The use of mathematical models for predicting rainfall is a practical and accurate way of determining this parameter to be applied to regions which do not have any precipitation data. Based on the intense rainfall equations and Gumbel's probability model for maximum daily precipitation of meteorological stations in Minas Gerais State, the objective of this work was to adjust the Bell's Method, with regional features, for rainfall, adjusting one model for each region. The regional parameters were estimated by non-linear multiple regression, using Gauss-Newton's method. The goodness of the models was evaluated by the coefficient of determination and mean errors of prediction as compared to the original data. Data from three meteorological stations in the Northern region, which were not used to adjust the respective model, were used for validation purposes. The most frequent precipitation was tested by the arithmetic mean, the weighted mean by the inverse-square-distance and the geo-statistical prediction (kriging). The models produced good statistical parameters, with low mean errors, showing their accuracy, specially when the kriging method for estimating the most frequent precipitation was used.
intense precipitation; hydraulic structures; regionalization; kriging
