We introduce a conditional Generative Adversarial Network (cGAN) approach to generate cloud reflectance fields (CRFs) conditioned on large scale meteorological variables such as sea surface temperature and relative humidity. We show that our trained model can generate realistic CRFs from the corresponding meteorological observations, which represents a step towards a data-driven framework for stochastic cloud parameterization.
@article{arxiv.2002.07579,
title = {Modeling Cloud Reflectance Fields using Conditional Generative Adversarial Networks},
author = {Victor Schmidt and Mustafa Alghali and Kris Sankaran and Tianle Yuan and Yoshua Bengio},
journal= {arXiv preprint arXiv:2002.07579},
year = {2020}
}
Comments
Code is available on Github: https://github.com/krisrs1128/clouds_dist