Model-based inference of conditional extreme value distributions with hydrological applications
share
Multivariate extreme value models are used to estimate joint risk in a number of applications, with a particular focus on environmental fields ranging from climatology, hydrology to oceanography. The semi-parametric conditional extreme value model of Heffernan and Tawn (2004) provides the most suitable of current statistical models in terms of its flexibility to handle a range of extremal dependence classes and its ability to deal with high dimensionality. The standard formulation of the model is highly inefficient as it does not allow for any partially missing observations, that typically arise with environmental data, to be included and it suffers from the curse of dimensionality as it involves a d-1-dimensional non-parametric density estimator in d-dimensional problems. A solution to the former was proposed by Keef et al. (2009) but it is extremely computationally intensive, making its use prohibitive if the proportion of missing data is non-trivial. We propose to replace the d-1-dimensional non-parametric density estimator with a model-based copula with univariate marginal densities estimated using kernel methods. This approach provides statistically and computationally efficient estimates whatever the degree of missing data or the dimension, d, and improvements in other inferential aspects. The methods are illustrated through the analysis of UK river flow data at a network of 46 sites.
READ FULL TEXT