Hierarchical Graph Networks for Accurate Weather Forecasting via Lightweight Training

2025-10-24

AI Summary

Weather forecasting faces challenges due to the complex coupling of physical processes across spatiotemporal scales; fixed-resolution models struggle to capture multiscale dynamics, while existing hierarchical graph neural networks (HGNNs) often lose global trends during downsampling, compromising physical consistency. To address this, we propose HiAntFlow—a multiscale graph neural network framework. Its key innovations are: (1) a Latent-Memory-Retention mechanism that preserves continuity of global climate dynamical features during downward information propagation; and (2) a Latent-to-Physics branch that explicitly incorporates PDE-derived physical fields for cross-scale physical embedding. Integrated with pretraining initialization and a lightweight ensemble training strategy, HiAntFlow achieves >5% average error reduction in 13-day forecasts and 5–8% improvement in extreme-event prediction (1st/99th percentile). It converges within a single training run, significantly reducing computational cost and carbon footprint.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search

Abstract

Climate events arise from intricate, multivariate dynamics governed by global-scale drivers, profoundly impacting food, energy, and infrastructure. Yet, accurate weather prediction remains elusive due to physical processes unfolding across diverse spatio-temporal scales, which fixed-resolution methods cannot capture. Hierarchical Graph Neural Networks (HGNNs) offer a multiscale representation, but nonlinear downward mappings often erase global trends, weakening the integration of physics into forecasts. We introduce HiFlowCast and its ensemble variant HiAntFlow, HGNNs that embed physics within a multiscale prediction framework. Two innovations underpin their design: a Latent-Memory-Retention mechanism that preserves global trends during downward traversal, and a Latent-to-Physics branch that integrates PDE solution fields across diverse scales. Our Flow models cut errors by over 5% at 13-day lead times and by 5-8% under 1st and 99th quantile extremes, improving reliability for rare events. Leveraging pretrained model weights, they converge within a single epoch, reducing training cost and their carbon footprint. Such efficiency is vital as the growing scale of machine learning challenges sustainability and limits research accessibility. Code and model weights are in the supplementary materials.

Problem

• Capturing multiscale weather dynamics missed by fixed-resolution methods
• Preserving global climate trends during hierarchical model downscaling
• Reducing computational costs while maintaining extreme event prediction accuracy

Innovation

• Hierarchical Graph Networks embed physics in multiscale framework
• Latent-Memory-Retention preserves global trends during downward traversal
• Latent-to-Physics branch integrates PDE solutions across diverse scales