| Abstract: | The phenomenon known as 'terrain' is a continuous surface. However, when a digital terrain representation is based on a regular raster (i.e. a DEM) the digital surface is commonly not continuous. This is the case for the derivation of variables such as slope, aspect, and curvature values as performed in today's Geographic Information Systems (GIS). Often, there is no surface specified at all, as, for instance, when flow lines or watersheds are constructed. The discrepancy between the phenomenon to be modelled and its digital representation causes the terrain analysis results to be less accurate than they could be. Furthermore, if more than one type of terrain information is derived the results are likely to be based on different specifications of the seemingly same terrain surface. The combined application of the derivation results will likely introduce inconsistencies. This paper suggests founding the specification of digital terrain representations on a careful analysis of the properties of the phenomenon. The paper details the reasons for, and advantages of, continuous surface representations and emphasises the importance of a comprehensive documentation of the conceptual models underlying digital terrain representations. A review of suitable interpolation approaches for the specification of terrain surfaces is given. The paper discusses how the resulting digital surfaces are analysed and how measurement uncertainty may be accounted for. |
