The objective of this paper is to compare three unvaried interpolation algorithms: inverse square distance, minimum curvature and ordinary kriging using observations of annual precipitation average of one thousand and twenty-seven climatic stations including all State of São Paulo, representing an area of approximately 248,808.8 km² (2,91% of the brazilian territory), from 1957 to 1997. The maps of spatial variability didn't present great visual differences, however when the comparison criterion, the mean square error of prediction was calculated for each technique, the great advantage was confirmed by using kriging interpolator. The difference among the observed and predicted values for kriging was much smaller than for the other two methods, indicating to be this the interpolator to be used in the spatial distribution of annual precipitation for this data.
ordinary kriging; geostatistic; semivariogram
