ABSTRACT
This chapter describes two sampling designs by which the sampling locations are evenly spread throughout the study area: regular grid sampling and spatial coverage sampling. The spatial coverage sampling design is used to fill in the empty spaces of an existing sample. Sampling on a regular grid is an attractive option for mapping because of its simplicity. The data collected on the grid nodes are not used for design-based estimation of the population mean or total. Commonly used grid configurations are square and triangular. Local undersampling with regular grids can be avoided by relaxing the constraint that the sampling units are restricted to the nodes of a regular grid. This is what is done in spatial coverage sampling or, in case of a sample that is added to an existing sample, in spatial infill sampling. Existing sampling units can easily be accommodated in the k-means algorithm, using them as fixed cluster centres.
