Resumo. The objectives of this study were to verify the possibility of application of the spatial dependence structurefor stratus definition in forest stands and to compare, based on the sampling error, Random Sampling Design (ACS) and Stratified Sampling Design (ACE), being the stratification realized by the spatial interpolatorsInverse of Square of Distance (IQD) and Kriging. The data was provided from conventional inventories conducted in the years of 2006 and 2007, in clonal eucalypt stands on projects located inCurvelo, João Pinheiro and Montes Claros Counties, in Minas Gerais state. The results showed that, of the 50 projects, 64 % presented medium or high spatial dependence degree, evidencing that geo-statistical methods reducethe error without increasing the costs in the inventory. The stratification based on IQD which was used inthe projects with low spatial dependence degree (DE) contributed for a mean reduction of 68.4 % in thesampling error. The use of the kriging in the projects with medium or high DE, contributed for a mean reduction of 47.0 % and 65.7 % in the sampling error, respectively. Even in the projects where the spatial dependence structure was detected, the stratification based on IQD generates improvementof the estimates in the ACE compared to the ACS.