Regional flood frequency analysis based on peaks-over-threshold approach: A case study for South-Eastern Australia

• Published in: Journal of hydrology. Regional studies 2023-06, Vol.47, p.101407, Article 101407
• Main Authors: Pan, Xiao, Rahman, Ataur, Haddad, Khaled, Ouarda, Taha B.M.J., Sharma, Ashish
• Format: Article
• Language: English
• Subjects: Annual maximum flood; Australia; Floods; Peaks over threshold; Regional flood frequency analysis; Very frequent floods
• ISSN: 2214-5818; 2214-5818
• DOI: 10.1016/j.ejrh.2023.101407

Southeast Australia Regional flood frequency analysis (RFFA) is a widely adopted statistical method to estimate design floods in ungauged catchments. Annual maximum flood (AMF) model is the most popular method in developing RFFA techniques. However, the AMF-based approaches are criticised for its limitations in the range of very frequent to frequent flood estimation. As an alternative, the peaks-over-threshold (POT) based approach has shown theoretical advantages in this flood range. POT based RFFA is currently underemployed internationally due to its complexity in modelling. This study develops POT-based RFFA techniques for south-eastern Australia using data from 151 catchments. A comparison is made between ordinary least squares (OLS) and weighted least squares (WLS) methods in developing POT-based RFFA techniques. The OLS based method is found to perform better than the WLS. The median relative error values of the developed prediction equations range 31–38%. The new POT-based RFFA technique overcomes the limitations of the current Australian Rainfall and Runoff, which does not have any RFFA technique for very frequent floods. It is expected that these new POT-based RFFA technique will be used in practice in south-east Australia. [Display omitted] •A new regional flood frequency analysis method is developed using peaks over threshold model.•The developed models provide regional floods estimates for south-east Australia accurately.•A comparison is made between ordinary least squares and weighted least squares methods.