We introduce POLS (shorthand for POLygon Sampling), a versatile GIS tool for extracting a random subset of polygons from a vector layer. POLS enables users to optionally (i) set the sampling intensity in terms of percent area of the layer, number of polygons, or both; (ii) specify different strata to be sampled with equal or different intensity; (iii) preclude the occurrence of adjacent polygons in the sample; (iv) ensure that the output sample is spatially balanced; (v) estimate empirically (through simulation) the inclusion probability of each individual polygon; and (vi) compute the Horvitz–Thompson Estimator (HTE) and its confidence interval for target variables measured in the sample polygons. POLS is specially suited for accuracy assessments of thematic maps that use polygons as sampling units, but it can also be applied to any probability-based survey that relies on GIS polygons. The option enforcing non-adjacency potentially increases sampling efficiency by reducing the effect of spatial autocorrelation. The spatial balance option ensures that the polygons in the sample are well distributed across the extent of the layer. When the non-adjacency constraint is used, the tool applies a novel random-selection algorithm that is designed to reduce the impact of this constraint on both the inclusion probability and the spatial distribution of sample polygons. We describe the tool and the algorithm behind it, compare the latter with two other methods that we previously tested, study the impact of the non-adjacency constraint and the spatial balance on the inclusion probability, and demonstrate the estimation of both the HTE and its variance for a sample target variable. The tool is freely available on the internet.
Read full abstract