ABSTRACT
Regional flood frequency analyses often include identification of (a) homogeneous regions and (b) a suitable frequency distribution for derivation of regional formulae. In this paper, annual maximum (AM) flood series (of 23 gauging stations) and peak over threshold (POT) series (of the selected 14 gauging stations) of 10–34 years data of River Brahmaputra (India) and its tributaries are analyzed using five L-moments-based statistics: (i) discordancy measure, (ii) cluster analysis, (iii) heterogeneity measure, (iv) L-moment diagram, and (v) Z-test. Data were screened using (i) outlier test, (ii) randomness test utilizing the first-order serial autocorrelation coefficient (r1), (iii) cross-correlation test, and (iv) discordancy measure. For the identified three homogeneous regions, GEV and General Pareto distributions are suggested as suitable distributions, flood generation mechanisms are detected, and regional formulae are derived.
