Wednesday, Aug 6: 2:00 PM - 3:50 PM
0743
Topic-Contributed Paper Session
Music City Center
Room: CC-210
It is undisputed that the Earth is experiencing a period of changing climate; as a consequence, large-scale extreme events (e.g., heat waves, floods, storm surges, hurricanes, or wildfires) are anticipated to become more frequent, severe, and interdependent. An important step towards global resilience to these extreme phenomena is to develop models that can accurately predict the frequency and severity of such widespread extremes. This requires a combined research effort between various scientific fields, namely, extreme-value theory, spatial statistics, scientific machine learning and statistical computing, and civil and environmental engineering. The Weather and Climate Extremes (WACE) workshop series has been bringing together leading researchers from the fields of Statistics, Atmospheric Sciences, and Hydrology/Hydroclimatology since 2016 for advancing research on Climate Extremes, with the latest iteration hosted by Clemson University in May 2023, and the next iteration slated to be at University of Missouri in June 2025. On a similar note, Colorado School of Mines hosted the Extremes2024 workshop in November 2024, bringing together academic and industry researchers to develop new collaborations, delineate mathematical and statistical challenges, and advance the state-of-the-science in the modeling and inference for high-dimensional spatiotemporal, compound extreme events. Both workshops have a strong focus on interdisciplinary research, and their programming is a mix of short courses, panels, talks and poster sessions, providing opportunities for participants at various career stages to get involved. This session will showcase some of the research presented at Extremes2024 and WACE 2025, by researchers who have been/will be participating in the workshops.
Applied
Yes
Main Sponsor
ASA Advisory Committee on Climate Change Policy
Co Sponsors
Section on Statistical Computing
Section on Statistics and the Environment
Presentations
Understanding spatial variation in extreme events is crucial for risk management, early warning systems, and policy-making. Spatial extreme datasets exhibit complex dependencies across geographic regions. Examples include maximum temperatures and wind speeds that increase the risk of wildfires, peak river discharges that lead to floods, and low soil moisture affecting crop yields,. While recent advances in spatial extreme models provide more realistic representations of joint tail dependencies, statistical inference remains computationally demanding, especially for large datasets over hundreds of locations. These challenges stem from costly matrix operations on precision matrices and numerical integration in marginal distributions. In this study, we investigate scalable alternatives to full likelihood inference, leveraging advances in spatial modeling, amortized learning, and density regression techniques. We evaluate these methods through simulation studies and apply them to a high-resolution surface skin temperature dataset from the North American Land Data Assimilation System. Our findings provide insights into efficient, data-driven approaches for modeling spatial extremes, paving the way for more practical applications in climate science, environmental risk assessment, and beyond.
Accurate weather and climate representations rely on effectively combining large-scale numerical climate models with fine-scale observational data. While weather and climate models capture broad-scale dynamics across various spatial and temporal scales, they often face challenges such as modeling biases, high computational costs, and difficulty in resolving local variability and extremes. On the other hand, fine-scale observations provide valuable, high-resolution insights into localized phenomena but are typically sparse and difficult to integrate into large-scale frameworks. This talk presents an innovative approach to address these challenges by utilizing reduced representations from physics-based model outputs and enhance them with observational information. By extracting reduced-order representations from large-scale models and integrating them with fine-scale data, this method refines model outputs, improving their resolution and reliability while allowing to work with dimensionally-reduced model outputs.
Extreme weather and climate change pose a large risk to critical energy systems. Uncertainty quantification of these negative impacts is important for developing resilience, especially during compound extreme weather events involving multiple climate variables. We create a modeling workflow that investigates simultaneous risks of extreme weather to the interdependent electricity and natural gas network systems. Our solution relies on fitting Bayesian Generative Additive models in moving-windows, which is embarrassingly parallel. A Gaussian copula accounts for the multivariate spatio-temporal dependence. Overall, the formulation is interpretable and provides uncertainty quantification from probabilistic simulations of weather variables during extreme events. This framework is invariant to the definition of extreme, and thus can be used in other case studies with similar objectives. We illustrate our methodology using Argonne EVS' high-fidelity climate model output of temperature, wind speed, and solar irradiance to assess the impact of compound hazards on critical energy systems in ISO New England.
This work aims to develop a rigorous and coherent statistical framework to enhance risk assessment methodologies for rare geophysical events. The primary objectives are: (1) to estimate the joint distribution of inputs characterizing rare events; (2) to develop efficient statistical emulators for generating large input-output datasets, enabling forward propagation to approximate output distributions; and (3) to comprehensively quantify uncertainties, addressing multiple sources and their implications for downstream extreme value analysis. The proposed framework will be applied to the study of storm surges and volcanic eruptions, demonstrating its practical utility in assessing the risks associated with geophysical extremes.