Title: L-moment and composite model approaches for regional flood frequency analysis

Authors: Kiran Yarrakula; Biswajit Samanta; Debasis Deb

Addresses: Centre for Disaster Mitigation and Management (CDMM), Vellore Institute of Technology (VIT University), Vellore-632014, Tamilnadu, India ' Mining Engineering Department, IIT Kharagpur, West Bengal, 721302, India ' Mining Engineering Department, IIT Kharagpur, West Bengal, 721302, India

Abstract: In the design of hydraulic structures such as dams, culverts and urban drainage systems, flood frequency analysis plays a significant role. In this study, regional flood frequency analysis based on L-moment and neural network approach based on composite model has been carried out for the Subernarekha River, India. To identify the homogeneous regions, heterogeneous test and screening of the data with discordance measure have been performed. For calculating heterogeneity, 500 simulations are conducted using four parameter Kappa distribution. PE-III distribution is identified as best distribution for the study area based on L-moment ratio diagram and |Zdist|-statistic. Error statistics of regional quantile estimates for different distributions and composite model have been performed. The R² values of composite model for the quantile range below 0.99 and above 0.99 are 0.995 and 0.993 respectively. The growth factors for composite model and PEIII for the 50 years return period are 2.10 and 2.3 respectively.

Keywords: best fit contribution; frequency analysis; artificial neural networks; ANNs; Z-statistic; homogeneity; L-moment; composite modelling; regional floods; flood frequency; flooding; hydraulic structures; structural design; India.

DOI: 10.1504/IJHST.2015.070097

International Journal of Hydrology Science and Technology, 2015 Vol.5 No.2, pp.95 - 118

Received: 13 Oct 2014
Accepted: 27 Mar 2015
Published online: 26 Jun 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article