Tropical Pacific trends under global warming: El Niño-like or La Niña-like?

Many state-of-the-art climate models project that the tropical Pacific sea surface temperature (SST) response to the greenhouse gas (GHG) forcing would be an ‘El Niño-like’ pattern, meaning that the eastern equatorial Pacific warms faster than anywhere else in the tropical Pacific ([1], and reference therein). Such a pattern would be associated with weakened trade winds, reduced upwelling in the eastern equatorial Pacific, and a flattened equatorial thermocline across the Pacific Ocean. It was further suggested that all these climatological changes will modify, or even significantly change, the characteristics of the El Niño-Southern Oscillation (ENSO) in the future ([2], and reference therein). For example, the ENSO-related warming tends to shift westward due to the flattened thermocline [3], and the ENSO amplitude might increase as the barrier to deep convection is reduced in the eastern equatorial Pacific [4]. These model results describe a vivid picture of how ENSO will respond to the future global warming if the mean-state change in the tropical Pacific is El Niño-like. Now a natural question to ask is how reliable the projected El Niño-like pattern might be in light of the available observations. If the models are unable to simulate the observed mean-state change since the Industrial Revolution, the era in which the global warming has already been taking place, we must exercise caution when interpreting their projected changes in the future. Fig. 1a shows the ensemble trend of the tropical Pacific SST from four different datasets for the period of 1880–2015. It appears that there was a cooling trend since 1880, especially in the eastern equatorial Pacific, indicating an intensified zonal SST gradient (SSTG) across the equatorial Pacific. Such a cooling trend has also been noticed in previous studies [5,6]. Given the discrepancy in representing internal variability in different datasets, it was argued that the change in SSTG may be partly associated with internal variability [7]. However, when using selected datasets that have trends out of the uncertainty range associated with internal variability ([8]; Supplementary Table S1), the cooling trend and the SSTG intensification are even more pronounced (Fig. 1d). Furthermore, the meridional SST gradient in the eastern equatorial Pacific, which plays a crucial role in determining the latitudinal location of the Intertropical Convergence Zone (ITCZ), also showed a trend of intensification over the same period in these datasets (Supplementary Fig. S1). It is clear that the available observations suggest that the global warming in the recent past tends to induce a ‘La Niña-like’ rather than ‘El Niño-like’ mean-state change in the tropical Pacific. Trends of tropical Pacific SST averaged over observational datasets (first row), and over model outputs under the historical GHG forcing (second row) and under the RCP85 GHG forcing (last row). In (a)-(c), all datasets are used to calculate the average trends. In (d)-(f), only selected datasets that pass our uncertainty test are used. The uncertainty test is based on the method of Lian et al. [8], which gives the uncertainty range of trend estimate in the presence of large natural variations. Here the test is not applied to every grid point, but to the trend of zonal SST gradient (SSTG) as defined by the difference between the average SSTs in the two rectangles shown in the figure (See Supplementary Information for more details). The time periods used to calculate the trends in observations, model historical runs, RCP85 runs are 1880–2015, 1861–2001 and 2006–2098, respectively. Unit is °C per century. On the contrary, the majority of the CMIP5 models [9] simulate a warming trend in the eastern equatorial Pacific and a weakened SSTG under the historical GHG forcing, though only four of these trend estimates pass our uncertainty test and thus may be considered as forced responses (Fig. 1b and Supplementary Table S2). The ensemble trend of these four models shows a mean-state change that closely resembles an ‘El Niño-like’ pattern (Fig. 1e), as noted in previous studies. Under the future GHG forcing scenario, many more CMIP5 models project a robust trend of weakening SSTG (Supplementary Table S3), and the multi-model ensemble trend shows a distinct ‘El Niño-like’ pattern (Fig. 1c and f). Clearly, for both the past and the future, the present climate models tend to produce an ‘El Niño-like’ mean-state change as a forced response to the GHG forcing. The contrasting trends between the observations and the historical simulations prompt us to rethink the ocean-atmosphere coupled system in the tropical Pacific (Fig. 2a) and the mechanisms that may control its long-term changes under the global warming. A schematic showing the