2.1 Global model ECHAM6.3–HAM2.3

ECHAM6.3–HAM2.3 is a global state-of-the-art aerosol–climate model (Tegen et al., 2019). In this study, we use the same model configuration as already detailed in Heinold et al. (2022). Global simulations were initialized on 1 October 2019 and integrated forward until 31 March 2020, i.e., 6 months ahead. The global model is set up and tuned for a horizontal resolution of approximately ${\mathrm{1.875}}^{\circ }\times {\mathrm{1.875}}^{\circ }$, which results from the spectral truncation at T63. In the vertical dimension, 47 levels are used from the ground to 0.01 hPa. Model layers stretch in the vertical dimension, having around 70 m layer thickness close to ground and more than 1 km in the lower stratosphere. This vertical resolution is quite coarse for resolving stratospheric processes, and a better vertical refinement of the middle atmosphere is envisioned for future studies.

The aerosol module HAM2.3 applies the aerosol microphysics model M7 (Vignati et al., 2004) to the prediction of several aerosol species, including sea salt, sulfate, mineral dust, organic carbon (OC), and black carbon (BC). The latter two are of special interest for the radiative impact of the investigated wildfire emissions. The aerosol microphysics scheme treats source processes (e.g., from fire emissions); transformations due to microphysics and atmospheric chemistry; and loss processes, e.g., due to scavenging by clouds and precipitation, which influence the mass and number concentration of the aerosols in the model. The aerosol microphysics is coupled with a double-moment ice- and liquid-cloud microphysical scheme (Lohmann et al., 2007) to allow aerosol particles to interact with clouds. Aerosol particles also impact the atmospheric radiation fluxes. The particle optical properties are interactively calculated during model run time, taking into account their respective size and chemical composition. The aerosol particles are treated as internal mixtures, and optical aerosol properties such as extinction coefficient, single-scattering albedo, and asymmetry parameter are input from a Mie pre-calculated lookup table.

Atmospheric radiation is computed within ECHAM6.3 using the PSrad/RRTMG (rapid radiative transfer model for GCMs; Iacono et al., 2008; Pincus and Stevens, 2013) radiation package. A total of 30 spectral bands are considered with 16 in the shortwave part of the spectrum, mainly influenced by solar radiation, and 14 in the longwave part of the spectrum, mainly influenced by terrestrial radiation. In addition to aerosol particles, the optical properties of various gaseous atmospheric constituents and different hydrometeor types are also considered. For the analysis of the instantaneous direct aerosol radiative forcing, there is a double call of the radiative scheme to exclude atmospheric adjustments in the diagnosis.

2.2 Setup of global simulation experiments

2.3 Circulation analysis

In the troposphere, the meridional overturning circulation is mainly fueled by strong meridional temperature gradients and is thus called thermally driven. In contrast, in the stratosphere the main driver that continuously induces a meridional transport of air is momentum deposition by breaking waves (Holton, 1997; Butchart, 2014). This is a key difference from the tropospheric circulation. As it turns out, the conventional Eulerian mean of motion does not sufficiently take into account this additional eddy transport (Butchart, 2014; Matsuno, 1980), also known as Stokes drift (Dunkerton, 1978; McIntyre, 1980).

As a result of this under-representation of eddy effects, the circulation, calculated with the Eulerian mean, deviates strongly from the Lagrangian circulation deduced from tracer and mass transport in the stratosphere. A well-established solution to this problem is the so-called transformed Eulerian mean (TEM), derived by Andrews and McIntyre (1976, 1978), as discussed in the review by Butchart (2014).

The TEM circulation can be understood as the part of the meridional Eulerian mean circulation that is not balanced by the Stokes drift, and it is, thus, also referred to as the residual circulation (Butchart, 2014). Defined in this way, the calculated residual circulation is in much closer agreement with the Lagrangian mean paths of tracers compared to the conventional Eulerian mean whenever eddy activity is important. This makes the TEM a very well-suited tool for the investigation of the circulation as well as possible circulation adjustments in the stratosphere. Following Peixoto and Oort (1992) and Grotjahn (1993), the residual wind speeds ${\stackrel{\mathrm{‾}}{v}}^{*}$ and ${\stackrel{\mathrm{‾}}{\mathit{\omega }}}^{*}$ are defined as

where $\stackrel{\mathrm{‾}}{v}$ and $\stackrel{\mathrm{‾}}{\mathit{\omega }}$ are the zonally and monthly averaged meridional and vertical wind speeds (Eulerian sense), respectively. p denotes atmospheric pressure, θ is the potential temperature, stability is related to $\mathrm{\Gamma }={\partial }_{\mathrm{p}}\stackrel{\mathrm{‾}}{\mathit{\theta }}$, φ is latitude, ∂_p and ∂_φ are shortcuts for partial derivatives, and a the radius of the Earth.

