Large-scale perturbation of the stratosphere
The Australian wildfire season 2019/2020 was marked by an unprecedented burn area of 5.8 million hectares (21% of Australia’s temperate forests)7 and exceptionally strong PyroCb activity in the south-east of the continent8. The strongest PyroCb outbreak occurred on the New Year’s Eve (Fig. 1a) and on the 1st of January the instantaneous horizontal extent of the stratospheric cloud amounted to 2.5 million km2 as inferred from nadir-viewing TROPOMI9 satellite measurement (Fig. 1b). On that day, an opaque cloud of smoke was detected in the stratosphere by the CALIOP space-based laser radar (lidar)10 at altitudes reaching 17.6 km (Fig. 2). Another PyroCb outbreak with stratospheric impact, although less vigorous, took place on 4 January 2020 and on 7 January, the horizontal extent of the stratospheric smoke cloud peaked at 6.1 million km2 (Fig. 1b) extending over much of the Southern midlatitudes (Fig. 2 and Supplementary Fig. 4c).
The high-altitude injections of smoke rapidly tripled the stratospheric aerosol optical depth (SAOD) in the southern extratropics. The SAOD perturbation has by far exceeded the effect on stratospheric aerosol load produced by the North American wildfires in 2017, putting the Australian event on par with the strongest volcanic eruptions in the last 25 years (Fig. 3), i.e. since the leveling off of stratospheric aerosol load after a major eruption of Mount Pinatubo in 199111. Three months after the PyroCb event, the SAOD perturbation has remained at the volcanic levels, gradually decreasing with a rate similar to the decay of stratospheric aerosol produced by moderate volcanic eruptions.
Using aerosol extinction profiles retrieved from the limb-viewing NASA OMPS-LP instrument12 we find the total aerosol particle mass lofted into the so-called stratospheric “overworld”13 (above 380 K isentropic level corresponding to ~12–17 km altitude) is 0.4 ± 0.2 Tg (Fig. 4), which is nearly three times larger than the estimates for the previous record-high North American wildfires2. The increase in the stratospheric abundance of the gaseous combustion products, derived from the NASA Microwave Limb Sounder (MLS) satellite observations14, is as remarkable as the aerosol increase. Figure 4 puts in evidence that the stratospheric masses of carbon monoxide (CO) and acetonitrile (CH3CN) bounded within the southern extratropics increase abruptly by 1.5 ± 0.9 Tg (~20% of the pre-event levels) and 3.7 ± 2.0 Gg (~5%), respectively, during the first week of 2020. The injected mass of water was estimated at 27 ± 10 Tg that is about 3% of the total mass of stratospheric overworld water vapour in the southern extratropics (see “Methods”).
The gases and particles injected by the PyroCbs were advected by the prevailing westerly winds in the lower stratosphere. The patches of smoke dispersed across all of the Southern hemisphere extratropics in less than two weeks with the fastest patches returning back over Australia by 13 January 2020, whereas the carbon-rich core remained bounded within midlatitudes as shown in Fig. 2. During the following months, most of the particulate material dwelled in the lower stratosphere, the larger and heavier particles sedimented to lower altitudes while the carbon-rich fraction ascended from 15 to 35 km due to solar heating of black carbon (Supplementary Fig. 1).
The large amount of aerosols produced a significant radiative forcing (RF), which we quantified using explicit radiative transfer modelling based on the measured aerosol optical properties (Supplementary notes 1). In the latitude band between 25 and 60° S, an average cloud-free reference monthly radiative forcing as large as about −1.0 W/m2 at the top of the atmosphere (TOA) and −3.0 W/m2 at the surface is found in February 2020 (Supplementary Fig. 2). This can be attributed to perturbation to the stratospheric aerosol layer by the Australian fires plumes. The area-weighted global-equivalent cloud-free RF is estimated (Supplementary Table 1) to values as large as −0.31 ± 0.09 W/m2 (TOA) and −0.98 ± 0.17 W/m2 (at the surface). It is important to notice that these estimations don’t take the presence of clouds into account and are to be taken as purely reference values. For typical average cloud cover in the area affected by the plume15, the surface all-sky RF can be reduced to ~50% and the TOA all-sky RF to ~30–50% of the clear-sky RF estimations16 (see Supplementary notes 1 for details). From the perspective of the stratospheric aerosol layer perturbation, the global TOA RF produced by the Australian fires 2019/2020 is larger than the RF produced by all documented wildfire events and of the same order of magnitude of moderate volcanic eruptions during the last three decades (that have an integrated effect estimated at17 −0.19 ± 0.09 W/m2, or smaller18). In contrast to the non-absorbing volcanic sulphates, the carbonaceous wildfire aerosols absorb the incoming solar radiation, leading to yet more substantial radiative forcing at the surface, due to the additional large amount of energy absorbed in the plume. This can be linked to the ascent of a smoke cloud in the stratosphere, which is dicussed in the next section.
