Modification of a digital elevation model (DEM) in a flat topographic area with respect to manmade features

Abstract

This study compares two Digital Elevation Models (DEMs) that are available free of charge: (1) the Consultative Group for International Agriculture Research Consortium for Spatial Information SRTM C-band CGIAR-CSI v4.1 (SRTM): 3 arc sec (approximately 92 m at the equator; originally 1 arc sec but only distributed with 3 arc sec) and (2) the Advanced Spaceborne Thermal Emission and Reflection Radiometer-Global Digital Elevation Model ASTER GDEM v2 (ASTER2): 1 arc sec (approximately 31 m at the equator). Additionally, the DEM was modified according to known topographic features in the study area. The first step was investigating whether there is a spatial shift between the different DEMs by using a very high resolution (VHR) satellite GeoEye image. Beside visual comparisons, statistical methods were applied to compare the elevation models. Reference data used in this study are the Ground Control Points (GCPs) collected in a previous investigation in the same study area. SRTM proved to be the better of two available free elevation models (SRTM and ASTER2). This conclusion is based on an assessment of the different investigated aspects such as morphologic details, reliability, completeness, and accuracy. The ability to modify the SRTM model with 92 m horizontal resolution from the Shuttle Radar Topography Mission is here discussed. The study area is located in Titas Upazila, Comilla district, Bangladesh and comprises manmade topographic features (e.g., road embankments and mounds that houses are built on above the monsoon flooding level), which are not or not completely represented in the DEM due to their small spatial extent. To represent these topographic features, the DEM was refined by dividing each pixel into 0.5 m pixel spacings. The elevated areas (roads and villages) were digitized using GeoEye satellite imagery and Google Earth. The pixels located in the elevated areas were given the proper elevation and rejoined to the original DEM raster. The effect of trees can be excluded because of their scarcity in the studied area, and because their existence is limited just to both sides of the artificially elevated streets and areas where people live. Furthermore, the bias in the SRTM model is eliminated by two steps: (1) the mean (value) of the differences between the GCPs and the corresponding points of the SRTM is subtracted from SRTM points, and then the root mean square error (RMSE) is diminished to 0.67 m; (2) the same mean (value) of the differences is subtracted from the whole SRTM model. The finally modified DEM represents the real terrain surface with the most important details of the study area. This modified elevation model may be used in studies to model groundwater flow driven by topography.