Using these residual velocities one can now define a residual meridional mass stream function to quantify and display the stratospheric meridional overturning circulation (Peixoto and Oort, 1992), e.g.,

where g is the acceleration due to gravity. The following section takes a closer look at the impact of the Australian wildfires on the atmospheric state, including the circulation variables ${\stackrel{\mathrm{‾}}{\mathit{\omega }}}^{*}$, ${\stackrel{\mathrm{‾}}{v}}^{*}$, and ${\stackrel{\mathrm{‾}}{\mathit{\psi }}}^{*}$.

3.1 Modeled distribution and optical properties of Australian smoke

The Australian fire outbreak led to a significant input of black (BC) and organic carbon (OC) aerosol into the lower stratosphere. Estimates range from 0.3 to 2.1±1.1 Tg (Khaykin et al., 2020; Peterson et al., 2021; Hirsch and Koren, 2021), which accommodates the total emitted amount of 0.8 Tg used in our study well. Once emitted from southeastern Australian fire hotspots, the smoke aerosol began to spread over the entire Southern Hemisphere, reaching the southern tip of America after around 10 d (Ohneiser et al., 2020). As a result, aerosol optical thickness (AOT) increased significantly in the otherwise relatively pristine Southern Hemisphere (e.g., Hirsch and Koren, 2021). So, for example, the ground-based sun photometer observations of the AErosol RObotic NETwork (AERONET) showed a significant increase in AOT after a large peak around mid-January and persistent high levels thereafter (Heinold et al., 2022). In Punta Arenas, at the southern tip of South America, a monthly average of 0.1 was measured for the 550 nm AOT for January to March, an increase of more than a factor of 2 compared to 2019. Similar values of 0.08–0.12 and 0.04–0.065 on monthly average were also measured in Amsterdam Island and at the Antarctic stations, respectively, during this period.

The ECHAM6.3–HAM2.3 simulations reproduce the evolution of the smoke plume with the pyroconvective injection heights prescribed. In the model, the stratospheric transport of the Australian wildfire smoke caused a monthly mean smoke AOT of up to 0.26 just downwind of the fire region in southeastern Australia in January 2020. Thereby, the plume dispersion is more confined and the smoke vortex is more pronounced in the nudged than in the free-running simulations. Persistently enhanced smoke AOT was simulated south of 30^∘ S with values of 0.01 to 0.03 until March 2020 (Fig. 1). A significantly increased smoke AOT is also simulated over Antarctica due to the southward transport of the stratospheric smoke during the 3-month period considered. In agreement with the observations of, e.g., Khaykin et al. (2020) and Ohneiser et al. (2020), these model results indicate a considerable disturbance of the radiation balance in the Southern Hemisphere associated with the wildfire smoke burden.

For the 2019–2020 Australian wildfire plume, the model calculates a particle single scattering albedo (SSA) of 0.82–0.85 at 550 nm for the modeled ratio of BC to total carbon (BC = $\mathrm{BC}/\mathrm{OC}$) of 0.05–0.08 at the height of maximum smoke extinction (≈16–24 km). The asymmetry parameter for the Australian smoke is about 0.6 at 550 nm, which is a typical value for wildfire aerosol (e.g., Reid et al., 2005). These parameters compare well with the values derived from the inversions of the ground-based multiwavelength lidar observations at Punta Arenas, Chile, in January 2020 by Ohneiser et al. (2020) and are also within the range of other aerosol models (e.g., Bellouin et al., 2020; Brown et al., 2021). For a detailed evaluation and discussion of the modeled smoke plume dispersal and optical properties we refer to the earlier study by Heinold et al. (2022).

Figure 1Maps of monthly mean smoke AOT for January to March 2020 calculated as the difference between the nudged ECHAM6.3–HAM2.3 simulation and a model run without the pyroconvective smoke injection in southeastern Australia. The location of the AERONET stations mentioned in the text is marked by colored symbols: Punta Arenas, Chile (53.14^∘ S, 70.89^∘ W), cyan triangle; Amsterdam Island (37.80^∘ S, 77.57^∘ E), light blue circles; Marambio (64.24^∘ S, 56.63^∘ W) and Vechernyaya hill (67.66^∘ S, 46.16^∘ E), magenta pentagons; the South Pole (90.00^∘ S), yellow hexagons.

