Forecasts. ERA-5 provides global hourly and monthly atmospheric data with a spatial resolution of 0.25°×0.25°at 37 pressure levels, which is widely used in various studies 140 Albergel et al., 2018;Wang et al., 2019). ERA-5 adds additional characteristics to ERA-Interim reanalysis, making it even richer in climate information (Albergel et al., 2018). In this study, monthly U-wind and V-wind from ERA-5 dataset were used for meteorology analysis.

Method
The relationships among AOD, fire count, fire radiative power (FRP), and total precipitation over Australia were 145 investigated by using MODIS AOD, fire hotspot count and fire radiative power data from MCD14ML with confidence level greater than 50%, and total precipitation data from ERA-5. The total, fine, coarse mode AOD, and AE from AERONET, and the AOD products from MODIS and MERRA-2 were utilized to explore their spatio-temporal variations associated with the fire events over Australia. A case of long-range transport of biomass burning aerosol was selected for case study based on the AERONET observations at Birdsville. The 72-h back trajectories for Birdsville site 150 with altitudes of 200 and 500 m above ground level (a.g.l.) were simulated to investigate the potential aerosol sources over the central Australia. The ERA-5 wind fields, the MODIS fire and AOD products, and CALIOP AOD products were used to show the regional transport of biomass burning aerosols from southeastern Australia to central Australia during the case period.
To evaluate the accuracy of MERRA-2 monthly AOD products, the monthly AERONET AOD values were 155 calculated based on the daily measurements. Several statistical variables were used in this study for MERRA-2 AOD performance, including the correlation coefficient (R), the RMSE, and relative mean bias (RMB). R was used to analyze the correlation between AERONET and MERRA-2 AOD. The RMSE (Eq. (1)) and RMB (Eq. (2)) were used to evaluate the uncertainty of MERRA-2 monthly AOD products.

Results and discussion
3.1 Long-term statistical aerosol properties associated with fire events in Australia

Variations and relationships of fire events, precipitation, and aerosol amount
The fire counts and FRP values provide information of the fire frequencies and emissions. The FRP is increasingly 165 used to quantify the regional and global biomass consumption, trace gas and aerosol amount. in spring and summer. Furthermore, the peak of the monthly mean AODs coincides with the FRP peak in the months of October-January each year. However, the correlation between the AOD and fire counts was slightly weaker (0.43) for the whole nineteen-year period. This is related to the intensive and frequent fire activities in northern savanna and southeastern forest regions in spring and summer. These results indicated similar emission from fires and regional transport of biomass burning aerosol in Australian continent, particularly during spring and summer. The finding was 180 consistent with that of Mitchell et al. (2013Mitchell et al. ( , 2017 and Yang et al. (2020b). It was worth mentioning that the frequent dust activities over Central Australia also contributed to the increase in AOD in spring and summer, which may weaken the correlation coefficient between AOD and FRP or fire counts. The total precipitation was found with high values during the period December-March, and low values in other months. In most cases, the peaks of AOD were observed earlier than the peaks of total precipitation during the period January 2002-May 2020, which is well consistent with the 185 fact that high AOD values (including the maxima) occur in the dry season (typically April -November).

