Quality assessment of national digital surface models

Figure 1: DOM/AV points with CIR ADS80 (Click to enlarge image)

Figure 2: DOM/AV point density [points/m2] (Click to enlarge image)

Figure 3: DOM/AV mean triangle area [m2] (Click to enlarge image)

National digital surface models are characterized by global spatial accuracy and density estimates like the official LiDAR dataset DOM/AV of swisstopo with an official average point density of 0.5 points/m2, a vertical accuracy of +/- 0.5 m in open areas and +/- 1.5 m in terrain with vegetation. Besides the given survey accuracy of datasets, for a variety of monitoring tasks knowledge about the local point distribution is often more important to assess the interpolated values of digital surface models. Most applications use interpolated surface models as grids with unknown local accuracy estimates. Due to long and complicated acquisition procedures of national datasets, the effective local point density can differ significantly from the given average estimates. Very often repeated flight acquisition and hilly terrain even emphasize already clustered point distributions.

Figure 1 shows a sample area (ca. 100x100m) in the Wallis (green dots are DOM/AV data points, color infrared image from ADS80 as background) with often encountered point clustering. To estimate the local point density a spatial grid of 10x10m has been calculated. 

Figure 2 demonstrates 9 cells with effective values in the middle row [points/m2]. The high point density of the cell in rank 1 overestimates the quality compared to ranks 2 and 3. To achieve a more robust estimate of the spatial point distribution, a local triangulation for each grid points is calculated and allows the comparison of the triangle areas of the cell point clouds. 

Figure 3 indicates the mean triangle area [m2]. The smaller the mean triangle area, the better are LiDAR points distributed within the grid cell. Therefore a different rank is achieved for the 3 cells in the middle row compared to Figure 2.

Statistics of triangle areas represent robust quality estimates compared to local point density and are a prerequisite for many analysis tasks in national monitorings and also for the integration of heterogenous digital surface models. The implementation is based on GDAL (Geospatial Data Abstraction Library) and CGAL (Computational Geometry Algorithms Library).

Publications

• Integration of heterogenous digital surface models (pdf)
  

Contact

• Ruedi Bösch