Stochastic Super-Resolution for Downscaling Time-Evolving Atmospheric Fields With a Generative Adversarial Network

This datasets supports the paper "Stochastic Super-Resolution for Downscaling Time-Evolving Atmospheric Fields with a Generative Adversarial Network" submitted to IEEE Transactions in Geoscience and Remote Sensing. A preprint of the paper can be found here: https://arxiv.org/abs/2005.10374. The code that uses these data is available at https://github.com/jleinonen/downscaling-rnn-gan. The file "goes-samples-2019-128x128.nc" contains the training dataset called "GOES-COT" in the paper, consisting of cloud optical depth measurements from the GOES-16 satellite. The files "gen_weights*.nc" contain the generator weights saved at different time steps during training for the two different datasets described in the paper.