ACE: A fast, skillful learned global atmospheric model for climate prediction

TL;DR

ACE is a fast, stable, and energy-efficient machine learning atmospheric emulator that accurately reproduces climate variables and conserves physical laws over a 100-year simulation, outperforming baseline models.

Contribution

We introduce ACE, a novel 200M-parameter ML model that emulates a global atmospheric model with long-term stability and physical consistency for climate prediction.

Findings

ACE nearly conserves column moisture without explicit constraints

ACE outperforms baseline on over 90% of tracked variables

ACE is 100x more energy efficient than the reference model

Abstract

Existing ML-based atmospheric models are not suitable for climate prediction, which requires long-term stability and physical consistency. We present ACE (AI2 Climate Emulator), a 200M-parameter, autoregressive machine learning emulator of an existing comprehensive 100-km resolution global atmospheric model. The formulation of ACE allows evaluation of physical laws such as the conservation of mass and moisture. The emulator is stable for 100 years, nearly conserves column moisture without explicit constraints and faithfully reproduces the reference model's climate, outperforming a challenging baseline on over 90% of tracked variables. ACE requires nearly 100x less wall clock time and is 100x more energy efficient than the reference model using typically available resources. Without fine-tuning, ACE can stably generalize to a previously unseen historical sea surface temperature dataset.