J. Washtell, S. Carver and K. Arrell
School of Geography, University of Leeds, LS2 9JT
Tel. (+44) 0113 3433343
Fax (+44) 0113 3433308
Viewshed analyses are a potentially useful way of classifying landscapes for the purposes of geomorphometry as well more common applications in landscape evaluation, scenic quality assessment and associated comparative analyses (Fisher et al., 2004; Germino et al., 2001; Wilson et al., 2008; ). It is theoretically possible to quantify a landscape based on openness and the cumulative visibility of a range of geomorphometric indices such as slope, aspect, curvature, roughness, ruggedness, altitudinal range, peakiness, etc (Wood, 2002, Fisher et al., 2004). This can provide the user with an indication of both how open the terrain is and what the precise characteristics of a landscape are that are visible from any point on a terrain surface. This could be implemented using standard, off-the-shelf viewshed analysis tools available in most proprietary GIS software, but the processing overheads, particularly in regard to the extremely long run times involved, where t = years rather than hours or days, make this an impractical proposition even with the most powerful processors. Of course, parallelisation of viewshed algorithms has been shown to be effective in reducing the time required for such analyses (Ware et al., 1996; Kidner et al., 1997; Ware et al., 1998) and distributed GRID computing offers still further reductions in overall run times (Rana and Sharma, 2006), but these methods do not lend themselves easily to most users of desktop GIS. Much effort has been put into developing faster and more efficient viewshed tools in GIS (Huanping et al., 2007; Izraelevitz, 2003) using a variety of algorithms such as tracking in, tracking out, approximation of line of sight, reference plane and block partitioning. This paper describes an alternative approach that utilises a novel voxel-based viewshed algorithm that can significantly reduce overall run times to acceptable levels and allow interactive, real-time evaluation of viewsheds on a standard desktop PC with large terrain models consisting of several million cells.