RN-14 : Regional flood frequency analysis

Abstract

The main objective of hydrological studies of floods is to develop appropriate procedures in order to arrive at desired design variable for the particular structure to be safe under extremes of floods without leading to overdesign and consequent increase in cost. The flood frequency analysis procedures are based on general statistical and probability concepts and the data used in such analysis has to satisfy the criteria of homogeneity, independence, randomness and time invariance. When only short records are available, regional analysis approach can provide estimates of floods for limited data and ungauged locations, through appropriate regional multiple regression relationships of statistical parameters of the flood series and catchment characteristics. The choice of catchment morphometric characteristics depends on the (i) judgement of the likely predictive success of variables, and (ii) numerical calculation of regression equations and the interpretation of their coefficients. However, these characteristics are often dodged by correlations, resulting in interdependence amongst independent parameters.

The most significant development in this area was bringing out of a manual by U.S. Geological Survey in 1960, which was followed by number of studies including U.K. Flood studies and other typical studies covering general probability considerations, use of historical information, criteria for regional homogeneity etc. Some of the important typical studies in India and abroad during 1960-1987 have been reviewed. Significant developments include (i) use of GEV, TCEV and Wakeby distributions, (ii) use of probability weighted moments and Bayesion methodology, (iii) consideration of effect of serial correlation and cross correlation in flood series at different sites, (iv) use of hirarchical approach and pale hydrological information, (v) use of geomorphologic parameters, rainfall area correlation structure alongwith flood characteristics to delimit hydrologically homogeneous regions, and (vi) development of criterion for selection of physiographic and meteoro1:gie factors.

The application of probability weighted moments technique parameter estimation appears to work well for situations where records are extremely short and streamflow histories are highly skewed and highly kurtotic. James and Stein estimators give a lower risk with respect to mean squared error than that for the minimum variance, unbiased maximum likelihood estimator. These new developments provide the necessary techniques for combining flood information from gauged sites within a region and using it to infer flood frequencies at ungauged sites.

Most of the regional flood studies in India are based on USGS approach with RDSO data. There is need for taking up systematic regional flood studies using large data base collected by different organisations and agencies. These would enable pooling together of vast information for use in hydrologic analysis and design.