Large biases and uncertainties remain in real-time predictions of El Niño–Southern Oscillation (ENSO) using process-based dynamical models; recent advances in data-driven deep learning algorithms provide a promising mean to achieve superior skill in the tropical Pacific sea surface temperature (SST) modeling. Here, a specific self-attention–based neural network model is developed for ENSO predictions based on the much sought-after Transformer model, named 3D-Geoformer, which is used to predict three-dimensional (3D) upper-ocean temperature anomalies and wind stress anomalies. This purely data-driven and time-space attention-enhanced model achieves surprisingly high correlation skills for Niño 3.4 SST anomaly predictions made 18 months in advance and initiated beginning in boreal spring. Further, sensitivity experiments demonstrate that the 3D-Geoformer model can depict the evolution of upper-ocean temperature and the coupled ocean-atmosphere dynamics following the Bjerknes feedback mechanism during ENSO cycles. Such successful realizations of the self-attention–based model in ENSO predictions indicate its great potential for multidimensional spatiotemporal modeling in geoscience.

Uncertainties in ocean-mixing parameterizations are primary sources for ocean and climate modeling biases. Due to lack of process understanding, traditional physics-driven parameterizations perform unsatisfactorily in the tropics. Recent advances in the deep-learning method and the new availability of long-term turbulence measurements provide an opportunity to explore data-driven approaches to parameterizing oceanic vertical-mixing processes. Here, we describe a novel parameterization based on an artificial neural network trained using a decadal-long time record of hydrographic and turbulence observations in the tropical Pacific. This data-driven parameterization achieves higher accuracy than current parameterizations, demonstrating good generalization ability under physical constraints. When integrated into an ocean model, our parameterization facilitates improved simulations in both ocean-only and coupled modeling. As a novel application of machine learning to the geophysical fluid, these results show the feasibility of using limited observations and well-understood physical constraints to construct a physics-informed deep-learning parameterization for improved climate simulations.

The frequent occurrences of the second-year surface cooling condition in the eastern equatorial Pacific, as observed in late 2021, are attributed to decadal changes in the thermocline depth, which determine the relative dominances of local cooling effect in the east and subsurface warming effect remotely from the west. Coupled models need to adequately represent these processes in a balanced way, thus being able to successfully predict the observed sea surface temperature evolution in late 2021.

The tropical Pacific is currently experiencing an El Niño event. Various coupled models with different degrees of complexity have been used to make real-time El Niño predictions, but large uncertainties exist in the intensity forecast and are strongly model dependent. An intermediate coupled model (ICM) is used at the Institute of Oceanology, Chinese Academy of Sciences (IOCAS), named the IOCAS ICM, to predict the sea surface temperature (SST) evolution in the tropical Pacific during the 2015–2016 El Niño event. One unique feature of the IOCAS ICM is the way in which the temperature of subsurface water entrained in the mixed layer (T_e) is parameterized. Observed SST anomalies are only field that is utilized to initialize the coupled prediction using the IOCAS ICM. Examples are given of the model’s ability to predict the SST conditions in a real-time manner. As is commonly evident in El Niño-Southern Oscillation predictions using coupled models, large discrepancies occur between the observed and predicted SST anomalies in spring 2015. Starting from early summer 2015, the model can realistically predict warming conditions. Thereafter, good predictions can be made through the summer and fall seasons of 2015. A transition to normal and cold conditions is predicted to occur in late spring 2016. Comparisons with other model predictions are made and factors influencing the prediction performance of the IOCAS ICM are also discussed.