Southern Hemisphere Volcanism Triggered Multi-year La Niñas during the Last Millennium

The worldwide catastrophic weather events caused by the recent three-year La Niña from 2020 to 2022 have garnered significant attention (Jones, 2022). Compared to a single-year La Niña, a multi-year La Niña is more likely to cause severe droughts, heatwaves, and wildfires in the southwestern United States (Cole et al., 2002; Williams et al., 2020). It also leads to floods in Australia and Southeast Asia and deadly hurricanes over the North Atlantic (Raj Deepak et al., 2019; Jones, 2022), primarily due to its stronger cumulative impact (Wang et al., 2023).

The occurrence of multi-year La Niña events was previously attributed to several factors, including significant upper-ocean heat discharge induced by a preceding extreme El Niño (Wu et al., 2019; Iwakiri and Watanabe, 2021), influences from higher latitudes of the North Pacific that cause meridionally broad easterly wind anomalies slowing the heat recharge of the equatorial Pacific (Park et al., 2021; Geng et al., 2023; Zhang, 2023), the negative phase of the North Pacific Meridional mode (Kim et al., 2023), and tropical inter-basin interactions involving the Indian Ocean Dipole and/or Atlantic Niño (Okumura et al., 2011; McPhaden, 2023). A recent study has also indicated that the prevailing multi-year La Niña events since 1970 have primarily been triggered by rapid onsets following extreme or central-Pacific El Niños (Wang et al., 2023).

Since 1920, there have been ten instances of multi-year La Niña events, with eight of them occurring after 1970 (Wang et al., 2023). The frequent occurrence of multi-year La Niñas after 1970 is likely associated with a faster warming of the western Pacific, particularly a strengthened sea surface temperature (SST) gradient towards the west. This warming contrast between the western and eastern equatorial Pacific enhances the zonal advective feedback for central-Pacific El Niño to multi-year La Niña events and the thermocline feedback for super El Niño to multi-year La Niña events. Whether this strengthened SST gradient towards the west is a result of internal variability or a response to external forcing remains an open question (Power et al., 2021; Hartmann, 2022; Lee et al., 2022; Heede and Fedorov, 2023).

In the future, anthropogenic forcing–induced global warming will intensify the maximum warming in the subtropical northeastern Pacific, enhancing the regional thermodynamic response to perturbations. As a result, anomalous easterlies of the first-year La Niña will be generated further northward compared with the present-day climate, leading to a slower heat recharge of the equatorial Pacific and therefore an increased frequency of multi-year La Niña events. Under a low-emission scenario, the projected frequency is expected to range from 19%±11%, while under a high-emission scenario, it is expected to range from 33%±13% Geng et al. (2023).

These studies have primarily focused on investigating the internal variability of air–sea interaction and its potential influences from climate change. However, an intriguing question to explore is whether external forcing can directly impact the occurrence of multi-year La Niña events, rather than solely altering the mean state. Investigating this aspect can provide valuable insights into the dynamics of multi-year La Niña events.

In a recent study by Fasullo et al. (2023), it was simulated that following the largest 2019 Indian Ocean Dipole, the increase in biomass aerosol from the 2019–20 Australian wildfire season could lead to an increase in cloud albedo and cooling of the southeastern subtropical Pacific Ocean. This cooling tendency resulted in a northward shift of the intertropical convergence zone, which favored the occurrence of strong La Niña events from 2020 to 2022. This work suggests that the sudden cooling in the Southern Hemisphere (SH) caused by biomass aerosol can trigger multi-year La Niña events. On longer time scales, a cooling trend in the eastern tropical Pacific was also simulated due to cooler temperatures over the Southern Ocean near Antarctica (Kang et al., 2023a, b, c). One of the factors contributing to this cooling trend was the onset of the Antarctic ozone hole since about 1980 (Hartmann, 2022).

All these simulations have demonstrated that a cold SH can lead to a drying of the boundary in the SH Pacific and a decrease in moist static energy. Consequently, this leads to a northward shift of the intertropical convergence zone and triggers a multi-year La Niña event or a La Niña-like trend (Fasullo et al., 2023; Kang et al., 2023a, b, c). However, it is worth noting that this mechanism currently lacks support from observations due to the limited number of available samples from wildfires or ozone holes. Therefore, it is important to seek evidence from paleo-proxies, which can provide valuable insights into past climate conditions.

