The generalized extreme value distribution (GEV) has had great acceptance for discretion of the maximum natural events. Previous studies show that GEV distribution fits values closer to reality when there is need for extrapolation of sampling data for longer periods of returns. In order to contribute to the planning of activities that are influenced by the intensity of precipitated water, we adjusted a model of probability distribution of heavy rain through GEV, using LH moments for estimating its parameters and statistical test proposed by Wang (1998) for checking the quality of the adjustments developed in the Matlab. We analyzed historical time series of maximum rainfall for different durations obtained from rain gauges in the city of Cascavel, Paraná State. Moreover, these series were fitted by Gumbel distribution for purposes of comparison with the GEV. This study shows that the Gumbel distribution underestimates the GEV distribution for large return periods.
heavy rain; GEV distribution; LH moments
