Satellite Monitoring of Anomalous Wildfires in Australia

Here we present the results of satellite monitoring of wildfires in Australia for the period of 2001–2020. Annual and monthly dynamics of wildfire areas and CO and CO2 carbon-bearing trace gas emissions from wildfires have been analyzed for the whole territory of Australia based on satellite data. It was found that anomalous fires occurred in the territory of New South Wales during the 2019–2020 fire season. Values of burned-out areas exceeded the values of previous years 3.5–25.8-fold. Annual mean volumes of carbon-bearing gas emissions in this region exceeded the values of previous years 4–59-fold for carbon monoxide CO and 4.6–50-fold for carbon dioxide CO2. The spatial distribution of the excess concentrations of CO from wildfires in New South Wales was recorded according to the monthly mean data of the AIRS instrument (Aqua satellite). At the same time, the excess of CO2 concentration in the atmosphere was estimated using the TANSO-FTS (GOSAT satellite) data. It was demonstrated that an anomalously high number of fires in this state of Australia was caused by extreme drought associated with abnormally high surface temperatures, low rainfall and humidity which created conditions for intense fires and emissions of carbon-bearing gases associated with the combustion of eucalyptus and tropical rain forests prevailing in this region.


INTRODUCTION
Wildfires are one of the main types of natural disasters that occur on Earth. Annually, vast territories of various countries are exposed to these dangerous phenomena (Korovin, 1996;Shvidenko et al., 2011). The strongest wildfires are usually caused by extremely high air temperatures and low relative humidity which in combination with strong wind create ideal conditions for fast fire spreading (Mokhov et al., 2003;Bondur, 2011;Bondur, 2016;Bondur et al., 2019a;Bondur et al., 2019b;Bondur et al., 2020b). Wildfire seasons become longer as the number of dry and hot days increases which leads to more frequent and severe fires (Mokhov et al., 2003).
Apart from direct impact on vegetation cover, wildfires significantly influence the atmospheric composition, being one of the most important natural sources of various trace gas and aerosol emissions (Tomshin et al., 2012;Bondur, 2016;Bondur and Ginzburg, 2016). Overall, the wildfires are the source of more than 20% of pollutants entering the planet's atmosphere (Isaev and Korovin, 1995). Harmful trace gases from wildfires spread far beyond contamination sources because of the atmospheric circulation. Emission plumes from large wildfires can spread for thousands of kilometers (Andreae and Merlet, 2001;Bondur, 2011;Bondur, 2016). Wildfires and the subsequent biomass burning lead to unfavorable climate changes as they are the cause of increased greenhouse gas emissions into the atmosphere (Matthews, et al., 2012;IPCC, 2013;Kulmala et al., 2015;Bondur and Ginzburg, 2016). The products of biomass combustion resulting from wildfires consist of many compounds with the main share of carbon-bearing trace gases (CO, CO 2 ) (Bondur, 2016). Carbon dioxide (CO 2 ) is the main biomass combustion product due to its high carbon content. Carbon monoxide (CO) is the by-product of incomplete biomass combustion and it is present in the wildfire smoke in high concentrations (Koppmann et al., 2005).
For the Australian territory, wildfires are widespread and regular phenomena that annually cause significant damage to the forest resources, infrastructure and has a serious impact on the environment (Filkov et al., 2020). Wildfires in Australia take place throughout the whole year; however, their intensity and seasonality vary depending upon the region (Matthews et al., 2012;Dowdy, 2018;BoM, 2019;Di Virgilio et al., 2019).
Australia is the third largest source of global emissions of carbon-bearing gases from biomass combustion (Desservettaz et al., 2017). Estimated contribution of emissions from wildfires in Australia into global carbon emissions is about 8% (Paton-Walsh et al., 2010;Desservettaz et al., 2017;Dong et al., 2020).
Vast areas and the inaccessibility of many territories exposed to wildfires complicate early fire detection and assessment of its effects. Therefore, taking into account the features and scales of these phenomena, satellite methods and technologies appear promising to address the problem of detecting sources of wildfires and assessment of their effects (Bondur, 2011;Bondur, 2016;Bondur et al., 2017;Bondur et al., 2019a;Bondur et al., 2019b;Bondur et al., 2020a).
