REGIONAL FLOOD FREQUENCY ANALYSIS OF NEPAL RIVERS

Abstract

Water resources utilization is abruptly rising with the development of civilization and rapidly growing inhabitants. For design and planning of the water related structures such as bridges, weirs, dams, barrages, spillways, canals, embankments, culverts etc, flood quantiles of different return period are required. Such extreme flood quantiles are popularly known as design flood. Due to highly varied spatial and temporal rainfall and runoff, estimation of such event is of great challenge for engineers. Deterministic and probabilistic approaches are used for determination of designed peak flood. The deterministic approach is always first choice for the estimation of designed flood peaks, but it requires large record of past flood events. Probabilistic .approach which is based on the frequency analysis is become very popular where past flood records are of very limited length. This study aims to form hydrological homogeneous regions of Nepalese territory from the statistical point of view. Observed values of annual maximum flow data (AFS) are collected for various stations. Regionalization is based geographical and physical characteristics, hydrological and meteorological parameters. L-moment based regional hydrologic homogeneity test and clustering are used for finalization of hydrological-homogeneous regions. Clustering has been done using K-mean clustering, Fuzzy C-mean, and Hierarchical clustering method. Finally the number of optimum clusters has been decided by using Kohonen Self organizing map. The regional growth parameters for all the .clusters have been drawn. The flood quantiles for different return periods have been calculated for all the gauged and ungauged sites of the basin. Estimated return period flood of the derived regional flood—frequency relationships have been compared with that of at-site and Water and Energy Commission Secretariat of Nepal (WECS) methods. The regional relationships were found to have far better estimation capabilities than the WECS method. This study is expected to be helpful for deriving flood quantiles for gauged and ungauged sites of Nepal.