Causal Climate Emulation with Bayesian Filtering

TL;DR

This paper introduces an interpretable climate model emulator that uses causal representation learning and Bayesian filtering to efficiently simulate climate dynamics while incorporating physical causal relationships.

Contribution

It presents a novel causal emulator framework with a Bayesian filter for stable long-term climate prediction, advancing interpretability and accuracy over existing methods.

Findings

Learns accurate climate dynamics

Demonstrates component importance on synthetic data

Validates on real climate models

Abstract

Traditional models of climate change use complex systems of coupled equations to simulate physical processes across the Earth system. These simulations are highly computationally expensive, limiting our predictions of climate change and analyses of its causes and effects. Machine learning has the potential to quickly emulate data from climate models, but current approaches are not able to incorporate physically-based causal relationships. Here, we develop an interpretable climate model emulator based on causal representation learning. We derive a novel approach including a Bayesian filter for stable long-term autoregressive emulation. We demonstrate that our emulator learns accurate climate dynamics, and we show the importance of each one of its components on a realistic synthetic dataset and data from two widely deployed climate models.