- Prof. Ben Shaby – Statistics Department – Penn State University – USA
- Fundação FCUL (Anfiteatro) – Bloco C1 Piso 3 – FCUL – Campo Grande – 14:30h
- Quarta-feira, 7 de Janeiro de 2015
- Referência Projeto: EXPL/MAT-STA/0622/2013
Abstract: Wildfires have the potential to inflict huge losses of life, infrastructure, and habitat. I will describe two projects related to wildfire risk, both of which employ spatial max-stable models. In the first, we develop a spatially-heterogeneous wire weather warning scheme for Australia based on composite fire index scores. The idea is that fire indices commonly used in wildfire research seem to be geographically variable in how high values need to be to be realistically considered extreme. In the second project, we look to the future, attempting to extract meaningful information about extreme fire weather in California from high-resolution weather model output. Here, the problem is that the weather model represents the tails of the fire weather distribution poorly, so instead of treating output as weather variables, we treat output as covariates and build a spatial extreme value regression model fit to observed data.