Extreme Precipitation Nowcasting Using Multitask Latent Diffusion Models

Chaorong Li, Xudong Ling, Qiang Yang, Mingxiang Chen, Fengqing Qin, Yuanyuan Huang

Abstract

Deep learning models have achieved remarkable progress in precipitation prediction. However, they still face significant challenges in accurately capturing spatial details of radar images, particularly in regions of high precipitation intensity. This limitation results in reduced spatial localization accuracy when predicting radar echo images across varying precipitation intensities. To address this challenge, we propose an innovative precipitation prediction approach termed the multitask latent diffusion model (MTLDM). The core idea of MTLDM lies in the recognition that precipitation radar images represent a combination of multiple components, each corresponding to different precipitation intensities. Thus, we adopt a divide-and-conquer strategy, decomposing radar images into several sub-images based on their precipitation intensities and individually modeling these components. During the prediction stage, MTLDM integrates these sub-image representations by utilizing a trained latent-space rainfall diffusion model (DM), followed by decoding through a multitask decoder to produce the final precipitation prediction. Experimental evaluations conducted on the MRMS dataset demonstrate that the proposed MTLDM method surpasses state-of-the-art techniques, achieving a critical success index (CSI) improvement of 13%–26%.
BibTeX
@article{li2025extreme,
  author = {Chaorong Li and Xudong Ling and Qiang Yang and Mingxiang Chen and Fengqing Qin and Yuanyuan Huang},
  title = {{Extreme Precipitation Nowcasting Using Multitask Latent Diffusion Models}},
  journal = {IEEE Trans. Geosci. Remote. Sens.},
  year = {2025},
  url = {https://openreview.net/forum?id=dxhG5oi5uL},
}

Loading related content...

Figure