Design-based strategies for sampling spatial units from regular grids with applications to forest surveys, land use, and land cover estimation

The purpose of this paper is to compare some spatial strategies for sampling polygons onto a grid partitioning a study area. Most of the schemes considered in the paper are aimed at avoiding the selection of neighboring polygons. When one or more auxiliary variables are similar or well correlated with the values of the survey variable, the auxiliary information is adopted at estimation level by means of the difference or the regression estimators, or at design level, using the values of auxiliary variables to determine the inclusion probabilities. Applications to large-scale forest inventories, land use estimation, and forest cover estimation are discussed. A simulation study is performed to compare the adopted strategies in terms of bias (if present), accuracy, and accuracy estimation. The simulation is designed to mimic forest inventories and forest cover estimation, starting from some real situations. An application to plan future surveys for land use estimation in Italy is reported.