3.2 Radiative perturbations

3.2.1 Indications of global adjustments to smoke–radiation interactions

The sudden increase in $\mathrm{OC}/\mathrm{BC}$ burden due to the Australian fire outbreak led to a significant rise in absorption of shortwave radiation in the middle atmosphere (MATM; in our case 200 hPa to TOA) that mainly localized over the southern Pacific Ocean during January 2020 (see Fig. 2a and d) with large-scale average values of shortwave radiation flux convergence larger than 5 W m⁻². Atmospheric mixing and transport led to a more equal redistribution of stratospheric smoke aerosol in the subsequent months (Fig. 2b and c). Simulations with nudged meteorology indicate that significant amounts of smoke have been transported polewards with the implications that the bright Antarctic surface may have been darkened by stratospheric smoke when viewed from TOA. However, the poleward transport pathway is not as clearly visible in the ensemble simulations (Fig. 2d–f). This could be explained by the fact that, firstly, the fire emissions in the variety of ensemble realizations were not necessarily carried out during suitable fire weather conditions but were imposed at externally determined times independently of large-scale conditions. Secondly, it is plausible that the coarse horizontal and vertical resolution of the climate model does not represent the formation and movement of the smoke-induced vortex well, as detailed in Khaykin et al. (2020), such that the flow field that is superimposed by nudging is able to more realistically represent the smoke transport. An extended discussion of this aspect can be found in Sect. 4.

The average longwave forcing of the middle atmosphere (see Fig. 2g–l) appears to be weaker in amplitude than the shortwave forcing. It is mainly the nudged model runs that show a global negative longwave radiation response. The maximum amplitude was reached in February 2020 (Fig. 2h). The ensemble simulations confirm a negative longwave response in the Southern Hemisphere that appears to be weaker than the corresponding shortwave forcing. Towards February 2020, spatial areas increase in size in which nudged as well as ensemble simulations agree in the negative sign of the stratospheric longwave response, also in the Northern Hemisphere. We interpret this signal as a fingerprint of an interhemispheric change in the stratosphere that can only be realized by changes in the global circulation pattern and thus need to be interpreted as global adjustments to stratospheric smoke–radiation interactions.

Figure 2Convergence of monthly average radiative flux perturbations (difference between FIRE1 and FIRE0) between TOA and 200 hPa, abbreviated as middle atmosphere (MATM) forcing. The upper two rows from (a) to (f) present shortwave MATM forcing; the lower two rows from (g) to (l) are longwave MATM forcing. Time is increasing from left to right as indicated in the sub-panel titles, from January to March 2020. Average results from nudged (first and third row) and ensemble (second and fourth row) simulations are compared. Hatching represents areas in which the ensemble mean perturbations are different from zero at the 95 % confidence level.

The situation becomes even more clear when larger smoke perturbations are considered (see Fig. 3). Increasingly larger radiative perturbations result from smoke emissions scaled with factors 1, 2, 3, and 5 (see Sect. 2.2.1). The time-averaged radiative flux convergence in the middle atmosphere is dominated by shortwave absorption in the Southern Hemisphere between 60 and 20^∘ S (Fig. 3d). The stratosphere responds with a negative longwave flux convergence that appears to be rather constant south of 45^∘ S (Fig. 3e). However, we even find negative values of longwave forcing up to around 60^∘ N. In total, a positive forcing still localizes at southern mid-latitudes, however, with a significantly smaller amplitude. This fact again indicates that the highly localized shortwave forcing is compensated for to a considerable degree by a non-local (with respect to the smoke location) longwave forcing. In comparison to the middle atmosphere, forcings originating from the lower atmosphere are relatively small (Fig. 3g–j). The only component that contains perturbation signals is the longwave forcing. Due to the high uncertainties, it remains speculative whether a positive longwave forcing from the lower atmosphere compensates for some of the negative stratospheric forcing in the southern mid-latitudes. At the Earth's surface, dimming of solar radiation due to extinction by smoke is predominant in the southern mid-latitudes and has a maximum amplitude between 40 and 50^∘ S. In total, the shortwave TOA flux is positive (Fig. 3a) because the absorbed flux in the middle atmosphere more than offsets the reduction in shortwave fluxes at the surface. In relative terms, this reduction appears to be much smaller than reported elsewhere in the literature (Hirsch and Koren, 2021; Yu et al., 2021; Liu et al., 2022); however, we show in the next section that the presence of clouds is of major importance for the quantification of surface and TOA forcing of the Australian smoke event. The negative longwave TOA flux compensates for the shortwave TOA forcing such that in total the radiative forcing cannot be differentiated from zero with confidence (Fig. 3b and c).

