Direct radiative forcing of biomass burning aerosols from the extensive Australian wildfires in 2019–2020

The monthly AOT at 0.55 μm, white-sky SA at 0.3–0.7 μm, active fires, and SST used in this study are Level-3 products from the MODIS satellite observations spanning the entire earth (data downloaded from https://neo.sci.gsfc.nasa.gov/). MODIS on Terra (Earth Observing System AM-1) and Aqua (Earth Observing System PM-1) are key payload imaging sensors that observe the entire Earth system every 1–2 d within visible and 36 infrared spectral bands. The Level-3 monthly MOD(MYD)AL2_M_AER_OD AOT at 0.55 μm, MCD43C3_M_BSA white-sky SA in the visible range of 0.3–0.7 μm are used mainly for estimating ARF. The SA for 2019–2020 has been estimated by a trend analysis from 2000 to 2017 (see figures S4 and S6 supplementary data), since the SA data is available from 2000 to 2017 from MODIS. The monthly averages of AOD and SA over the course of the last two decades (from 2019 to 2020) are provided in the supplementary data (figure S6). The active fire count (MOD14A1_M_FIRE) encompasses the number of fires in a 1000 km² area observed by MODIS Terra using the standard fire and thermal anomalies algorithm and is useful for monitoring the spatial distribution of active fires in Australia. The burned area data are obtained from the product of MCD64CMQ Grid Burned Area that based on the MCD64A1 algorithm (Giglio et al 2018). The MCD64A1 product wildly used for estimating fire emission (Giglio et al 2010, 2013). The long-term time-series of SST observed from MODIS, i.e. MYD28 daytime and nighttime SST derived from the 1- and 4-micron channels, are valuable for understanding the influence of climate on Australian climate and propensity for fires. The intensity of positive and negative IODs is estimated by the difference in SST anomalies between the western equatorial IO (50° E–70° E and 10° S–10° N) and the southeastern equatorial IO (90° E–110° E and 10° S–0° N), and is also known as the DMI. A positive phase of IOD means a DMI higher than 0. To analyse DMI, averages of SST for spring (September—November, 2002–2019) and anomalies of SST for 2019 were calculated and are shown in the supplementary data (figures S1 and S2). We also used the 2019 Australian landcover distribution (figure S3) obtained from the MODIS-IGBP file (available via http://lpdaac.usgs.gov) to illustrate the fire behaviour associated with fuel type. It is based on landcover data observed by TERRA and AQUA MODIS and is classified into 17 types of landcover (i.e. 16 vegetation types and water) using the International Earth-Biosphere Program (IGBP) classification algorithm (Belward 1996). For more information on the definition of IGBP classification, see Strahler et al (1999).

The ARF of biomass-burning aerosols is estimated from monthly mean MODIS satellite data, i.e. AOT at 0.55 μm and SA, using an empirical relationship between AOT, SA, and ARF (see figure 1). The empirical relationship is derived from data obtained from the AERONET stations representing typical biomass-burning aerosols (Yoon et al 2019). The magnitude of ARF is associated with the aerosol loading (AOT) and the direct aerosol radiative forcing efficiency, (FE; in ${\text{W}}\;{{\text{m}}^{ - 2}}$). In general, a large aerosol loading (i.e. high AOT) and higher FE (i.e. high absolute value of FE) enhance ARF (i.e. ARF = FE × AOT). The FE is defined as the ARF (${\text{W}}\;{{\text{m}}^{ - 2}}$) scaled by AOT (i.e. when AOT = 1, FE is identical to ARF); it depends on the aerosol type (e.g. biomass-burning, dust, urban-industrial aerosols); it is also strongly influenced by the SA, i.e. FE = α × SA + β, where α and β are coefficients that vary depending on aerosol type. Therefore, ARF can have different values depending on the reflectance of aerosols relative to that of the surface, which means that the same aerosol type and loading can impose different radiative effects depending on SA, i.e. warming forcing at relatively high SA and cooling forcing at relatively low SA. This is because albedo affects the light-absorbing properties of biomass-burning aerosols, which play an important role in determining the direct aerosol radiative FE (Stohl 2007). For biomass-burning aerosols, ARF is predicted to exert a cooling effect when SA is lower than 0.461, while it is predicted to exert a warming effect at relatively high SA (>0.461). A detailed description can be found in Yoon et al (2019).

