Investigating the Robustness of Extreme Precipitation Super-Resolution Across Climates

2025-07-12

AI Summary

Coarse-resolution climate models are inadequate for direct extreme precipitation risk assessment. This paper proposes a novel distribution-parameter super-resolution framework that bypasses the need for large ensemble simulations and instead learns spatial mappings from coarse-grid climate fields and topographic data to the location, scale, and shape parameters of the generalized extreme value (GEV) distribution. The method employs vector generalized linear/additive models incorporating both spatial autocorrelation and inter-variable dependencies. A key innovation is the introduction of the “robustness gap” metric to quantify model generalizability under climate change, alongside the first theoretical derivation of the physical upper bound on super-resolution capability. Applied to summer hourly extreme precipitation over Switzerland, the framework achieves high-accuracy downscaling and demonstrates robustness under pseudo–global-warming scenarios. The theoretical maximum super-resolution factor is ~4×.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Reasoning under Uncertainty: Graphical Models
• Planning, Routing, and Scheduling: Model-Based Reasoning

Application Category

• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

The coarse spatial resolution of gridded climate models, such as general circulation models, limits their direct use in projecting socially relevant variables like extreme precipitation. Most downscaling methods estimate the conditional distributions of extremes by generating large ensembles, complicating the assessment of robustness under distributional shifts, such as those induced by climate change. To better understand and potentially improve robustness, we propose super-resolving the parameters of the target variable's probability distribution directly using analytically tractable mappings. Within a perfect-model framework over Switzerland, we demonstrate that vector generalized linear and additive models can super-resolve the generalized extreme value distribution of summer hourly precipitation extremes from coarse precipitation fields and topography. We introduce the notion of a "robustness gap", defined as the difference in predictive error between present-trained and future-trained models, and use it to diagnose how model structure affects the generalization of each quantile to a pseudo-global warming scenario. By evaluating multiple model configurations, we also identify an upper limit on the super-resolution factor based on the spatial auto- and cross-correlation of precipitation and elevation, beyond which coarse precipitation loses predictive value. Our framework is broadly applicable to variables governed by parametric distributions and offers a model-agnostic diagnostic for understanding when and why empirical downscaling generalizes to climate change and extremes.

Problem

• Improving robustness of extreme precipitation super-resolution across climates
• Assessing model robustness under climate-induced distributional shifts
• Identifying upper limits on super-resolution factors for precipitation

Innovation

• Super-resolving distribution parameters directly
• Using vector generalized linear models
• Introducing robustness gap for diagnostics