Developing regional flood frequency models for Victoria, Australia by using Index Flood (IF) approach
Date
2009
Authors
Masan, M.N.
Hewa, A.G.
Editors
Anderssen, S.R.
Advisors
Journal Title
Journal ISSN
Volume Title
Type:
Conference paper
Citation
World IMACS Congress and MODSIM09 International Congress on Modelling and Simulation: Interfacing Modelling and Simulation with Mathematical and Computational Sciences, Proceedings, 2009 / Anderssen, S.R. (ed./s), pp.3407-3413
Statement of Responsibility
Conference Name
18th World IMACS Congress and International Congress on Modelling and Simulation: Interfacing Modelling and Simulation with Mathematical and Computational Sciences, MODSIM09 (13 Jul 2009 - 17 Jul 2009 : Cairns, Australia)
Abstract
The Index Flood (IF) method is a widely used procedure in regional flood frequency analyses (RFFA) to make flood predictions for ungauged catchments. The IF method assumes that a group is hydrologically homogeneous and this homogeneity is exploited to produce quantiles which have been shown to be more reliable than at-site estimates (Cunnane 1988). The IF method has been widely applied in a number of studies and found to be a reliable approach if the groups are hydrologically homogenous. Hence, the RFFA by the IF method is often accomplished in three steps: delineation of homogeneous groups, determination of regional growth curves, and the development of regional prediction models for the IF. The selection of an appropriate regional frequency distribution is of prime importance in RFFA and has been investigated for many decades, but hydrologists still debate the relative merits of different distributions. For instance, Engineers Australia (1987) recommends Log Pearson Type III (LPIII) fitted by the Method of Moments (MOM), LPIII/MOM for Australian catchments. However, many studies (Vogel et al. 1993; Wang 1997; Rahman et al. 1999) have found that the Generalized Extreme Value (GEV) distribution fitted by L moments (GEV/L) method yields more reliable flood quantile estimates than LPIII/MOM. Although direct statistical analyses of at-site and regional floods have been conducted, no regional growth curves have been derived for general use in Victoria, Australia. The aim of this study was to identify the most appropriate regional flood frequency distribution for Victoria and to conduct RFFA by using the IF method. Hence the study reported in this paper consists of: (1) assessing the suitability of the GEV fitted by L2 moments (GEV/L2) for RFFA for Victoria, Australia; (2) delineation of homogenous groups; (3) development of regional growth curves for the defined groups; (4) development of regional prediction models for the IF and; (5) making predictions for ungauged (test) catchments. 72 catchments were selected and assigned into groups using cluster analysis; homogeneity of each group was also assessed. Regional growth curves for the defined groups were developed and RFFA was carried out by using the Index Flood (IF) method. A total of 72 unregulated catchments were selected and the extracted Annual Maximum series (AM) of each station was modelled by LPIII/MOM, LN/MOM, GEV/L and GEV/L2 methods. The results indicated GEV/L2 as the most appropriate method for at-site flood frequency analyses in Victoria as the quantiles made by the GEV/L2 exhibited smaller bias and mean square errors than those of the other models. Suitability of GEV/L and GEV/L2 for RFFA was assessed by constructing corresponding moment ratio diagrams. These show that L2 moment ratios of the 72 catchments lie symmetrically around the theoretical GEV curve while the majority of the L moment ratios are scattered below the theoretical GEV curve indicating that GEV/L2 is therefore more appropriate than GEV/L for RFFA in Victoria. (Figure Presented).
School/Discipline
Dissertation Note
Provenance
Description
Access Status
Rights
Copyright 2009 The Modelling and Simulation Society of Australia and New Zealand