Generative Modeling for Atmospheric Convection

TL;DR

This paper demonstrates that a Variational Autoencoder can effectively generate and analyze small-scale atmospheric convection patterns, offering a computationally efficient approach to improve climate model representations of storms.

Contribution

The study introduces a VAE-based generative model trained on global data to replicate, cluster, and identify anomalies in convective storm structures, enhancing climate modeling capabilities.

Findings

VAE accurately reconstructs convection structures

Unsupervised clustering reveals convective regimes

Model identifies anomalous storm activity

Abstract

While cloud-resolving models can explicitly simulate the details of small-scale storm formation and morphology, these details are often ignored by climate models for lack of computational resources. Here, we explore the potential of generative modeling to cheaply recreate small-scale storms by designing and implementing a Variational Autoencoder (VAE) that performs structural replication, dimensionality reduction, and clustering of high-resolution vertical velocity fields. Trained on ~6*10^6 samples spanning the globe, the VAE successfully reconstructs the spatial structure of convection, performs unsupervised clustering of convective organization regimes, and identifies anomalous storm activity, confirming the potential of generative modeling to power stochastic parameterizations of convection in climate models.