Main Article Content
The stands stratification for successive forest inventory is usually based on stands cadastral information, such as the age, the species, the spacing, and the management regime, among others. The size of the sample is usually conditioned by the variability of the forest and by the required precision. Thus, the control of the variation through the efficient stratification has strong influence on sample precision and size. This study evaluated: the stratification propitiated by two spatial interpolators, the statistician one represented by the krigage and the deterministic one represented by the inverse of the square of the distance; evaluated the interpolators in relation to simple random sampling and the traditional stratification based on cadastral data, in the reduction of the variance of the average and sampling error; and defined the optimal number of strata when spatial interpolators are used. For the generation of the strata, it was studied 4 different dendrometric variables: volume, basal area, dominant height and site index in 2 different ages: 2.5 years and 3.5 years. It was concluded that the krigage of the volume per hectare obtained at 3.5 years of age reduced in 47% the stand average variance and in 32% the inventory sampling error, when compared to the simple random sampling. The volume interpolator IDW, at 3.5 years of age, reduced in 74% the stand average variance and in 48% the inventory sampling error. The less efficient stratificator was the one based on age, species and spacing. In spite of the IDW method having presented high efficiency, it doesn t guarantee that the efficiency be maintained, if a new sampling is accomplished in the same projects, contrarily to the geostatistic krigage. In forest stands that don t present spatial dependence, the IDWmethod can be used with great efficiency in the traditional stratification. The less efficient stratification method is the one based on the control of age, species and spacing (STR), contributing with 17% of the average variability reduction and with 13% of the sampling error reduction. The optimal number of strata that minimizes the variance is 6 for both interpolators.