Impact of volcanic eruptions on extratropical atmospheric circulations: review, revisit and future directions

Robock and Mao (1992) and Perlwitz and Graf (1995) were one of the most pioneering studies, reporting boreal-winter (December to next year February, DJF) surface warming in the NH extratropics after historical volcanic eruptions. Such warming was suggested to occur via strengthened westerly winds resulting from an enhanced pole-to-equator temperature gradient in the upper troposphere and lower stratosphere. Supporting their findings, Stenchikov et al (2006) found an enhanced Arctic lower stratospheric polar vortex (hereafter Arctic polar vortex), positive NAO and Eurasian warming in the two consecutive winters after volcanic eruptions from Met Office Hadley Centre/Climatic Research Unit global surface temperature version 2 (Jones and Moberg 2003), Hadley Centre Sea Level Pressure dataset (HadSLP) version 1 (Basnett and Parker 1997), and National Centers for Environmental Prediction/National Center for Atmospheric Research reanalysis (NCEP1; Kalnay et al 1996) data. Driscoll et al (2012) later found the same results from 20th Century Reanalysis version 2 (20CRv2; Compo et al 2011) and 45 year second-generation reanalysis from the European Centre for Medium-Range Weather Forecasts (ERA40; Uppala et al 2005). By examining the 1883 Krakatau and 1991 Pinatubo eruptions with the 20CRv2 and ERA40, Zambri and Robock (2016) also reported an enhanced Arctic polar vortex, positive NAO and northern Eurasian warming, which are statistically significant at 10% level, in the first DJF following volcanic eruptions. Based on climate reconstruction dataset, Fischer et al (2007) provided the evidence of positive NAO and corresponding northern European surface warming and wetting in the DJF following volcanic eruptions in the last half-millennium period. Ortega et al (2015) also showed the emergence of positive NAO in response to strong volcanic eruptions in the last millennium. However, Swingedouw et al (2017) indicated that NAO response to volcanic eruptions in the instrumental era (1883 Krakatau, 1902 Santa María, 1963 Agung, 1982 El Chichón and 1991 Pinatubo eruptions) is not significantly detected, raising the uncertainty of tropospheric circulation responses to volcanic eruptions.

Given the importance of AO/NAO to atmospheric variability in the extratropics, the AO/NAO responses to volcanic eruptions are briefly updated in figure 1 using the latest observations and reanalysis data. The HadSLP version 2 (Allan and Ansell 2006), the Cowtan and Way data version 2.0 (Cowtan and Way 2014), and ERA5 reanalysis (Hersbach et al 2020) are utilized for sea level pressure, near-surface air temperature, and 10 hPa zonal wind, respectively. These datasets have advantages in both spatio-temporal coverage and reliability than the datasets used in previous studies. The ERA5 for instance presents better performance in several aspects with higher resolution than NCEP1, 20CRv2 and ERA40 (Hersbach et al 2020).

Zoom In Zoom Out Reset image size

Figures 1(a) and (b) illustrate the time series of AO and NAO indices in DJF from pre- to post-eruption years for individual volcanic eruptions in the historical period and their average. Here, AO is defined as the first empirical orthogonal function mode of detrended DJF sea level pressure anomalies over the north of 20°N using HadSLP version 2. The NAO is defined same as AO but over 20°W–80°E and 20–80°N. These statistical approaches, instead of simple methods based on sea level pressure difference between the two points, are adopted to consider large-scale atmospheric circulation pattern. Results show the emergence of positive AO and NAO in the first DJF after volcanic eruptions, except for 1963 Agung and 1982 El Chichón eruptions which exhibit only positive AO (insignificant NAO) and NAO (insignificant AO) response, respectively. These results present more robust AO/NAO responses to the historical volcanic eruptions than those reported by Swingedouw et al (2017), and highlight the importance of datasets and analysis methods to improve the signal-to-noise ratio.

