Abstract: Contour curvature is one of the fundamental topographic indicators that reflects the convexities in the horizontal direction of the topographical surface and expresses the divergence and confluence patterns of the motion of the surface matters. Calculating the contour curvature can measure the information conversion rate of the DEM interpolated by the vector contours, and provide a reference for setting the appropriate DEM resolution. This study based on vector contour data and DEM of Xiannangou watershed in Ansai County, Shaanxi Province, China, using the circle fitting algorithms (adjacent three-point method, interval three-point method and least square method) and surface fitting models (Evans model, Shary model and Zevenbergen model) to calculate the contour curvature respectively. The comparative analysis between the real topography of the site and the computational results indicated as follows: (1) The results from the three-point fitting method are different from the results using the least squares method, and the former describes the spatial pattern distribution of the contour curvature more accurately. (2) The frequency distribution of the results calculated by the least squares method is the highest, and the difference between the frequency curves of the two three-point fitting methods is small. (3) In the calculation results of the DEM, the results from the S-model display a larger difference than that by the E-model and the Z-model in the spatial pattern. The calculation results based on the E-model have a better effect on the contour curvature description. The difference presented by this paper when using different algorithms to calculate the contour curvature provides important reference in retrieving the contour curvature information in the practical digital terrain analysis.

Key words: digital terrain analysis, digital elevation model (DEM), contour curvature, circle fitting algorithms