Figure 3Ensemble and time zonal mean perturbations (JFM average) of radiative fluxes at TOA (first row) and the surface (fourth row) as well as respective flux convergences between TOA and 200 hPa (second row), indicated as middle atmosphere (MATM) forcing, and between 200 hPa and the surface (third row), indicated as lower-atmosphere (LATM) forcing, as a function of latitude. The colors represent deviations from the reference simulation (FIRE0) with different fire emission strengths scaled by a factor of 1 (FIRE1: blue), 2 (FIRE2: green), 3 (FIRE3: orange), and 5 (FIRE5: red). The sum of shortwave (left column) and longwave (middle column) contributions gives the net contribution (right column). Thick lines represent the ensemble mean value, whereas light shadings show the 95 % confidence interval.

Download

3.2.2 Effective radiative forcing of smoke and the role of clouds

In the following, we quantify and discuss the effective radiative forcing of the Australian smoke event based on a composite of all global simulation experiments where rescaled perturbations from different fire emission cases have been averaged. For this, linearity of the atmospheric responses as assumed and perturbations divided by the respective fire scaling factor are averaged. The composite average across all experimental setups allows for a much more robust assessment of underlying effects, although non-linear responses may become more likely when fire emission strength is successively increased.

The effective radiative forcing and its components, instantaneous radiative forcing (IRF) and adjustments (Adjs), are presented in Fig. 4a for all-sky conditions. In the Southern Hemisphere, the shortwave ERF is around 0.5 W m⁻² and is dominated by shortwave IRF. Adjustments become more and more important in the shortwave effects over time and contribute to around 30 % of the total ERF in March 2020. In contrast, longwave ERF is dominated by adjustments and maximizes between February and March 2020. In comparison with the nudged simulations, the shortwave IRF appears to be smaller in the ensemble simulations due to less aerosol transport towards the southern high latitudes (see also Fig. 2a vs. d). In the temporal evolution of shortwave ERF, this mismatch seems to be compensated for by stronger and more positive shortwave adjustments in the ensemble runs. In the tropics, ERF and its components are weak and uncertain in January 2020. In later months, longwave adjustments become more evident and contribute to a global negative longwave forcing. However, the net ERF in the tropics, for which substantial negative values were obtained in the nudged simulations, remains uncertain in the ensemble runs. In the Northern Hemisphere, the disagreement in ERF between nudged and ensemble simulation is the largest, which makes conclusions about the actually realized ERF considerably more uncertain. While the nudged simulations show similar longwave adjustments there as in the other regions, the ensemble simulations show a much weaker signal.

Figure 4(a) All-sky forcings and (b) cloud effects averaged over specified geographical regions (outer interval on x axis), over all months (inner interval on x axis) and over all fire strength perturbations. The different components of the effective radiative forcing (ERF) are represented by differently colored bars. ERF is split into instantaneous radiative forcing (IRF) and adjustments (Adj) and is also decomposed into shortwave and longwave components. Each bar represents the average of ensemble mean values for the four normalized fire perturbation experiments. Values for which the standard deviation among the four perturbation experiments is larger than the actual mean value are indicated by light colors. Data from the nudged experiments are shown as small horizontal gray lines for comparison and are also calculated as the average over all normalized fire perturbations. Regions are split at $\mathit{\phi }=\mathrm{20}{}^{\circ }$ S and $\mathit{\phi }=\mathrm{20}{}^{\circ }$ N, where φ represents latitude.

