Accuracy Considerations in the Analysis of Depressions in Medium-Resolution Lidar DEMs

Abstract

The purpose of this study is to determine the effect of vertical accuracy of a Digital Elevation Model (DEM) on the occurrence of topographic depressions. Stochastic depression modeling of a medium-resolution lidar DEM for a low-relief study area was carried out using Monte Carlo simulation of a range of Root Mean Square Error (RMSE) values for vertical error. Depth and size of observed depressions were compared to the stochastic modeling results in order to separate artificial from real depressions. Small and shallow depressions were more likely to be artificial than large and deep depressions, but the use of single threshold values for surface area, mean depth, or maximum depth to distinguish artificial from real depressions results in many incorrect classifications, and further empirical field validation is required. Stochastic error modeling of DEMs was effective in determining the reliability of a complex unconstrained terrain derivative such as the occurrence of topographic depressions. However, stochastic approaches do not properly account for large systematic errors common in lidar DEMs. As lidar data become more widely used and the accuracy expectations for terrain derivatives increase as a result, a more rigorous characterization and/or removal of these systematic errors will become necessary.