Bayesian Hierarchical Modeling of Extreme Low Temperatures in Northern Finland

• Activity Number: 32
• Type: Contributed
• Date/Time: Sunday, August 9, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Emeric Thibaud* and Anthony C. Davison
• Companies: Colorado State University and Ecole Polytechnique Fédérale de Lausanne
• Keywords: Extreme event ; Temperature modeling ; Max-stable process ; Bayesian hierarchical model

Abstract

In Northern Finland, extreme low temperatures below -36ºC prevented the spread of the moth Epirrita Autumnata which can be responsible of large defoliations in birch forests. Global change may affect this equilibrium and it is essential to model extreme low temperatures over these regions to assess the risks of outbreaks. Brown-Resnick models have proven to be well-suited for modeling extremes of complex environmental processes, but their full density function cannot be calculated in general, preventing the widespread use of these models in Bayesian inference. In this talk we consider one particular case for which the full likelihoods of Brown-Resnick processes can be calculated, exploiting the occurrence times of componentwise maxima. We propose the construction of a Bayesian hierarchical model for extreme low temperatures in Northern Finland. The resulting approach allows both the complex modeling of marginal distributions of extremes and an appropriate treatment of extremal dependence. Our results show that minimum temperatures may not be as extreme in the future as in current times, with greater probabilities for large outbreaks of Epirrita Autumnata.