Regional flood frequency analysis using l-moments for geographically defined regions: an assessment in Brazil

Abstract

The devastating effects of floods, combined with scarce data sets, have stimulated the development of hydrological regionalisation techniques. The present study proposed an evaluation of the L‐moments based index‐flood procedure, coupled with watershed grouping based on geographical convenience for regionalisation of maximum streamflows. A pioneer analysis for South America, addressing the geographical classification method adopted by National Water Agency of Brazil (ANA), was conducted considering over 100 watersheds in southern Brazil. Nonstationary and discordant maximum annual streamflow (MAS) series were removed with the aid of the Mann–Kendall test and discordancy measure, whereas the heterogeneity measure was used to check regional homogeneity. The best regional distribution was identified by the Z DIST goodness‐of‐fit measure. Finally, multiple nonlinear regressions, considering morphological and meteorological watershed characteristics, were performed to obtain better index‐flood estimates. It was concluded that: (a) the proposed methodological strategy provided satisfactory estimation of design floods; (b) area, mean slope, stream gradient, and flow length, were the most satisfactory explanatory variables; (c) the fitted equations stand out as a state‐of‐art alternative for the scarce hydrological monitoring in southern Brazil; and (d) the hydrological boundaries defined by ANA might not be the most adequate approach from a regional point of view.