Dynamics for El Niño-La Niña asymmetry constrain equatorial-Pacific warming pattern

The El Niño-Southern Oscillation (ENSO) is a climate phenomenon arising from coupled ocean-atmosphere interactions in the equatorial Pacific. While current climate models generally reproduce the overall spatial pattern and temporal evolution of ENSO, they often fail to capture its inherent nonlinearities, particularly the asymmetry between El Niño (warm) and La Niña (cold) phases. This asymmetry, often measured by sea surface temperature (SST) skewness, is characterized by more intense and frequent El Niño events compared to La Niña events in the eastern Pacific. The poor representation of this asymmetry in climate models is a significant deficiency, potentially leading to inaccurate projections of future climate change. Previous studies have suggested several possible causes, including Pacific mean-state SST biases affecting the atmospheric Walker Circulation and state-dependent noise in ENSO excitation. However, the dominant nonlinear dynamical process responsible for this asymmetry remains unclear. Oceanic nonlinear dynamical heating (NDH), particularly in the subsurface ocean along the equatorial thermocline and Equatorial Undercurrent (EUC), is a deterministic advective process that could significantly contribute to ENSO asymmetry. While its role in the surface mixed layer has been investigated, the impact of subsurface NDH in climate models hasn't been thoroughly explored. This study aims to address this gap by evaluating the relationship between subsurface NDH, ENSO asymmetry, and associated atmospheric nonlinearities in state-of-the-art climate models, ultimately assessing the consequences for future climate projections.

Numerous studies have highlighted the discrepancies between observed and simulated ENSO asymmetry in climate models. An et al. (2005) and Zhang and Sun (2014) documented this deficiency in CMIP simulations. Bellenger et al. (2014) and Bayr et al. (2018, 2019) analyzed the role of atmospheric feedbacks and error compensation between dynamic and thermodynamic processes in contributing to these biases. Other research has investigated the influence of Pacific mean-state SST biases (Sun et al., 2006; Bayr et al., 2020), state-dependent noise (Levine et al., 2016), and the nonlinear rectification effect of ENSO on the mean climate state (Sun and Zhang, 2006; Liang et al., 2012; Sun et al., 2014). Hayashi and Jin (2017) specifically focused on the role of subsurface NDH in enhancing ENSO asymmetry through observational studies, providing the foundation for the current research. Studies on the impact of ENSO changes on future climate projections (Yeh et al., 2009; Collins et al., 2010; Watanabe et al., 2012; Kim et al., 2014; Kohyama et al., 2017, 2018; Karamperidou et al., 2017; Cai et al., 2018; Hu and Fedorov, 2018; Brown et al., 2020) often show a large spread, indicating the need to refine our understanding of the underlying ENSO dynamics.

This study utilized multiple reanalysis datasets (ORAS3, ORAS5, SODA331, GODAS, ERA-interim, ERA5, TropFlux, OAFlux, CERES EBAF Ed4.0, GEWEX SRB version 3, ISCCP-FH, GPCP version 2.3, NOAA ERSST version 5) and climate model outputs from the Coupled Model Intercomparison Project Phase 5 (CMIP5) and Phase 6 (CMIP6) to investigate ENSO asymmetry and its relationship to subsurface NDH. The CMIP datasets included monthly fields of SST, zonal wind stress, near-surface temperature, surface heat fluxes, precipitation, potential temperature, and three-dimensional ocean currents. Data availability varied across models; a subset of CMIP5 (25) and CMIP6 (18) models provided the necessary ocean data for NDH calculations. ENSO asymmetry was quantified using the skewness of detrended Niño-3 SST anomalies. Subsurface NDH was calculated from nonlinear temperature advective terms in a specific equatorial eastern-Pacific box (100°W–180°, 1°S–1°N, 50–150 m). The study also examined atmospheric nonlinearities, including shortwave (SW) surface heat flux feedback asymmetry and central-Pacific zonal wind stress feedback asymmetry. CMIP models were categorized into groups based on their subsurface NDH efficiency (σNDHsub/σENSO). To assess future climate responses, the study compared historical simulations with future scenario simulations (CMIP5 RCP8.5 and CMIP6 SSP5-8.5) for 2051-2100. The El Niño-likeness of tropical Pacific warming was defined as the spatial correlation between SST trend patterns in scenario simulations and the ENSO-regressed SST anomaly pattern in historical simulations. Statistical significance was assessed using two-tailed Student's t-tests and Welch's t-tests.

