VMU-Diff: A Coarse-to-fine Multi-source Data Fusion Framework for Precipitation Nowcasting

2026-05-14

AI Summary

This study addresses the significant challenge of precipitation nowcasting, which is inherently difficult due to the chaotic nature of atmospheric systems. Existing approaches typically rely solely on radar data, with deterministic models prone to producing blurry forecasts and probabilistic models suffering from artifacts and computational inefficiency. To overcome these limitations, this work proposes a coarse-to-fine two-stage framework: the first stage fuses radar and multi-band satellite observations using a Vision Mamba U-Net to capture global motion dynamics, while the second stage employs a residual conditional diffusion model to generate high-fidelity fine-scale details. By innovatively integrating multi-source remote sensing data, Vision Mamba’s state-space modeling, and a residual diffusion mechanism, the proposed method achieves superior performance on the Jiangsu SWAN dataset, demonstrating notably higher accuracy and enhanced realism in short-term precipitation forecasts.

Abstract

Precipitation nowcasting is a vital spatio-temporal prediction task for meteorological applications but faces challenges due to the chaotic property of precipitation systems. Existing methods predominantly rely on single-source radar data to build either deterministic or probabilistic models for extrapolation. However, the single deterministic model suffers from blurring due to MSE convergence. The single probabilistic model, typically represented by diffusion models, can generate fine details but suffers from spurious artifacts that compromise accuracy and computational inefficiency. To address these challenges, this paper proposes a novel coarse-to-fine Vision Mamba Unet and residual Diffusion (VMU-Diff) based precipitation nowcasting framework. It realizes precipitation nowcasting through a two-stage process, i.e., a deterministic model-based coarse stage to predict global motion trends and a probabilistic model-based fine stage to generate fine prediction details. In the coarse prediction stage, rather than single-source radar data, both radar and multi-band satellite data are taken as input. A spatial-temporal attention block and several Vision mamba state-space blocks realize multi-source data fusion, and predict the future echo global dynamics. The fine-grained stage is realized by a spatio-temporal refine generator based on residual conditional diffusion models. It first obtains spatio-temporal residual features based on coarse prediction and ground truth, and further reconstructs the residual via conditional Mamba state-space module. Experiments on Jiangsu SWAN datasets demonstrate the improvements of our method over state-of-the-art methods, particularly in short-term forecasts.

Problem

• precipitation nowcasting
• multi-source data fusion
• deterministic model
• probabilistic model
• spatio-temporal prediction

Innovation

• multi-source data fusion
• Vision Mamba
• conditional diffusion model
• precipitation nowcasting
• coarse-to-fine framework