Download

Literature values of the net TOA radiative forcing from Australian wildfires are mostly negative (see Table 1). At a first glance, existing knowledge seems to be in conflict with our results. However, previous studies have mainly been limited to reporting forcing for clear-sky situations; forcing that ignores atmospheric adjustments; or forcing excluding the bright, snow-covered surfaces in the polar regions. Previously, Heinold et al. (2022) had shown that the high total cloud cover over the Southern Ocean increases the planetary albedo so much that the dominance of scattering by smoke is lost. Thus, the original brightening of the dark ocean surface becomes a darkening of the bright, cloud-covered ocean, and the negative shortwave forcing turns into a positive one (Bellouin et al., 2020). It is a known effect that increasing cloud cover also increases the instantaneous radiative forcing of absorbing aerosols that lie above the clouds (Chand et al., 2009). This effect is especially pronounced over the oceans. It becomes apparent from Fig. 4b in which the individual components of the clear-sky forcing have been subtracted from the respective all-sky forcings, in agreement with the definition of cloud radiative effects (Stephens, 2005). The similarity between all-sky forcing and cloud effects is striking. It means that the clouds are the main driver of the positive forcings obtained in the Southern Hemisphere simply because of their presence, not regarding any cloud changes. In the tropics, the cloud effects significantly contribute to negative longwave adjustments because the cold cloud-top temperatures of the abundant cirrus increase the contrast of the warm perturbation in the lower stratosphere. As an intermediate conclusion, it can be stated that just the presence of clouds has a significant impact on ERF estimates and turns shortwave ERF into positive values in contrast to previously reported clear-sky values.

Sellitto et al. (2022)Heinold et al. (2022)Yu et al. (2021)Liu et al. (2022)Khaykin et al. (2020)Sellitto et al. (2022)Chang et al. (2021)Fasullo et al. (2021)Hirsch and Koren (2021)

Table 1Reported values of radiative forcing by Australian wildfire smoke are compared to our simulation results (first four rows). Our ERF estimates have been derived from the ensemble runs, including the confidence interval, with all values rescaled by the fire scaling factor. Literature values have been sorted in descending order.

Download Print Version | Download XLSX

3.3 Adjustments in the stratosphere

Shortwave heating by the extreme Australian smoke event triggers local and non-local temperature increases in the stratosphere. To illustrate the consequences for the energy balance, Fig. 5 shows the net heating rate, i.e., radiative heating due to convergence of shortwave and longwave fluxes, for the globally averaged lower stratosphere. A maximum between 0.03 and 0.04 K (d)⁻¹ per fire scaling factor is reached after around 1 week after smoke injection. The nudged simulations seem to reach higher heating maxima for smaller fire strengths. In the following adjustments, a chain of processes acts to reduce the net stratospheric heating rate down to zero again. The turning point is reached equally fast for each fire strength after around 9 weeks for the ensemble simulations or after 6 weeks for the nudged simulations. After the early peak in net heating, the decay approximately follows an exponential curve with decay constants of 14 and 21 d for the nudged and ensemble simulations, respectively. The nudged simulations appear to fall off faster as some of the smoke-induced shortwave perturbation leaves the lower stratosphere due to stronger self-lofting upward (see next paragraph) and as the longwave response also appears to be stronger.

Figure 5Net heating rate perturbation (i.e., the combination of shortwave and longwave radiation flux convergences) averaged globally for stratospheric altitudes between $\mathrm{380}<\mathit{\theta }<\mathrm{500}$ K as a function of time. Colors indicate the different fire strength perturbations, similar to Fig. 3. Thick solid lines are ensemble mean values, and thin solid lines represent averaged nudged data. The light gray horizontal line marks the zero line for orientation.

Download

As part of the stratospheric adjustments, a positive temperature perturbation emerges (see Fig. 6). Initially the temperature perturbation develops in the southern mid-latitudes in both the nudged and the ensemble simulations and subsequently spreads towards north and south in the following weeks. While the expansion of the positive temperature perturbation towards the south, i.e., in the polar direction, can mainly be attributed to the transport of smoke aerosol and is therefore much more pronounced in the nudged runs, the expansion towards the tropics is related to the changes in the stratospheric circulation. The positive temperature perturbation reaches northern mid-latitudes after 10 to 14 d. Progression further north can also be seen in the nudged runs. However, the considerable variability in the polar vortex in northern winter possibly hinders the attribution of any temperature effects in the ensemble simulations. A warmer Northern Hemisphere during 2020 was also identified in Rieger et al. (2021) based on observational data. Based on our analyses, we wonder if the observed warming may be explained not only as a coincidence but also in part by smoke-induced circulation changes. Another difference between the nudged and ensemble runs is that the temperature perturbation is also more confined in the vertical direction in the ensemble simulations. This could indicate that the vertical self-lofting is more pronounced in the nudged runs, consistent with Fig. 2a–f in which significantly larger shortwave forcings were found for the nudged runs.