Rising bubble of smoke
The primary patch of smoke originating from the New Year’s Eve PyroCb event followed an extraordinary dynamical evolution. By the 4 January 2020, en route across the Southern Pacific, the core plume started to encapsulate into a compact bubble-like structure, which was identified using CALIOP observations on 7 January 2020 as an isolated 4-km tall and 1000 km wide structure (Fig. 5a). Over the next 3 months, this smoke bubble crossed the Pacific and hovered above the tip of South America for a week. It then followed a 10-week westbound round-the-world journey that could be tracked until the beginning of April 2020 (Supplementary notes 2 and Supplementary Fig. 3), travelling over 66,000 km.
The large amount of sunlight-absorbing black carbon contained in the smoke cloud provided a localized heating that forced the air mass to rise through the stratosphere. With an initial ascent rate of about 0.45 km/day, the bubble of aerosol continuously ascended during the three months with an average rate of 0.2 km/day. While remaining compact, the bubble was leaking material from its bottom part, leaving an aerosol trail that was progressively dispersed and diluted, filling the whole mid-latitude austral stratosphere up to 30 km (Supplementary Fig. 1, see also Fig. 10b). The rise ceased in late March 2020 when the top of the bubble reached 36 km altitude (Fig. 5b and Supplementary Fig. 3). This is substantially higher than any coherent volcanic aerosol or smoke plume observed since the major eruption of Pinatubo in 1991.
Along with the carbonaceous aerosols, the bubble entrained tropospheric moisture in the form of ice aggregates injected by the overshooting PyroCbs. In the warmer stratosphere, the ice (detected by MLS sensor as high as 22 km, cf. Figure 5b) eventually evaporated, enriching the air mass with water vapour. This led to extraordinary high water vapour mixing ratios emerging across the stratosphere within the rising smoke bubble (Fig. 5b). The decay of CO within the bubble (Fig. 5c) was faster than that of water vapour, reflecting the fact that, unlike water vapour, the carbon monoxide was subject to photochemical oxidation whose efficiency increases sharply with altitude19.
Temperature profiles from GNSS radio occultation sensors exhibit a clear dipolar anomaly within the bubble with a warm pole at its bottom and a cold pole at its top (Fig. 5d). Although counterintuitive from the pure radiative transfer perspective, the observed temperature dipole within the heated cloud represents an expected thermal signature of a synoptic-scale vortex.
The compact shape of the smoke bubble could only be maintained through an efficient confinement process. The meteorological analysis of the real-time operational ECMWF integrated forecasting system (IFS)6 reveals that a localized anticyclonic vortex was associated with the smoke bubble during all its travel, moving and rising with it (Figs. 5, 6a, b and Supplementary Fig. 5). With a peak vorticity of 10−4 s−1 (Figs. 6c, 7b) and a maximum anomalous wind speed of 13 m s−1 (Fig. 6d) during most of its lifetime, the vortex had a turnover time of about 36 h. It has therefore survived about 60 turnover times demonstrating a remarkable stability and resilience against perturbations. The ascent was surprisingly linear in potential temperature at a rate of 5.94 ± 0.07 K day−1 (Fig. 7a). This corresponds to a heating rate dT/dt which varies from about 3 K day−1 at the beginning of January to 1.5 K day−1 at the end of March. The altitude rise is from 16 to 33 km for the vortex centroid and from 17 to 36 km for the top of the bubble according to CALIOP. The upper envelope of the OMPS detection of the bubble is seen as the cyan curve on Fig. 7a. This envelope is always above the top detected from CALIOP. Such a bias is expected as OMPS-LP is a limb instrument that scans a much wider area than the narrow CALIOP track. Both CALIOP and OMPS-LP detect that the top of the bubble rises initially faster than the vortex core, by about 10 K day−1. This period corresponds to the initial travel of the bubble to the tip of South America. In terms of altitude ascent, the rates 10 and 5.94 K day−1 translate approximately as 0.45 and 0.2 km day−1.