Aerosol optical depth and size distribution
Generally, the seasonality of MODIS AOD and FRP over Australia showed that the high mean AODs coincided with the FRP during the period September-February every year. Moreover, many studies suggested that active fire events were frequent and intense, and biomass burning aerosol was the dominant type over Australia during the period 190 September-February. For example, Mitchell et al. (2013) showed that high AOD values (>0.1) were observed at three sites (i.e. Lake Argyle, Jabiru, and Darwin) in northern Australia during the months September-February with maximum values of~0.28 in October. In addition, higher AOD values were also commonly observed during the months September-February in southeastern Australia (Mitchell et al., 2017;Yang et al., 2020b). Therefore, we defined the periods of September-February as Biomass burning (BB) period, while the periods of March-August as 195 non-Biomass burning (non-BB) period. Figure 3 shows the temporal variations of annual mean AERONET and MODIS AOD during BB period and non-BB period at ten sites. In general, Aqua MODIS DB AOD and AERONET AOD showed the similar annual variations during both the BB period and non-BB period at most sites. However, there were differences in the trends of MODIS AOD and AERONET AOD at a few sites (e.g., Adelaide Site 7, Birdsville, Lake Lefory), which is mainly due to the missing observations of AERONET at those sites. (i.e., Jabiru and Lake Argyle) had higher averaged AOD values both during BB period and non-BB period compared to other sites, which was related to the wildfire activities during BB period and active sea salt production driven by strong winds during non-BB period. Furthermore, the two sites exhibited significant interannual variations during BB period, 210 especially at Lake Argyle. Additionally, a significant increase in AOD was observed during the BB period of 2019/2020 at sites in southeastern Australia, such as Tumbarumba, Fowlers Gap, and Adekaide Site 7, which can be explained by the intensive and frequent fire activities in this region during that period. 230 Figure 5 shows the monthly averaged aerosol volume-size distribution at nine sites over Australia. In general, the aerosol particle size distributions showed a bimodal lognormal pattern with radius smaller than 0.6 μm as fine mode aerosol and with radius larger than 0.6 μm as coarse mode aerosol (Dubovik et al., 2002;Zheng et al., 2017;Zheng et al., 2019). It is clear that the volume sizes with peak volume concentrations for both fine mode and coarse mode aerosols were higher during BB period than during non-BB period at ten sites. In northern Australia, the peak volume 235 concentrations of fine mode aerosols at Jabiru and Lake Argyle were the highest in November and October with values of 0.029 and 0.046 μm 3 /μm 2 , while the peak volume concentrations of coarse mode aerosols were the highest in December and January with values of 0.041 and 0.032 μm 3 /μm 2 , respectively. The result suggested that both fine mode and coarse mode volumes significantly increased during BB period, which was mostly related to the biomass burning and dust aerosols from wildfires. Note that the sea salt aerosols from ocean would also contribute to the differences in 240 volume size distributions among various sites, or between BB and non-BB periods. For example, the peak volume concentration values of fine mode (coarse mode) aerosols at Jabiru was lower (higher) than that at Lake Argyle. This was because of Jabiru's closer proximity to the coast, where sea salt aerosols have a greater impact on atmosphere during the wet season (typically from November to April) (Radhi et al., 2012). In western Australia, the peak values of fine mode (coarse mode) aerosol volume concentration at Learmonth and Lake Lefory were the highest in September 245 https://doi.org/10.5194/acp-2020-1139 Preprint. Discussion started: 9 November 2020 c Author(s) 2020. CC BY 4.0 License.
(October), and the lowest in January (June). The coarse mode volume concentrations were obviously larger at Learmonth than at Lake Lefory, peaking at 0.02 to 0.03 μm 3 /μm 2 in October, compared with~0.005 to 0.015 μm 3 /μm 2 at Lake Lefory. The higher coarse mode volume concentration was associated with the site's location in the North-Western dust pathway from the Australian interior deserts (e.g. the Gibson Desert and Great Victoria Desert). In central Australia, there was a significant increase in volume concentrations of coarse mode aerosols during BB period at 250 Birdsville, Fowlers Gap, and Adelaide Site 7, due to the frequent dust activities during the period. Meanwhile, increases in volume concentrations of fine mode aerosols were observed at the three sites, which may result from long-range transport of biomass burning aerosols (Yang et al., 2020b). In eastern Australia, the coarse mode aerosols were dominant in almost all seasons at Lucinda. However, the volume concentration of fine mode aerosols increased during BB period with peak value of 0.012 μm 3 /μm 2 in November. The fine mode and coarse mode volume concentrations 255 were both higher in December-January at Canberra during BB period, which was mostly related to the fine smoke particles, fire-induced dust particles from forest fires, and sea salt particles from ocean. During the BB period of 2019/2020, the significant impact of the huge fires on aerosol particle size distributions in southeastern Australia was evident. Significant increase in fine mode aerosol concentrations can be seen at Tumbarumba, which was located in the area with high frequency and intensity of fires, as well as at Fowlers Gap and Birdsville due to the regional transport of 260 biomass burning aerosol. Additionally, a significant increase in coarse mode aerosol concentrations were found at Birdsville. Furthermore, during the BB period of 2019/2020, the peak value of coarse mode aerosol volume concentration at Birdsville was much higher (~0.058 μm 3 /μm 2 ) than the multi-year averaged maximum in January (~0.025 μm 3 /μm 2 ).    were generated by the biomass burning. Additionally, the dominance of fine mode aerosols was observed at Birdsville and Fowlers Gap, which was primarily due to the long-range transport of biomass burning aerosols from eastern and southeastern Australia and will be discussed in Section 3.3. Figure 12 shows the occurrence frequency profile of each aerosol type in each month during the period September 2019-February 2020 from CALIPSO observations over the selected domain in southeastern Australia (gray shadow in Fig.1(b)). During this period, polluted dust was abundant at 350 heights roughly from 0.5 to 4 km, with peak occurrence frequency at heights from 0.5 to 2 km. High occurrence frequency of elevated smoke was observed at heights from 2 to 4 km. During the strongest biomass burning period (i.e., December), smoke was the dominant aerosol type detected at heights 2.5-12 km. Further, smoke was the dominant aerosol detected at heights roughly from 6.2 to 12 km in January. The result was consistent with the finding of Ohneiser et al. (2020), who reported that smoke injected over the source regions (i.e. Southeastern Australia) at heights below 10 355 km widely remained in the troposphere, and then was injected into higher heights (12-20 km) and transported to South America by the eastward advection. In general, the dust occurrence frequency was higher at the heights of 0-2 km and decreased with the increase of height. However, the occurrence frequency of dust increased at heights roughly from 2 to 5 km in November, January, and February. This increase also explained the relatively high coarse mode aerosol volume concentrations at the sites in southeastern Australia (Fig.5), which was mostly related to the fire-induced dust emissions 360 caused by the pyro-convection during extreme fire events. Clean marine aerosol was the dominant type detected below 0.5 km, due to the proximity of this domain to the West Pacific Ocean. It was evident that the peak occurrence frequencies of clean marine decreased during the period December-January, which was also related to the strong fire activities.

