Zero-Shot Statistical Downscaling via Diffusion Posterior Sampling
This work addresses the challenge of generalizing climate downscaling to Global Climate Models for climate scientists, representing a novel zero-shot approach rather than an incremental improvement.
The paper tackles the problem of zero-shot statistical downscaling for climate models without paired training data, achieving significant improvements in 99th percentile errors and successfully reconstructing complex weather events like tropical cyclones across heterogeneous models.
Conventional supervised climate downscaling struggles to generalize to Global Climate Models (GCMs) due to the lack of paired training data and inherent domain gaps relative to reanalysis. Meanwhile, current zero-shot methods suffer from physical inconsistencies and vanishing gradient issues under large scaling factors. We propose Zero-Shot Statistical Downscaling (ZSSD), a zero-shot framework that performs statistical downscaling without paired data during training. ZSSD leverages a Physics-Consistent Climate Prior learned from reanalysis data, conditioned on geophysical boundaries and temporal information to enforce physical validity. Furthermore, to enable robust inference across varying GCMs, we introduce Unified Coordinate Guidance. This strategy addresses the vanishing gradient problem in vanilla DPS and ensures consistency with large-scale fields. Results show that ZSSD significantly outperforms existing zero-shot baselines in 99th percentile errors and successfully reconstructs complex weather events, such as tropical cyclones, across heterogeneous GCMs.