Monthly Diffusion v0.9: A Latent Diffusion Model for the First AI-MIP

TL;DR

Monthly Diffusion v0.9 is a climate emulator utilizing a latent diffusion model with a CVAE architecture to efficiently simulate atmospheric variability at monthly scales with modest computational resources.

Contribution

It introduces a novel latent diffusion approach combined with a CVAE architecture for climate modeling, enabling efficient monthly-scale simulations in data-sparse regimes.

Findings

Successfully models low-frequency atmospheric variability.

Operates efficiently at monthly timesteps with modest computational requirements.

Provides initial promising results in climate emulation.

Abstract

Here, we describe Monthly Diffusion at 1.5-degree grid spacing (MD-1.5 version 0.9), a climate emulator that leverages a spherical Fourier neural operator (SFNO)-inspired Conditional Variational Auto-Encoder (CVAE) architecture to model the evolution of low-frequency internal atmospheric variability using latent diffusion. MDv0.9 was designed to forward-step at monthly mean timesteps in a data-sparse regime, using modest computational requirements. This work describes the motivation behind the architecture design, the MDv0.9 training procedure, and initial results.