Author(s): M. del Jesus; P. Camus; I. J. Losada
Keywords: No keywords
Abstract: In standard hydrologic applications, the return period of any variable (river discharge, fo instance) is directly related to the return period of its driver (in this case precipitation). This assumption greatly simplifies the characterization of derived magnitudes, for which there is normally no measured record and whos behavior must be numerically modeled. However, this simplification may lead to significant errors when th dynamics analyzed are the composition of different factors, i.e. when the magnitude is multivariate. In this work we present a method that makes use of data from Atmosphere-ocean general circulation models (AOGCMs) to analyze large-scale climate variability (long-term historical periods, future climate projections) and to obtain a series of representative synoptic states on which to condition measured values. The proposed method can b considered a hybrid approach, which combines a probabilistic weather type downscaling model with a stochasti weather generator component. Predictand distributions are reproduced modeling the relationship with AOGCM predictors based on a physical division in weather types (Camus et al. 2014). The multivariate dependenc structure of the predictand (extreme events) is introduced linking the independent marginal distributions of th variables by a probabilistic copula regression (Ben Alaya et al. 2014). This hybrid approach is applied for th conditioning of daily precipitation and maximum sea water level to AOGCM data in the isle of Tenerife, Spain Results are then used to obtain the set of hydraulic boundary conditions needed to determine the water leve reached under flood conditions in a storm stream outlet.