Learning Hierarchical Causal Representations of the Effects of Forcings on Temperature in Climate Models

2026-09-25

AI Summary

This study addresses the limited trustworthiness and lack of causal attribution in machine learning climate simulators arising from their black-box nature. To this end, we introduce hierarchical causal representation learning into sea surface temperature field modeling for the first time. The proposed method explicitly decouples internal variability from external forcing responses, disentangling atmospheric dynamical interactions from the forcing effects of greenhouse gases and aerosols. Trained on Earth system model data, the framework accurately predicts both global mean and regional temperature evolution under unseen future climate scenarios, demonstrating physically consistent perturbation responses. By bridging causal inference with climate dynamics, this approach significantly enhances the interpretability and predictive accuracy of machine learning-based climate simulations.

Abstract

Machine learning (ML) emulators provide a fast and cost-effective method to simulate climate change scenarios after being trained on Earth System Models projections. However, the black-box nature of those data-driven approaches limit the usability and trustworthiness of their outputs and in particular their use as causal attribution tools. Here, we develop a hierarchical causal representation learning framework applied to sea surface temperature fields from a state-of-the-art global climate model. As a key advance over previous work, our framework explicitly models both atmospheric dynamical interactions arising from internal climate variability and forced responses due to changes in atmospheric greenhouse gas and aerosol concentrations. When trained on future climate change scenarios, our method accurately predicts the long-term global mean and regional temperature evolution and shows physically realistic responses to perturbations in greenhouse gas and aerosol concentrations when evaluated on unseen scenarios. Our results underline the potential of causal representation learning frameworks for advancing climate model emulation.

Problem

• climate model emulation
• causal representation learning
• black-box machine learning
• causal attribution
• internal climate variability

Innovation

• Causal Representation Learning
• Climate Model Emulation
• Hierarchical Framework
• Forced Response
• Internal Climate Variability