Development of Large Topographic Datasets for Hydrologic Analysis

Stream and watershed delineation based on Digital Elevation Models (DEM) is currently standard practice in hydrologic studies. For the United States, DEMs with a horizontal resolution (i.e., cell size) of 30 meters is publicly available from the National Elevation Dataset (NED) developed and distributed by the U.S. Geological Survey (USGS); and, for some areas of the country, USGS has also developed 10-meter DEMs. Efforts to develop DEMs of even higher resolutions continue taking place, although the advantages of increasing the accuracy of the results are partially offset by the increased file size, difficulty to handle them, slow screen rendering and increase computational effort. Among these efforts, those based on the use of Light Detection And Ranging (LiDAR) pose the problem that traditional GIS-based interpolation techniques (IDW, Spline, Kriging and TOPORASTER, among others) for developing DEMs from point data have difficulty processing large amounts of data. Terrain is an alternative format for storing topographic data that intelligently decimates data points and creates a simplified, yet equally accurate for practical purposes, DEMs or Triangular Irregular Networks (TIN). This study evaluates the impact of the decimation method and interpolation technique, used to create the Terrain, on the resulting DEM and also on the streams and watersheds delineated from it. Two case studies were considered for assessing the effect of the different methods and techniques. One of them consisted of dendridic topography in Williamson Creek, Austin, Texas, and the other of deranged topography in Hillsborough County, Florida. LIDAR data with resolution of three and five feet, respectively, and additional supporting breaklines were available and used to develop DEMs employing the different methods and techniques as follows: (1) the LiDAR points were converted to raster format by traditional interpolation methods, (2) the LiDAR points were converted to TIN format and then to raster format, and (3) the LiDAR points were converted to Terrain. In all cases, the DEMs were validated with a set of points that were removed from the original dataset before processing it. Likewise, all DEMs were used to delineate streams and watersheds with the eight-direction pour point algorithm (D8). The results were compared using a standardized metric that measures the accuracy of the stream lines and watershed polygons, the size of the files, and computational effort. It is expected that the conclusions of this study will serve as guidelines for Terrain creation for water resources applications.