Topographic accuracy assessment of bare earth lidar-derived unstructured meshes

This study is focused on the integration of bare earth lidar (Light Detection and Ranging) data into unstructured (triangular) finite element meshes and the implications on simulating storm surge inundation using a shallow water equations model. A methodology is developed to compute root mean square error (RMSE) and the 95th percentile of vertical elevation errors using four different interpolation methods (linear, inverse distance weighted, natural neighbor, and cell averaging) to resample bare earth lidar and lidar-derived digital elevation models (DEMs) onto unstructured meshes at different resolutions. The results are consolidated into a table of optimal interpolation methods that minimize the vertical elevation error of an unstructured mesh for a given mesh node density. The cell area averaging method performed most accurate when DEM grid cells within 0.25 times the ratio of local element size and DEM cell size were averaged. The methodology is applied to simulate inundation extent and maximum water levels in southern Mississippi due to Hurricane Katrina, which illustrates that local changes in topography such as adjusting element size and interpolation method drastically alter simulated storm surge locally and non-locally. The methods and results presented have utility and implications to any modeling application that uses bare earth lidar.

► A method of computing topographic error and accuracy in a digital elevation model and finite element mesh is developed.
► A novel cell area averaging method for interpolating a lidar-derived DEM to an unstructured finite element mesh is presented.
► The cell area averaging method is developed to minimize topographic elevation error in an unstructured finite element mesh.
► Simulated inundation extent and maximum water levels were affected by local changes in terrain representation.