Figures 1(c)–(h) present spatial patterns of sea level pressure, near-surface temperature, and 10 hPa zonal wind responses averaged over all five historical eruptions or two strongest eruptions (i.e. 1883 Krakatau and 1991 Pinatubo eruptions), relative to five year averages before the eruptions. They robustly reveal positive AO/NAO, northern Eurasian warming, and Arctic polar vortex enhancement in the first DJF after volcanic eruptions, confirming previous studies (e.g. Stenchikov et al 2006, Zambri and Robock 2016). Nevertheless, these responses are mostly insignificant when analyzing only two strongest cases (figures 1(d), (f) and (h)), although stronger responses than other cases are anticipated (Bittner et al 2016a, Zambri and Robock 2016, ). These results indicate the presence of the uncertainty or nonlinearity of extratropical circulation responses to volcanic eruptions, possibly due to the small sample size (Bittner et al 2016a).

Many studies have attempted to detect and understand NH-winter AO/NAO response to volcanic eruptions. Based on the Atmospheric Model Intercomparison Project model analyses, Mao and Robock (1998) suggested that warming over the NH high-latitude continents in response to the 1982 El Chichón eruption likely resulted from the enhanced Arctic polar vortex and its downward propagation to the troposphere. Otterå (2008) and Marshall et al (2009) also found similar results for the 1991 Pinatubo eruption from two different models. Their results were consistent with Stenchikov et al (2002) who reported a close relationship between the positive AO response to the 1991 Pinatubo eruption and the strengthened Arctic polar vortex in the perturbation model experiments.

The driving mechanism of Arctic polar vortex intensification after strong volcanic eruptions (e.g. figure 1(g)) has been examined by conducting climate model simulations. Stenchikov et al (2002) proposed that polar stratospheric cooling by ozone depletion through photochemical reactions can strengthen the Arctic polar vortex. They also suggested that the reduced meridional temperature gradient in the troposphere, which results in weak planetary-scale wave activities, may help the polar vortex intensification. In contrast, Toohey et al (2014) highlighted the importance of the enhanced stratospheric residual circulation in the polar vortex intensification after the 1991 Pinatubo eruption. Bittner et al (2016b) attributed the polar vortex enhancement after the 1815 Tambora eruption to anomalous subtropical westerlies which induce an equatorward wave propagation and Eliassen-Palm (EP) flux divergence (strengthening of polar vortex) in the NH high-latitudes.

The above mechanisms are briefly updated in figure 2 using the latest climate model simulations. Both the coupled model intercomparison project phase 6 (CMIP6; Eyring et al 2016) and the last millennium simulations of the Community Earth System Model Last Millennium Ensemble (CESM-LME; Otto-Bliesner et al 2016) are utilized. While CMIP6 provides multiple climate model simulations for the historical period (1850–2014), CESM-LME allows to analyze multiple tropical volcanic eruptions occurred during the last millennium with a large ensemble size. A total of 192 ensembles from 26 CMIP6 models are analyzed (table S1). The CESM-LME simulation consists of 17 ensemble members, 12 and 5 of them being integrated with all external forcings and volcanic forcing only, respectively.

Zoom In Zoom Out Reset image size

Figure 2 illustrates EP flux anomalies in the first October–November and DJF following the two strongest tropical volcanic eruptions in the historical period (i.e. 1883 Krakatau and 1991 Pinatubo eruptions; top row) from CMIP6 model simulations and those following the five strongest tropical volcanic eruptions in the last millennium period from CESM-LME simulation (bottom row). Here, only five tropical volcanic eruptions are selected in the last millennium by excluding northern- and southern-extratropical volcanic eruptions (see figures S1(a) and (b), Fasullo et al 2019), which are defined as the eruptions presenting the largest interhemispheric shortwave perturbation differences (see below).

