paper
2020
Hybrid physics-ML & parameterization
Observations & remote sensing New model or method
GAN or VAE Preprint
Code available 2020-02-10
022
Feb 2020
Victor Schmidt, Mustafa Alghali, Kris Sankaran et al.
arXiv
Conditional GAN generates realistic cloud reflectance fields from sea surface temperature and humidity, a step toward stochastic cloud parameterization.
Hybrid physics-ML & parameterization
Modeling Cloud Reflectance Fields using Conditional Generative Adversarial Networks
In one sentence. Conditional GAN generates realistic cloud reflectance fields from sea surface temperature and humidity, a step toward stochastic cloud parameterization.
Abstract
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.
Labels
- Topic
- Hybrid physics-ML & parameterization, Observations & remote sensing
- Type
- New model or method
- Method
- GAN or VAE
- Domain
- Global
- Data
- Satellite, ERA5 / reanalysis
- Status
- Research
- Identifiers
- arXiv:2002.07579 DOI:10.13016/m2yzsj-c41r OpenAlex:W3006293290
- Added
- 2026-10-22
Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.