Emulating the Forced Response of Climate Models with Flow Matching

2026-05-16

AI Summary

This study addresses the computational burden of global climate models, which limits their ability to generate large ensembles needed to account for internal variability and scenario uncertainty. The authors propose the first deep learning emulator based on flow matching, trained across multiple Shared Socioeconomic Pathways (SSPs) and incorporating key external forcings—including carbon dioxide, methane, nitrous oxide, sulfate aerosols, and ozone. This approach successfully generalizes to unseen scenarios and represents the first application of flow matching to multi-forcing climate response modeling. The results demonstrate that accurately capturing the synergistic effects of multiple forcings is essential for reproducing long-term temperature trends. The emulator’s simulated land surface temperature responses show high fidelity and strong agreement with those from the established statistical emulator MESMER-M, highlighting its potential for efficient, high-resolution climate projections.

Abstract

Global climate models are essential tools to simulate past and potential future pathways of climate change, as well as associated climate impacts. Shared Socioeconomic Pathways (SSPs) describe a range of future scenarios of global economic and demographic development. These SSPs are intrinsically linked to changes in climate forcings, the external drivers, such as greenhouse gas and aerosol emissions, which in turn lead to the human impact on the energy balance of the Earth over time. These forcings are fundamental boundary conditions in climate models in order to gain insight into the potential climatic impacts of these changes described by each SSP. Running a climate model, however, is extremely computationally expensive, conflicting with the need for large ensembles of simulations for each model to give, e.g., more robust estimates in the presence of internal variability (the inherent, chaotic fluctuations within the climate system) and scenario uncertainty. Recent research has demonstrated the ability to capture climate model dynamics using machine learning when conditioned on forcings from different climatic scenarios. We here train a Deep Learning (DL) model on multiple SSPs and successfully generate scenarios unseen during training. Our emulator is validated against MESMER-M, a statistical emulator of land surface temperature. Our research demonstrates the capacity to generate such changing climate states in response to a variety of simultaneous climate forcings (e.g., carbon dioxide, methane, nitrous oxide, sulphate aerosols, and ozone). In particular, our ablation studies underline a need to include a range of different forcings to represent long-term atmospheric trends with a DL emulator.

Problem

• climate models
• Shared Socioeconomic Pathways
• climate forcings
• computational cost
• internal variability

Innovation

• flow matching
• climate emulator
• deep learning
• climate forcings
• Shared Socioeconomic Pathways