A NEW FRAMEWORK FOR ACCURACY ASSESSMENT OF LIDAR-DERIVED DIGITAL ELEVATION MODELS

Abstract. Existing approaches to accuracy assessment of LiDAR-derived DEM are typically based on statistical methodology and focus on vertical accuracy. This paper presents a new framework which calls for assessment of not only a DEM’s vertical accuracy, but also its ability to preserve a point’s elevation rank in the bare earth topographic surface. New methods to assess each aspect are presented. For DEM’s vertical accuracy, approximation theory from numerical analysis is used to quantify the total error at each DEM point as well as its three error components – sensor error, ground error, and interpolation error. For DEM’s elevation order which is critical to model terrain structure, the concept of isomorphism in set theory is drawn as the mathematical rationale and Kendall’s rank correlation efficient is used to quantify the accuracy. The new framework is illustrated using a DEM derived from LiDAR for sea-level rise vulnerability assessment of a tidal salt marsh. Compared to conventional methods based on statistics, the new framework and methods produce detailed mapping of error distribution thus enable the identification of the main sources of error and where improvement is needed most. Results call for re-evaluation of the current practice of assessing filtering accuracy in LiDAR data processing and further research on relative elevation and isomorphism.