Geostatistics

Variables in Space

Spatial data, the kind that are handled in geostatistics, consist of locations and a value for one or more variables at each location. Each variable is used to represent a parameter of interest. For example, precipitation data from rain gauges might include the latitude and longitude; elevation; and rain amount, in millimeters, at each rain gauge. The data can be collected at systematic or random locations but need not come from a statistical sampling design. A critical aspect of geostatistics that is often absent in other approaches is the explicit recognition of the size of the spatial unit being characterized by the data. This spatial unit size is called the spatial “support” of a measurement or variable. In the case of rain gauge data, the rain amount may be described as belonging to a unit that is quite small, nearly a point. For data on plant density, the support would be the size of the plot within which the count of individuals is made. Another example is [Page 207]the volume of soil used for an extraction of a particular mineral. Generally, data to be analyzed geostatistically are expected to have a support whose area or volume does not change over the region of interest. The reason for this is that as support changes, overall variation and spatial patterns always change. A familiar example of this phenomenon comes from agriculture, where soil moisture variation from field to field may not capture the range of wetness and drought that is experienced by individual plants. Therefore, the required duration of irrigation based on mapping average field moisture may not supply enough water to small, dry patches within a field.

...