Probabilistic models have played an important role for stochastic modeling of the extreme events. They have been widely used in hydrological studies to predict, for a confidence level, expected values in a given return period. Several probabilistic models have been applied based on different adjustment methods which estimate the statistical moments of them. The aim of this study was to verify the adherence of Generalized Extreme Values (GEV) and Gumbel probabilistic models at two frequency of occurrence of heavy rainfall events in the Upper Rio Grande region (UGR), Southern Minas Gerais. UGR region was divided into three sub-regions with similar physiographic features in which four rain-gauging stations were studied: Plateau of Campos das Vertentes (PCV), Mantiqueira Range (SM) and a transition sub-region (TR). For estimating the distribution parameters, the Moment Method, Maximum Likelihood and L-moments were tested. The adherence of probabilistic models was evaluated by both Filiben and Chi-square tests. L-moments method showed to be the best method for all sub-regions and models. In addition, for the SM and TR sub-regions Gumbel model produced better results while for the PCV, GEV model was more appropriate.
asymptotic distributions; statistical hydrology; statistical inference
