Regionalization of catchments for flood frequency analysis for data scarce Rift Valley Lakes Basin, Ethiopia

Abstract

Study regionRift Valley Lakes Basin, Ethiopia Study focusWe performed regionalization of catchments using K-means method based on variety of catchment characteristics and tested hydrological homogeneity of the regions using flood statistics. Following that, flood frequency analysis (FFA) for the identified regions was computed using regional flow data. New hydrological insights for the regionFour hydrologically homogeneous regions were identified. Generalized extreme value (GEV), Lognormal (LN2), Wakeby, and Generalized Pareto (GP) were the best fitted distribution models for regions; one up to four respectively. Maximum likelihood was chosen as the most efficient parameter estimation method for regions two, three, and four, whereas the method of moment was chosen for region one. Region one contained one gauged catchment, therefore regression equation was not developed for this region. The linear regression between mean annual flood (MAF) and catchment characteristics performed well (R2 = 0.827, 0.899 and 0.994) for regions two, three and four respectively. The relative errors between observed and estimated MAF in the pseudo ungauged catchments resulted 0.511, 0.039 and 0.166 for regions two, three and four respectively. Hence, the developed regional frequency curves and regression equations can be used for flood estimation at the required return period (T) in the homogeneous regions of the basin.