Abstract
The regional flood frequency analysis (RFFA) technique is widely used to estimate design floods in ungauged catchments. This study presents development and testing of an ordinary kriging based RFFA technique using peaks-over-threshold (POT) model. Ordinary kriging offers significant advantages in RFFA by providing unbiased, optimal interpolation of spatial data, leading to more accurate flood quantile estimation. This study uses flood and catchment characteristics data from 419 stream gauging stations in eastern Australia. It has been found that the median relative error values are found to be in the range of 29 and 33%, which is more accurate than regression based RFFA models recommended in Australian Rainfall and Runoff (national guideline). The POT based kriging model is found to provide more accurate design flood estimates in coastal areas than the inner part of Australia. Further study should focus on developing a bias correction method to minimize tendency of over-estimation by the new kriging based RFFA model.