Low-resolution satellite data from Terra, Aqua, NOAA, Suomi NPP, FengYun-3, Meteor-M, and other satellites are the main sources of fresh data on wildfires in large areas (Bondur, 2016;Bondur et al., 2019a;Bondur et al., 2020a). The use of such higher frequency data allows to detect fires quickly even on remote territories. This is especially important for the monitoring of wildfires and assessment of volumes of wildfire emissions into the atmosphere in order to take timely measures to ensure safety and prevent negative effects of these natural phenomena (Pu et al., 2004;Bondur, 2011Bondur, , 2016Giglio et al., 2016;Bondur et al., 2017, Bondur andGordo, 2018;Bondur et al., 2019b;Bondur et al., 2020a).
The primary objective of this study is to obtain spatiotemporal distributions of long-term values of burned-out areas and volumes of carbon-bearing gas (CO and CO 2 ) emissions in Australia to identify the annual and seasonal dynamics of wildfires and their effects.
One of the key problems addressed in this study was to obtain the most accurate data on the volumes of emissions of carbonbearing gases (CO, CO 2 ), achieved through the integrated use of satellite data of various spatial resolution, as well as vegetation maps, taking into account the variability of vegetation cover every 5 years.
Satellite data from Terra and Aqua satellites (MODIS instrument) were used in this work. Based on these data, the burned-out areas were calculated, and the volumes of emissions of carbon-bearing gases (CO, CO 2 ) caused by combustion of biomass in Australia from 2001 to September 2020 were assessed. The 2019-2020 fire season was analyzed in more detail in the state of New South Wales. Variations in surface temperature and changes in CO concentrations were analyzed using data from the AIRS instrument (Aqua satellite). Through the use of GOSAT (TANSO-FTS) data, the share of carbon dioxide in the atmosphere was estimated for the period of 2019-2020 wildfires in the state of New South Wales.
This study can serve as a basis for comparing the values of the concentrations of carbon-bearing trace gases detected by satellite and data calculated on the basis of emission factors to assess the contribution of wildfire emissions to atmospheric pollution against other sources.

MATERIALS AND METHODS
A technique described in (Bondur, 2011;Bondur, 2016;Bondur et al., 2017;Bondur and Gordo, 2018) was used to monitor wildfires in Australia. MODIS (Terra, Aqua satellites) data were used during wildfire monitoring as main data for the assessment of areas of burned-out territories and the detection of fire source boundary changes. To calculate the wildfire areas, we used 1 km spatial resolution Level-2 data processing MOD14 product (Giglio et al., 2016).
In this study the calculation of the total matter mass emitted into the atmosphere as a result of a wildfire, including CO and CO 2 emissions, was carried out using the Seiler-Crutzen equation (Seiler and Crutzen, 1980) with the introduction of a correction factor obtained to clarify the burned-out area (Bondur, 2016;Bondur and Gordo, 2018).
Where A is the wildfire area based on MOD14 data (m 2 ); B is the biomass density within the burned-out territory (kg/m 2 ); C is the biomass combustion completeness (%); D is the emission factor (the matter mass emitted into the air during the burning of 1 kg of the biomass (g/kg); k is the correction factor; E is the total matter mass emitted into the atmosphere as a result of a fire (g). The correction factor k was obtained according to the technique described in (Bondur and Gordo, 2018), the technique is based on comparing the results of calculations of wildfire areas obtained from satellite data of low (Terra) and medium (Landsat-8) spatial resolution for selected test sites. As a result, the values of the correction factor k were obtained, varying in the range of 0.65-0.91. The use of the correction factor k allowed to obtain more reliable values of burned-out areas using satellite data of low spatial resolution, and, therefore, to assess of the volumes of emissions of carbonbearing gases CO, CO 2 in a more accurate way.
The values of emission factors B, C and D in Eq. 1 are presented in Wiedinmyer et al., 2011). Each factor corresponds to specific type of burned vegetation. For the selected study area, the type and characteristics of the vegetation cover were determined using the satellite data of the MODIS instrument (Terra and Aqua satellites) data, the Land Cover Type 1 (MCD12Q1) International Geosphere-Biosphere Program (IGBP) Slassification product (Version 6) with 500 m spatial resolution (Friedl et al., 2010).