The confining properties of the vortex are confirmed by the co-located anomalies in tracers and aerosol from the TROPOMI instrument. The satellite observations (Supplementary notes 3.3 and Supplementary Fig. 4a, b) reveals an isolated enhancement of the aerosol absorbing index and of the CO columnar content, as well as the presence of a deep mini ozone hole, depleted by up to 100 DU. All three features were captured in the same position by the ECMWF analyses (Fig. 6e and Supplementary Figs. 4 and 7).
It is worth noticing that the vortex was not a single event. It had several companions, albeit of smaller magnitude and duration, also caused by localized smoke clouds. The most noticeable lasted one month and travelled the hemisphere. Another one found a path across Antarctica where it was subject to the strong aerosol heating of permanent daylight and rose up to 27 km.
The second vortex is borne from the smoke cloud that found its way to the stratosphere during the PyroCb event of 4–5 January 2020. This cloud initially travelled north east passing north of New Zealand before taking a south easterly direction crossing the path of the cloud emitted on 31 December 2019 and the main vortex. Fig. 8a, b shows that a vortex-like structure can be spotted as early as 7 January, coinciding with the location of a compact bubble according to CALIOP. Subsequently, the bubble crossed the path of the first vortex on 16 January while rising and intensifying (see Fig. 8c, d) and travelled straight eastward crossing the Atlantic and the Indian Ocean until it reached the longitude of Australia by the end of January where it disappeared after travelling all the way round the globe. During this travel, the altitude of the vortex centroid rose from 15 to 19 km and the top of the bubble, as seen from CALIOP, reaches up to 20 km.
The third bubble has been first detected by CALIOP on 7 January at 69° S/160° W. It then moved over Antarctica (Fig. 9a) until the end of January where it spent a week over the Antarctic Peninsula before moving to the tip of South America, shortly after this region was visited by the main vortex. It eventually moved to the Atlantic where it dissipated by 25 February (Fig. 9b). The bubble was accompanied by a vortex during its whole life cycle as seen in Fig. 9c. Although the magnitude of this vortex was modest compared to the main one and even the second one in terms of maximum vorticity (Fig. 9e), it performed a very significant ascent from 18 to 26 km (Fig. 9d). We attribute this effect to the very effective aerosol heating received during the essentially permanent daylight of the first period over Antarctica. The simultaneous rise of the main vortex and the third vortex is very clear from the OMPS-LP latitude-altitude sections in Fig. 10c, d.
The combined trajectories of the main vortex and its two companions are shown in Fig. 10e. Figure 10a shows how the trajectories of the vortices form the skeleton of the dispersion path of the smoke plume in the stratosphere.
Figure 10b shows that OMPS-LP follows closely the evolution of the top altitude of all the vortices under the shape of well-defined branches in a longitude-time Hovmöller diagram. It is worth noticing that both the main and the third vortex spent some time wandering in the vicinity of the Drake passage at the beginning of February. Such a stagnation situation is prone to sensitivity. The IFS forecasts during the end of January predicted that the main vortex would cross to the Atlantic, while instead it did not and began moving westward over the Pacific as it reached a higher altitude where easterly winds prevail. Several secondary branches that seem to separate from the main one followed by the main vortex are also visible in this diagram. A detailed inspection reveals that they are indeed associated with patches left behind by the main bubble as it moved upward. It is apparent from several of the panels of Fig. 7a that the top part of the bubble remained always compact while the bottom part was constantly leaking material. Figure 10c, d shows latitude–altitude cross sections of OMPS aerosols on 16 January and 1 February. The fast rise of the main vortex during that period is visible near 50S while the second vortex is located at lower altitude and the third vortex corresponds to the towering structure by 75–80S.