Both CMIP6 and CESM-LME simulations show an equatorward wave propagation in the stratosphere, which potentially results from the enhanced westerly in the subtropical stratosphere (figures S1(d) and (e)). The volcanic aerosols-induced heating in the tropical lower stratosphere strengthens pole-to-equator temperature difference in the upper troposphere and lower stratosphere (gray contours in figures S1(d) and (e)) which in turn strengthens the subtropical westerlies through the thermal wind balance. The enhanced subtropical westerly allows Rossby waves to propagate deep into the tropics by shifting the critical latitude equatorward. The resultant EP flux divergence in NH high-latitudes (figure 2) may contribute to Arctic polar vortex strengthening. This result supports the physical mechanism suggested by Bittner et al (2016b). Since many CMIP6 models have a high-top configuration with the model top above 50 km (table S1; Chen et al 2021, IPCC 2021), the same analyses are repeated with high-top models which may better represent stratospheric processes. High-top models show essentially the same results as those of full models (figure S2). It should be, however, noted that previous studies have found little improvement for high-top models in simulating extratropical circulation responses to volcanic eruptions (Marshall et al 2009, Charlton-Perez et al 2013, Polvani and Camargo 2020). The analyses of the 1815 Tambora eruption, to be compared with Bittner et al (2016b), also present similar results (figure S3), demonstrating the robustness of the results. It is noteworthy that Stenchikov et al (2002) indicated that the increased pole-to-equator temperature difference in the upper troposphere and lower stratosphere may not be a necessary condition for Arctic polar vortex intensification. They alternatively proposed a bottom-up mechanism relating the polar vortex intensification with a weakened vertical wave propagation in the NH mid-latitudes. However, EP flux in figure 2 does not show a weakened vertical wave propagation in the middle troposphere, not supporting Stenchikov et al (2002).

The Arctic polar vortex intensification following tropical volcanic eruptions could drive a positive AO in the troposphere through the downward coupling on the sub-seasonal time scale (Baldwin et al 2021). Daily events with strong Arctic polar vortex occur more frequently at the first winter after historical tropical volcanic eruptions compared to pre-eruption winters (figure 3(a)) in CMIP6 model simulations. Composite analysis reveals that strong polar vortex anomalies tend to propagate downward from the lower stratosphere to the troposphere (figure 3(b)), as discussed in Baldwin and Dunkerton (2001) and Thompson et al (2002). It explains a significant correlation between 10 hPa zonal wind and AO index anomalies (p-value = 0.01) (figure 3(c)). In figure 3(c), many winters from 26 CMIP6 models are found in the first quadrant. This result, i.e. anomalously strong Arctic polar vortex and positive AO, suggests that the historical volcanic eruptions may affect the NH mid- to high-latitude surface climate by modulating the Arctic polar vortex and its downward coupling in the first DJF after the eruptions as proposed by Robock and Mao (1992) and Perlwitz and Graf (1995). Consistent results are also obtained from CESM-LME simulation (figures 3(d)–(f)) and subset of high-top CMIP6 models (figure S4), supporting the reliability of the physical pathway.

Zoom In Zoom Out Reset image size

There is a large uncertainty in the NH surface climate response to volcanic eruptions in climate models, possibly due to inter-model difference in volcanic forcings as well as dynamical-physical processes. Stenchikov et al (2006) reported that CMIP3 models (Meehl et al 2007) show too weak AO and Eurasian surface air temperature responses. Driscoll et al (2012) also reported that CMIP5 models (Taylor et al 2012) do not robustly reproduce the observed NH mid- to high-latitude climate responses to volcanic eruptions. Charlton-Perez et al (2013) presented the consistent result by analyzing the subset of high-top CMIP5 models. Using the multi-models that participated in the Paleoclimate Modeling Intercomparison Project phase 3 (PMIP3; Braconnot et al 2012), Swingedouw et al (2017) reported that the models do not consistently show positive NAO anomalies after strong volcanic eruptions during the last millennium. Unlike these studies, Zambri and Robock (2016) found positive AO response in the first DJF to the two strongest historical volcanic eruptions, i.e. 1883 Krakatau and 1991 Pinatubo, in CMIP5 models. Barnes et al (2016) also reported Arctic polar vortex enhancement to the 1991 Pinatubo eruption in CMIP5 models.

