The Effect of Scale-Related Issues on the Geostatistical Analysis of Ordnance Survey(R) Digital Elevation Data at the National Scale

Abstract

Digital Terrain Models (DTMs) are used widely for environmental applications (good introductions to applications of DTMs are Moore et al., 1991 and Weibel and Heller, 1991). DTMs may be used, for example, in modelling processes such as erosion or for relating spatial variation in topography to variation in other variables. To utilise DTMs effectively it is necessary to ensure that they are sufficiently accurate to meet the requirements of a specific application. The techniques of geostatistics are well suited to assessing accuracy but to use geostatistics effectively one must know whether or not to model the spatial variation as stationary across the region in concern. Decisions concerning stationarity become increasingly complex when the region of interest is very large (for example, the whole of the UK). The ultimate aim of the present research is to develop efficient approaches to handling Ordnance Surveys(R) (the national mapping agency for Great Britain) digital elevation data at the national scale. This paper presents the results of a preliminary investigation into a subset of the British dataset. An approach is presented for the classification of spatial variation in Ordnance Survey digital elevation data, and the subsequent computation and modelling of class-specific variograms. The effect that such decisions have on the estimation of unknown values using kriging is considered and discussed.