Dear All,
We have the pleasure, thanks to the support of the ESSEC IDS dpt, Institut des Actuaires, LabEx MME-DII, and **the group BFA (SFdS)**, to invite:
**Prof. Holger ROOTZÉN**
Chalmers University of Technology, Gothenburg, Sweden
**Date and place**: **Monday, October 21, at 12:30 pm, EEE - ESSEC La Défense room 202**
**“ Is climate change making extreme rains
** more frequent, or bigger, or more dangerous? ”
*Records can become bigger because the underlying distribution changes, or because one makes more tries, or both. Dangers of flooding is determined by the interplay between changes in record sizes and the spatial extents of extreme rainfall events. We use NOAA data with more than 60 years of observation and with observations going up to at least 2010 to study large rainstorms in northeastern USA. Data on daily rain at individual gauge stations provides the most direct path to understanding the how climate change affects rainstorms. However, annual maxima data are more widely available, and often of higher quality. We hence use both kinds of data, together with the close relation between the PoT method with generalized Pareto distributed excesses and the annual maxima method with the generalized extreme value distribution. A preliminary answer is that record rainstorms in northeastern USA are becoming larger because extreme rainfalls are becoming more frequent, but not because the distribution of sizes of individual rainstorm is increasing. To model changes in spatial extent of rainfall events we derive new representations of the GP distributions associated with the stable mixture GEV distributions. These models inherit the appealing properties of the stable mixture GEV models which provide components of variance models, time series models, and spatial and continuous parameter models for annual maxima. Additionally, they allow for simple explicit formulas for densities, and hence for easier ML estimation than competing models. An important challenge is that extremes, say rainstorms, may hit a some, but not all, spatial locations, so that some, but not all, components are extreme. For GP modelling this means that models must be able to accommodate cases where some of the components have mass on their lower boundary, which may be finite or equal to minus infinity.*
Kind regards,
Jeremy Heng, Olga Klopp and Marie Kratz
**http://crear.essec.edu/working-group-on-risk** |