Assessing the Geographic Generalization and Physical Consistency of Generative Models for Climate Downscaling

2025-10-15

AI Summary

Climate downscaling generative models suffer from poor geographical generalization and lack standardized metrics for evaluating physical consistency. Method: We propose a physics-informed evaluation and modeling framework featuring (i) a power spectral density–based frequency-domain loss function to enhance reconstruction of small-scale atmospheric structures, and (ii) a physical consistency diagnostic system centered on conservation laws—specifically divergence and vorticity. Contribution/Results: We systematically reveal, for the first time, that state-of-the-art models exhibit substantial degradation in generalization performance outside Europe and severe inaccuracies in predicting key dynamical variables. Our spectral loss reduces small-scale structural reconstruction error by 23–37%, significantly improving cross-regional generalization and physical interpretability. This work establishes a new paradigm for trustworthy, AI-driven high-resolution climate simulation.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications

Abstract

Kilometer-scale weather data is crucial for real-world applications but remains computationally intensive to produce using traditional weather simulations. An emerging solution is to use deep learning models, which offer a faster alternative for climate downscaling. However, their reliability is still in question, as they are often evaluated using standard machine learning metrics rather than insights from atmospheric and weather physics. This paper benchmarks recent state-of-the-art deep learning models and introduces physics-inspired diagnostics to evaluate their performance and reliability, with a particular focus on geographic generalization and physical consistency. Our experiments show that, despite the seemingly strong performance of models such as CorrDiff, when trained on a limited set of European geographies (e.g., central Europe), they struggle to generalize to other regions such as Iberia, Morocco in the south, or Scandinavia in the north. They also fail to accurately capture second-order variables such as divergence and vorticity derived from predicted velocity fields. These deficiencies appear even in in-distribution geographies, indicating challenges in producing physically consistent predictions. We propose a simple initial solution: introducing a power spectral density loss function that empirically improves geographic generalization by encouraging the reconstruction of small-scale physical structures. The code for reproducing the experimental results can be found at https://github.com/CarloSaccardi/PSD-Downscaling

Problem

• Evaluating deep learning models' geographic generalization in climate downscaling
• Assessing physical consistency of downscaling models using physics-inspired diagnostics
• Addressing failures in capturing second-order atmospheric variables accurately

Innovation

• Benchmarks deep learning models for climate downscaling
• Introduces physics-inspired diagnostics for model evaluation
• Proposes spectral density loss to improve generalization