The probable minimum precipitation associated to the period, location and frequency of occurrence represents an important tool to support agricultural practices. The objective of this work was to map the probable minimum monthly and fortnightly precipitation for Southern Minas Gerais State, applying geostatistical procedures. The probability level considered was 75% and the period evaluated was the rainy season. The values of probable minimum precipitation were estimated by Log-Normal Probability Distribution, following Kolmogorov-Smirnov adequacy test and using long-term daily precipitation data set of 69 pluviometric stations in the region and its surroundings. Exponential and spherical semi-variogram models, adjusted by Maximum Likelihood (ML) and Minimum Weighted Square (MWS), were modelled to choose the best to represent the experimental semi-variogram. Based on Spatial Dependence Degree and Cross-Validation, the exponential model adjusted by MWS presented better performance. The probable minimum precipitation maps have shown considerable variability during the rainy season. The greatest values were estimated for the region close to Mantiqueira and Canastra Ranges.
geostatistic; spatial continuity; irrigation planning