Zoom In Zoom Out Reset image size

We estimated the perturbation of surface air temperature in response to biomass-burning aerosols using a relationship between the surface air temperature and radiation budget at the TOA, following Gregory and Forster (2008). This is based on the key assumption that changes in surface air temperature (ΔT_S) are positively correlated with changes in the net radiation forcing at the TOA (ΔF), i.e. ΔF∝ ρ ΔT_S (Cubash et al 2001, Collins et al 2013). ρ has been suggested as a 'transient climate response parameter' (TCRP, K W⁻¹ m²) in the study of Gregory and Forster (2008), and TCRPs can be used for any climate change scenario in which changes in global mean surface air temperature are calculated as a function of changes in radiative forcing. TCRPs are thus measures of the response of global mean surface air temperature to radiative perturbations. The TCRP can be conveniently applied for estimation of the surface air temperature response without the need to actually run and analyse model simulation results (Gregory and Forster 2008). The parameter can vary from one model to another, but within each model it is invariant (WMO 1986). In the one-dimensional radiative-convective model, a 1 W m⁻² change in radiative forcing corresponds to a 0.5 K change in surface temperature (i.e. ρ = 0.5 K W⁻¹ m²; Ramanathan et al 1985). In this study, we have found the TCRP, i.e. ρ = 0.248 K m² W⁻¹, which is driven by linear changes in radiative forcing and surface air temperature simulated by the global ECHAM5/MESSy Atmospheric Chemistry—Climate (EMAC) model (see supplementary data; Chang et al 2021). We used this model to derive changes in temperature caused by changes in ARF estimated by observational data. Note that it is only a rough estimation of the direct radiative forcing effect in EMAC simulations. It only inferred how much the surface temperature is reduced by decreased incoming solar energy due to biomass burning aerosols. The actual effects of aerosol–radiation–climate interactions are spatiotemporally varied, and therefore need to be careful consideration of the heterogeneity of physicochemical aerosols distribution and surface reflectance for more accurate estimation, which differs from estimating the effects of well-mixed greenhouse gases.

The number of active fires, burned area data, and landcover distribution were analysed using MODIS derived products to understand the spatial distribution of fires and their characteristics. The AOT retrieved from Terra/Aqua MODIS represents the amount of aerosols generated by wildfires and is used to estimate the ARF. Figure 2 shows the seasonal distribution of the burned areas, active fires counts, and AOT at 0.55 μm for the four states of Australia (NT, QLD, NSW, and VIC) during the past two decades (2000–2020). The burned areas are largely distributed in both historically fire frequent regions in North Australia (NT, QLD) and unusually fire spots in Southeast Australia (NSW, VIC). The frequency of active fires shows that historically Australia is a region prone to recurrent wildfires, specifically in northern parts of the country during the spring season from September to November (Giglio et al 2013) and in southern parts of Australia during the summer season from December to February (Clarke et al 2011). The burned area and number of active fires in 2019–2020 in NSW and VIC show that the wildfires of this summer were unusual compared to the number of active fires in the last 20 years (figure 2). The number of active fires, burned areas and AOT show a positive relationship in general. However, the number of active fires does not fully explain the magnitude of ways wildfires affect climate. The analysis of burned areas (figures 3 and S6) are closely related to fuel types (figure S3). The AOT helps in estimating how extensively biomass is consumed during fires.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Wildfires are generally affected by both geographical climate type and weather conditions. Potential fuel types and fuel availability are determined by climate type and weather conditions (e.g. drought, wind, extreme heat) affect fuel moisture levels that are critical to fuel flammability and affect fire spread conditions as well. This is depicted in figure 2 as the differences of fire behaviours in forests between the northern states (NT and QLD), and the southern states (NSW and VIC); the latter showed corresponding changes of the AOT values and the number of active fires but with a greater variation for AOT values, whereas the former depicted an opposite tendency, i.e. a high number of active fires and large burned areas often met with smaller AOT values. A large portion of NT and QLD lies in the tropical savanna climate zones (BOM, bom.gov.au/climate; Beringer et al 2015) where fires quickly consume available light fuel (e.g. grass and shrubs; see figure S3), rapidly burn out, and therefore emit rather small amounts of aerosols, i.e. AOT < 0.1, with high numbers of active fires and large burned areas (figures 2 and 3). Most regions of VIC and NSW have temperate climates (BOM, bom.gov.au/climate), despites smaller burned areas, but they consume vast amounts of biomass since more flammable materials are allowed by unusually dry and hot weather conditions due to unprecedented strong positive IOD and release of a large amount of aerosols into the atmosphere (e.g. AOT > 0.2).

