A study of the El Niño phenomenon over the past 21,000 years suggests that El Niño responded in complex ways to a changing climate, with several competing factors playing a part in its varying strength. See Letter p.550
The episodic warming and cooling of the tropical Pacific Ocean's surface waters, known as the El Niño–Southern Oscillation (ENSO), is responsible for large year-to-year variations in global climate, and causes widespread effects that include droughts (Fig. 1), floods and fires. It is therefore important to know how ENSO may change in the future. To address this question, researchers can look to past climates for clues. In this issue, Liu et al.1 (page 550) present a set of climate- model experiments investigating the evolution of ENSO over the 21,000 years since the peak of the most recent glacial period. Their experiments show that ENSO varied on a range of timescales, producing a complex history with epochs of weaker and stronger activity.
ENSO is a natural fluctuation of climate that arises from interactions between the atmosphere and ocean in the tropical Pacific, with El Niño events (warming) occurring about every three to seven years. The strength of these events may vary as a result of natural, internal feedback processes2 as well as through external factors such as human-induced global warming3. Over the past 21,000 years, the global and regional climate changed in response to variations in incoming solar radiation driven by slow changes in Earth's orbit around the Sun; the melting of continental ice sheets; and natural fluctuations in greenhouse gases. In their study, Liu and colleagues consider how ENSO was modified by these competing factors.
Previous studies of past ENSO behaviour using complex global models have been restricted to simulations of 'time-slices' of key periods such as the Last Glacial Maximum (21,000 years ago) and mid-Holocene (6,000 years ago)4. These time-slices typically consist of only hundreds to a few thousand years because of the computer resources needed to run such experiments. Now, for the first time, Liu et al. use a complex global climate model — the Community Climate System model version 3 (CCSM3) maintained at the US National Center for Atmospheric Research — to simulate the full 21,000 years from the last glacial. This has allowed the sensitivity of ENSO to multiple influences (forcing) to be investigated in a continuously evolving climate. The model's forcing includes changes in Earth's orbital parameters, the concentrations of greenhouse gases and continental ice sheets, as well as ocean freshwater input from melting ice. The authors also carried out experiments to explore the sensitivity of ENSO to each of these factors individually.
Liu et al. find that ENSO gradually became around 15% stronger during the Holocene (the past 11,000 years). This strengthening occurred in response to altered incoming solar radiation due to orbital changes, which led to warming of the tropics and increased feedbacks between the atmosphere and upper layers of the ocean. The role of orbital forcing is confirmed by the experiment that involved orbital changes only, which reproduces the gradual trend in ENSO amplitude.
The model simulates a slightly weakened ENSO at the Last Glacial Maximum. In the subsequent 'deglacial' period, as the climate warmed and continental ice sheets melted, ENSO amplitude varied on millennial timescales as a result of changes in ocean freshwater input from melting ice, which modified the circulation in the Atlantic Ocean. Liu et al. identify a set of mechanisms for this response, ultimately pointing to the varying magnitude of the annual cycle of temperature in the equatorial Pacific, which is strong when ENSO is weak and vice versa5. During the same period, increasing atmospheric carbon dioxide concentrations tended to weaken ENSO, whereas the impact of retreating continental ice sheets on atmospheric circulation led to a strengthening.
The authors compare these results with reconstructions of ENSO variability from proxy records such as corals, mollusc shells and lake and ocean sediments6,7,8,9,10. Several proxy records from the early to mid-Holocene (11,000 to 5,000 years ago) indicate that ENSO was weaker than it is now, with reductions of around 30–50%, although the timing of the minimum varies between records6,7,8,9. Another reconstruction does not show a significant change in ENSO variability at this time10. Climate-model simulations of the mid-Holocene4 show a 10–15% weakening of ENSO (relative to the pre-industrial climate), which is consistent with Liu and co-workers' results but smaller than most proxy records suggest. The apparent underestimate of the change by models may reflect limitations of the proxies, or may highlight insufficient model sensitivity.
There are fewer proxy records of ENSO from the period before the Holocene, and the available records do not provide a clear picture of ENSO in glacial and deglacial climates6,9. Model simulations of the Last Glacial Maximum also fail to show a consistent change in ENSO variability4. Liu et al. propose that this lack of agreement between models may be due to the competing effects of large continental ice sheets and low atmospheric CO2 levels, which influence ENSO in opposing ways during this time. Disagreement between proxy records suggests that the spatial pattern of ENSO impacts across the Pacific may also have changed.
A great strength of this study is its use of multiple simulations using individual forcing to confirm the role of each factor in ENSO changes. However, the results are based on a single climate model, and it is well known that models differ in their simulation of ENSO. In particular, models disagree about changes in ENSO strength in both past climates such as the mid-Holocene and the Last Glacial Maximum4 and in projections of future climate with increased concentrations of greenhouse gases3. For this reason, it is imperative that the authors' 21,000-year experiments are repeated using other climate models, ideally including the full set of sensitivity experiments.
Another challenge is to develop proxy reconstructions of ENSO that can provide a clearer picture of past variability than is currently available, and to improve methods of comparing models and proxy records10,11. In particular, the large unforced natural variability of ENSO implies that records spanning many decades or longer may be required to identify changes in strength that are outside this natural range2,10. As Liu et al. conclude, to provide a robust basis for comparison between models and proxy reconstructions, an expanded set of proxy records is needed, particularly from the equatorial Pacific region, which is most sensitive to ENSO.
Liu and colleagues' study constitutes a major step towards understanding the complex history of this crucial phenomenon. At the same time, their results suggest that ENSO may respond in opposing ways to different regional and global influences, highlighting the challenge of predicting its future activity.