Theme-5-8-Relative performance of regional flood prediction models.

Abstract

Prediction of extreme hydrologic events, such as, peak flood magnitude for design decision in water resources constitutes a serious problem throughout the world. The accuracy attained is limited by the available data at a given site and lack of complete understanding of the underlying mechanism originating the extreme flood events.

Regional flood frequency techniques are used mainly to estimate a specific return period flood flow at a gauging site or to provide site specific estimates based on limited single site data. Regionalisation technique associates annual flood characteristics with physiographic and climatic causative factors. Based on the distributional assumption, degree of spatial heterogeneity and intersite correlation, many regional approaches have been suggested in the recent past. These include estimation of N-dimensional location parameter by James-Stein estimator subject to Lindley modification, use of Wakeby distribution, coupling the Index-Flood method with estimation by probability weighted moments ,(PWM), and regionalising the parameter of Box-Cox transformation.

For a data sparse country like India, regional frequency analysis can be of great value. Hence identification of a suitable technique from amongst the available techniques reported in literature is desirable. In this paper an attempt has been made to evaluate the relative performance of techniques that use (i ) Index-Flood method with PWM estimators, (ii) Wakeby distribution with James-Stein estimators for corrected means, (iii) Log-Boughton distribution, and (iv) Method of power transformation, in estimating flood quantiles.