Topographic attributes computed from digital elevation models are depe
ndent on the resolution of the elevation data from which they are comp
uted. A regular rectangular grid is not an ideal representation of top
ographic surfaces for the study of scale effects. Spectral and wavelet
techniques are obvious alternatives but have several deficiencies, pa
rticularly in their use of oscillatory basis functions. The positive w
avelet representation has very attractive properties of localisation a
nd feature representation. Preliminary application to one-dimensional
topographic data (profiles) yields useful results, including the ident
ification of changes in topographic structure with scale. Extension to
two-dimensional analysis will allow quantification of characteristic
shapes, scales and orientations in the landscape.