Extending Precipitation Nowcasting Horizons via Spectral Fusion of Radar Observations and Foundation Model Priors

TL;DR

This paper introduces PW-FouCast, a spectral fusion framework that enhances precipitation nowcasting by integrating weather foundation model priors in the frequency domain, significantly extending forecast horizons.

Contribution

The novel spectral fusion architecture effectively combines radar data and meteorological priors, achieving state-of-the-art results in precipitation nowcasting.

Findings

Achieves state-of-the-art performance on SEVIR and MeteoNet benchmarks.

Effectively extends forecast horizon while maintaining structural fidelity.

Introduces spectral priors to improve long-term precipitation predictions.

Abstract

Precipitation nowcasting is critical for disaster mitigation and aviation safety. However, radar-only models frequently suffer from a lack of large-scale atmospheric context, leading to performance degradation at longer lead times. While integrating meteorological variables predicted by weather foundation models offers a potential remedy, existing architectures fail to reconcile the profound representational heterogeneities between radar imagery and meteorological data. To bridge this gap, we propose PW-FouCast, a novel frequency-domain fusion framework that leverages Pangu-Weather forecasts as spectral priors within a Fourier-based backbone. Our architecture introduces three key innovations: (i) Pangu-Weather-guided Frequency Modulation to align spectral magnitudes and phases with meteorological priors; (ii) Frequency Memory to correct phase discrepancies and preserve temporal evolution; and (iii) Inverted Frequency Attention to reconstruct high-frequency details typically lost in spectral filtering. Extensive experiments on the SEVIR and MeteoNet benchmarks demonstrate that PW-FouCast achieves state-of-the-art performance, effectively extending the reliable forecast horizon while maintaining structural fidelity. Our code is available at https://github.com/Onemissed/PW-FouCast.