Hierarchical Graph Networks for Accurate Weather Forecasting via Lightweight Training

TL;DR

This paper introduces hierarchical graph neural networks with physics embedding for weather forecasting, achieving improved accuracy and efficiency in modeling multiscale climate dynamics, especially for rare events.

Contribution

It proposes novel HGNN architectures with memory and physics integration mechanisms, enabling accurate, fast, and sustainable weather predictions across multiple scales.

Findings

Flow models reduce forecast errors by over 5% at 13-day lead times.

Models improve reliability for extreme weather events by 5-8%.

Pretrained models converge within a single epoch, lowering training costs.

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.