A semi-automated tool for reducing the creation of false closed depressions from a filled LIDAR-derived digital elevation model

Closed depressions on the land surface can be identified by ‘filling’ a digital elevation model (DEM) and subtracting the filled model from the original DEM. However, automated methods suffer from artificial ‘dams’ where surface streams cross under bridges and through culverts. Removal of these false depressions from an elevation model is difficult due to the lack of bridge and culvert inventories; thus, another method is needed to breach these artificial dams. Here, we present a semi-automated workflow and toolbox to remove falsely detected closed depressions created by artificial dams in a DEM. The approach finds the intersections between transportation routes (e.g., roads) and streams, and then lowers the elevation surface across the roads to stream level allowing flow to be routed under the road. Once the surface is corrected to match the approximate location of the National Hydrologic Dataset stream lines, the procedure is repeated with sequentially smaller flow accumulation thresholds in order to generate stream lines with less contributing area within the watershed. Through multiple iterations, artificial depressions that may arise due to ephemeral flow paths can also be removed. Preliminary results reveal that this new technique provides significant improvements for flow routing across a DEM and minimizes artifacts within the elevation surface. Slight changes in the stream flow lines generally improve the quality of flow routes; however some artificial dams may persist. Problematic areas include extensive road ditches, particularly along divided highways, and where surface flow crosses beneath road intersections. Limitations do exist, and the results partially depend on the quality of data being input. Of 166 manually identified culverts from a previous study by Doctor and Young in 2013, 125 are within 25 m of culverts identified by this tool. After three iterations, 1,735 culverts were identified and cataloged. The result is a reconditioned elevation dataset, which retains the karst topography for further analysis, and a culvert catalog.