Large Enhancements in Southern Hemisphere Satellite‐Observed Trace Gases Due to the 2019/2020 Australian Wildfires

The 2019/2020 Australian wildfires emitted large quantities of atmospheric pollutant gases and aerosols. Using state‐of‐the‐art near‐real‐time satellite measurements of tropospheric composition, we present an analysis of several emitted trace gases and their long‐range transport, and compare to the previous (2018/2019) fire season. Observations of carbon monoxide (CO) show that fire emissions were so intense that the distinct Australian fire plume managed to circumnavigate the Southern Hemisphere (SH) within a few weeks, with eastward propagation over the South Pacific, South America, the South Atlantic, Africa, and the Indian Ocean. Elevated atmospheric methane levels were also detected in January 2020 fire plumes over the Pacific, defined using CO as a plume tracer, even though sampling was restricted spatially by aerosols and clouds. Observations also show significant enhancements of methanol (CH3OH) from the fires, where CH3OH:CO enhancement ratios increased within the aged plume downwind over the South Pacific indicating secondary in‐plume CH3OH formation.

In this study, we use state-of-the-art satellite retrievals to provide detailed analysis of the spatial and temporal evolution of several trace gas distributions sourced from the 2019/2020 Australian fires. While available for the stratosphere, such capabilities were in their infancy when previous major burning events occurred in Australia (e.g., February 2009).

Fire Data Sets
We use two different satellite-derived fire activity datasets: fire radiative power (FRP) from the Global Fire Assimilation System (GFAS vn1.2; Kaiser et al., 2012) and BA from the Fire INventory from NCAR (FINN near-real-time (NRT) vn1.0; Wiedinmyer et al., 2011). Both products are provided at a daily temporal resolution and are based on direct Moderate Resolution Imaging Spectroradiometer (MODIS) measurements (e.g., FRP and thermal anomalies). These quantities are merged with secondary information (including land surface type and emission factors) to derive top-down emissions for trace gases and aerosols (Kiely et al., 2019;Wooster et al., 2018).

Trace Gas Data Sets
In this study we use trace gas retrievals from the Infrared Atmospheric Sounding Interferometer (IASI). IASI is a Michelson interferometer which observes the spectral range 645-2,760 cm −1 with spectral sampling of 0.25 cm −1 (Illingworth et al., 2011). It measures simultaneously in four fields of view (FOV, each circular at nadir with a diameter of 12 km) which are scanned across track to sample a 2,200 km-wide swath . IASI is one of a suite of nadir-sounders flying on Eumetsat's MetOp-A, -B and -C satellites in sun-synchronous polar orbits with local overpass times of 9.30 (day) and 21.30 (night). Here we use carbon monoxide (CO), methanol (CH 3 OH), and methane (CH 4 ) data from MetOp-B produced by NRT processing systems developed by the Rutherford Appleton Laboratory (RAL). CO profiles are co-retrieved with column amounts of CH 3 OH, other trace gases and dust in an extended version of RAL's Infrared-Microwave-Sounding (IMS) scheme, which is described in the Supporting Information Text S1. IMS was developed originally to retrieve temperature, water vapor, ozone, surface spectral emissivity and cloud jointly from co-located measurements by IASI, the Microwave Humidity Sounder (MHS) and the Advanced Microwave Sounding Unit (AMSU-A) on MetOp-A (RAL Space, 2015). CH 4 data are retrieved by an improved version of the IASI scheme reported by RAL Space (2015), which is detailed by Siddans et al. (2017). Data are available from the Centre of Environmental Data Analysis (CEDA, Siddans et al., 2020). CO and CH 3 OH have been quality filtered for a geometric cloud fraction of 0.5 or less (0.1 or less for CH 4 given the greater sensitivity to interference from cloud/aerosol) and a cost value of 1,000.0 or less (120 or less for CH 4 plus a convergence flag equally 1.0). For CO and CH 3 OH, we experimented with a stricter geometric cloud fraction threshold of <0.2. We found this had negligible impact on the scientific results, but did reduce the spatial coverage, making the fire signals noisier. Hence we used the <0.5 cloud fraction threshold. We have also investigated more localized responses in the short-lived nitrogen dioxide (NO 2 ) utilizing tropospheric column NO 2 (TCNO 2 ) data from the TROPOspheric Monitoring Instrument (TROPOMI) on-board the ESA's Sentinel-5 Precursor (S5P) satellite (Veefkind et al., 2012). S5P was launched in October 2017 into a sun-synchronous polar orbit with a local overpass time of approximately 13.30. TROPOMI is a nadir-viewing instrument with spectral ranges of 270-500 nm (UV-Vis), 675-775 nm (near-infrared, NIR), and 2305-2385 nm (shortwave-infrared, SWIR). The TROPOMI retrievals represent the highest resolution of any current tropospheric trace gas sensor with a nadir horizontal resolution of 3.5 × 7.0 km in the UV-Vis-NIR and 7.0 × 7.0 km in the SWIR. The data were filtered for a radiance cloud fraction of <0.5 and a quality flag >0.75.