Aerosol properties over Australia during the 2019 huge fire events
Observations from MODIS and AETONET revealed that the aerosol loading over Australia continent, especially in

Case study of long-range transport of biomass burning aerosol
Many studies showed that dust is the major type of atmospheric aerosol over central Australia (Qin and Mitchell, 2009;Mehta et al., 2018;Mukkavilli et al., 2019). However, the central Australia is also affected by the long-range transport of biomass burning aerosols during BB period. During the BB period of 2019/2020, large amounts of smoke 385 plumes produced by southeastern Australian fire were found transported to the South Pacific and even South America by the prevailing westerly winds (Ohneiser et al., 2020). Meanwhile, those smoke plumes may be transported to the inland and therefore affect the regional aerosol properties over central Australia. From the Sections 3.1 and 3.2, it is evident that the aerosols over central Australia were significantly affected by the fine mode aerosols during the 2019/2020 fire events, implying the potential contribution form fire events. Here, we use a case study to illustrate the 390 long-range transport of biomass burning aerosols, which occurred on 18-26 December 2019. These results suggested that Birdsville, which is located at the Lake Eyre Basin, was likely significantly affected by the fine mode aerosol transported from a long distance.

400
We next investigated how the fine mode aerosol was produced and transported to the Birdsville site. Figure 15 shows the spatial distributions of fire spots and wind fields over Australia ( values and the fine particle aerosol contribution significantly increased. This result was confirmed by the strong correlation between FRP and AOD over Australia. Moreover, the result also demonstrated that one reason for the strong correlation of AODs among widely separated sites over Australian continent, which were reported by Mitchell et al. (2013Mitchell et al. ( , 2017, was the regional transport of biomass burning aerosols, especially during BB period every year.

435
In this study, the impact of long-term fire events on aerosol properties in Australia were investigated by using MODIS fire products, ground-based and satellite-based aerosol products. Further, the spatiotemporal variations of aerosol properties during the 2019/2020 great fire events and a case study of long-range transport of biomass burning aerosols over Australia were investigated. The main findings are as follows.  https://doi.org/10.5194/acp-2020-1139 Preprint. Discussion started: 9 November 2020 c Author(s) 2020. CC BY 4.0 License. https://doi.org/10.5194/acp-2020-1139 Preprint. Discussion started: 9 November 2020 c Author(s) 2020. CC BY 4.0 License.    https://doi.org/10.5194/acp-2020-1139 Preprint. Discussion started: 9 November 2020 c Author(s) 2020. CC BY 4.0 License.  Fig.1(b)).