DEM application and qualification with regard to terrain analysis, land use classification and watershed modeling

This thesis addresses removing the errant elevation values from computer-extracted Digital Elevation Model (DEM) and their application as an alternative to Light Detection And Ranging (LiDAR) on slightly undulated landscape near the eastern shore of Lake Huron, Ontario, Canada. The effect of DEM resolution on the computation of terrain attributes, the role of DEM in land cover mapping and estimating of runoff, and sediment yield using different DEM resolutions are also studied.;Removing non-ground points using an orthophoto does not require predefined thresholds values, compared to complex filtering algorithms. Photo-based DEM represented better horizontal accuracy while LiDAR showed best vertical accuracy, when compared together.;There was not significant differences between vertical residual of DEM, showing that the photo- based DEM can be a good substitute, for computing terrain attributes whenever LiDAR are not affordable in this landscape.;Classification accuracy was 60% in pixel-based classification, less than the object-oriented approach (80%), due to using only spectral information of image. The object- oriented approach showed that overall accuracy will increase to 94% by using LiDAR thematic layers and have positive effects on the segmentation process and applying membership functions. Furthermore, statistical analysis did not show significant differences between reference and classified data and the Kappa indicated that the best land cover map is produced by object-oriented approach using DEM thematic layer.;Large differences were observed between 1 and 10 m resolutions on all topographical and hydrological attributes, showing that 10 m DEM is too coarse for this landscape, as the trend of plan curvature indicated that at least a 4 m DEM can capture the grain (longest significant relief wavelength) in this landscape.;Normalized runoff and sediment yield at Lucknow hydrometric station was significantly different compared to Water Erosion Prediction Project (WEPP) predicted values. The WEPP model with 1 m DEM resolution did not predict runoff or sediment yield significantly better than simulations using 5 and 10 m resolutions in this particular landscape. This conclusion is important and shows that it is not quite necessary to extract 1 m DEM resolution for watershed modeling using WEPP in this landscape, in future studies.