Fire Activity
During the 2019/2020 fire season (November-December-January, NDJ), satellite observations detected substantially larger fire activity on the Australian south-eastern coastline around highly populated regions such as Sydney compared to the average of the previous 10 seasons (NDJ 2009(NDJ -2019. GFAS FRP suggests that on average (NDJ climatology) ( Figure 1a) there was limited fire activity over south-eastern Australia (i.e., 10-20 mW/m 2 ). The peak activity was more widespread across the north-western territories with FRP typically between 20 and 30 mW/m 2 . However, in NDJ 2019/2020 the entire Australian south-eastern coastline experienced large-scale fires with intensities well above 50 mW/m 2 (Figure 1b). This is supported by the FINN BA (Figures 1c and 1d) with fire events peaking above 10 km 2 widespread across the south-east coast. In contrast, the fire events in the NDJ climatology predominantly occurred over the north-western territories and northern coastlines, with lower BA between 3.0 and 7.0 km 2 . Figure 1e shows the time-series of the total daily Australian FRP and BA for the climatological (median daily totals, 2009-2019) and 2019/2020 fire seasons. Between November 1, and January 31, the climatological FRP has a small range (i.e., 25th-75th percentiles) between approximately 0 and 10 GW. Climatological BA ranges between approximately near-zero and 1,700 km 2 . However, the 2019/2020 FRP (BA) is typically between 20 (100) and 150 (1,700) GW (km 2 ), but with peaks in late December (FRP = ∼320 GW, BA = ∼2,200 km 2 ) and early January (FRP = ∼580 GW, BA = ∼3,000 km 2 ). The peak FRP and BA values sit well outside the variability of the climatologies, highlighting the extreme fire activity experienced in the 2019/2020 fire season.

Nitrogen Dioxide
The response of TCNO 2 to the NDJ 2019/2020 Australian fires is less pronounced relative to the other trace gases investigated here due to the relatively short NO 2 lifetime of a few hours (Alvarado et al., 2010;Logan, 1983). In December 2018 (Figure 2a), there are clear TCNO 2 hotspots (0.5-1.5 × 10 15 molecules/ cm 2 ) in the north-western territories which clearly overlap with GFAS FRP (Figure 1a). Similar relationships are seen near the coast in Queensland. Clear urban signals exist such as Sydney (over 4.0 × 10 15 molecules/cm 2 ), Melbourne (over 4.0 × 10 15 molecules/cm 2 ), and Brisbane/Adelaide (1.0-2.0 × 10 15 molecules/cm 2 ). However, in December 2019 (Figure 2b), there was a large spatial increase in TCNO 2 values  (d) and (e). Panel (d) shows time series of regional November-December-January (NDJ) TCNO 2 for 2018/2019 (red) and 2019/2020 (blue) weighted by regional fire radiative power (FRP). Dashed lines represent the 3-months average. Panel (e) represents regional 2018/2019 and 2019/2020 TCNO 2 weighted by 2019/2020 and 2018/2019 FRP, respectively. > 4.0 × 10 15 molecules/cm 2 around Sydney. Again, these TCNO 2 hotspots are spatially correlated with fire activity (Figure 1) suggesting that fires have degraded the air quality in urban regions. Figure 2c shows the difference between December 2019 and 2018, where there are mixed TCNO 2 differences across the northern states, with enhancements (0.0-1.0 × 10 15 molecules/cm 2 ) in the Western Australia territory and along the south-eastern coastline (2.0 × 10 15 molecules/cm 2 ). To rule out other sources (e.g., urban) driving the TCNO 2 enhancement along the south-eastern coastline, TCNO 2 values have been weighted by FRP over the region (black box, Figure  . Therefore, the large 2019/2020 TCNO 2 values are highly likely to be driven primarily by fire activity along the south-east coastline, especially around Sydney.

