Optimal sampling strategies for raster-based geographical information systems

Many biogeographical properties must be entered into raster-based geographical information systems (GIS) as samples. For example, properties of vegetation such as biomass, species abundance and species distribution, and of soils such as organic content, must be sampled discretely and it is these sample data that are entered into and manipulated within the GIS. Kriging, an optimal technique for linear unbiased estimation, can be used to design an optimal strategy to sample the property of interest in the field and subsequently to map the property with known minimum or kriging error within the GIS environment. Once the data are in the GIS kriging can be used to estimate optimally at any location and over any support or cell as and when necessary. Optimal sampling strategies were designed for mapping the percentage cover of a crop of spring barley, estimating the mean biomass of a grassland and mapping the reflectance of a grassland measured with a field radiometer.