Stratification based on Slovenian forest typology has been proposed for sampling forest land-use/ land-cover data and changes over time. Using land-use map and forest typology map we evaluated geographic scales of variance for landscape-level indices and compared stratification by administrative units and regions. As forest prevails in 10 out of 12 statistical regions in terms of its surface area, these regions cannot be effective stratification tool for sampling and mapping land-use. Statistical regions accounted for 21 percent of the total variance of percent agriculture and 17 percent of total variance in amount of forest. Through the classification of quadratic 1 square kilometre tiles according to the typology of forest site types, somewhat greater proportion of total variance has been explained by stratification than on the spatial scale of statistical regions, although on the account of the high number of groups (29 strata). The great differences in fragmentation indices for these forest groups illustrate that the spatial scale for the formation of strata is smaller than the regional one.