Fast, Scale-Adaptive, and Uncertainty-Aware Downscaling of Earth System Model Fields with Generative Machine Learning
This addresses the need for accurate, high-resolution climate impact assessments by providing a faster and more adaptable downscaling tool for researchers and policymakers.
The paper tackles the computational expense and poor generalization of existing downscaling methods for Earth system models by proposing a consistency model that downscales arbitrary simulations in a zero-shot manner without retraining. It outperforms state-of-the-art diffusion models at a fraction of the computational cost and generalizes to unseen climate states.
Accurate and high-resolution Earth system model (ESM) simulations are essential to assess the ecological and socio-economic impacts of anthropogenic climate change, but are computationally too expensive to be run at sufficiently high spatial resolution. Recent machine learning approaches have shown promising results in downscaling ESM simulations, outperforming state-of-the-art statistical approaches. However, existing methods require computationally costly retraining for each ESM and extrapolate poorly to climates unseen during training. We address these shortcomings by learning a consistency model (CM) that efficiently and accurately downscales arbitrary ESM simulations without retraining in a zero-shot manner. Our approach yields probabilistic downscaled fields at a resolution only limited by the observational reference data. We show that the CM outperforms state-of-the-art diffusion models at a fraction of computational cost while maintaining high controllability on the downscaling task. Further, our method generalizes to climate states unseen during training without explicitly formulated physical constraints.