Volcanic eruptions, as natural external forcings, have the potential to alter our climate on various time scales, ranging from subseasonal (Xing and Liu, 2023), interannual (Adams et al., 2003), multidecadal (Mann et al., 2021), and centennial (Miller et al., 2012) time scales. This is achieved through the injection of sulfur dioxide into the stratosphere (Robock, 2000). The formation of volcanic sulfate aerosols occurs through the reaction between sulfur dioxide and hydroxide or water vapor, and these aerosols persist in the stratosphere for approximately one to two years (Timmreck, 2012). They have a significant impact on global climate by scattering incoming solar radiation, ultimately resulting in global surface cooling and monsoon changes (Iles and Hegerl, 2014; Paik et al., 2020; Singh et al., 2020; Liu et al., 2022b). Frequently erupting volcanoes offer us an opportunity to study the response of multi-year La Niñas to the Southern Ocean cooling. In comparison to the extensively researched El Niño response to tropical volcanic eruptions caused by the suppression of the African monsoon and the resultant tropical westerly anomaly (Khodri et al., 2017; Liu et al., 2022a; Pausata et al., 2023), the El Niño–Southern Oscillation (ENSO) response to high-latitude volcanic eruptions has received less attention.

To investigate the response of ENSO to volcanic eruptions in the tropics, Northern Hemisphere, and SH, our previous work conducted a reconstruction analysis covering the period from 900 to 2000 AD (Liu et al., 2018b). In this analysis, the focus was on the SH eruptions that could cool the SH rather than the tropics and Northern Hemisphere (Haywood et al., 2013; Pausata et al., 2015; Colose et al., 2016; Stevenson et al., 2016; Liu et al., 2018b; Erez and Adam, 2021), and multiproxy data were used. Specifically, tree ring records in southwestern North America were utilized as reliable indicators of ENSO variability, due to the stable Pacific–North American teleconnection over the last millennium (Li et al., 2011; Han et al., 2023).

During the period 900–2000 AD, for which ENSO reconstructions were available (Fig. 1a), a total of 26 SH volcanic eruptions were recorded in the newest volcanic reconstruction (Sigl et al., 2015). The two most recent eruptions for which reliable observations were available occurred in 1931 and 1979 (Fig. 1a). By conducting a superposed epoch analysis (SEA; Haurwitz and Brier, 1981) on all available reconstructions for these 26 eruptions, a three-year negative anomaly in the boreal-winter Niño-3 index was observed (Fig. 1b). Notably, a significant anomaly (at the 90% confidence level) was observed during the eruption winter and the third winter following the eruptions.

Figure 1.  Three-year La Niña response to SH volcanic eruptions in paleo-proxies: (a) Three sets of long-term ENSO indices reconstructed based on tree rings from southwestern North America covering the periods 1408–1978 AD, 1300–2000 AD, and 900–2000 AD (gray lines), as well as instrumental October–March Niño-3 index from 1920–2000 AD (red line). Blue triangles and dotted vertical lines denote cold seasons (defined by the beginning year of the cold season of the NH) following 26 SH eruptions back to 900 AD. In the inset, the 1931 and 1979 SH eruptions are also marked. (b) Composite response of three Niño-3 reconstructions to 26 SH volcanic eruptions. Green and orange colors mark the pre- and post-eruption composites, respectively. “0” denotes the first cold season following the eruptions. Confidence intervals (90%, 95%, 99%) are marked by horizontal dashed lines.

High-latitude volcanic eruptions, which are known for their significant cooling effect in the hemisphere of eruption, can induce a migration of the intertropical convergence zone towards the comparatively warmer hemisphere (Haywood et al., 2013; Pausata et al., 2015; Liu et al., 2016; Colose et al., 2016; Stevenson et al., 2016; Liu et al., 2018b; Erez and Adam, 2021). This migration is primarily driven by the atmospheric energy balance (Schneider et al., 2014). Simulations have shown that volcanic eruptions in the SH can cause a northward shift in the intertropical convergence zone (Haywood et al., 2013; Liu et al., 2016; Stevenson et al., 2016). As a result, this northward migration of the intertropical convergence zone leads to the occurrence of a multi-year La Niña event, as simulated by Fasullo et al. (2023).