The peaks of AOT at 0.55 μm reflect the temporal variation in the number of active fires (figure 2). In VIC, the two peak values correspond to bushfires in the summers of 2006–2007 and 2019–2020. During the two fire seasons, the positive IOD is also reflected in the variability of the DMI that corresponds to unusually strong positive IOD phases (see figure S2 in the supplementary data). In 2003, lightning (natural ignition) triggered massive wildfires in the Alpine region and northeastern VIC, according to the Goverment of Victoria. If fires break out in very difficult-to-access geographic locations, they have more potential to spread out into a larger area as there are more chances of being out of control.

Figure 3 shows the distribution of accumulated burned areas and active fire counts over 6 months (September 2019–February 2020) as gleaned from MODIS/Terra data. The burned areas are marked with yellow, red, and dark red pixels that represent the area burned i.e. 10²–10³, 10³–10⁴, and more than 10⁴hectares, respectively. The locations of active fires are marked with brown, yellow, and red pixels that represent the number of active fires, i.e. 1–5, 5–30, and more than 30 pixels/1000 km² d⁻¹, respectively. The 2019–2020 wildfires occurred in southeastern Australia, which is mainly composed of temperate broadleaf areas (see figure S3), as shown by the large number of red pixels in figure 3 (i.e. meaning more than 30 fires in 1000 km² d⁻¹). This could be explained by copious fuels (i.e. dense wood and broadleaf trees), which are sorts of slow-burning materials that take a longer time to burn and spread fires. On top of that, dry and hot weathers allow more flammable conditions as reducing fuel moisture levels (Davies and Legg 2011). These overall conditions lead to high numbers of active fires with relatively small burned areas and release of a huge amount of biomass burning aerosols from wildfires as shown by high AOT values in NSW and VIC. A large number of active fires are also indicated in northern Australia (e.g. NT, QLD), a region with a historically high frequency of fires but these wildfires in northern Australia are less severe than those in southeastern Australia (e.g. NSW, VIC). This demonstrates that fuel availability and condition are crucial to determine the level of fire activity in Australia, a factor that is closely associated with the type of climate. The ignition and persistence of fires are also closely related to weather conditions as well as climate type. The interaction between climate and available fuel may explain the variability in fire incidence across different Australian regions (Williams et al 2009).

The radiative forcing of aerosols is highly dependent on the optical thickness of the aerosols, indicating that wildfires in southern Australia (NSW and VIC) have a much greater impact on climate than previous wildfires in northern and central Australia (NT and QLD)—see AOT and ARF in figure S6. Although the lifetime of biomass-burning aerosols is short, i.e. 7–10 d (Schum et al 2018), wildfires are very difficult to extinguish, with the result that fires last for several weeks to several months under fire-prone conditions (dry and windy days). Figure 4 shows the AOT and ARF distributions for 1–8 January 2020, demonstrating the vast amounts of aerosols emitted from extensive wildfires and the strong cooling effects of these aerosols over southern Australia, large parts of the Pacific Ocean, and even South America. The NASA/CNES CALIPSO (Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation) satellite observed that Australian wildfire-smoke approached the stratosphere, i.e. ∼25 km above the surface (NASA Earth Sciences Disasters Program, cited 2020). These strong radiative forcings are also reflected in the monthly average values in figure S6, which shows the monthly averages for ARF (W m⁻²), AOT, and SA from September 2019 to February 2020. Severe individual fires may not necessarily be reflected in an average monthly value. Nevertheless, the highly enhanced AOT in southeastern Australia demonstrates the significance of the fires that occurred in 2019–2020, especially those from November 2019 to January 2020 (figure S6). The heavy aerosol loading was transported through the Pacific Ocean and cooled the atmosphere, an effect that lasted longer than 3 months.

