Séance Séminaire

Séminaire de Probabilités et Statistique

"Modeling spatial dependence for extremes in climate studies"

For a wide range of scientific applications in climate research, the observations are scattered in space, either on a regular grid or at irregularly spaced locations. For example, climatological or pollutant data are recorded at different locations and measurements classically exhibit some degree of spatial dependence. While the mean behavior of most spatial processes such as daily temperatures or wind fields is well modeled and understood by the statistical and scientific communities, our understanding of how to measure the spatial dependence for extreme events is still incomplete from a statistical perspective. During this talk, we present two possible directions to explore this problem of spatial dependencies. One approach is based on a Baysiean hiearchical model for extremes and the other one focused on stationary max-stable fields. The advantages and drawbacks of these two approaches are discussed. We illustrate our results by studying climate fields such as precipitation and temperatures.