Improving the Predictability of the Madden-Julian Oscillation at Subseasonal Scales with Gaussian Process Models

2025-05-21

AI Summary

Quantifying uncertainty in Madden–Julian Oscillation (MJO) subseasonal prediction remains a fundamental challenge due to the nonlinear, multivariate, and inherently stochastic nature of MJO dynamics. Method: This study proposes a nonparametric probabilistic forecasting framework based on Gaussian processes (GPs), featuring two key innovations: (i) empirical correlation calibration to align model covariance with observed spatiotemporal dependencies, and (ii) a novel posterior covariance correction mechanism enabling consistent joint modeling of multivariate MJO indices. Contribution/Results: Relative to conventional artificial neural network (ANN) baselines, the method achieves significantly higher forecast skill for the first five days. Its reliable probabilistic forecasts extend the effective predictability horizon to over 22 days—surpassing state-of-the-art methods by more than three weeks. Crucially, uncertainty quantification is both physically interpretable and statistically well-calibrated. To our knowledge, this is the first end-to-end, fully calibrated, GP-driven probabilistic forecasting system for MJO, establishing a new paradigm for subseasonal predictability research.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

The Madden--Julian Oscillation (MJO) is an influential climate phenomenon that plays a vital role in modulating global weather patterns. In spite of the improvement in MJO predictions made by machine learning algorithms, such as neural networks, most of them cannot provide the uncertainty levels in the MJO forecasts directly. To address this problem, we develop a nonparametric strategy based on Gaussian process (GP) models. We calibrate GPs using empirical correlations and we propose a posteriori covariance correction. Numerical experiments demonstrate that our model has better prediction skills than the ANN models for the first five lead days. Additionally, our posteriori covariance correction extends the probabilistic coverage by more than three weeks.

Problem

• Enhancing MJO prediction accuracy at subseasonal scales
• Providing direct uncertainty levels in MJO forecasts
• Extending probabilistic coverage beyond three weeks

Innovation

• Uses Gaussian Process models for MJO prediction
• Calibrates GPs with empirical correlations
• Proposes posteriori covariance correction method