Figure 6Temporal evolution of the potential temperature perturbation for the FIRE1 case with respect to the FIRE0 case. Shown is the 1 K contour of the average potential temperature perturbation for the (a) nudged simulations and (b) ensemble simulation that is color coded by time (see legend). Each second day is presented as beginning with 1 January 2020. Note that the pressure at the vertical axes ranges between 300 and 10 hPa.

Download

The stratospheric circulation changes become apparent from Fig. 7. In the southern tropics at around 10^∘ S, an upward motion lofts air towards the tropopause region. Between 100 and 200 hPa, the mean flow divides into a northward- and southward-directed branch of the global atmospheric circulation. The northward-directed branch reaches high into the stratosphere and is known as the Brewer–Dobson circulation (Butchart, 2014). Nudged and ensemble simulations agree with confidence that due to the effects of absorbing wildfire smoke, a positive anomaly of the residual streamfunction develops in southern mid-latitudes and in the tropics. This leads to a reduction in the strength of the poleward circulation in the Southern Hemisphere and to an increase in the circulation strength in the upper northern hemispheric branch. This increase propagates much further northward in the nudged simulations than in the ensemble simulations that are highly impacted by the variability in the winter vortex. Both the nudged and the ensemble runs also agree on a small weakening of the lower northward branch, but this finding is highly uncertain as well.

Figure 7Average perturbations in residual circulation for (a) nudged simulations and (b) ensemble simulation as a function of latitude and height. Contour lines indicate the state of the reference simulation FIRE0 with solid lines for positive values and dashed lines for negative values. Perturbations of the residual circulation in FIRE1 are presented in colored shading. Both contours and shading have logarithmic spacing. Hatched regions have the same meaning as in Fig. 2.

Download

Changes in the global circulation impact the energy budget of the stratosphere and especially how local diabatic cooling or heating by radiation is compensated for by adiabatic heating or cooling from the global circulation. This aspect is visualized in Fig. 8. On average, lofting of air in the tropics leads to adiabatic cooling (dashed contours in Fig. 8) that counterbalances local diabatic heating by radiation and convection. Towards the poles, the average circulation acts to heat the stratosphere, thereby compensating for the dominant longwave cooling. This cooling is much more pronounced in the winter hemisphere, which is why a stronger circulation is needed there for compensation. The circulation response (colored shading in Fig. 8) develops such that the excess of shortwave heating by smoke in the Southern Hemisphere is partly compensated for due to adiabatic cooling. Moreover, the strengthening of the northward circulation causes additional adiabatic heating in the tropics and the northern mid-latitudes, thus redistributing the initial smoke-induced perturbation on a global scale. Such a temperature-broadening mechanism by circulation changes has recently been analyzed in detail for volcanic impacts on tropical dynamics (Brown et al., 2022). It is again highly uncertain how far this heating perturbation reaches northwards.

Figure 8Similar to Fig. 7 but for the average perturbations in adiabatic heating due to circulation changes.

