Seven probability distributions were analysed: Exponential, Gamma, Log-Normal, Normal, Weibull, Gumbel and Beta, for monthly and annual rainfall in the south-central region of Ceará, Brazil. In order to verify the adjustments of the data to the probability density functions, the non-parametric Kolmogorov-Smirnov test was used with a 5% level of significance. The rainfall data were obtained from the database at SUDENE, recorded from 1913 to 1989. For the total annual rainfall, adjustment of the data to the Gamma, Gumbel, Normal and Weibull distributions was satisfactory, and there was no adjustment to the Exponential, Log-normal and Beta distributions. Use of Normal distribution is recommended to estimate the values of probable annual rainfall in the region, this being a procedure of easy application, performing well in the tests. The Gumbel frequency distribution was the one that best represented rainfall data over a monthly period, with the highest number of adjustments during the rainy season. In the dry season, rainfall data were best represented by an exponential distribution.
Rains; Agricultural climatology; Precipitation (meteorology) - prediction