Carbon Monoxide
The 2019/2020 Australian fire season produced extensive quantities of emitted CO, as observed by IASI. Figure 3c shows a large total-column CO (TCCO) plume originating over south-eastern Australia and propagating across the entire Pacific, reaching South America. Here, the TCCO ranges between approximately 16-21 × 10 17 molecules/cm 2 , peaking over the Australian coastline and midway between continents in the NDJ period 2019/2020. In the previous burning season (NDJ 2018/2019), TCCO was considerably lower and ranged between 10-13 × 10 17 molecules/cm 2 , with no obvious fire signal over Australia (Figure 3a). Figure 3e shows the difference between fire seasons to be widespread and large (1.5-7.5 × 10 17 molecules/cm 2 ) across the Pacific. Peak inter-year differences of over 5.0 × 10 17 molecules/cm 2 occur over south-eastern Australia and the mid-Pacific, as the plume propagates eastwards. The inter-year differences are significant over large areas (99% confidence level based on the Student t-test and where absolute mean differences are greater than 1.0 × 10 17 molecules/cm 2 ) as shown by the green polygon-outlined regions.
During the 2018/2019 season, fire plumes of limited extent were detected, with moderate CO outflow from the east coast in November 2018 peaking at approximately 19.0 × 10 17 molecules/cm 2 (Figure 3b). In NDJ 2019/2020, there is large and frequent CO outflow throughout the entire season ( Figure 3d). TCCO peaks at more than 30.0 × 10 17 molecules/cm 2 and persists across the 20°-40°S band. Figure 3f shows that the inter-year difference exceeds 15.0 × 10 17 molecules/cm 2 and the larger quantities of CO propagating out into the Pacific (Figure 3c). CO outflow from the peak fire activity (Figure 1e) in late December 2019/early January 2020 went on to circumnavigate the entire SH (Figure 4). Between December 27, 2019 and January 9, 2020, the fire emissions led to large quantities of CO (>25 × 10 17 molecules/cm 2 ), which formed largescale plumes propagating toward South America. Pyroconvection during this period uplifted plumes to altitudes where CO and other trace gases could more easily be detected by IASI (i.e., colder temperatures and above clouds). In the following fortnight, January 10-23, 2020, these plumes reached South America (peak TCCO > 20.0 × 10 17 molecules/cm 2 ) and started propagating into the South Atlantic (15-18 × 10 17 molecules/cm 2 ). A week later, the TCCO plume reached southern Africa (no local fire sources apparent there, unlike e.g., November 15-28, 2019) with values still between 15 and 18 × 10 17 molecules/cm 2 . The TCCO quantities were also enhanced to 14-16 × 10 17 molecules/cm 2 over the Indian Ocean, reaching the Australian west coastline and fully encircling the SH. Over south-eastern Australia (black box in Figure 2c We can also use IASI TCCO measurements to investigate the emission of CO from these fires. Figure 5 represents the daily CO burden for three regions, labeled Boxes 1-3 in Figure 5, and highlights step changes in the burden with time associated with fresh CO emission. For Box 1, there are two large step changes (Day 60 to Day 61 and Day 64 to Day 65) of approximately 4.0 and 7.0 Tg, respectively. A few days later as the plume enters Box 2, the step changes are approximately 8.0 Tg in both cases (i.e., Day 61 to Day 63 and Day 65 to Day 67). As the CO plume propagates eastwards, it is subject to chemical processing and entrainment of clean air, but as the plume rises, it likely enters an altitude of increased IASI vertical sensitively allowing for more CO to be detected. As the CO plume enters Box 3, the step changes are approximately 3.0 Tg (Day 62 to Day 65) and 4.0 Tg (Day 77 to Day 79). Therefore, between the December 30, 2019 and January 4, 2020, there were two large pulses of CO emission from the Australian fires, each in the order of approximately 6-8 Tg. Though it should be noted that this approach assumes limited chemical processing of the emitted CO as it propagates across the South Pacific and that there is limited meridional transport out of the boxes, especially Boxes 1 and 2.