One of the important factors influencing the frequency and intensity of wildfires is the temperature of the Earth's surface. To study the seasonal variations in the temperature we used monthly mean data for 2003-2020, obtained from the AIRS instrument (Aqua satellite), 1°× 1°spatial resolution AIRS3STM v006 product (Tian et al., 2013).
Carbon-bearing trace gases such as CO and CO 2 constitute a main share of wildfire emissions from biomass combustion. Their impact on the climate on the regional and planetary scales is long lasting (Bondur, 2016;Bondur and Ginzburg, 2016;Bondur et al., 2020b). To study changes in the concentrations of carbon monoxide (CO), the excess of the monthly mean values Frontiers in Earth Science | www.frontiersin.org January 2021 | Volume 8 | Article 617252 3 (product AIRS3STM) was calculated in comparison with the data of previous years starting from 2003. The study of carbon dioxide (CO 2 ) concentration changes was carried out using TANSO-FTS (GOSAT satellite) data in the near-IR region of the electromagnetic spectrum, 2.5°× 2.5°spatial resolution product (Yoshida et al., 2011). The excess CO 2 concentrations over monthly mean concentrations for previous years since 2009 were calculated.
In the course of the study, the total burned-out area values and the volumes of emissions of carbon-bearing trace gases caused by fires for the entire territory of Australia were obtained. New South Wales was selected for more detailed study. Figure 1 shows distributions of monthly wildfire areas for the whole territory of Australia from January to December for the period of 2001-2020 (A) and annual areas of burned-out territories for this period (B) obtained from satellite monitoring data.

RESULTS
The analysis of Figure 1A has shown that the largest wildfire areas on the Australian territory were detected in spring and summer months (for the southern hemisphere), i.e., from September to December. Maximum burned-out area values  Figure 1 has shown that the values of carbonbearing gas emissions were approximately proportional to the areas of burned-out territories.
As an example, Figure 2A presents spatial distribution of wildfire sources detected in Australia in November-December 2019. As can be seen from Figure 2A, wildfire sources were registered in all the states of Australia. However, New South Wales has experienced the most intense wildfires resulting in burning more areas than in other fire seasons in the region over the previous 20 years. The spatial distributions of obtained temperature anomalies are shown in Figure 3.
The rise in surface temperature from September 2019 to January 2020 ( Figure 3) was above average compared to past years since 2003, increasing the need for evaporation. A longterm rainfall deficit in the territory of New South Wales according to the data of (BoM, 2019) seriously influenced the occurrence and spread of wildfires. Low humidity levels prepared the vegetation cover (Nolan et al., 2020), dominated by eucalyptus and tropical forests, for the burning of biomass caused by wildfires, which contributed to the emergence of significant volumes of emissions of various trace gases, including carbonbearing gases, into the atmosphere. Figure 4 shows annual distributions of wildfire areas (A), as well as CO (B) and CO 2 (C) emission volumes from wildfires in New South Wales for the time period of 2001-2020. From the analysis of Figure 4A, it follows that the total areas of wildfires in this territory until 2019 were quite stable, with minor peaks in 2002 (14.9 thous. km 2 ) and in 2003 (11.7 thous. km 2 ). Extremely high values of wildfire areas in the territory of this state were registered in 2019 and reached 37.6 thous. km 2 . Total fire areas in New South Wales in 2019 exceeded those in previous years (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018) 3.5-25.8-fold.   From the analysis of Figure 5A, it follows that wildfires in New South Wales from November 2019 to February 2020 significantly exceeded the data for past 9 years starting from 2001. Australian wildfires of unprecedented scale in the 2019-2020 fire season burned predominantly temperate broadleaf forests in eastern Australia.
After analyzing the results published in ) and comparing them with our data on the wildfire areas, we observed a discrepancy of about 4 thous. km 2 . This gap in results may be due to the fact that the end of the study presented in  fell on the beginning of January 2020, while the data presented in our study includes a full month.
Analysis of Figures 5B,C showed that abnormally high values of the volumes of carbon-bearing gases CO, CO 2 were detected in November 2019 (4.3 mln. t CO and 75.5 mln. t CO 2 ), December 2019 (5.9 mln. t CO and 100.8 mln. t CO 2 ) and January 2020 (3.1 mln. t CO and 46.6 mln. t CO 2 ). The highest values of emissions were from the combustion of the sub-tropical broadleaved deciduous forest vegetation type of Wollemi National Park.