The analysis of paleo-reconstructions provides evidence that volcanic eruptions in the SH can induce a multi-year La Niña event through the ocean cooling effects. However, it is crucial to note that, for a specific event, the ENSO anomaly can be influenced by both internal variability and external forcing (Liu et al., 2018a; Zanchettin et al., 2019). For instance, after the eruption in the SH in 1931, a cooling signal persisted for three years, whereas no discernible signal was observed during the winter of 1979 eruption (Fig. 1a). The most powerful SH volcanic eruption in recent decades, known as the Hunga Tonga-Hunga Ha'apai eruption on 15 January 2022, had a volcanic explosivity index of five, comparable to the Krakatoa eruption in 1883 (Carn et al., 2022). However, this eruption only released 0.5–1.5 Tg of SO₂ into the stratosphere, even though it injected 146 Tg or approximately 10% of the total stratospheric water vapor prior to the eruption (Millán et al., 2022). As a result, the decrease in SH SST was relatively small (Zuo et al., 2022; Schoeberl et al., 2023).

Our results were primarily derived from ENSO reconstruction using tree rings in southwestern North America. However, it is important to note that different paleoclimate reconstructions can yield divergent responses. For example, ENSO reconstructions based on coral records did not exhibit a significant multi-year La Niña response (not shown), similar to their inability to capture an El Niño response following tropical volcanic eruptions (Dee et al., 2020), against other reconstructions (Adams et al., 2003; Robock, 2020). This raises an interesting yet challenging issue of identifying which reconstructions are more reliable, and further research in this area is warranted.

Although both wildfires (Fasullo et al., 2023) and SH volcanic eruptions (Fig. 1) can contribute to the occurrence of a multi-year La Niña, they cool the Southern Ocean through different processes. For the Australian wildfires, tropospheric processes—specifically, the aerosol–cloud feedback—was identified as the mechanism responsible for cooling the Southern Ocean. On the other hand, for SH volcanic eruptions, the stratospheric aerosols play a role in cooling the Southern Ocean through the reflection of solar radiation. The strong zonal winds in the stratosphere result in a uniform distribution of volcanic aerosols, independent of the longitude of the eruption location. However, it is crucial to investigate the impact of the eruption season on the evolution of ENSO, as discussed by Stevenson et al. (2017). Unfortunately, the Last-Millennium volcanic reconstruction cannot provide information about the specific season of volcanic eruptions.

In this study, we present compelling evidence from paleo-proxies, utilizing a substantial sample size of 26 SH volcanic eruptions, to support the hypothesis that ocean cooling in the SH can lead to a multi-year La Niña event (Fasullo et al., 2023). Moreover, these findings emphasize the importance of incorporating the biomass aerosol process as a crucial component of the new ENSO cycle, in addition to the traditional air–sea interaction (Jin, 1997). Furthermore, this study highlights the necessity for greater focus on the Southern Ocean, as most climate models currently struggle to accurately simulate the Pacific response driven by the Southern Ocean. This can primarily be attributed to excessively weak stratocumulus cloud feedback in these models (Kang et al., 2023a).

The long-term volcanic eruption reconstruction, covering the period from 500 BC to 2000 AD, was utilized to expand the sample size for SH volcanic eruptions (Sigl et al., 2015). Additionally, ENSO indices reconstructed mainly from tree rings in southwestern North America were used, which covered the periods of 1408–1978 AD (D'Arrigo et al., 2005), 1300–2006 AD (Cook et al., 2008), and 900–2002 AD (Li et al., 2011), respectively. Instrumental SST data from the HadISST1 dataset were also used as a reference (Rayner et al., 2003). These ENSO reconstructions primarily capture boreal-winter Niño-3 SST changes and exhibit a high correlation with the instrumental Niño-3 index (Liu et al., 2018b).

In order to assess the impact of volcanic forcing, SEA (Haurwitz and Brier (1981) was employed to composite the ENSO response to volcanic eruptions. For the composite analysis, an 11-year window was utilized, including five years before and five years after the eruptions. To determine the significance, we employed the bootstrapped resampling method. The SEA was repeated 10 000 times using random draws from the studied period.

Acknowledgements. We acknowledge the useful discussion with Tao GENG and Yonggang LIU on an early version of the manuscript. The work was supported by the National Key Research and Development Program of China (Grant No. 2020YFA0608803) and the National Natural Science Foundation of China (Grant Nos. 41975107, 41875092 and 42005020).