Geostatistics in Physical Geography. Part I: Theory

Abstract

Geostatistics grew from meteorologists' desire to interpolate weather characteristics from sparse data, and from mining engineers who wanted to estimate quantities of minerals in bodies of rock from drill cores. In both cases the properties of interest behaved as spatially correlated random variables. Practice in mining is now underpinned by sound theory, the theory of regionalized variables. It is widely recognized that the theory is properly applicable in many other branches of earth, atmospheric and marine science. Part I of this paper reviews the more elementary aspects of the theory and its assumptions. It describes the method of local estimation embodied in regionalized variable theory known as kriging. The central tool of geostatistics is the variogram, which expresses quantitatively and succinctly spatially correlated variation, and its estimation is described. The more common forms of variogram are illustrated, and authorized models for them are listed. There are sound procedures for choosing and fitting models to variograms now using well-tried computer programs such as MLP and Genstat. Regionalized variable theory provides geographers with a concise and coherent methodology for describing and analysing spatially distributed data.