A Digital Terrain Model (DTM) is a digital representation of the bare-earth ground surface—the topography with vegetation, buildings, bridges, and other above-ground features stripped away. It is most commonly stored as a regular raster grid where each cell holds an elevation value.
Why it matters
The bare-earth surface is what most terrain analysis actually needs. Slope, aspect, hillshade, contours, drainage, and viewshed all assume you are modelling the ground, not the tree canopy or rooftops. Running these analyses on a surface that still contains buildings and forests produces spurious slopes, false ridges, and broken flow paths.
DTM vs DSM vs DEM
A DSM (Digital Surface Model) records the top of everything—canopy, roofs, structures. A DTM records the bare ground. DEM (Digital Elevation Model) is the umbrella term and is often used loosely; depending on the source it may mean either a DTM or a DSM, so always check the documentation. Subtracting a DTM from a DSM yields a normalized height model (e.g. a canopy height model).
A concrete example
LiDAR is the classic DTM source: the point cloud is classified into ground and non-ground returns, and only the ground points are interpolated to a grid—for example a 1 m bare-earth DTM. Many national mapping agencies publish 1 m or 2 m LiDAR DTMs. Photogrammetric and global products (SRTM, Copernicus GLO-30) are nearer to DSMs because they capture reflective surfaces rather than true ground.
Common pitfall
Treating any DEM as a DTM. Global "DEMs" like SRTM include canopy and building height in vegetated and urban areas, so using them where a bare-earth surface is required introduces systematic elevation bias.