Scale and the spatial structure of landform: optimising sampling strategies with geostatistics

Information on the scale or frequency of spatial variation in properties such as land-form is of value in a wide variety of contexts including classification of land-form types and as an input to environmental modelling applications. This paper utilises this information to demonstrate how producers of digital terrain data sets may ascertain the best approach to employ and the nature and configuration of data that would be required to fulfil a particular user's requirements in terms of information and accuracy. The approach presented is applicable whether data are sampled on the ground or by remote sensing. The research centres around an examination of Ordnance Survey(R) Land-Form PROFILE (TM) contour data and the particular focus is the potential of the application of geostatistics to the improvement of the National Height Dataset (NHD). The research illustrates that to apply geostatistics it was necessary to ensure homogeneity of spatial variation across the region in concern. That is, it was necessary to classify spatial variation. Once classification of spatial variation was achieved it was possible to quantify the variation (using the variogram) and identify dominant forms of spatial variation for particular regions to aid the design of optimal sampling strategies. The results demonstrate that significant gains in efficiency can be obtained by adapting (i) the geostatistical approach and (ii) classifying the spatial variation prior to the application of geostatistics.