Download

Due to adjustments, a new balance of the stratospheric state is established. Geostrophy is the strongest constraint for the zonally and monthly averaged dynamics. The positive temperature perturbation in the Southern Hemisphere causes the slope of the geopotential surfaces to increase toward the South Pole; i.e., the perturbation of ${\partial }_{\mathit{\phi }}\overline{\mathrm{\Phi }}/a$, where φ denotes latitude and $\overline{\mathrm{\Phi }}$ the zonally and monthly averaged geopotential, is getting more negative. The resulting pressure-gradient force is nearly exactly balanced by the meridional component of the Coriolis force $-f\stackrel{\mathrm{‾}}{u}$. Since the Coriolis parameter f has negative values in the Southern Hemisphere, the perturbation of the mean zonal wind $\stackrel{\mathrm{‾}}{u}$ also becomes negative and reaches values of a few meters per second in 100 hPa (not shown). The ageostrophic wind components ${\stackrel{\mathrm{‾}}{v}}^{*}$ and ${\stackrel{\mathrm{‾}}{\mathit{\omega }}}^{*}$ describe the zonal mean residual circulation (see Sect. 2.3). As shown in Fig. 9a, ${\stackrel{\mathrm{‾}}{v}}^{*}$ is getting more positive in the Southern Hemisphere; thus the magnitude of the average poleward motion is reduced, and the southern hemispheric branch of the mean circulation is slowed down (see also Fig. 7 again). The sign of $\mathrm{\Delta }{\stackrel{\mathrm{‾}}{v}}^{*}$ is less clear in the tropics and in the Northern Hemisphere, and especially in the former region the response depends on the affected circulation branch. The significant positive correlations in Fig. 9a–c make it clear that the adjustment of the mean meridional wind is accompanied by an adjustment of the large-scale waves measured by the so-called Eliassen–Palm flux defined after Edmon et al. (1980). The effect is only present in the freely running ensemble simulation, whereas nudging destroys this important relation between the mean flow and the wave activity (not shown). The vertical component of the residual circulation is impacted differently by radiative heating depending on the region under consideration (Fig. 9d–f). In the Southern Hemisphere, the additional heating by shortwave smoke forcing induces increased upward motion (given by negative $\mathrm{\Delta }{\stackrel{\mathrm{‾}}{\mathit{\omega }}}^{*}$) that is the strongest in January 2020 before longwave cooling starts to balance shortwave heating in the later months (Fig. 9d). Thus, the response of the Southern Hemisphere is thermally direct. The circulation change is responsible for an adiabatic cooling induced by $\approx \left|\mathrm{\Gamma }\right|\mathrm{\Delta }{\stackrel{\mathrm{‾}}{\mathit{\omega }}}^{*}$ with $\mathrm{\Gamma }={\partial }_{\mathrm{p}}\overline{\mathit{\theta }}$ that compensates for the smoke perturbation. However, an indirect relationship between radiative heating and circulation adjustments is found in the tropics (Fig. 9e). Thus, circulation changes generate additional adiabatic heating, which further amplifies the radiation-induced increase in temperature. Circulation changes remain uncertain in the Northern Hemisphere (Fig. 9f) and appear to be unrelated to radiative perturbations.

Figure 9Relationship between components of the residual circulation and the forcing by large-scale waves (a, b, c) or the radiative heating (d, e, f). Symbols mark monthly and regionally averaged anomalies with colors indicating fire strength from FIRE1 (blue) to FIRE5 (red) and symbol style for different months: January (circle), February (square), and March (triangle). Gray error bars indicate the 95 % confidence interval. Regions are split by $\mathit{\phi }=\mathrm{20}{}^{\circ }$ S and $\mathit{\phi }=\mathrm{20}{}^{\circ }$ N, where φ represents latitude.

Download

3.4 Adjustments in the troposphere

In the following, tropospheric adjustments due to the Australian wildfire smoke are considered. Smoke-induced heating affects not only the lower stratosphere as described above, but also the upper troposphere at around 200 hPa. This is visualized by vertical profiles of radiative heating, potential temperature, relative humidity, and cloud cover in Fig. 10. Again, the heating by smoke induces a positive temperature perturbation. This temperature anomaly not only travels upwards by self-lofting (as, e.g., seen in Fig. 6), but also progresses downwards and subsequently enters the upper troposphere over the course of several months, as can clearly be seen in Fig. 10b. The fact that the average temperature increases at 150 hPa and below despite the decreasing heating rates (see Fig. 10a) is an indication that effects due to changes in dynamics rather than sedimenting aerosol play an important role here. The warming of upper tropospheric layers leads to reduced relative humidity for which negative perturbations reach anomalous values of a few percent at 200 hPa (Fig. 10c). The reduction in supersaturated regions in the upper troposphere also reduces the coverage of the high clouds in the model, e.g., by about 1 % at 200 hPa (Fig. 10d). Similar aerosol effects on cirrus clouds were described in Samset and Myhre (2015), who applied perturbations by black carbon at different atmospheric layers. Furthermore, Samset and Myhre (2015) pointed out that instantaneous forcing and adjustments differ the most when absorbing aerosol perturbations are introduced higher up in the atmosphere.

Figure 10Globally averaged vertical profiles of anomalies of (a) net radiative heating, (b) potential temperature, (c) relative humidity, and (d) cloud cover as a function of pressure. Similar to Fig. 5, thick colored lines represent ensemble mean perturbations with respect to the FIRE0 case. Temporal evolution from January 2020 (leftmost sub-panels) to March 2020 (rightmost sub-panels) is shown. Humidity is shown relative to saturation over liquid water. Note that the x scale is non-linear to improve visibility.

