PixelFlowCast: Latent-Free Precipitation Nowcasting via Pixel Mean Flows

2026-05-11

AI Summary

This work addresses the challenge of balancing fidelity and inference efficiency in precipitation nowcasting by proposing a two-stage probabilistic forecasting framework. The approach first generates a coarse global trend using a deterministic model, then employs KANCondNet to extract spatiotemporal conditional features that drive Pixel Mean Flows—a pixel-level generative method requiring no latent-space compression and only a few refinement steps. Integrated with an x-prediction mechanism, the framework enables end-to-end efficient high-fidelity prediction. Evaluated on the SEVIR dataset, the method significantly outperforms existing approaches, achieving superior accuracy and speed—particularly in long-horizon forecasts—and demonstrates strong potential for operational deployment.

Abstract

Precipitation nowcasting aims to forecast short-term radar echo sequences for extreme weather warning, where both prediction fidelity and inference efficiency are critical for real-world deployment. However, diffusion-based models, despite their strong generative capability, suffer from slow inference due to multi-step sampling trajectories, limiting their practical usability. Conditional Flow Matching (CFM) improves efficiency via straightened trajectories, but relies on latent space compression, which inevitably discards high-frequency physical details and degrades fine-grained prediction quality. To address these limitations, we propose PixelFlowCast, a two-stage probabilistic forecasting framework that achieves both high-efficiency and high-fidelity prediction without latent compression. Specifically, in the first stage, a deterministic model first produces coarse forecasts to capture global evolution trends. In the subsequent stage, the proposed KANCondNet extracts deep spatiotemporal evolution features to provide accurate conditional guidance. Based on this, a latent-free, few-step Pixel Mean Flows (PMF) predictor employs an $x$-prediction mechanism to generate high-quality predictions, effectively preserving fine-grained structures while maintaining fast inference. Experiments on the publicly available SEVIR dataset demonstrate that PixelFlowCast outperforms existing mainstream methods in both prediction accuracy and inference efficiency, particularly for long sequence forecasting, highlighting its strong potential for real-world operational deployment.

Problem

• precipitation nowcasting
• inference efficiency
• high-fidelity prediction
• latent compression
• fine-grained structures

Innovation

• Pixel Mean Flows
• Latent-Free
• Conditional Flow Matching
• Precipitation Nowcasting
• KANCondNet