Stratification of an eucalyptus plantation through geostatistical interpolators and remote sensing

Abstract

The objective of this work was to evaluate the use of geostatistical interpolators and remote sensing images for the stratification of Eucalyptus sp. stands, at harvesting age. Estimates by stratified random sampling and systematic sampling were compared for the timber volume variable. The study area was a clonal plantation of Eucalyptus sp. in the municipality of Lagoa Grande, in the state of Minas Gerais, Brazil. The forest inventory was performed in 2011, considering four sampling situations and different plot numbers. The area stratification was carried out based on kriging of the basal area (target variable), on the reflectance values in the TM5 band (auxiliary variable) extracted from a Landsat 5 TM image, and on the cokriging of these variables. Basal area and reflectance in the TM5 band showed a spatial dependence structure. Stratification reduced the sampling error up to 40%, in comparison with that generated from systematic sampling. Remote sensing images and kriging and cokriging interpolators are efficient to define the strata in Eucalyptus sp. stands, as a consequence of the higher precision of the stratified random sampling estimators, which allow more precise volume estimates of the forest inventory than those of systematic sampling.