The study found that ENSO asymmetry, specifically the positive skewness of eastern-Pacific temperature anomalies, is poorly reproduced in most CMIP5 and CMIP6 models. However, a strong linear relationship exists between ENSO SST skewness (γENSO) and the relative strength of subsurface NDH variability to ENSO amplitude (σNDHsub/σENSO) across the CMIP models (correlation coefficient of 0.78, p < 0.00001). This implies that the intensity of simulated subsurface NDH is a key determinant of ENSO asymmetry. Most CMIP models exhibit too-weak subsurface NDH and too-weak linear dynamical ocean-atmosphere coupling. Analyzing CMIP models based on their subsurface NDH efficiency revealed that models with higher efficiency (group H and HH) better simulate the observed spatial patterns of SST skewness and mean equatorial NDH, resembling the patterns seen in reanalysis data. In contrast, models with low NDH efficiency (group L) displayed a westward extension of the cold tongue, indicating a biased climate mean state. These group L models show too-weak anomalous central-Pacific zonal wind stress responses to ENSO SST anomalies and inadequate anomalous westward EUC responses to westerly wind anomalies. This indicates deficiencies in the linear dynamical coupling processes from SST to winds and from winds to the EUC. The study further revealed that the projected future change in the long-term mean subsurface NDH is linearly related to the change in ENSO amplitude in group H models. However, even within group H, a spread in projected ENSO amplitude exists, with some models showing ENSO strengthening and others showing weakening. Crucially, the El Niño-likeness of tropical Pacific warming (defined as the spatial correlation between the SST trend and ENSO-regressed SST anomaly) is strongly correlated with ENSO amplitude change only in group H models (r = 0.57, p = 0.042; r = 0.89, p = 0.0073 for the subset group HH). The models in group H that project ENSO strengthening (H-sEN) show more eastern-Pacific warming, contrasting with those projecting weakening (H-wEN). This relationship was absent in group L models, suggesting that realistic simulation of ENSO nonlinear dynamics is essential to understand the impact of ENSO changes on future climate.

The findings highlight the importance of subsurface NDH as a key dynamical factor controlling ENSO asymmetry and its implications for future climate projections. The poor simulation of ENSO asymmetry in most climate models is attributed to biases in linear dynamical ocean-atmosphere coupling, where errors in SST-wind coupling are often compensated by errors in thermodynamic radiative feedbacks. This error compensation masks the relationship between ENSO amplitude change and the equatorial-Pacific warming pattern in response to greenhouse gas forcing. Only when climate models accurately simulate both linear dynamical coupling and thermodynamic feedbacks, thus reducing these error compensations, does the relationship between ENSO amplitude and warming patterns become apparent. The results demonstrate that accurately simulating ENSO nonlinear dynamics, especially the subsurface NDH, is crucial for projecting future equatorial Pacific warming patterns. Furthermore, the differences in the response to ENSO amplitude change between models with realistic NDH (group H) and those without (group L) highlight the critical role of ENSO nonlinearity in modulating the mean state and the resulting surface warming pattern.

This study demonstrates that the accurate simulation of subsurface nonlinear dynamical heating (NDH) is a critical factor for realistically representing El Niño-Southern Oscillation (ENSO) asymmetry in climate models. Models that accurately capture subsurface NDH exhibit a clear relationship between ENSO amplitude changes and the resulting pattern of equatorial Pacific warming under future climate scenarios. Biases in linear dynamical ocean-atmosphere coupling, specifically in the relationship between SST and winds and between winds and the Equatorial Undercurrent, significantly contribute to the inability of many climate models to represent ENSO asymmetry. This work underscores the need for improved model representation of both linear and nonlinear ENSO dynamics to reduce uncertainty in future climate projections. Future research should focus on identifying the mechanisms underlying the biases in linear dynamical coupling and improving their representation in climate models.

The study's reliance on CMIP5 and CMIP6 models might limit the generalizability of findings to other model frameworks. The availability of necessary ocean data for NDH calculations varied across models, restricting the number of models included in certain analyses. The focus on subsurface NDH might overlook other contributing factors to ENSO asymmetry. Future research could explore other potential mechanisms and incorporate a wider range of climate models to improve the robustness of conclusions.