Monthly Diffusion v0.9: A Latent Diffusion Model for the First AI-MIP
career value
198K/year
🤖 AI Summary
This study addresses the challenge of simulating low-frequency internal atmospheric variability at monthly timescales under data-sparse conditions. The authors propose a latent diffusion model that integrates Spherical Fourier Neural Operators (SFNOs) with a Conditional Variational Autoencoder (CVAE). This approach introduces SFNOs into a latent diffusion framework for the first time, enabling efficient forward simulation of global climate states at monthly timesteps while balancing modeling accuracy and computational efficiency on 1.5° grid-resolution data. Experimental results demonstrate that the model reliably captures key characteristics of low-frequency atmospheric variability even under low-data constraints, confirming its feasibility and novelty for monthly-scale climate simulation.
Application Category
📝 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.
Problem
Research questions and friction points this paper is trying to address.
climate emulation
low-frequency atmospheric variability
monthly time-step
data-sparse regime
latent diffusion
Innovation
Methods, ideas, or system contributions that make the work stand out.
latent diffusion
spherical Fourier neural operator
climate emulator
conditional variational auto-encoder
low-frequency atmospheric variability
🔎 Similar Papers
No similar papers found.
💼 Related Jobs
