Â鶹´å

April 17, 2026 - HYBRID EVENT

, University of Missouri

Location: Zoom, Posner 151

Title: Flexible and Efficient Spatial Extremes Estimation and Emulation via Extreme-
Aware Variational Autoencoders

Abstract: The world is full of extreme events. For example, a central question in public health planning might be to assess the likelihood of extreme exposures (meteorological conditions, air pollution, social stress, etc.). Similarly, complex phenomena such as atmospheric turbulence and large wildfires can be viewed as extremes. Such extreme events typically occur in spatial and/or temporal clusters. Yet, the standard methodologies that statisticians deal with spatially dependent processes (Gaussian processes and Markov random fields) are not suitable for complex tail dependence structures. More flexible spatial extremes models exhibit appealing extremal dependence properties but are often exceedingly prohibitive to fit and simulate from in high dimensions. Here I present recent work where we develop a new spatial extremes model that has flexible non-stationary extremal dependence properties, and we integrate it in the encoding-decoding structure of a extremes-aware variational autoencoder (XVAE), whose parameters are estimated via variational Bayes combined with deep learning. The XVAE is amortized and can be used to efficiently analyze high-dimensional data or as a spatio-temporal emulator that characterizes the distribution of data or mechanistic model output states and produces outputs that have the same statistical properties as the inputs, especially in the tail. Through extensive simulation studies, we show that our XVAE is substantially more time-efficient than traditional Bayesian inference for such models, while also outperforming many spatial extremes models with a stationary dependence structure. We demonstrate our method applied to a high-resolution satellite-derived dataset of sea surface temperature in the Red Sea and to a high-resolution simulation model of a turbulent plume, such as one would find in a wildfire. We present some current extensions. This is joint work with Likun Zhang and Xiaoyu Ma (University of Missouri), Raphael Huser (KAUST), and Kiran Bhaganagar (University of Texas-San Antonio).

Bio: Christopher K. Wikle is Curators’ Distinguished Professor of Statistics at U. Missouri (MU), with additional appointments in Soil, Environmental and Atmospheric Sciences and the Truman School of Public Affairs. He is currently the Director of the College of Arts and Science Center for Spatio-Temporal Statistics and AI. He obtained his Ph.D. from Iowa State University in 1996 and has been on the faculty at MU for 28 years. His research specialty is spatio-temporal statistics, with primary applications to geophysical processes, complex biological processes, and the environment. He focuses on developing computationally efficient deep hierarchical Bayesian dynamic spatio-temporal models motivated by scientific principles, with more recent work at the interface of deep neural modeling and statistics. He is Fellow of the ASA, IMS, ISI, and AAAS and has published 2 award winning books in spatio-temporal statistics. Dr. Wikle is Associate Editor for several journals and is one of six inaugural members of the Statistics Board of Reviewing Editors for the AAAS flagship journal, Science.