83rd Annual

Monday, 10 February 2003: 4:45 PM
Hydrologic assessment: application of extreme value theory for climate extreme scenarios construction
Jeanna Goldstein, EC, Saint-Laurent, QC, Canada; and M. Mirza, D. Etkin, and J. Milton
Poster PDF (138.7 kB)
Human and socio-economic systems are vulnerable to climate extremes in Canada. Recent analyses demonstrate that hydrological disasters are on the rise in Canada and that densely populated provinces are highly exposed. Substantial research works on the extremes has been carried out, analyses of extremes in terms of future climate change is currently receiving growing attention at national and international levels. The Intergovernmental Panel on Climate Change (IPCC) of the United Nations concluded about possibility of increased frequency and intensity of extremes. It is therefore necessary to construct scenarios of climate extremes and supply them to the hydrology impact researchers/modellers in order to facilitate vulnerability, impact and adaptation assessments. However, currently hydrology impact studies in Canada experience a great necessity for climate extreme scenarios. Extreme Value Theory (EVT) technique is proposed to simulate current and future climate extreme scenarios to be used for hydrologic impact models. Different forms of the EVT, evidence for EVT forms applications and parameter estimation procedures are described. The EVT is a branch of applied statistics which specifically deals with the facts of frequency increase of extreme events. The EVT offers to converge a series of maxima (minima) series of the meteorological data to the Generalized Extreme Value (GEV) distribution, to model the behaviour of the excess over a given threshold basing on the generalized Pareto distribution (GP) and to produce probabilistic statistics of climate extremes. The L-moment and Maximum Likelihood methods are considered to estimate the parameters of the GEV distribution. The goodness-of-fit (GOF) test is presented to verify the GEV distribution fitness to the samples.

Supplementary URL: