Wang et al. (2026) Skillful and interpretable ENSO prediction via a glass-box dynamical-deep learning model

Identification

• Journal: npj Climate and Atmospheric Science
• Year: 2026
• Date: 2026-08-14
• Authors: Mingyu Wang, Jianping Li
• DOI: 10.1038/s41612-026-01513-6

Research Groups

• Frontiers Science Center for Deep Ocean Multi-spheres and Earth System (DOMES)/Key Laboratory of Physical Oceanography/Academy of Future Ocean/College of Oceanic and Atmospheric Sciences/Center for Ocean Carbon Neutrality, Ocean University of China, Qingdao, China
• Laboratory for Ocean Dynamics and Climate, Qingdao Marine Science and Technology Center, Qingdao, China

Short Summary

This study presents a novel approach to ENSO prediction using a dynamical-deep learning model that achieves skillful predictions up to 19 months ahead while rendering all model terms explicit.

Objective

• Investigate the feasibility of applying dynamical system deep learning (DSDL) for interpretable ENSO prediction

Study Configuration

• Spatial Scale: Global, focusing on the Pacific Ocean and its surrounding regions
• Temporal Scale: Interannual climate variability with a focus on predicting ENSO events up to 19 months ahead

Methodology and Data

• Models used: Dynamical system deep learning (DSDL) model
• Data sources: Historical climate data from satellite observations, reanalysis datasets, and in-situ measurements

Main Results

• The DSDL model achieves skillful predictions of ENSO events up to 19 months ahead, outperforming state-of-the-art models.
• Physical interpretation reveals that the "glass-box" DSDL model functions as a multi-basin synergistic prediction framework.

Contributions

• This study represents a fundamental advance in interpretable climate forecasting by transforming the "black box" into a physically accountable dynamical system.
• The proposed approach has the potential to improve long-term ENSO predictions and provide valuable insights into the underlying physical mechanisms.

Funding

• National Natural Science Foundation of China (42288101)
• Laoshan Laboratory (No. LSKJ202202600)
• Fundamental Research Funds for the Central Universities (202242001, 202515028)

Citation

@article{Wang2026Skillful,
  author = {Wang, Mingyu and Li, Jianping},
  title = {Skillful and interpretable ENSO prediction via a glass-box dynamical-deep learning model},
  journal = {npj Climate and Atmospheric Science},
  year = {2026},
  doi = {10.1038/s41612-026-01513-6},
  url = {https://doi.org/10.1038/s41612-026-01513-6}
}