A digital elevation model, commonly abbreviated to DEM, is a digital representation of elevation across a geographic area. It normally represents the height of the Earth's surface using a regular raster grid in which each cell contains an elevation value, although the terminology is sometimes used more broadly for digital representations of terrain. DEMs are fundamental geospatial datasets because elevation influences water flow, slope, visibility, climate patterns, accessibility, construction and many physical geographic processes.
A DEM can be used to derive terrain characteristics such as slope, aspect, hillshade, contours, drainage networks, watershed boundaries and relief. Elevation values can be measured relative to a reference surface and can use different vertical datums, so the meaning of the elevation must be understood before analysis. The term DEM is sometimes used interchangeably with terms such as digital terrain model and digital surface model, but technical distinctions are important.
A digital terrain model generally aims to represent bare earth terrain, whereas a digital surface model can include buildings, vegetation and other objects above the ground. Usage of these terms varies between datasets and organisations, so metadata should be consulted rather than assuming a universal definition.
A DEM is usually stored as a raster in which each grid cell represents an area of ground and contains an elevation value. The horizontal resolution determines the size of the cells and therefore the level of terrain detail that can be represented. A high resolution DEM can represent small gullies, embankments and other local terrain variations that a coarse DEM may smooth out.
Increasing resolution, however, does not automatically increase accuracy because a fine grid can still contain inaccurate elevation measurements. DEMs can be generated from several technologies. Traditional surveying can provide elevation points that are interpolated into surfaces. Photogrammetry can derive elevation from overlapping aerial or satellite photographs. LiDAR can measure terrain and surface features using laser pulses. Radar based methods can produce elevation products over large areas. The source technology affects accuracy, resolution, vegetation penetration, vertical datum and whether the resulting model represents terrain or surface features. Consequently, two DEMs covering the same area can produce different analytical results because they were generated differently.
The distinction between elevation models is important in geographic analysis. A digital terrain model, or DTM, generally represents the elevation of the bare ground after removing vegetation, buildings and other above ground objects. A digital surface model, or DSM, represents the upper surface and may include trees, buildings and other structures. A DEM may be used as a general term covering one or more of these concepts, depending on the source.
A LiDAR derived surface model can contain elevation values corresponding to rooftops and tree canopies, while a terrain model derived from classified ground returns attempts to represent the underlying ground. This distinction affects applications. Flood modelling generally requires an appropriate representation of terrain and drainage, while urban visibility analysis may require buildings and vegetation. Telecommunications studies may require surface elevation plus structures. Therefore, users should not select an elevation dataset solely because it is labelled DEM. They should examine what the values actually represent and what vertical reference system and processing methods were used.
DEMs are commonly used for terrain derivatives. Slope measures the rate of elevation change across distance and can be expressed as an angle or percentage. Aspect describes the direction in which a slope faces. Hillshade creates a simulated illumination effect that makes terrain structure easier to visualise. Contours connect locations with equal elevation and can be generated from a DEM for topographic mapping.
Flow direction algorithms estimate the direction water would travel across the terrain, while flow accumulation estimates areas receiving drainage from upstream cells. These calculations can be used to identify channels, watersheds and drainage networks. Curvature measures aspects of the shape of the terrain surface. Viewshed analysis determines which areas are visible from a specified observation location based on terrain elevation. These derivatives transform a simple elevation surface into a collection of geographic variables that can support environmental, engineering and planning analyses.
Hydrology is one of the most important applications of DEMs. Water generally flows from higher to lower elevations, so terrain structure provides a basis for modelling drainage. GIS software can process DEMs to identify depressions, calculate flow direction and accumulation, delineate drainage basins and estimate stream networks.
Flood modelling can incorporate terrain elevation to estimate areas that may be inundated under specified water levels, although accurate flood modelling also requires information about river hydraulics, structures, surface roughness, drainage systems and other factors. DEMs can support watershed management, erosion studies and catchment analysis. The quality of hydrological results depends strongly on the resolution and accuracy of the elevation data. Small errors or missing terrain features can alter the modelled flow path. Bridges, culverts, walls, roads and artificial drainage systems can also complicate automated terrain based hydrological modelling because the surface represented by a DEM may not correspond directly to the pathways water actually follows.
DEMs are also used in geology, geomorphology, ecology, engineering and urban planning. Geomorphologists can analyse landforms, relief and erosion patterns. Geologists can use terrain derivatives to investigate structural features and drainage relationships. Ecologists can examine how elevation, slope and aspect influence habitat. Engineers can evaluate terrain for roads, pipelines, buildings and infrastructure. Telecommunications planners can use elevation models for radio line of sight and coverage modelling.
Renewable energy studies can use slope and aspect to investigate potential solar or wind development conditions. Archaeologists can use high resolution elevation models to identify subtle landscape features that may be difficult to see in conventional aerial imagery. In each case, the usefulness of the DEM depends on resolution, vertical accuracy, terrain complexity and whether vegetation or buildings have been included. The same elevation dataset can therefore be appropriate for one application and unsuitable for another.
Vertical reference is a technical issue that is often overlooked. Elevation can be expressed relative to an ellipsoid, a geoid or a physical vertical datum. GNSS receivers commonly produce ellipsoidal heights, while topographic and engineering applications frequently use orthometric heights related to a geoid or national height datum. The difference between these height concepts can be significant. A DEM assembled from datasets using inconsistent vertical references can contain systematic offsets even if the individual measurements are accurate.
Horizontal and vertical units must also be considered. Some datasets use metres while others use feet. Vertical accuracy can vary across the area because of vegetation, terrain, sensor geometry and processing. Metadata should therefore document the source, acquisition method, resolution, vertical accuracy, coordinate reference system, vertical datum and processing history.
DEMs are models rather than perfect copies of the Earth's surface. They generalise terrain according to their sampling density, measurement method and processing. Raster resampling can smooth or alter elevation values, while interpolation can create surface characteristics that were not directly observed. Voids, artefacts, noise and incorrect classification can affect terrain derivatives. A coarse DEM may fail to represent narrow valleys, embankments or small drainage channels.
A very high resolution model may contain surface detail that is irrelevant to a regional analysis. Users should therefore match DEM resolution and accuracy to the geographic question. A DEM is particularly valuable because it converts elevation into a computational geographic surface from which many other variables can be derived. Its significance extends beyond displaying mountains and valleys. It provides a quantitative representation of terrain that allows GIS and other geospatial systems to model processes involving height, slope, visibility, drainage and relief.
