Explore the Ideology of Deep Learning in ENSO Forecasts

2026-01-05

arXiv.org

AI Summary

This study addresses the limited scientific trust and operational adoption of deep learning in ENSO prediction due to its lack of interpretability. The authors propose an interpretability framework grounded in bounded variation functions, which enhances model expressiveness by reactivating “dead” neurons in the saturation regions of activation functions. Through controlled experiments, they elucidate the geographic origins of ENSO predictability signals and the mechanisms underlying the spring predictability barrier. Both theoretical analysis and empirical results demonstrate that ENSO predictability is primarily governed by the tropical Pacific, and that the spring barrier largely stems from the absence of key predictive variables. The proposed method achieves robust forecasting performance while maintaining physical consistency, offering a novel pathway to overcome the spring predictability barrier in ENSO prediction.

Technology Category

• Machine Learning: Deep Learning Theory
• Planning, Routing, and Scheduling: Model-Based Reasoning
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Social Networks and Social Media: Influence propagation, information diffusion, and the prediction on networks

Abstract

The El Ni{~n}o-Southern Oscillation (ENSO) exerts profound influence on global climate variability, yet its prediction remains a grand challenge. Recent advances in deep learning have significantly improved forecasting skill, but the opacity of these models hampers scientific trust and operational deployment. Here, we introduce a mathematically grounded interpretability framework based on bounded variation function. By rescuing the"dead"neurons from the saturation zone of the activation function, we enhance the model's expressive capacity. Our analysis reveals that ENSO predictability emerges dominantly from the tropical Pacific, with contributions from the Indian and Atlantic Oceans, consistent with physical understanding. Controlled experiments affirm the robustness of our method and its alignment with established predictors. Notably, we probe the persistent Spring Predictability Barrier (SPB), finding that despite expanded sensitivity during spring, predictive performance declines-likely due to suboptimal variable selection. These results suggest that incorporating additional ocean-atmosphere variables may help transcend SPB limitations and advance long-range ENSO prediction.

Problem

• ENSO prediction
• deep learning interpretability
• Spring Predictability Barrier
• climate forecasting
• model opacity

Innovation

• interpretable deep learning
• bounded variation
• dead neuron recovery
• ENSO predictability
• Spring Predictability Barrier