A Comparison of Interpolation Techniques in Producing a DEM from the 5 m National Geospatial Institute (NGI) Contours

Continuous elevation surfaces, which are commonly referred to as Digital Elevation Models (DEM), are vital sources of information in flood modelling. Due to the multitude of interpolation techniques available to create DEMs, there is a need to identify the best suited interpolation techniques to represent a localised hydrological environment. This study investigated the accuracies of commonly applied interpolation techniques including Inverse Distance Weighting (IDW), Nearest Neighbour (NN), Kriging, Spline and Topo to Raster interpolation techniques as applied to a 5-m interval elevation contours as a precursor to simulate a flood zone in the Roodepoort region in Johannesburg, South Africa. A 50 cm resolution DEM derived from aerial Light Detection and Ranging (LiDAR) point cloud was used as a reference to compare the five interpolations techniques. The Topo to Raster results were not significantly different from the reference data (P = 0.79 at 95% confidence level), where elevation values were on average underestimated by 0.93 m. In contrast, the spline interpolation showed the highest significant difference from the reference data (P = 0.00 at 95% confidence level), with an average underestimation of the elevation by 69.84 m. Outlier identification using standardized residual analysis flagged significant elevation outliers that were produced in the interpolation process, and it was noted that most of the outliers across all techniques coincide with areas that showed frequent topographical changes. Specifically, the largest significant differences using the Topo to Raster technique were overestimations of the elevation that occur in the upstream section of the tributary. The Spline technique in contrast showed significant underestimations of the elevation throughout the river system. Overall, the results indicate that the Topo to Raster technique is preferred to accurately represent the topography around a river system of the study area.