STLDM: Spatio-Temporal Latent Diffusion Model for Precipitation Nowcasting

TMLR Paper5706 Authors

Abstract: Precipitation nowcasting is a critical spatio-temporal prediction task for society to prevent severe damage owing to extreme weather events. Despite the advances in this field, the underlying complex and stochastic nature of this task still poses challenges to previous approaches. Specifically, deterministic models produce blurry predictions while generative models suffer from poor accuracy. In this paper, we present a simple yet effective model architecture termed STLDM, which learns the latent representation from end to end alongside both the Variational Autoencoder and the conditioning network. Experimental results across multiple radar datasets demonstrate that the proposed STLDM is more effective and superior to the state of the art.

Submission Length: Regular submission (no more than 12 pages of main content)

Assigned Action Editor: ~Andreas_Lehrmann1

Submission Number: 5706