AI-empowered next-generation multiscale climate modelling for mitigation and adaptation by Eyring et al. 2024

Summary

This research paper from Nature Geoscience proposes a novel approach to climate modelling that leverages artificial intelligence (AI) to enhance existing models. The authors argue that current Earth system models (ESMs) are limited by their coarse resolution and reliance on parameterizations for subgrid-scale processes, leading to systematic errors in climate projections. They propose a multiscale approach, which combines high-resolution climate models with improved coarser-resolution hybrid ESMs, which incorporate essential Earth system processes and feedbacks. This integration, they suggest, can be achieved by implementing machine learning (ML) algorithms to represent these subgrid-scale processes. The paper further calls for modernized infrastructures to support this approach, including increased use of Earth observations and operationalized policy-relevant simulations. This vision aims to create a step change in climate modelling, allowing for more accurate and timely projections to inform mitigation and adaptation strategies in a rapidly changing world.

https://www.nature.com/articles/s41561-024-01527-w