A Digital Elevation Model (DEM) is a digital raster (or grid) representation of elevation values of land surface shapes and features, where each grid cell takes a single elevation value with reference to a certain vertical datum. A DEM can be global, regional or local in scope, and can be used to characterize the dry land surface (topography) or submerged surfaces (bathymetry). Since a DEM cannot contain information of shapes and features under overhanging structures, it is often referred to as 2.5D instead of truly 3D.
A digital elevation model is an overarching term for either a digital surface model (DSM) or digital terrain model (DTM). A DSM includes elevations of surface features such as trees, buildings, bridges and artificial objects such as poles, power lines, cars etc., and thus contains always the highest elevations of any feature for any given raster cell. A DTM does not include such features but reflects the elevation of bare land surface shapes, excluding elevated or overhanging features.
DEMs can be obtained using active or passive measurements. Active measurements involve the generation of electromagnetic signals towards a surface and timing the reception of the (return) signal(s). This can be achieved through laser scanning (LiDAR) using visible or infrared light pulses for bathymetric or topographic measurements respectively, radio waves (SONAR) used in bathymetric measurements, or microwaves (synthetic aperture radar, SAR) used in topographic mapping. The most widely known active remotely sensed global DEM is derived from the Shuttle Radar Topography Mission (SRTM) obtained by a SAR mounted on the space shuttle Endeavour, offering 30 m resolution with a vertical accuracy typically between 5 and 20 m, covering 80% of Earth’s surface.
Passive measurements detect reflection of sun light, or energy radiated from the surfaces. Their distance to the detector can then be inferred from the measurement of angles. Historically, line scanning imagers were used, but nowadays, these are replaced by acquisitions of overlapping 2D frame images. On the images, corresponding land surface features are detected which act as tie-points. The distance between the sensor and the tie-points is calculated in a process called photogrammetry. The most widely known spaceborne passive remotely sensed global DEM is derived from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data onboard the Terra satellite. It offers similar resolution and accuracy compared to SRTM, but with 99% coverage.
Only LiDAR can generate both accurate DSMs and DTMs from the same data acquisition, by using multiple returns from a single emitted pulse. All other techniques generate DSMs, from which elevated features can be identified and filtered out in postprocessing to create DTMs, however with typically lower accuracy and more artefacts.