Earth-systems data often exhibit highly irregular and complex patterns, for example caused byextreme weather events . Here, we propose a novel GAN-based approach forgenerating spatio-temporal weather patterns conditioned on detected extreme events . Our approach augments GAN generator and discriminator with an encodedextreme weather event segmentation mask . These segmentation masks can becreated from raw input using existing event detection frameworks . As such, ourapproach is highly modular and can be combined with custom GAN architectures . We highlight the applicability of our proposed approach in experiments withreal-world surface radiation and zonal wind data . We hope to use this approach in

Author(s) : Konstantin Klemmer, Sudipan Saha, Matthias Kahl, Tianlin Xu, Xiao Xiang Zhu

Links : PDF - Abstract

Code :
Coursera

Keywords : approach - weather - event - gan - patterns -

Leave a Reply

Your email address will not be published. Required fields are marked *