A Digital Elevation Model (DEM) is a gridded representation of terrain elevation where each cell stores a height value. DEMs come in two main variants: Digital Terrain Models (DTM) represent the bare-earth surface with vegetation and buildings removed, while Digital Surface Models (DSM) include the tops of all features. DEMs are derived from multiple sources -- stereo photogrammetry, InSAR, LiDAR, and GPS surveys -- each with different accuracy, resolution, and coverage characteristics. DEMs underpin slope analysis, watershed delineation, viewshed computation, flood modeling, orthorectification of imagery, and countless other spatial analyses.
Elevation data is the third dimension that transforms 2D mapping into 3D understanding of the landscape. A DEM encodes the shape of the terrain -- ridgelines, valleys, slopes, flat plains -- in a regular grid that computers can analyze systematically.
The source technology determines DEM characteristics. LiDAR produces the highest-accuracy DEMs (5-15 cm vertical accuracy) with the ability to separate bare earth from vegetation and buildings, but coverage is limited and acquisition is expensive. InSAR provides moderate-accuracy global DEMs -- SRTM (30 m, ~10 m vertical accuracy) and TanDEM-X (12 m, ~2 m vertical accuracy) cover most of Earth's land surface, but these are DSMs that include canopy. Stereo photogrammetry from optical satellites generates DEMs from parallax between images, with accuracy depending on the baseline and image resolution.
Terrain derivatives computed from DEMs include slope (the rate of elevation change), aspect (the compass direction a slope faces), curvature (how slope changes across the surface), hillshade (simulated illumination for visualization), and hydrological products like flow direction, flow accumulation, and watershed boundaries. These derivatives are often more useful than raw elevation for analysis -- slope determines erosion potential and construction suitability; aspect controls solar exposure and microclimate; watershed boundaries define the fundamental units of hydrological management.
DEM quality assessment requires understanding the error characteristics of the source. LiDAR accuracy degrades in dense vegetation (fewer ground returns), steep slopes (larger footprints), and areas with low point density. InSAR DEMs have systematic biases in vegetation (canopy elevation, not ground), urban areas (layover effects), and steep terrain (shadow and layover). Knowing these error patterns is essential for choosing the right DEM for each application and interpreting results appropriately.