skip to content
 

Generalized additive modelling of hydrological sample extremes

Presented by: 
V Chavez-Demoulin ETH Zürich
Date: 
Thursday 31st October 2013 - 14:50 to 15:25
Venue: 
INI Seminar Room 1
Abstract: 
Co-authors: Anthony Davison (EPFL, Lausanne), Marius Hofert (ETHZ, Zurich)

Estimation of flood frequencies and severities is important for many water management issues. We present a smoothing extreme value method fitted by penalized loglikelihood. Spline smoothing is used to estimate the parameters of the frequency and size distributions of extremes, depending on covariates in a non- or semiparametric way. The frequency process of high level extremes is modelled by a Poisson process, either homogeneous or non-homogeneous. The extreme sizes are considered to follow a generalized Pareto distribution. Being given by two parameters, the method of spline smoothing is not straightforward to apply. An efficient fitting algorithm based on orthogonal reparametrisation is developed to achieve this task. The method is applied to the daily maximum flows of an hydrological station in Switzerland and is used to estimate 20-year return levels.
The video for this talk should appear here if JavaScript is enabled.
If it doesn't, something may have gone wrong with our embedded player.
We'll get it fixed as soon as possible.
Presentation Material: 
University of Cambridge Research Councils UK
    Clay Mathematics Institute The Leverhulme Trust London Mathematical Society Microsoft Research NM Rothschild and Sons