PA-Net: Precipitation-Adaptive Mixture-of-Experts for Long-Tail Rainfall Nowcasting

2026-03-14

AI Summary

This work addresses two major challenges in precipitation nowcasting: the high computational cost of modeling spatiotemporal dynamics at million-scale resolution and the scarcity of extreme rainfall events due to their long-tailed distribution. To tackle these issues, we propose PA-Net, a precipitation-adaptive Transformer framework featuring three key innovations: a novel precipitation-adaptive mixture-of-experts (PA-MoE) mechanism that dynamically activates experts based on local rainfall intensity; a decomposed spatiotemporal attention module combining dual-axis compressed latent-variable attention with convolutional dimensionality reduction to lower computational overhead; and an intensity-aware training strategy that enhances modeling capability for heavy rainfall events. Evaluated on the ERA5 dataset, PA-Net consistently outperforms existing methods, achieving particularly significant improvements in forecasting performance for extreme precipitation scenarios such as heavy storms.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Computer Vision: Large Vision Models

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Precipitation nowcasting is vital for flood warning, agricultural management, and emergency response, yet two bottlenecks persist: the prohibitive cost of modeling million-scale spatiotemporal tokens from multi-variate atmospheric fields, and the extreme long-tailed rainfall distribution where heavy-to-torrential events -- those of greatest societal impact -- constitute fewer than 0.1% of all samples. We propose the Precipitation-Adaptive Network (PA-Net), a Transformer framework whose computational budget is explicitly governed by rainfall intensity. Its core component, Precipitation-Adaptive MoE (PA-MoE), dynamically scales the number of activated experts per token according to local precipitation magnitude, channeling richer representational capacity toward the rare yet critical heavy-rainfall tail. A Dual-Axis Compressed Latent Attention mechanism factorizes spatiotemporal attention with convolutional reduction to manage massive context lengths, while an intensity-aware training protocol progressively amplifies learning signals from extreme-rainfall samples. Experiment on ERA5 demonstrate consistent improvements over state-of-the-art baselines, with particularly significant gains in heavy-rain and rainstorm regimes.

Problem

• precipitation nowcasting
• long-tailed distribution
• extreme rainfall
• spatiotemporal modeling
• computational cost

Innovation

• Precipitation-Adaptive MoE
• Long-Tail Rainfall Nowcasting
• Dual-Axis Compressed Latent Attention
• Intensity-Aware Training
• Transformer-based Nowcasting