Aplicabilidade e ajuste da distribuição log-normal a 3 parâmetros em estudo de precipitação máxima anual na bacia do Rio Verde

Abstract

Probabilistic estimates of extreme events are of great relevance for the planning of human activities and also for the management of water resources. In this context, we aimed to analyze the performance of the probability distribution, Log Normal to 3 parameters, which were adjusted by the Moment (MM) and Maximum Likelihood (MV) methods applied to historical series of maximum annual daily precipitation of 7 rainfall stations in the Rio Verde watershed, in Minas Gerais, obtained from HIDROWEB / ANA with at least 30 years of observation. The catchment area of the Verde River has an area of 6,891.4 km2 and about 450 thousand inhabitants. In order to verify the best estimation of the method parameters used, the adhesion tests of Komolgorov Smirnov and Chi-square were applied. The Chisquare test was more rigid in Kolmogorov-Smirnov, classifying as inadequate 43% of the models. Taking this test as base, identify greater adequacy (88%) in the estimation by the MV method. In a precision analysis, a historical Varginha series adjusted by the MV method presented the lowest value, showing a better fit. In studies with maximum annual rainfall, in this region, a Lognormal distribution of 3 parameter estimates by the MV method may be more appropriate.