DTM vs DSM, explained for engineers

If you've scoped enough drone LiDAR projects you've watched both sides of a conversation talk past each other, the engineer asking for the DTM and the surveyor providing a DSM without anyone catching the mismatch until the drainage model produces flow paths through tree canopies. The conflation is common and consequential.

This article is the one-pager that puts the distinction beyond recovery. Definitions, geometry, the math behind the canopy height model, hydro-enforcement, and the spec details that catch out scopers — all in one place.

Definitions

A DTM — Digital Terrain Model is the bare-earth surface of the site. Imagine a perfectly contoured carpet stretched across the landscape, draped over the ground but passing under every tree, building, vehicle, fence and structure above. That carpet is the DTM.

A DSM — Digital Surface Model is the surface of everything-as- captured. The same site with the canopy tops, roof ridges, parapets, power-pole crossarms and fence lines all included as if you covered the whole thing in plastic-wrap. That is the DSM.

Both are commonly delivered as a raster (GeoTIFF), a TIN (LandXML) or a 12d / DWG surface. The file format and the resolution can be identical between them. What differs is which points from the underlying cloud got used to build the surface.

How each is computed

Both surfaces start from the same captured LiDAR cloud. The difference is in classification:

DTM construction:

1. Classify each return as ground / not-ground using algorithms like Cloth Simulation Filter, progressive TIN densification, or morphological filters
2. Apply manual review at edge cases (steep batters, bridges, dense canopy, water bodies)
3. Use only ground-classified returns as input to surface generation
4. Interpolate between ground returns to produce a continuous raster or TIN

DSM construction:

1. Take the highest return per cell (or for a TIN, the topmost return per node)
2. Combine ground points with canopy / structure tops
3. Generate the surface from this combined set

The DTM is the surface that survives ground classification. The DSM is the surface that doesn't filter anything out.

When you want a DTM

The bare-earth DTM is the engineering surface. You want it for basically every civil and hydrological workflow:

• Earthworks and cut/fill — design surfaces are compared against the existing ground; you need the ground itself, not the trees on top of it
• Hydrology and drainage modelling — water flows on bare earth, not on canopy. TUFLOW, HEC-RAS, MIKE all need DTMs (or hydro-enforced DTMs — more below)
• Contour generation — engineering drawings show contours of the ground, not contour lines wrapping every tree
• Volumetric calculations — pile volumes, excavation quantities, fill balances all reference the ground beneath features
• Road and rail design — formation grades, batter ties, drainage invert levels all sit on bare earth
• Property feasibility — buildable area, slope class, drainage paths, easement constraints

If your downstream user is an engineer, planner, hydrologist or designer, the deliverable they need is almost certainly the DTM.

When you want a DSM

The DSM has real uses too — they're just narrower:

• Vegetation work — canopy height, biomass, fuel-load class, clearance compliance. You can't measure trees from a DTM.
• Shadow studies — solar access, overshadowing analysis. The obstructions that cast shadow are everything-above-ground.
• Line-of-sight analysis — where can you see from / to? Driven by the DSM, not the bare ground beneath.
• Building / asset inventory — roof shapes, structure heights, parapet positions. All sit in the DSM.
• 3D visualisation — for stakeholder briefings, marketing, planning approvals, the DSM is what makes the site look like the site.

If your downstream user is an arborist, ecologist, asset manager, fire planner, urban designer or visualisation specialist, the DSM is probably what they want — sometimes in addition to a DTM, often not instead.

The canopy height model — DSM minus DTM

If you have both, you can subtract one from the other:

CHM = DSM − DTM

The result, the Canopy Height Model, is a per-cell measure of how much stuff sits above the ground at every point of the site. It's the dominant output in vegetation and fire-management workflows:

• Tree-height inventory at the cell level
• Biomass estimation calibrated against species and density
• Fuel-load class for fire-behaviour modelling
• Encroachment distance from canopy to powerline conductors
• Vegetation change detection between epochs

A capture that produces both DTM and DSM gives you the CHM for free — it's a simple raster subtraction in QGIS or ArcGIS Pro.

Note that "canopy" in CHM is a slight misnomer: the model includes everything-above-ground, vegetation and structures alike. For pure vegetation work, you typically also produce a separate building mask and subtract that out.

