Accounting for shape reliability in modeling contour-derived topographic properties for use in soil-terrain correlation

Elevation contours are known to be poor quality data for digital terrain modeling, but they are often the only available topographic information at national scale, especially in developing countries. We investigated several methods to derive elevation and slope data from contours for two contiguous watersheds in Burundi. Two key issues in digital terrain modeling were addressed: (1) finding the 'best' elevation interpolator, and (2) assessing the related uncertainty and its propagation to slope models. The key validation criterion was the reproduction of the terrain shape as inferred from the pattern of contours, which is more important than absolute accuracy in soil-terrain correlation. A method using a triangulated irregular network (TIN) and four grid-based methods were compared and combined. The most satisfactory results were achieved by combining the TIN-based method with a grid-based method. Treating contours as inequality constraints proved useful in simulating the elevation uncertainty. The Zevenbergen and Thorne and the Evans-Young slope algorithms were compared based on their sensitivity to the elevation uncertainty. Outputs from simulation were filtered to produce realistic alternative elevation models. In that case, the slope variance values were similar for the two algorithms, suggesting similar performances. Checking for shape reliability was found critical for the validation of topographic models.