Comparison of Different Strategies for Determining Raster‐Based Least‐Cost Paths with a Minimum Amount of Distortion

The conventional raster‐based least‐cost path technique, in which the connectivity graph is constructed by treating each raster cell as a node and linking each node to its adjacent neighbors, has been widely used to determine optimum paths in space. Unfortunately, paths calculated with this method are subject to distortions that stem from the imposed node‐link structure. Due to the distortions, the calculated paths may be longer than the true optimum paths in continuous space, and because of this, they may also look highly unrealistic when displayed on a map. This study examines how the quality of the raster‐based paths can be improved by using larger connectivity patterns on the one hand, and placing nodes on the sides of the cells of a raster instead of at their centers, on the other. In addition, the utility of the quadtree structure is examined. The different methods are evaluated against three different datasets using the cost distance of the calculated paths and processing times as evaluation criteria. The results indicate that raster‐based paths can be improved by using the evaluated techniques, although the degree of improvement is also dependent on the properties of the underlying cost raster.