Impact of catchment homogeneity on the accuracy of design flood estimation : a case study for NSW

![CDATA[Regional Flood Frequency Analysis (RFFA) is widely adopted to estimate design floods in ungauged catchments. In RFFA, homogeneity of region is a fundamental concept, which is the state of being similar flood responses, mostly the reflection of similar catchment and flood characteristics. In general, it can be said that the more the homogeneity for a region the more accurate the flood quantile estimation is by the developed RFFA model. The main objective of this study is to examine the impact of heterogeneity on the accuracy of flood quantile estimation in RFFA. This study selected 88 gauged catchments across New South Wales (NSW), Australia. Based on Hosking and Wallis (HW) test, it was found that a few stations were discordant, and H-statistics ranged from 6.62 to 14.23, which indicated the region was highly heterogeneous. Selected catchments were grouped into two regions based on drainage divisions. Similarly, HW test statistics were calculated and developed RFFA model to estimate flood quantiles. A log-log model was used to develop prediction equations using ordinary least squares regression (OLS). Median absolute relative errors (REs) were estimated and compared with H-statistics. The leave-one-out (LOO) validation was adopted to check the relative accuracy of the developed RFFA techniques. It was found that degree of heterogeneity does not have a direct impact on the RE. Further study is needed to better understand the link between regional heterogeneity and the precision of quantile estimation.]]