Nowadays, irrigation is one of the most important agricultural technique. Therefore, this technique can not be the only source to supply water for crops, because the irrigation system may be over designed, increasing installation costs. One of alternatives to solve this problem is to analyze the probability of rainfall, decreasing costs and easing the irrigation management. This study purposes to characterize probable rainfall for Madre de Deus Village, comparing four (4) probability distribution models (Gama, Normal, Log-normal at 2 and 3 parameters). Daily rainfall depths were totalized in consecutive periods of 10, 15 and 30 days, being evaluated for 13 different probability levels, using historical series of 57 year observation, between 1942 and 1999. Kolmogorov-Smirnov statistical test was applied to evaluate the adequacy of the adjusted probability models, towards to the most adequate model for each historical series. Log-normal at 3 parameters was the most adequate for monthly periods and for fortnight and decennial periods, Gama distribution was the best model.
Irrigation; probability distribution; hydrology
