RainShift: A Benchmark for Precipitation Downscaling Across Geographies

2025-07-07

AI Summary

To address the poor generalization of global precipitation downscaling models across hemispheres—where geographic distribution shifts severely degrade performance—this paper introduces RainShift, the first benchmark explicitly designed for evaluating cross-geographic distribution shift. RainShift integrates multi-source, high-resolution observational datasets to systematically assess the transferability of deep learning super-resolution methods—including GANs and diffusion models—across distinct geographic domains. Key contributions include: (1) the first empirical demonstration that existing models suffer widespread and severe performance degradation in cross-hemispheric settings; (2) the proposal and validation of spatial generalization enhancement strategies, notably data alignment, which significantly improves prediction accuracy in data-scarce regions; and (3) evidence that expanding training coverage only partially mitigates distribution shift, whereas data alignment delivers substantial gains. This work establishes a new, rigorous benchmark and methodological foundation for fair and robust global high-resolution precipitation modeling.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search

Abstract

Earth System Models (ESM) are our main tool for projecting the impacts of climate change. However, running these models at sufficient resolution for local-scale risk-assessments is not computationally feasible. Deep learning-based super-resolution models offer a promising solution to downscale ESM outputs to higher resolutions by learning from data. Yet, due to regional variations in climatic processes, these models typically require retraining for each geographical area-demanding high-resolution observational data, which is unevenly available across the globe. This highlights the need to assess how well these models generalize across geographic regions. To address this, we introduce RainShift, a dataset and benchmark for evaluating downscaling under geographic distribution shifts. We evaluate state-of-the-art downscaling approaches including GANs and diffusion models in generalizing across data gaps between the Global North and Global South. Our findings reveal substantial performance drops in out-of-distribution regions, depending on model and geographic area. While expanding the training domain generally improves generalization, it is insufficient to overcome shifts between geographically distinct regions. We show that addressing these shifts through, for example, data alignment can improve spatial generalization. Our work advances the global applicability of downscaling methods and represents a step toward reducing inequities in access to high-resolution climate information.

Problem

• Assess generalization of downscaling models across regions
• Address performance drops in out-of-distribution geographic areas
• Reduce inequities in high-resolution climate data access

Innovation

• Deep learning super-resolution for ESM downscaling
• Evaluating geographic generalization with RainShift
• Data alignment improves spatial generalization