Robustness analysis of regional flood frequency models: a case study

Abstract

Robust models which allow flood estimation for basins with short records have been investigated. For this purpose a river basin in Turkey with an area of 36000 km2 is chosen, which has 11 gaging stations with records 18–47 years long. It is shown that the basin has two regions which arc homogeneous within themselves but significantly different from each other. Various probability distribution functions were fitted to observed data in each region. In order to test robustness of different models, flood series arc generated at each station for each of the assumed flood populations. Regional curves are obtained using the generated data and compared with the regional curve assumed for the population with regard to bias and variability at different return periods. It is found that the performance of the models with few parameters are the best. It is shown that the performance of regional methods arc much better than average performances at the sites, and that the application of regional analysis by dividing the basin into homogeneous subregions significantly reduces the variability of estimates.