Methanol and Methane
Focusing on January 3-16, 2020, when the fire plumes were most spatially extensive (see TCCO in Figures 4  and 6a), CH 3 OH and CH 4 both showed substantial increases from the fires (Figures 6b and 6c). Total column 7 of 13 CH 3 OH (TCCH 3 OH), relative to January 3-16, 2019, show peak enhancements of over 15 × 10 15 molecules/ cm 2 co-locating with TCCO enhancements of over 15 × 10 17 molecules/cm 2 (Figure 6a). Though CH 3 OH infrared absorption features are much weaker than those of CO and low background CH 3 OH abundances (e.g., over the ocean) are difficult to detect in individual soundings (i.e., large estimated errors; see Text S2, Figure S5), robust signals are detected in large sources such as the Australian fire plume (i.e., >15 × 10 15 molecules/cm 2 , Figure 6b, Text S2, Figure S6). CH 4 , on spatial and temporal scales observable by satellite fluctuates by only a few percent of its global mean value, unlike the order of magnitude of variability in CO and CH 3 OH. To retrieve CH 4 perturbations at the percent level requires careful handling of cloud and, in the case of the Australian wildfire plumes, also smoke (and dust). As shown in Text S3 and Figure S9, IASI detects daily CO plumes, but the corresponding CH 4 distribution is less well sampled due to stringent filtering necessary for cloud. To detect the methane signal associated with Australian wildfires in this particular season in the presence of other sources of inter-year variability is at the percent level. However, the column average CH 4 (CACH 4 ) trend-adjusted anomaly for the January 3-16, 2020 was determined, with reference to the de-trended and de-seasonalized multiannual mean for January   (Figure 6c). This anomaly is positive at the >0.01 ppmv level over much of the region, and some features (e.g., 0.015-0.03 ppmv) cohere with prominent structure (e.g., over the central South Pacific) in the CO and CH 3 OH plumes (see Figures 6a  and 6b). Therefore, we use CO as an in-plume tracer to quantify the much lower amplitude CH 4 signals from the fires. In this case, the TCCO data is subject to the stricter cloud filtering as applied to TCCH 4 and a given IASI sounding is used only when retrievals of both species are available, to reduce sampling errors. Figure 6d shows the time evolution of in-plume column average CO (CACO) and CACH 4 averaged over 150°E−90°W, 50-20°S (black box in Figure S9). The in-plume thresholds for CACO and CACH 4 were subjectively set at 0.07 ppmv and 1.75 ppmv. Investigation of the sensitivity of the CACH 4 (CACO) threshold between 1.73 (0.065) and 1.77 (0.075) ppmv highlights a limited impact on the results in Figure 6d. Both thresholds had to be met, along with stringent cloud screening and other quality control for CH 4 , in order for the pair of CO and CH 4 soundings at a given location to be sampled.