These studies suggest that surface responses to volcanic eruptions, either via direct effect or indirect effect through the Arctic polar vortex intensification, are still uncertain in climate models. Such uncertainty was further emphasized by recent studies with the latest climate models including CMIP6 models. For instance, Polvani and Camargo (2020) reported the absence of Eurasian winter warming after the 1883 Krakatau eruption in the low-top and high-top models with fine vertical resolution in the stratosphere. By analyzing large ensemble simulations, they indicated that the observed Eurasian winter warming to the 1883 Krakatau eruption may not be related to volcanic eruption. Polvani et al (2019) also showed that Arctic polar vortex enhancement and Eurasian winter warming after the 1991 Pinatubo eruption are not well captured by climate models due to strong internal variability. Analyzing 1991 Pinatubo eruption simulations from six climate models participating in the Model Intercomparison Project on the climatic response to Volcanic forcing (VolMIP; Zanchettin et al 2016), Zanchettin et al (2022) reported that climate models generally do not show positive NAO after volcanic eruptions. In contrast to these studies, Zambri et al (2017) found Eurasian winter warming in the first DJF to ten strongest last millennium volcanic eruptions in PMIP3 multi-models and CESM-LME. The ensemble mean of five decadal climate prediction models also showed positive NAO and associated northern Eurasian winter warming to the 1963 Agung, 1982 El Chichón and 1991 Pinatubo eruptions, albeit weaker than the observations (Hermanson et al 2020). Recently, Liu et al (2020) also found lower stratospheric tropical warming and Arctic polar vortex enhancement after the five largest historical tropical volcanic eruptions in CMIP6 models.

To understand varying model responses to volcanic eruptions, the possible factors that determine the uncertainty need to be explored. The varying magnitude of volcanic eruptions (Dhomse et al 2020), differing model mean state (Zanchettin et al 2012), and small ensemble size (Bittner et al 2016a) could possibly induce different climate responses across the models, enlarging discrepancies between the observations and models. In regard of volcanic forcing, Azoulay et al (2021) proposed a threshold behavior, sulfur injections of 10 Tg, approximately corresponding to the 1991 Pinatubo eruption, being a threshold for Arctic polar vortex and surface climate responses. DallaSanta and Polvani (2022) further suggested that statistically significant Eurasian winter warming can be obtained with sulfur injections of at least 20 Tg. However, they showed that the forced signal barely exceeds the internal variability even with 160 Tg of sulfur injected, which is considerably greater than any volcanic eruptions known to have occurred in the past 2000 years. The large-scale environments could also play a critical role in shaping the circulation response to volcanic eruptions. Coupe and Robock (2021) for instance suggested that Eurasian winter warming after volcanic eruption can be explained only when observed El Niño-like equatorial Pacific Ocean variations are considered in the model.

The varying circulation responses to historical volcanic eruptions in NH winter are examined in figure 4 by using CMIP6 models. All available models (table S1) are utilized here unlike in previous studies where only selected models were used. Figure 4(a) shows that CMIP6 models simulate the observed positive AO/NAO response in the first DJF after the 1883 Krakatau and 1991 Pinatubo eruptions as in Zambri and Robock (2016). It leads to northern Eurasian warming (figure 4(b)). The multi-model ensemble mean also qualitatively well capture the observed Arctic polar vortex enhancement (figure 4(c)), with a significant inter-model relationship with AO index variations (figure 3(c)). These results suggest the downward influence of the stratosphere to the troposphere after the 1883 Krakatau and 1991 Pinatubo eruptions. Although a few models do not simulate the observed responses, a strong relationship between Arctic polar vortex enhancement and positive AO response (figure 3(c)) and a consistent result derived from high-top CMIP6 models (figures S4(c) and S5) suggest that varying results in the previous studies are partly due to the small ensemble size.

