Spatial sampling strategies

A practical method to identify a spatial sampling strategy for efficient estimation of the spatial mean of an attribute in a selected area is presented. The (relative) efficiency of a spatial sampling strategy depends on the spatial (variability) correlation structure among the population elements. For undertaking spatial sampling, the following sequence of steps is considered: (i) computing the empirical variogram (or correlogram) using data from a pilot sample, (ii) fitting to these data a theoretical model of the variogram, (iii) computing average correlations between pairs of elements under the fitted model, and (iv) computing the expected efficiencies of the considered sampling strategies under the fitted model. Using data from an agricultural land use map, the method is tested. In this test, the expected efficiency of the sample mean under random sampling, stratified random sampling and systematic random sampling is computed and compared with the observed efficiency. The results are satisfactory.