Precipitation Downscaling with Spatiotemporal Video Diffusion

arXiv:2312.06071v2 Announce Type: replace-cross
Abstract: In climate science and meteorology, high-resolution local precipitation (rain and snowfall) predictions are limited by the computational costs of simulation-based methods. Statistical downscaling, or super-resolution, is a common workaround where a low-resolution prediction is improved using statistical approaches. Unlike traditional computer vision tasks, weather and climate applications require capturing the accurate conditional distribution of high-resolution given low-resolution patterns to assure reliable ensemble averages and unbiased estimates of extreme events, such as heavy rain. This work …

abstract applications arxiv climate climate science computational computer computer vision costs cs.cv cs.lg diffusion low meteorology physics.ao-ph precipitation prediction predictions rain science simulation statistical stat.ml tasks type video video diffusion vision weather workaround