Methods for constructing a sampling design for large area forest inventories are presented. The methods, data sets used, and the procedures are demonstrated in a real setting: constructing a sampling design for the first national forest inventory for Tanzania. The approach of the paper constructs a spatial model of forests, landscape, and land use. Sampling errors of the key parameters as well as the field measurement costs of the inventory were estimated using sampling simulation on data. Forests and land use often vary within a country or an area of interest, implying that stratified sampling is an efficient inventory design. Double sampling for stratification was taken for the statistical framework. The work was motivated by the approach used by The Food and Agriculture Organization of the United Nations (FAO) in supporting nations to establish forest inventories. The approach taken deviates significantly from the traditional FAO approaches, making it possible to calculate forest resource estimates at the subnational level without increasing the costs.