Zoom In Zoom Out Reset image size

Compared to the average of the 20 year data, the severity of the Australian wildfires in 2019–2020 is clear. Figure 5 compares the regional mean values of AOT at 550 μm, SA, and direct ARF at the TOA for 2000–2018 and for 2019–2020. A high number of active fires over this 20 year period were concentrated in northern Australia (NT); such wildfires were extinguished rapidly, generated a small amount of aerosol, and therefore rather small radiative forcing. However, for the wildfires occurred in southern Australia in 2019–2020 (VIC), the number of active fires was relatively small, but AOT was very large, and therefore their radiative forcing was much stronger. The regional variability of AOT during the fire seasons shows the dependency of aerosol on the climatic conditions (figures 3–5). The highest AOT values of 2019–2020 (AOT > 0.3) in VIC reflect more flammable conditions (i.e. unusual drought and hot weather) caused by unprecedented strong positive IODs that allow substantial amounts of available fuel. It makes fires more persistent, resulting in higher emissions of aerosols to be released into the atmosphere. In general, aerosol loading positively correlates with the intensity of ARF, as shown in figure 5. These AOT values are more than twice as large as those (i.e. AOT < 0.1) caused by biomass-burning aerosols emitted from the tropical savanna climate zones in western and northern Australia, where fires occurred most frequently in the country.

Zoom In Zoom Out Reset image size

According to the regional average of ARF, the most severe fires in NT, QLD, and NSW were those occurring from November 2019 to January 2020, followed by intense fire activity in VIC one month later (December 2019 to February 2020). The estimated ARF of 2019–2020 wildfires in NT and QLD (−12.1 to −8.6 W m⁻²) during the summer are only half of the values in NSW and VIC (about −21.1 to −16.2 W m⁻²). The unprecedented cooling effects caused by the recent wildfires are estimated at about −5.9 to −1.9 W m⁻² for Australia as a whole on average and −17.7 to −14.8 W m⁻² for southeastern Australia on average over 6 months. Over the ocean, the direct radiative forcing due to massive biomass burning aerosols transported from wildfires is estimated to be −16.9 to −9.2 W m⁻² on average over 6 months. Based on the TCRP derived by the 3D-climate chemistry EMAC model, this perturbation of radiation forcing could decrease the surface air temperature by −1.5 °C to −0.3 °C in Australia as a whole and by −4.4 °C to −3.7 °C in southeastern Australia, respectively.

The global annual averaged radiative forcing caused by the 2019–2020 Australian wildfires is estimated to be about −0.5 to −0.2 W m⁻² in this study and the corresponding change in surface air temperature is −0.13 °C to −0.04 °C. This is comparable to the radiative cooling effect of short-lived aerosols released from volcanic eruptions, i.e. global-mean cooling of 0.1 °C–0.3 °C calculated from the aerosol-ocean general circulation models of the Coupled Model Intercomparison Project Phase 5 (CMIP5) summarized in Gregory et al (2015). The estimated ARF of Australian wildfires is even higher than the 7 year average of radiative forcings from volcanic sulfate aerosols from 2005 to 2011 (i.e. about −0.1 W m⁻²) estimated by Ge et al (2016). This is consistent with the report by IPCC AR5 (2014) that the radiative forcing of volcanic eruptions for the years 2008–2011 is about −0.11 (−0.15 to −0.08 W m^–2) due to the lack of any major volcanic eruptions since 1991 (such as the eruption of Mr. Pinatubo). The IPCC climate models have considered the radiative forcing of volcanic eruptions to be the dominant cause of rapid and dramatic natural forcing, although volcanic eruptions are highly episodic.

The aerosol direct effects due to global fires have been estimated to be about −1.51 to −1.21 W m⁻² for the pre-industrial, recent and end-of-century periods (i.e. the year of 1850, 2000, and 2100) using the Community Atmosphere Model 5.0 (CAM5) (Ward et al 2012). The estimated radiative forcing, i.e. from trends in associated AOT, is rather small in magnitude reflecting a minor effect of global fires on climate change. However, the length of the wildfire season has increased by 18.7% globally from 1979 to 2013 due to global warming (Jolly et al 2015). The underestimated effects of fires are a common limit for most models as discussed in Koch et al (2009). General circulation models generally underestimate emissions from fires compared to terrestrial AERONET observational data (Matichuk et al 2008, Chin et al 2009, Tosca et al 2010, Johnston et al 2011). The climate response of fires should be considered in climate projections as reliable magnitudes. Although biomass-burning aerosols have relatively short lifetimes in the troposphere compared to volcanic sulfate aerosols in the stratosphere, they are emitted consistently for periods ranging from several weeks to several months until fires are extinguished. Furthermore, both wildfire frequency and severity have increased with climate change (Flannigan et al 2012, Jolly et al 2015, Dorre and Satín 2016), and therefore wildfires should be considered as an important radiative forcing factor.