mechanisms that control the long-term SST change in the tropical Pacific under the global warming. (a), The normal climate conditions in the tropical Pacific, including the zonal SST gradient and the associated Walker Circulation, the mean positions of deep convection and low-stratus cloud, and the upwelling. (b), The climate change from an atmospheric perspective. The Walker Circulation gets weakened and shrinks with the slowdown of global atmospheric overturning circulations; deep convection moves eastward and weakens; low-stratus cloud in the eastern tropical Pacific is reduced; SST in the eastern equatorial Pacific warms faster than the west counterpart and resembles the ‘El Niño-like’ pattern. (c), Climate change from the oceanic perspective. The SST in the eastern equatorial Pacific warms less faster than the west counterpart due to the strong upwelling in eastern tropical Pacific and resembles the ‘La Niña-like’ pattern; Walker Circulation gets strengthened and expands eastward; deep convection increases in the western tropical Pacific; low-stratus cloud in the eastern tropical Pacific increases. From an atmospheric perspective, the weakened global hydrological cycle under GHG forcing is suggested to be mainly responsible for the ‘El Niño-like’ response in the tropical Pacific [10]. Model experiments indicate that the GHG-forced global warming would make the global water vapor amount increase at a rate of ∼7% K−1, but the precipitation increase at a rate of merely ∼2% K−1. Therefore, there must be a slowdown of global atmospheric overturning circulations. While some studies argued that the convective mass flux is not closely related to the strength of the Walker Circulation [11], it is generally accepted that the slowdown of the global atmospheric overturning could weaken the Walker Circulation [10], which would warm the eastern tropical Pacific SST via the Bjerknes feedback. In addition, the low stratus clouds off the west coast of the South America would decrease with the increasing SST, thus permitting more solar radiation into the eastern equatorial Pacific Ocean and further enhancing the warming there. Both of these feedback processes favor an ‘El Niño-like’ mean-state change (Fig. 2b). From an oceanic perspective, on the other hand, the heat input to the eastern equatorial Pacific Ocean would be largely compensated by the enhanced upwelling there due to increased surface-layer stratification, which would strengthen the zonal SSTG and, through the Bjerknes feedback, result in a ‘La Niña-like’ mean-state change (Fig. 2c). This mechanism was referred to as an ‘ocean dynamical thermostat’ because of its modulating effect on the global warming. While the slow deep-ocean warming advected along the subsurface branch of the subtropical cells may gradually warm the equatorial thermocline [12], the effect of the ocean dynamical thermostat may be a permanent feature of the equilibrium climate [13]. It seems that this oceanic mechanism and the aforementioned atmospheric mechanism are both physically viable but are completely different in their consequences, and the latter is clearly playing a dominant role in the models. So the question boils down to which mechanism is actually winning in reality. Our analysis of observational data shows that the global warming has led to a ‘La Niña-like’ mean-state change since 1880, in contrast to the ‘El Niño-like’ change in the state-of-the-art CMIP5 models. The implication is that the oceanic mechanism is overly suppressed by the atmospheric mechanism in the models, perhaps due in part to a too diffusive thermocline in the eastern equatorial Pacific [14]. Other common biases may also contribute to the ‘El Niño-like’ pattern in the models [15]. For example, the well-known cold tongue bias could cause an overestimated heat flux into the eastern equatorial Pacific, and the underestimated negative feedback between SST and cloud cover could make the SST warming not sufficiently damped by the reduced solar radiation. It is thus necessary to correct these model biases in order to resolve the discrepancy between observed and modeled trends. One way to proceed is to improve model simulations of the equatorial undercurrent and inter-basin interactions [16,17]. At present, given the important impact of the mean-state change to natural variations such as ENSO, it is necessary to reevaluate many aspects of the past and future climate variability that were based on the simulated/projected ‘El Niño-like’ trends. This work was supported by grants from the China Ocean Mineral Resources Research and Development Association program (DY135-E2-3-01), the National Natural Science Foundation of China (41690121, 41690120, and 41730535) and the National Program on Global Change and Air–Sea Interaction (GASI-IPOVAI-04).

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.