In the first two weeks of January 2020, the in-plume domain-averaged daily CACO ranges between 0.09 and 0.12 ppmv, and the standard deviation indicates large spatial variability. In the second two weeks of January, the spatially averaged CACO is substantially lower, ranging between 0.07 and 0.08 ppmv, with much smaller spatial variability, consistent with mixing with neighboring CO-poor airmasses in the east and less intense incoming plumes in the west. In contrast, for CACH 4 the spatially averaged value drops only slightly and spatial variability is unchanged in the second two weeks, attributable to mixing with high CH 4 tropical airmasses in the east. The peak average CACH 4 values (1.78-1.79 ppmv) occur in the first few weeks and then January 27-29. Temporal correlation of the spatially averaged CO and CH 4 time series is 0.63 for the whole month and 0.86 between January 1 and 20. The daily spatial correlation in this period is between 0.2 and 0.5, indicative of a substantial fire-induced component of CH 4 . Here, these CO:CH 4 correlations are based on the Pearson correlation coefficient and are significant at 95% confidence level or higher. In the latter period of January (21-31) the correlation drops below 0.0 indicating the fire-induced component to be less significant for CH 4 in comparison to other processes. Overall, the relationship between CO and CH 4 in data produced from IASI on MetOp-B demonstrates that the plumes emitted from the Australian fires contained CH 4 as well as CO. Emission factors of CH 4 from vegetation fires have been shown to be similar in magnitude to those of CH 3 OH on a mass emitted per mass of dry matter consumed basis . It is worth noting that given the long lifetime of CH 4 against hydroxyl radical (OH) oxidation (∼9 years (McNorton et al., 2016)), any enhancement in CH 4 due to OH suppression in the plume on the transport timescale of a few days would be minimal (∼0.1% over 3 days under complete suppression of OH oxidation).

Enhancement Ratios
We investigate how the TCCH 3 OH:TCCO enhancement ratio changes as the fire plume traverses the Pacific across the four regional boxes defined in Figure 7a between the January 1 and 17 2020 (i.e., period of large fire plumes). Box 1 covers the primary fire region and Boxes 2-4 cover the downwind outflow. In all cases, we see TCCO-TCCH 3 OH correlation values (R in Figures 7c-7f) above 0.63, peaking at 0.92 in Box 3, indicative of common origins. The in-plume TCCO and TCCH 3 OH values are defined based on a subjectively chosen threshold (TCCO = 18.0 × 10 17 molecules/cm 2 and TCCH 3 OH = 5.0 × 10 15 molecules/cm 2 , where both criteria have to be met for the two species to be co-sampled), but sensitivity analysis of these thresholds (Text S4, Table S1), and the time period in which the plume(s) are sampled, show that our results are robust and relatively insensitive to our choices of these parameters.
In Box 1, the TCCH 3 OH:TCCO enhancement ratio (M), based on a simple linear least-squares fit, is 0.0036 ± 5.98%, which is similar to the FINN CH 3 OH:CO fire emissions ratio (0.0031 ± 0.04%). This suggests that the satellite observations of atmospheric enhancements close to the fires are consistent with freshly emitted fire pollution. As the plume propagates eastwards over the Pacific, there is an increase in the TCCH 3 OH:TCCO enhancement ratio. In Boxes 2, 3, and 4 M is 0.0059 ± 3.13%, 0.0091 ± 1.28%, and Figure 6. (a) Infrared Atmospheric Sounding Interferometer (IASI) two-week (January 3-16) 2020-2019 difference total-column carbon monoxide (TCCO) (10 17 molecules/cm 2 ) and (b) total column methanol (TCCH 3 OH, 10 15 molecules/cm 2 ). Gray regions represent missing satellite data (i.e., average values with error terms >15.0 × 10 15 molecules/cm 2 ). (c) Two-week (January 3-16, 2020) column average methane (CH 4 ,ppmv) anomaly with respect to the deseasonalized and de-trended multi-annual mean for January (2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017). (d) Daily time series of spatially averaged in-plume (150°E−90°W, 50-20°S; black box in Figure S9) IASI-observed CO (red) and CH 4 (blue) column average mixing ratios for January 2020. In-plume data are defined where CO and CH 4 values are both larger than the corresponding thresholds of 0.07 and 1.75 ppmv. Dashed lines represent the uncertainty range (average ± standard deviation). The black line represents daily spatial correlations between in-plume CO and CH 4 . R_all, R_1-20 and R_21-31 are the CO-CH 4 time series correlations for all of January, January 1-20 and January 21-31, respectively. 