Managing Low Flows in Flood Frequency Analysis in Eastern Australia

Flood frequency analysis (FFA) is needed in designing hydraulic infrastructure and many other water resources management tasks. This paper investigates the impact of censoring low flows in annual maximum flood (AMF) series in FFA using Multiple Grubbs-Beck (MGB) test. Two commonly used distributions in FFA, log Pearson type 3 (LP3) and Generalised Extreme Value (GEV) are compared. A total of 582 stream gauging stations across Eastern Australia are selected for this study. Results indicate that over two-third stations require low flow censoring and fitting of the LP3 distribution significantly improves with censoring. GEV-L moments method does not require censoring. Overall, GEV-L moments method is found to be more robust than the LP3 distribution in dealing with low flows in FFA. A regression analysis is carried out to explore the relationship between the number of censoring data points and catchment characteristics, which reveals a modest correlation between the number of censored data points and mean annual rainfall and catchment shape factor for South Australian catchments. These findings emphasise the importance of censoring in FFA and highlight the effectiveness of the MGB test in identifying low flows in Australian AMF data. The findings would help practicing engineers to implement FFA techniques recommended in the Australian Rainfall and Runoff (ARR) 2019.