In January 2003, recorded wildfire areas were 5.6 thous. km 2 and emissions from wildfires were 3.3 mln. t CO and 41.6 mln. t CO 2 . The largest volume of emissions was due to the combustion of the sub-tropical broadleaved evergreen forest vegetation type in the territory of the national parks -Kosciuszko National Park (KNP) and Cabramurra.
Initial and average CO 2 emissions on the east coast of Australia were estimated in (Bowman et al., 2020a) to be approximately 670 mln. t, while other data ranged from 550 to 850 mln. t. Comparing these results with the emissions obtained in this work, it can be concluded that the values obtained in (Bowman et al., 2020a) exceed those obtained by us. This discrepancy may be due to the use of different approaches to calculating wildfire emissions. The spatial distribution of atmospheric composition changes during severe fires in New South Wales were analyzed using monthly mean AIRS data (CO concentration) and TANSO-FTS data (CO 2 concentration). These changes are shown in Figures 6A,B.
Monthly mean CO concentrations from September 2019 to February 2020 were calculated using AIRS (Aqua satellite) data and compared to CO concentrations for the period of previous years starting from 2003 ( Figure 6A). The analysis of the obtained results allowed to detect anomalous excess of carbon monoxide concentration by 9×10 17 moles/cm 2 in December 2019 over the values of previous years in the territory of New South Wales.
Monthly mean carbon dioxide (CO 2 ) concentrations in the atmosphere in 2019 were calculated using TANSO-FTS (GOSAT) data. These concentrations for each month were compared with the data for previous years starting from 2009. The results are provided in Figure 6B. Their analysis revealed 11.5 ppm excess in carbon dioxide (CO 2 ) concentrations in northeastern New South Wales in October 2019. In December 2019, an anomalous excess of carbon dioxide concentrations in Frontiers in Earth Science | www.frontiersin.org January 2021 | Volume 8 | Article 617252 8 the atmosphere (by 15 ppm) was revealed compared to previous years (2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018). The amount of CO 2 that enters the atmosphere as a result of severe wildfires tends to accumulate and persist in the atmosphere for a period of time. Probably for this reason, in January 2020, in the territory of New South Wales, excess CO 2 concentrations were observed compared to the data of 2009-2019 for this month ( Figure 6B).
Studying the impact of carbon-bearing emissions from wildfires on the overall carbon balance in the long term is essential for understanding the damage to the atmosphere. These studies require a much more consistent and comprehensive evidence base, including coordinated work involving laboratory and field measurements of emissions, natural and classical field experiments examining the effects of fire severity and forestry practices, integrated into forest demography, carbon cycle and economic modeling (Bowman et al., 2020a;Bowman et al., 2020b). Given the features of the vegetation cover in New South Wales which is dominated by eucalyptus and tropical rain forests, the burning of biomass caused by wildfires contributed significant emissions of various trace gases into the atmosphere. Annual mean volumes of carbon-bearing gas emissions in 2019 exceeded previous values more than 4-59-fold (CO) and 4.6-50-fold (CO 2 ).
Excess concentrations of CO and CO 2 revealed from satellite data and calculated volumes of emissions of carbon-bearing trace gases from wildfires will be in demand for analyzing the effects of fires and assessing the contribution of emissions to atmospheric pollution.
The obtained results demonstrate high efficiency of satellite data use for assessing the spatiotemporal dynamics of fire sources, burned-out areas, and volumes of harmful gas emissions from wildfires, as well as for conducting research in the field of the atmospheric composition change during the spreading of strong fires and their impact on the planetary climate.

AUTHOR CONTRIBUTIONS
VB, KG, OV chose the topic of the research, test sites, satellite data used for the research. VB, KG developed a method for estimation of the wildfire areas and the resulting CO and CO 2 emissions. AZ, KG performed calculations and statistical analysis of the obtained results. OV carried out studies of seasonal variations in surface temperature in the studied area, as well as changes in CO and CO 2 concentrations. All authors have read and approved the final manuscript.