The topological viewshed: embedding topological pointers into digital terrain models to improve GIS capability for visual landscape analysis

The effort to develop a Digital Earth has made dramatic progress in terms of visualisation and visual data integration for use-cases which demand semantically rich analysis. To provide this analysis and ensure legitimate representations of the spatial data from which visualisation are derived, it is necessary to provide more comprehensive analytical capabilities of the view. Questions of aesthetic valuation of landscape require a richer analytical response than simply ‘whether and possibly how much of’ an object or area of land can be seen. It requires interrogation of the scene as it appears and to distinguish between transient visual effects and those locally invariant to view point change. This paper explores a data structure to support scene analytics. As such, it first reviews the existing techniques from the fields of GIS and computer graphics as to their potential and limitations in providing a qualitatively more nuanced visual analysis. It then introduces a new method of encoding visually apparent relationships into terrain models. A prototype implementation is presented based on the Quad-Edge Triangular Irregular Network, though it is believed that raster or vector implementation would be possible. Although developed primarily with landscape analysis in mind, the method could have wider applicability.