Deep learning the sources of MJO predictability: a spectral view of learned features

2025-10-03

AI Summary

This study addresses the scale origins of Madden–Julian Oscillation (MJO) predictability. We propose a scale-decomposition framework integrating deep convolutional neural networks (DCNNs) with spectral analysis to quantitatively identify the spatial scales driving MJO predictability. Results demonstrate that large-scale circulation modes constitute the dominant source of MJO predictability, while small-scale signals—though intrinsically unpredictable—exhibit envelope structures that, when reconstructed, significantly enhance large-scale forecast skill. This reconciles single-scale and multi-scale MJO theories. Our model achieves prediction skill (anomaly correlation ≥ 0.5) of 21 days for the Real-time Multivariate MJO (RMM) index and 33 days for the Regional MJO (ROMI) index, matching state-of-the-art subseasonal-to-seasonal models such as NCEP’s. These findings validate the primacy of large-scale dynamics in MJO predictability and establish a new paradigm for interpretable AI-based meteorological modeling.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Planning, Routing, and Scheduling: Model-Based Reasoning
• Cognitive Modeling & Cognitive Systems: Computational Creativity

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

The Madden-Julian oscillation (MJO) is a planetary-scale, intraseasonal tropical rainfall phenomenon crucial for global weather and climate; however, its dynamics and predictability remain poorly understood. Here, we leverage deep learning (DL) to investigate the sources of MJO predictability, motivated by a central difference in MJO theories: which spatial scales are essential for driving the MJO? We first develop a deep convolutional neural network (DCNN) to forecast the MJO indices (RMM and ROMI). Our model predicts RMM and ROMI up to 21 and 33 days, respectively, achieving skills comparable to leading subseasonal-to-seasonal models such as NCEP. To identify the spatial scales most relevant for MJO forecasting, we conduct spectral analysis of the latent feature space and find that large-scale patterns dominate the learned signals. Additional experiments show that models using only large-scale signals as the input have the same skills as those using all the scales, supporting the large-scale view of the MJO. Meanwhile, we find that small-scale signals remain informative: surprisingly, models using only small-scale input can still produce skillful forecasts up to 1-2 weeks ahead. We show that this is achieved by reconstructing the large-scale envelope of the small-scale activities, which aligns with the multi-scale view of the MJO. Altogether, our findings support that large-scale patterns--whether directly included or reconstructed--may be the primary source of MJO predictability.

Problem

• Investigating MJO predictability sources using deep learning methods
• Identifying dominant spatial scales for Madden-Julian oscillation forecasting
• Determining how large-scale patterns drive MJO predictability mechanisms

Innovation

• Deep convolutional neural network forecasts MJO indices
• Spectral analysis identifies large-scale dominant patterns
• Small-scale signals reconstruct large-scale envelope for prediction