Predicting hazards of climate extremes: a statistical perspective

2025-05-23

AI Summary

This study addresses the EU’s climate resilience needs by forecasting tail risks—specifically economic losses and fatalities—from short- to medium-term extreme climate events (floods, storms, heatwaves). Method: We propose a time-varying proportional tail model integrated with empirical Bayesian inference, grounded in extreme value theory and the peaks-over-threshold (POT) approach. The model employs the generalized Pareto distribution (GPD) and its discrete extension to quantify uncertainty in nonstationary extremes and conduct counterfactual worst-case scenario analysis. Contribution/Results: Validated on EU disaster data (1980–2023), the model reveals pronounced nonstationarity in extreme losses. It delivers high-accuracy, interpretable tail-risk warnings—enabling prudent adaptation decisions and evidence-based resilience policy formulation—while representing the first application of such a framework to quantify uncertainty and simulate plausible worst-case outcomes under nonstationary extreme event regimes.

Technology Category

• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty
• Machine Learning: Calibration & Uncertainty Quantification
• Game Theory and Economic Paradigms: Adversarial Learning

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Web Mining and Content Analysis: Models for Web evolution

Abstract

Climate extremes such as floods, storms, and heatwaves have caused severe economic and human losses across Europe in recent decades. To support the European Union's climate resilience efforts, we propose a statistical framework for short-to-medium-term prediction of tail risks related to extreme economic losses and fatalities. Our approach builds on Extreme Value Theory and employs the predictive distribution of future tail events to quantify both estimation and aleatoric uncertainty. Using data on EU-wide losses and fatalities from 1980 to 2023, we model extreme events through Peaks Over Threshold methodology and fit Generalised Pareto (GP) and discrete-GP models using an empirical Bayes procedure. Our predictive approach enables a 'What-if' analysis to evaluate hypothetical scenarios beyond observed levels, including potential worst-case outcomes for a precautionary risk assessment of future extreme episodes. To account for a time-varying behavior of extreme losses and fatalities we extend our predictive method using a proportional tail model that allows to handle heteroscedastic extremes over time. Results of our analysis under stationarity and non-stationary settings raise concerns, reinforcing the urgency of integrating predictive tail risk assessment into EU adaptation strategies.

Problem

• Predicting short-to-medium-term tail risks of climate extremes
• Quantifying estimation and aleatoric uncertainty in extreme events
• Modeling time-varying extreme losses and fatalities for EU adaptation

Innovation

• Extreme Value Theory for tail risk prediction
• Generalised Pareto models with empirical Bayes
• Proportional tail model for heteroscedastic extremes