The interpolation problem

DTM and DSM both interpolate between the captured points to produce a continuous surface. The interpolation behaviour matters more than people typically realise.

For a DSM, interpolation is usually fine — the surface points are densely sampled (canopy and structures are right where the laser hits first) and gaps are small.

For a DTM, interpolation is the loaded question. Ground returns beneath dense canopy may be sparse — 1–2 returns per square metre where the canopy is heavy, vs 100+ returns per square metre on bare ground. The DTM surface across those sparse areas is interpolated across larger gaps, with corresponding loss of detail and increased uncertainty.

Three implications worth knowing about:

1. Local DTM accuracy varies with canopy cover. A capture that reports 20 mm RMSE on bare-ground checkpoints may genuinely be 60–80 mm under dense canopy. A reputable vendor reports residuals stratified by cover class.

2. Small surface features can disappear. Drainage swales, table-drain edges, small ridges and erosion features need ground returns close enough together to resolve them. Where the ground point density drops, these features get smoothed out of the DTM.

3. Higher capture point density rescues sparse areas. This is why projects that need engineering-grade DTM under canopy fly at higher point density than equivalent bare-ground captures — you're spending pulses on the expectation that most of them won't reach the ground.

Hydro-enforcement — a DTM-specific concept

A standard DTM has a problem at structures. Take a culvert: water should flow through the culvert (the ground level inside the culvert is the invert, well below the road surface above). A standard DTM-from-LiDAR sees the road deck above the culvert and records it as ground (because it is ground — the road surface). The DTM, used as-is in a hydrology model, would route flood water over the road instead of through the culvert.

The fix is hydro-enforcement: explicit modification of the DTM at structures to honour the actual flow paths:

• Culverts represented as breaklines along the invert
• Bridges represented with deck heights for clearance analysis but hydrologic flow at the natural ground beneath
• Drainage pipe inlets and outlets explicit
• Weirs, sluices and other hydraulic structures called out as features

A hydro-enforced DTM (sometimes labelled "DEM" rather than "DTM" in hydrology software, confusingly) is what 1D and 2D models actually need. A capture that produces a "DTM" without hydro-enforcement is fine for civil design but wrong for flood work. Worth specifying explicitly at scoping.

Adjacent acronyms

The terrain-surface space has more acronyms than makes sense. Quick clarifying glossary:

• DEM — Digital Elevation Model. A generic term that can mean either DTM or DSM depending on context and discipline. Hydrology uses "DEM" to mean what surveyors call "DTM". Always ask which one a "DEM" actually is.
• nDSM / CHM — normalised DSM, same as the Canopy Height Model. Height above ground rather than above sea level.
• TIN — Triangulated Irregular Network. A way of representing a DTM or DSM as a mesh of triangles rather than a regular raster. Better for engineering, larger files, slower to display.
• bare-earth model — informal synonym for DTM, common in vegetation and environmental work.

How to specify the deliverable correctly

Three things worth being explicit about at scoping to avoid mismatch:

1. DTM, DSM or both — and which file format (GeoTIFF for raster, LandXML for TIN, 12da for 12d, DWG for AutoCAD/Civil 3D)
2. Grid resolution — 0.1 m, 0.25 m, 0.5 m, 1 m, etc. Smaller resolution = larger files, longer processing, no automatic improvement in accuracy beyond what the cloud supports
3. Hydro-enforcement — if any flood, drainage, stormwater or hydrology work will run from the DTM, specify hydro-enforcement at scoping with a breakline list

A clean scoping note for a typical civil capture might read: "Deliverables: hydro-enforced DTM at 0.5 m grid in GeoTIFF + LandXML; DSM at 0.5 m in GeoTIFF for visualisation. Hydro-enforce all culverts and bridges within the project boundary."

That sentence eliminates 80% of the ambiguity that drives delivery re-work.

TL;DR

| Surface | What's in it | Engineering use | | ------- | ------------------------ | ------------------------------ | | DTM | Bare-earth ground only | Civil, hydrology, contours, volumetrics | | DSM | Ground + canopy + structures | Vegetation, shadow, line-of-sight, visualisation | | CHM | DSM − DTM | Tree heights, biomass, fuel load, clearance |

Most engineering projects need the DTM, sometimes hydro-enforced. Vegetation, environmental and visualisation projects need both. Specify clearly at scoping and the deliverable lands right.