0.0081 ± 1.94%, respectively. When using the full range of the IASI TCCH 3 OH retrievals (i.e., the retrieval values ± the random errors) and recalculating the enhancement ratios, we find that M is perturbed by approximately 10%, which is relatively small, and the regional ratios follow the same tendency. This increase of TCCH 3 OH with distance, and therefore time, from the fires is suggestive of in-plume chemical production of CH 3 OH. This is similar to Coheur et al. (2009) andHolzinger et al. (2005), who used aircraft and satellite data respectively, to investigate fire plume CH 3 OH:CO enhancement ratios suggesting signs of secondary CH 3 OH production when studying southern European fire plumes. The magnitude of the CH 3 OH/ CO enhancement ratios derived here are consistent with other studies (e.g., Christian et al., 2003;Holzinger et al., 2005;Karl et al., 2007;Singh et al., 2004;Yokelson et al., 1999). Our results imply that IASI detected such secondary formation of CH 3 OH within the fire plume, but with a robust enhancement across the large portion of the Pacific, on a scale previously not discussed in the literature, to the best of our knowledge. Mixing with background air would likely dilute the plume counteracting some of the downwind increase in CH 3 OH enhancement relative to CO. This suggests that the in-plume production of CH 3 OH is likely larger than that suggested by the observed increase in the CH 3 OH:CO ratio alone. Such large-scale enhancements in CH 3 OH may have an important influence on the CH 3 OH budget, impact the oxidative capacity of the remote atmosphere, and potentially the CH 4 lifetime (Read et al., 2012). Such secondary methanol production could be driven by the self-reaction of methylperoxy (CH 3 O 2 ), the reaction of CH 3 O 2 with higher order peroxy (RO 2 ) radicals (Jacob et al., 2005) and OH (Müller et al., 2016), or possibly by less well-established oxidation of organics in the fire plume (Holzinger et al., 2005).

Conclusions
The 2019/2020 Australian fires ("black summer") constituted some of the largest regional wildfires in recent decades and produced large quantities of smoke, aerosols, and trace gases. Peak fire activity occurred on eucalyptus forest vegetation in south-eastern Australia during December and January. Data from the MetOp-B satellite produced by RAL's NRT processing system show that carbon monoxide (CO) emitted from the fires circumnavigated the entire Southern Hemisphere (SH). Compared with the 2018/2019 fire season, to levels of CO from the fire plumes were substantially (and significantly, 99% confidence level) larger by approximately 30%-70% over the South Pacific in the November-December-January average. Methanol (CH 3 OH), which is difficult to detect in normal circumstances due to its weak absorption signature, was in sufficient abundance to retrieve in-plume column amounts with a good precision. Satellite-observed enhancements in total column CH 3 OH relative to total column CO show a substantial increase downwind from the fires, over the Pacific Ocean. This is strongly suggestive of CH 3 OH production within the plume, on a scale not previously reported, as far as we are aware, with potentially important implications for the methanol budget and oxidative capacity of the remote atmosphere. Elevated levels of CH 4 were also detected in association with the fire-plumes during peak activity in early to mid-January 2020, even though the CH 4 emission rate is considerably lower than for CO and perturbations from uniform mixing less than 2%. We also detected more localized enhancements in NO 2 in fire identified regions in South-Eastern Australia.
With future climate and land-use change it is expected that wildfires are going to become more frequent and intense. Therefore, Earth observation (EO), as presented here, is going to be a vital resource to help monitor and understand future wildfire events globally. These EO capabilities will improve with the planned launches of advanced infrared and shortwave spectrometers such as IASI Next Generation and Sentinel 5 on the MetOp Second Generation in polar orbit (ESA, 2020) and the Infrared Sounder and Sentinel 4 on Meteosat Third Generation in geostationary orbit.