Implication of partial duration series on regional flood frequency analysis

ABSTRACT

In the regional flood frequency analysis, the selection of important site characteristics and a suitable number of clusters have a significant role in finding homogeneous regions. Generally, it is difficult to determine the exceedance threshold value or the number of average peaks per year (λ) in Partial Duration Series (PDS), and this difficulty is intensified in the context of regional frequency analysis, though, there are only limited studies to report this difficulty. In the present study, the Mahanadi basin is divided into homogeneous regions by using fifteen effective variables by applying different clustering techniques. Factor analysis has been introduced in finding suitable site variables. Out of fifteen variables, nine variables are selected based on factor analysis loadings. Two hard clustering techniques i.e. Hierarchical (HC), K-Means Clustering (KM) along with the Fuzzy C-Mean Clustering (FCM) techniques are used for finding out the possible number of clusters. To validate the results we used cluster validation indices to locate the optimal number of clusters. Results of the homogeneous region found from the FCM and KM hard clustering approach are more accurate and more robust than HC. The suitability of many frequency distributions to be used in the regional frequency is inspected. The results confirmed that the Generalized Pareto (GP) best described the PDS in the study area and also, show that the best PDS/GP performance is found in almost all the values of λ (2, 2.5, and 3). To test the homogeneity and to identify the best-fit frequency distribution, a regional L-moment algorithm is used. A unique regional flood frequency equation is developed which estimates the flood quantiles of ungauged catchments for each region of the basin and an index flood is also specified concerning catchment characteristics by using the multiple linear regression approach for designing small and medium hydraulic structures in the basin. The results are compared with the previous studies of flood frequency in this basin.

KEYWORDS:

• Regional flood frequency analysis
• partial duration series
• homogeneity
• clustering
• probability distribution

Acknowledgements

The author would like express their thanks to the National Institute of Technology, Rourkela for providing the research environment.

Disclosure statement

No potential conflict of interest was reported by the author(s).