Zero-Shot Statistical Downscaling via Diffusion Posterior Sampling

TL;DR

This paper introduces ZSSD, a zero-shot climate downscaling method that uses a physics-informed prior and coordinate guidance to improve generalization and physical consistency across different climate models.

Contribution

The paper presents ZSSD, a novel zero-shot downscaling framework that incorporates a physics-consistent prior and unified coordinate guidance to address domain gaps and gradient issues.

Findings

Outperforms existing zero-shot methods in 99th percentile error metrics.

Successfully reconstructs complex weather phenomena like tropical cyclones.

Ensures physical validity and robustness across diverse GCMs.

Abstract

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.