Zoom In Zoom Out Reset image size

The sea level pressure anomalies in the first DJF of the individual historical volcanic eruptions are further illustrated in figures 4(d)–(h). They reveal a wide range of surface response. The 1883 Krakatau and 1991 Pinatubo eruptions show positive AO/NAO responses, with greater intensity for the former presumably because of stronger radiative perturbation (figure 4(i)). These two cases contrast to the 1902 Santa María and 1982 El Chichón eruptions which are not followed by positive AO/NAO (figures 4(e) and (g)). Such difference likely results from rather weak radiative forcing compared to the 1883 Krakatau and 1991 Pinatubo eruptions (figure 4(i)) (Sato et al 1993, Jacobson et al 2020). The 1963 Agung eruption has asymmetric shortwave inflow with much greater intensity in the SH than in the NH (figure 4(i)) (Sato et al 1993, Jacobson et al 2020) and presents significant positive AO in high-latitudes (figure 4(f)) with comparable sea level pressure decrease over the Arctic to the 1991 Pinatubo eruption (figure 4(h)) but no change over the North Atlantic Ocean to western Europe (figure 4(f)). This result implies that the meridional structure of volcanic forcings likely plays a role in NH-winter climate response to volcanic eruptions. Thus, it would be valuable to examine NH circulation responses to extratropical volcanic eruptions which has received less attention.

During the last millennium, many volcanic eruptions occurred in mid- to high-latitudes (Gao et al 2008), reducing incoming solar radiation especially in the relevant hemisphere (Stevenson et al 2016). This hemispherically-asymmetric forcing causes surface cooling in the hemisphere of eruption and produces distinct climate responses, such as Intertropical Convergence Zone shift, interhemispherically-asymmetric summer monsoon precipitation decrease, and ENSO (e.g. Oman et al 2005, Liu et al 2016, Stevenson et al 2016, Zuo et al 2018, 2019a, 2019b, 2021, Pausata et al 2020).

For high-latitude volcanic eruptions, Oman et al (2005) reported rather minor responses of Arctic polar vortex and unclear surface anomalies following the Katmai eruption which occurred at 58°N. In contrast, Zambri et al (2019) showed negative NAO after the 1783 Laki eruption which occurred at 64°N. Sjolte et al (2021) additionally presented the appearance of positive and negative NAO to low- and NH high-latitude volcanic eruptions, respectively, during the last millennium from climate reconstructions. In general, there is little clarity as to how the latitude of volcanic eruptions affects the Arctic polar vortex, lower tropospheric circulation and Eurasian winter surface climate.

Figure 5 compares atmospheric circulation responses to tropical and northern extratropical volcanic eruptions. Here, CESM-LME simulations are utilized as they allow to analyze multiple tropical and northern extratropical volcanic eruptions in the last millennium. Five tropical and five northern extratropical volcanic eruptions are analyzed in figure 5. Following Fasullo et al (2019), northern extratropical volcanic eruptions are defined as top five volcanic eruptions which have the strongest interhemispheric difference in shortwave radiation reduction with a greater perturbation in the NH than in the SH (see figures S1(a) and (c)). The northern extratropical volcanic eruptions exhibit weak and non-robust NH mid- to high-latitude climate responses (figures 5(d)–(f)) as previously reported (Oman et al 2005). They contrast to the tropical volcanic eruptions which present positive AO and significant warming over the northern Eurasia in the first DJF following volcanic eruptions (figures 5(a) and (b)) with an enhanced Arctic polar vortex (figure 5(c)) as discussed above.

Zoom In Zoom Out Reset image size

This result indicates that extratropical circulation and surface climate responses to volcanic eruptions are dependent on the latitude of volcanic eruptions. A further analysis based on PMIP3 last millennium simulations shows largely consistent results, reaffirming the crucial role of the latitude of volcanic eruptions (figure S6). This finding suggests that the uncertainty of NH circulation response to volcanic eruptions can be partly caused by differing latitudes of volcanic eruptions.