Digital elevation models: DEM, DTM and DSM — differences, sources and when to use each in Colombia

Almost any serious geospatial project eventually needs to know the height of the ground: to compute slopes, model floods, plan a road or estimate volumes. That's where three acronyms get thrown around as if they were interchangeable —DEM, DTM and DSM— when they are not. Confusing them leads to expensive mistakes: a flood analysis run on the wrong model can underestimate how much water reaches a neighborhood. This note clears up the difference, reviews the sources available for Colombia, and offers guidance on when to use each.

The three acronyms, without ambiguity

An elevation model is a numerical representation of the height of a surface, usually as a grid of cells where each cell stores an altitude value. The key question is: the height of which surface?

| Acronym | What it measures | Includes buildings and trees | |---|---|---| | DSM (Digital Surface Model) | The surface as the sensor "sees" it from above | Yes — rooftops, tree canopy, infrastructure | | DTM (Digital Terrain Model) | Bare earth, the ground beneath vegetation and structures | No — ground only | | DEM (Digital Elevation Model) | A generic/umbrella term; in practice often means terrain | Depends on the source |

The confusing one is DEM. In technical literature it's an umbrella term covering both DSM and DTM. In everyday Colombian usage —and in many catalogs— "DEM" usually means the terrain model. The sensible habit: when someone hands you a "DEM", ask whether it's surface or terrain, because the difference between the two can be 20 or 30 meters in a forested area.

Why the difference matters in practice

Picture a rural parcel with a 25-meter-tall forest stand.

The DSM records the canopy: ~25 m above the real ground.
The DTM records the soil beneath the forest.

If you use the DSM to model where rainwater flows, the model will "believe" there's a 25-meter hill where there are only trees, and divert the flow unrealistically. For hydrology, slopes, roads and physical cadastre you use a DTM. For building-height analysis, line-of-sight, telecommunications or shadows, you use a DSM. The difference between the two (DSM − DTM) is precisely the object-height model (nDSM, or CHM for vegetation), which is very useful for estimating crop height, biomass or urban volumetrics.

Sources available for Colombia

Not every project needs —or can afford— a LiDAR survey. The good news is there's a ladder of sources, from free and global to proprietary and extremely precise.

Free global sources

SRTM (30 m). The NASA classic. Near-complete coverage of Colombia at 1 arc-second (~30 m). It's a DSM (it captured canopy and rooftops), though it's used as a terrain approximation at regional scales. Good for watershed analysis, not for parcel detail.
Copernicus DEM — COP30 / COP90. Derived from TanDEM-X, today the best free global option: cleaner than SRTM, with good consistency. It's a DSM. COP30 (~30 m) is the general-purpose tier.
ALOS World 3D — AW3D30 (30 m). From JAXA, also a DSM, sometimes with sharper edge definition than SRTM in mountainous terrain. Worth comparing for Andean areas.
NASADEM. A modern reprocessing of SRTM with better quality; a natural replacement for the original SRTM.

For much of Colombia, the mountain ranges and persistent cloud cover make it worth comparing two or three 30 m sources before choosing, because none is uniformly better across the whole country.

National sources

The IGAC and the Colombian Spatial Data Infrastructure (ICDE) publish elevation models derived from national photogrammetric flights, with better resolution than global sources in the areas already covered. For cadastral and land-planning projects it's always worth checking first what official inputs exist for the municipality, because they can save you a survey.

High-resolution (proprietary) sources

Drone photogrammetry. Produces very high-resolution DSMs (centimeters) and, with proper processing, a DTM. Cost-effective over areas from tens to a few thousand hectares. Limited under dense vegetation: the camera can't see the ground.
LiDAR (airborne or drone). The gold standard for a DTM under forest, because the laser pulse penetrates the vegetation and lets you classify "ground" vs "non-ground". It's the only reliable way to get bare earth under jungle. More expensive, justified when vertical accuracy is critical.

How to choose: a quick guide by use case

| Need | Model | Typical recommended source | |---|---|---| | Watershed / regional analysis | DTM/DEM | COP30, NASADEM (free) | | Urban flood modeling | High-resolution DTM | LiDAR or drone photogrammetry | | Slopes for rural cadastre | DTM | IGAC / photogrammetry | | Preliminary road design | DTM | Drone photogrammetry + ground control | | Building height / volumetrics | DSM (and nDSM) | Drone photogrammetry, LiDAR | | Crop height / biomass | CHM (DSM − DTM) | LiDAR, multitemporal drone | | Line-of-sight / telecom | DSM | COP30 (regional), LiDAR (detail) |

The practical rule: start with the most suitable free source and step up only when the case justifies it. A lot of money is wasted commissioning LiDAR for analyses a COP30 would have solved, and conversely, risk decisions get made on a 30 m SRTM when the problem demanded a centimeter-grade DTM.

Common mistakes worth avoiding

Using a DSM as if it were a DTM in hydrology or slope analysis. The most common —and the most costly.
Ignoring the vertical reference system. In Colombia the horizontal frame is MAGNA-SIRGAS, but heights may come as ellipsoidal or orthometric (above the geoid). Mixing them introduces errors of several meters. Always confirm the vertical datum before combining sources.
Comparing resolutions without validating real accuracy. A 30 m model isn't always worse than a resampled "12 m" one; cell resolution is not the same as accuracy.
Not reporting the model's date. Terrain changes: mining, urbanization, landslides. A DEM from ten years ago may no longer represent reality.

GeoSAT's role

In our photogrammetry, cadastre and territorial-analysis projects, choosing the right elevation model is one of the first technical decisions, and often the one with the biggest impact on the result. Over 30 years of experience we've integrated everything from free global sources to LiDAR and drone surveys, according to what each project actually needed —no more, no less.

If you're weighing which elevation model to use for a flood, road, cadastre or agriculture project, the most useful first step is to define the vertical accuracy the decision requires. Get in touch and we'll size it with you.