Abstract:
For design of various types of hydraulic structures as well as for taking different
flood management measures flood frequency estimates are required. Regional flood frequency
analysis is carried out employing L-moments and soft computing techniques viz. artificial
neural network (ANN) and fuzzy inference system (FIS) for the lower Godavari subzone 3(f)
of India. The study area covers an areal extent of 174,201 km2 and annual maximum peak
flood data of 17 catchments ranging in size from 35 to 824 km2 are used. The data screening is
carried out employing L-moments based Discordancy measure (Di) and regional homogeneity
is examined based on the heterogeneity measure (H). On the basis of the L-moment ratio
diagram and Zi
dist –statistic criteria, Pearson Type III (PE3) distribution is chosen as the suitable
frequency distribution for the region. For the region under study, a relationship is developed
between mean annual maximum peak flood and area of the catchment using the Levenberg-
Marquardt (LM) iteration and the same is coupled with the PE3 based regional flood frequency
relationship developed for estimation of floods of various frequencies for the ungauged
catchments of the region. The regional flood frequency relationships developed based on Lmoments
and soft computing techniques are compared.
