Computing uncertainty of physiographic features extracted from multiscale digital elevation models

Abstract

In this paper, it is proposed that the mapping of uncertainties of the three predominant physiographic features of terrains, which are mountain, basins and piedmont slopes, using variation in the spatial resolution over which these landforms are defined, can be performed with fuzzy classification. The proposed methodology allows for the generation of fuzzy certainty maps which assign high levels of uncertainty to regions with high levels of change across scales. This paper demonstrates that fuzzy certainty maps provide a better quantification of landform character than Boolean landform maps alone. In terms of sensitivity to noise, the methodology is able to identify narrow bridges, and spurious landforms, and assign these errors with low certainty values. However, it is unable to identify spurious modifications to landform shape, with these errors being assigned high certainty values. Ground truth maps are required to identify these errors.