The objective of this study was to determine the level of soil spatial variability in an area consisting of the land uses: Quesungual slash and mulch agroforestry system with less than two years (QSMAS<2), Slash-and-burn traditional system (SB) and Secondary forest (SF). Soil samples were taken in three parallel transects of 102 m in length, separated 9 meters. The profile was sampled in the depths from 0 to 5 cm, 5 to 10 cm, 10 to 20 cm and 20 to 40 cm in 6 points (09, 11 am and 05) during 9 days. Coefficient of variation for soil properties varied for bulk density (0.76 and 15.1%), organic carbon (30.4 and 54.3%), volumetric moisture (9.5 and 23.5%), sand (12.8 and 22.5%) and clay (14.0 and 29.2%). The geo-statistical analysis showed that the random component of the spatial dependence was predominant over the nugget effect. The functions of semivariograms, structured for each variable were used to generate maps of interpolated contours at a fine scale. The Moran (I) autocorrelation indicated that sampling ranges less than 9 m would be adequate to detect spatial structure of the volumetric moisture variable.