Author(s): Andrea Ramsay; Ousmane Seidou; Ioan Nistor
Linked Author(s): Ioan Nistor
Keywords: No Keywords
Abstract: Future distributions of extreme hydrologic events represent a major concern in the context of climate change and its potential impacts. The accurate forecasting of their probability distribution is crucial for the safe and economic design of hydraulic structures such as dam spillways. So far, two approaches have been used to infer the distributions of extreme events in a changing climate scenario: (1) the simulation of daily flows relying on the downscaled daily time-scale output of General Circulation Models (GCM), and (2) the use of a non-stationary probability distribution, which assumes that one parameter (usually the location parameter) is a linear or quadratic function of time. The method proposed in this paper attempts to bridge the gap between these two methods. It uses a non-stationary probability distribution which links the variation of the parameters to the monthly outputs of the climate models. It takes advantages of the robustness of monthly GCM outputs and the flexibility of non-stationary frequency analysis. The proposed method is applied in three steps: (1) calibration of a probability distribution of maximum seasonal flow, conditional on the seasonal mean flow, (2) building of a transfer model capable of forecasting the seasonal mean flow, using monthly climate model outputs, and (3) the use of predicted monthly climate variables to forecast future mean flows, the probability density function of the maximum seasonal flow, and finally, the risk associated with extreme flow events. This paper focuses on the first step by examining the calibration of the non-stationary Generalized Extreme Value (GEV) flood frequency distribution. Bayes Factors and Monte Carlo Markov Chains (MCMC) are used to demonstrate that the proposed model is superior to stationary flood distribution models as well as non-stationary distribution models assuming a linear variation of the location parameter and/or the scale parameter of the GEV distribution.
Year: 2009