Lifting of time- and space-evolving winds for the determination of extreme hydraulic conditions

摘要

A semi-parametric method based on the theory of max-stable processes is proposed that can be used to determine the time- and space-evolving wind fields associated with a given return value of wind speed at a specified reference location. The method uses wind velocity time series over a grid of locations and their location-specific extreme value distributions in order to 'lift' observed extreme events (storms) into yet unobserved and even more extreme events. Because the lifted fields are needed to force hydrodynamic models used in the determination of extreme hydraulic conditions, they need to be continuous in time and space and to extend over a few days before the peak of the storm in order to allow the spin up of the hydrodynamic models. Thus, even if the original values of the wind fields are not above the threshold used to determine the location-specific extreme value distributions at a given location and instant (outside the validity of the method) they still need to be lifted. This can be achieved by augmenting the location-specific extreme value distributions by the empirical distributions of the observations and lifting using the same method also the values below the thresholds. Mainly as a consequence of this procedure, the resulting lifted wind fields show low temporal and spatial wind speed gradients away from the peak of the storm. These gradients have raised concerns about the validity of the fields and it has been suggested that the lifting by means of a simple scaling factor may be preferable. To address these concerns, a thorough validation of the lifted fields has been carried out by 1) assessing the characteristics of the lifted fields and 2) assessing the ability of the lifted fields to reproduce extreme hydraulic conditions. The main recommendation of this study is that the proposed method be considered further for the determination of temporally and spatially evolving hydraulic conditions. (C) 2016 Elsevier B.V. All rights reserved.