Abstract:The procedures for (a) revealing the optimally growing initial errors (OGIEs) in a deep learning ENSO forecasting model and (b) evaluating the results in numerical models. El Niño‐Southern Oscillation (ENSO) is the dominant atmosphere–ocean coupled mode of year‐to‐year variations in the tropical Pacific. It shows diverse spatiotemporal characteristics and casts major influences on seasonal predictions of global weather–climate extrema. Despite numerous dynamical and statistical models for ENSO prediction and predictability studies, they are commonly subjected to one‐to‐three issues among less skillful simulation of El Niño diversity, huge requirements of computational resources and a low robustness in statistics. Here, an efficient deep‐learning model involving nonlinear coupling of multiple variables is independently developed to study the predictability of two types of El Niño events related to initial uncertainty, which is the first kind of predictability problem. The model can skillfully simulate statistically robust features of observed El Niño diversity in terms of periodicity, amplitude, and seasonal phase‐locking. Using this model, we have revealed mathematically several new types of fastest‐growing initial errors in two types of El Niño predictions based on a novel concept of conditional nonlinear optimal perturbation (CNOP), especially including one that can strengthen central Pacific types of events, which is rarely investigated before. Moreover, CNOPs are superimposed into a numerical model, GFDL CM2p1, for comprehensive validation and growth mechanism mining, which demonstrates the consistent dynamical evolution of initial errors in both numerical and AI models. Our study represents the first attempt to explore the first kind of ENSO predictability problem from perspectives of nonlinear error‐evolving dynamics using a data‐driven model. This is of great importance as it offers us sufficient confidence to perform ENSO‐related (such as the Madden–Julian Oscillation, etc.) mechanisms and predictability studies in the future without strongly relying on dynamical numerical models.

ENSO‐Related Precursor Pathways of Interannual Thermal Anomalies Identified Using a Transformer‐Based Deep Learning Model in the Tropical Pacific

Incorporating heat budget dynamics in a Transformer-based deep learning model for skillful ENSO prediction

Prediction of ENSO Using Multivariable Deep Learning

Real-time predictions of the 2023–2024 climate conditions in the tropical Pacific using a purely data-driven Transformer model

A self-attention–based neural network for three-dimensional multivariate modeling and its skillful ENSO predictions

Spatial-temporal transformer network for multi-year ENSO prediction

Echo State Network Based on Phase Space Reconstruction: El Niño 3 Index Forecasting

ENSO Forecasts with Spatiotemporal Fusion Transformer Network

Quantifying the Relative Contributions of the Global Oceans to ENSO Predictability With Deep Learning

An investigation into improving El Niño-Southern Oscillation prediction based on temporal transformer architecture

A Transformer‐Based Deep Learning Model for Successful Predictions of the 2021 Second‐Year La Niña Condition

ResoNet: Robust and Explainable ENSO Forecasts with Hybrid Convolution and Transformer Networks

Tendency-and-attention-informed deep learning for ENSO forecasts

Using Deep Learning to Identify Initial Error Sensitivity for Interpretable ENSO Forecasts

Explainable deep learning for insights in El Niño and river flows

Transformer for EI Niño-Southern Oscillation Prediction

Deep learning with autoencoders and LSTM for ENSO forecasting

Explained predictions of strong eastern Pacific El Niño events using deep learning

The first kind of predictability problem of El Niño predictions in a multivariate coupled data‐driven model

Understanding the Low Predictability of the 2015/16 El Niño Event Based on a Deep Learning Model