Spatial Statistics in Landscape Ecology

* Final gross prices may vary according to local VAT.

Get Access


In ecology, especially in plant ecology, it is commonplace to identify and quantify spatial patterns (Greig-Smith 1952, 1964; Kershaw 1964). In fact, most ecological data are inherently composed of several levels of spatial structure: large-scale trends (species responses to climate conditions, to migration process, etc.), small scale patterns, patchiness (weather conditions, physical conditions, dispersal mechanisms, predation, competition, etc.), and local random noise. Therefore, the notion of spatial autocorrelation implies that “Everything is related to everything else, but near things are more related than distant things” (Tobler 1970) and “with spatial data, there is a better than random chance that one can predict attribute values from a given areal unit from those taken on by its juxtaposed areal units …” (Haining 1980). Griffith (1992) provides several other definitions of spatial autocorrelation, which include “a diagnostic tool for spatial model misspecification; a surrogate for unobserved geographical variables; a nuisance in applying conventional statistical methodology to spatial data series; an indicator of the appropriateness of, and possible artifact of, areal unit demarcation.”