LiDAR, DEM, DTM and DSM: Choosing the Right Terrain Data

Quick answer

LiDAR is a measurement technology that can produce three-dimensional point clouds. DEM is a broad term for digital elevation data; DSM commonly represents the top surfaces of buildings, trees, and other objects, while a bare-earth terrain product attempts to represent the ground beneath removable surface features. DTM usage varies by organization and country, so the metadata—not the acronym alone—must establish what a dataset contains. The right product depends on the decision, required scale, dates, vegetation, accuracy, datum, and processing history.

What the topic means

A LiDAR sensor measures ranges to reflecting surfaces and records points with spatial coordinates and attributes. Classification may identify ground, vegetation, buildings, water, noise, and other categories. A gridded terrain surface can be derived from selected ground points; a surface model may retain trees and structures. Photogrammetry and other sensors can also create elevation models, so not every DEM is LiDAR-derived and not every LiDAR point cloud is a ready-to-use terrain model.

When it may be relevant

• Terrain, slope, drainage, watershed, corridor, earthwork, flood, erosion, or landform screening.
• Planning field traverses, survey control, geophysical lines, boreholes, monitoring, or access.
• Comparing topographic change where datasets have compatible reference, accuracy, date, and surface definition.
• Creating hillshade, contours, slope, aspect, flow-path, profile, or other spatial derivatives at an appropriate scale.

Useful information and inputs

• Dataset provider, product and version, acquisition date, sensor or production method, and license.
• Horizontal coordinate reference system, vertical datum, epoch where relevant, horizontal and vertical units, and geoid or transformation information.
• Point density or raster cell size, classification scheme, return information, breaklines, nodata, water treatment, and accuracy report.
• Project boundary, target feature size, required output scale, land cover, terrain steepness, and intended decision.
• Independent survey checkpoints, known benchmarks, imagery, field observations, and previous terrain products.

How the method or assessment generally works

The source is preserved and its metadata are checked before processing. The team identifies which points or surface definition fit the task, then defines the target grid, coordinate reference, vertical reference, cell size, extent, interpolation, nodata, and water or breakline handling. Reprojection, resampling, classification edits, gap filling, filtering, mosaicking, and smoothing are recorded because they change the output.

Derivatives such as hillshade, slope, flow direction, contours, profiles, or change surfaces are calculated only after the base surface passes reference, seam, void, and accuracy checks. Interpretation of a landform, hazard, drainage mechanism, or engineering condition is a separate step requiring suitable evidence and field correlation.

Typical outputs

• Source and processing manifest with dataset lineage.
• Controlled point-cloud subsets, bare-earth or surface grids, and map-ready previews.
• Hillshade, contours, slope, aspect, drainage, profile, or terrain-change derivatives.
• Coverage, void, uncertainty, and scale-limitation maps.
• Coordinate-aware figures and GIS layers for screening and field planning.

How the outputs should be interpreted

Cell size is not the same as positional or vertical accuracy, and resampling to a smaller cell does not create new information. A bare-earth model reflects a classification and interpolation process; it may not represent the true ground reliably beneath dense vegetation, water, buildings, bridges, or complex terrain. A DSM may be appropriate when roofs or canopy matter and misleading when ground drainage or slope geometry is the target.

QA/QC and evidence checks

• Verify CRS, horizontal and vertical datum, units, transform, extent, dimensions, resolution, nodata, and acquisition date.
• Inspect ground and non-ground classification, isolated points, vegetation remnants, bridge handling, water surfaces, voids, seams, and edge effects.
• Compare elevations with independent checkpoints or survey control appropriate to the required use.
• Preserve native resolution and document every reprojection, resampling, interpolation, fill, filter, and derivative algorithm.
• Inspect outputs numerically, spatially, and visually at the intended map and decision scale.

Limitations and common misunderstandings

LiDAR does not automatically see through all vegetation, water, roofs, or the ground surface. A hillshade can make noise or interpolation artifacts look like real landforms. Two elevation datasets cannot be subtracted defensibly unless their datums, surface definitions, dates, registration, accuracy, and uncertainty are compatible. Automated flow paths, lineaments, scarps, or change candidates are not field-verified conditions.

What may be needed for confirmation

Confirmation may include survey checkpoints, GNSS or total-station work, field mapping, drainage inspection, boreholes or test pits, repeat acquisition, higher-resolution data, alternative terrain processing, and review of source point classifications. Terrain-derived interpretations should be checked on the ground where they influence consequential decisions.

What to prepare before contacting HydroGeo

• The decision, project boundary, required scale, preferred output coordinate system, and target feature size.
• Original point-cloud or raster files with metadata rather than screenshots alone.
• Available survey control, benchmarks, imagery, previous maps, and field photographs.
• Known vegetation, water, access, date, resolution, datum, and accuracy constraints.