Avaliação de metodologias de regionalização e definição de regiões homogêneas na estimativa de vazões máximas anuais

Abstract

Floods are natural phenomena that occur in watercourses and have potential to cause serious economic and social problems. Estimates of maximum stream flows have a great value in the planning and adoption of measures that reduce the impacts caused by this type of event. However, sufficient data are not always available for estimation because they are scarce or nonexistent. Thereby, it is necessary to use hydrological regionalization methods. This work analyzes the performance of the Index-Flood/L-Moments method with those obtained by at-site methodology and direct multiple regression and dimensionless curve regionalization methods. These different methods were applied to Doce river basins and basins located predominantly in Espírito Santo. Three different Index-Flood/L-Moments applications were made: utilizing series for a defined base period (1940-2005), utilizing series presenting the totality of available data including some that were not submitted to consistence analyses and adopting the criteria proposed by the software RH 4.1. for the definition of homogeneous regions. Flows were estimated for the return periods of 2, 10, 20, 50 and 100 years.The results obtained from the 1940-2005 base period series, when compared to the obtained by at-site methodology,showed that the Index-Flood method/L-Moments reached better performance than those from the other regionalization methods.The different Index-Flood/L-Moments applications presented close results. However, the one that considered the totality of data presented better results for shorter return periods. Several regions classified as homogeneous by the methodology adopted by the software RH 4.1 would be considered heterogeneous if the Heterogeneity measure (H) was considered. Several stations were removed from the analysis by all applied methodologies because they presented some discordancy or did not fit in any region. This fact indicates possible inconsistency in their streamflow series.