Digital elevation models represent terrain height as a raster surface — but the umbrella term hides a crucial split. A Digital Surface Model (DSM) records the top of everything: canopy, rooftops, bridges. A Digital Terrain Model (DTM) records bare earth with those features removed. Confusing the two corrupts flood models, visibility studies and volume calculations, so the distinction is anything but pedantic.
DSM, DTM and their derivatives
DSMs come naturally from first-return LiDAR or dense image matching; DTMs require filtering and editing to strip vegetation and structures down to ground. Subtracting DTM from DSM yields a normalised height model — canopy height for foresters, building heights for urban analysts.
From either surface flow standard derivatives: slope, aspect, curvature, hillshade, contours — each inheriting the parent surface’s character and errors.
How elevation models are produced
Sources span airborne and drone LiDAR (the accuracy benchmark), photogrammetric matching from stereo imagery, InSAR (the near-global SRTM and TanDEM-X datasets), and ground survey for local truth. National programmes like the USGS 3DEP are systematically replacing legacy models with LiDAR-derived terrain.
Production is edit-intensive: bridge decks removed or retained per specification, artefacts cleaned, voids filled, and water bodies treated deliberately.
Hydro-enforcement and fitness for purpose
Hydrological modelling needs terrain where water flows correctly: lakes flattened, rivers monotonically descending, culverts cut through road embankments so digital water is not dammed by them. This hydro-enforcement is a defined editing discipline, and its absence is the classic reason flood models misbehave on otherwise excellent DTMs.
Fitness for purpose is the buyer’s question: resolution, vertical accuracy at checkpoints, surface type and enforcement status — stated in metadata, verified in QA.