Abstract:
Magnitude Prediction of likely occurrence of flood is of a great importance for solution of
variety of water resources problems for gauged, poorly gauged and ungauged catchments. Due to lack of
robustness in at-site frequency analysis, the L-moments based regional flood frequency analysis (RFFA)
has been found more applicable for flood analysis. In the present study, RFFA has been applied using less
bias and more efficient standardized L-moments and probability weighted moments (PWM) for Parvati
River in Central India. Different distributions including Wakeby-4, Wakeby-5, extreme value-I (EV-I),
generalized extreme value (GEV), logistic (L), generalized logistic (GL), generalized pareto (GP), normal
(NOR) and log normal (LNOR) have been applied in RFFA. Various tests attempted to detect randomness,
trends and homogeneity indicated that Annual Flood Series (AFSs) of all three sites of Parvati basin used
in analysis are distributed randomly having no trends and region may be considered as hydrologically
homogeneous. From the analysis of L-moment based goodness of fit tests such as L-moment ratio
diagram, L-kurtosis and Z-statistics, it has been observed that the GEV distribution with L-moments as
parameter estimator is the most robust distribution for estimation of floods of different return periods
for Parvati River system. The regional parameters of GEV distribution have been computed as u = 0.751,
a = 0.334 and k = –0.148. The regional equations developed in this study can be used for flood estimation
in gauged and ungauged basins of the region. The performance of at-site regional approach has been
compared with at-site approach with the help of D-index and it has been observed that at-site RFFA
offers better results than at-site analysis for estimation of floods.