Download

The simulated reduction in cirrus clouds impacts longwave as well as shortwave radiation. Due to the smaller amount of cirrus, a smaller amount of terrestrial longwave radiation is trapped in the troposphere. Thus, more longwave radiation is able to escape to space, leading to a negative forcing of around −1 W m⁻² for a cloud-cover reduction of 1 % (shown for the tropics in Fig. 11a). The compensation between shortwave and longwave radiative perturbations happens in such a way that the net cloud radiative effects remain rather unchanged (Fig. 11b), in line with known neutral to slightly positive net effects of cirrus (e.g., Chen et al., 2000; Stephens, 2005; Senf et al., 2020). The reduction in ice-containing clouds appears to be accompanied by a reduction in surface precipitation. However, this aspect is subject to considerable uncertainty (the majority of data points in Fig. 11c shows a negative precipitation anomaly, but a clear sorting with fire strength is missing). Over land, mixed-phase and ice-phase cloud microphysical processes play an extremely important role in precipitation formation (Mülmenstädt et al., 2015). It is therefore not implausible that a reduction in cirrus cover is partially accompanied by a reduction in precipitation. The precipitation behavior is linked to a dynamical response of the Earth system (see Fig. 11c). For $\mathrm{\Delta }{\stackrel{\mathrm{‾}}{\mathit{\omega }}}^{*}$ at mid-tropospheric levels of 500 hPa, a negative correlation with precipitation is obtained (−0.9). A larger $\mathrm{\Delta }{\stackrel{\mathrm{‾}}{\mathit{\omega }}}^{*}$ means that the upward motion in the tropics is reduced. This dynamical response causes the mean circulation to contribute less to the adiabatic cooling of the tropical troposphere, producing a net dynamical warming of $Q_{\mathrm{dyn}}\approx \left|\mathrm{\Gamma }\right|\mathrm{\Delta }{\stackrel{\mathrm{‾}}{\mathit{\omega }}}^{*}$. The greater this circulation-induced warming by Q_dyn, the more atmospheric cooling Q_atm is compensated for and the less precipitation must contribute to the warming of the atmospheric column (see Muller and O’Gorman, 2011). In general, the analysis of the energetic perspective of precipitation changes has proven to be very useful for understanding atmospheric adjustments to various forcing agents (Richardson et al., 2018), but a freely evolving model setup is needed because nudging destroys the link between precipitation and circulation changes. In the context of stratospheric aerosol intervention strategies, Haywood et al. (2022) could show that continuous injection of absorbing aerosols in the tropical lower stratosphere does not only impact stratospheric temperatures, water vapor, and ozone concentrations, but also affects the troposphere. Similar to our findings, absorbing aerosol in the stratosphere caused a downward influence on the troposphere, leading to a reduction in surface precipitation.

Figure 11Cirrus effects on TOA radiation and energetic constraints for precipitation changes in the tropical belt set here to $\mathit{\phi }\in (\mathrm{20}{}^{\circ }\mathrm{S},{\mathrm{20}}^{\circ }\mathrm{N})$. Symbols and colors have the same meaning as in Fig. 9.

Download

As seen above, not only cirrus cloud cover but the entire hydrological cycle is affected by the Australian fires, although the perturbations to evaporative and precipitation fluxes are quite uncertain (as indicated by the gray bars in Fig. 12a–c). Simulated total precipitation and evaporative fluxes are reduced by $-\mathrm{0.24}\pm \mathrm{0.13}\mathit{\%}$ and $-\mathrm{0.27}\pm \mathrm{0.15}\mathit{\%}$, respectively, for February and March 2020 on average where the interval provides bounds for confidence at the 95 % level. For the different water phases, the clearest signal occurs for the ice water path, which shows a systematic reduction of up to 1 % in the FIRE5 run. Again, the plausible underlying cause is the temperature increase localized near the tropopause together with increased downwelling of longwave radiation. These cause a slight perturbation of ice formation conditions such that fewer areas are supersaturated with respect to ice.

Figure 12Globally and monthly average perturbations of (a) surface evaporation fluxes, (b) convective precipitation, (c) large-scale precipitation, (d) total column water vapor Q_v, (e) liquid water path Q_c, and (f) ice water path Q_i as a function of time. All perturbations are computed relative to the FIRE0 run as reference and are presented in percent. Colors from blue to red indicate perturbation resulting from fire perturbation runs FIRE1, FIRE2, FIRE3, and FIRE5. Bright colors indicate data from ensemble simulations, whereas light colors represent nudged data shown side by side for the respective runs.

Download