A DEM, or Digital Elevation Model, is a digital representation of the Earth’s surface elevation stored as a grid of height values. Each cell in the grid holds a single number representing the elevation at that location, and together these cells create a three-dimensional picture of the terrain. DEMs are one of the most fundamental datasets in GIS, used for everything from mapping flood risk to planning roads.
How a DEM Works
Think of a DEM as a spreadsheet draped over the landscape. The surface is divided into equally spaced squares (pixels), and each square stores one elevation value. The size of those squares is the spatial resolution. A 30-meter DEM, for example, assigns one elevation reading to every 30-by-30-meter patch of ground. A 12.5-meter DEM captures more detail because each pixel covers a smaller area. Higher resolution generally means better accuracy, especially in rugged or heavily vegetated terrain, though it also means larger file sizes.
Every elevation value in a DEM is measured relative to a vertical datum, which is essentially an agreed-upon zero point. Some DEMs reference an ellipsoid, a smooth mathematical model of Earth’s shape. Others use the geoid, a surface based on Earth’s gravity field that closely approximates mean sea level. The geoid can differ from the ellipsoid by up to 100 meters in places, so knowing which datum a DEM uses matters when you’re combining datasets or need precise heights.
DEM vs. DSM vs. DTM
DEM is often used as a catch-all term for any elevation data, but in practice there are important distinctions between three related models:
- Digital Elevation Model (DEM): Contains bare-earth elevations only, with buildings, trees, and other surface objects removed.
- Digital Surface Model (DSM): Captures the elevation of the highest surface a sensor encounters, including rooftops, tree canopies, and other structures. This is the “first return” surface.
- Digital Terrain Model (DTM): A refined version of a DEM, often manually corrected and sometimes enriched with additional features like breaklines along ridges and streams for higher precision.
If you’re analyzing urban building heights or forest canopy, you want a DSM. If you’re modeling water flow across the ground or calculating slope for construction, you need a DEM or DTM that strips away surface objects. The difference between a DSM and a DEM at the same location can be tens of meters in a city or dense forest.
Raster Grids and TINs
Most DEMs you’ll encounter are raster grids: a uniform matrix of square cells, each holding an elevation value. This format is simple, widely supported, and works well for large-area coverage. The tradeoff is that every cell is the same size regardless of how complex the terrain is, so flat plains get the same detail as mountain ridges.
The alternative is a Triangulated Irregular Network, or TIN. Instead of a uniform grid, a TIN connects irregularly spaced elevation points into a mesh of triangles. This lets you pack more detail into complex areas (steep slopes, narrow valleys) while using fewer points in flat areas. TINs tend to produce more realistic-looking terrain visualizations, but they can create artificial flat spots along ridgelines and valley floors where triangle faces meet at awkward angles. Raster grids generally produce more accurate slope measurements.
How DEMs Are Created
Three main technologies produce elevation data today, each with different strengths.
LiDAR (light detection and ranging) fires rapid laser pulses from an aircraft or satellite and measures how long each pulse takes to bounce back. It’s the gold standard for accuracy and can penetrate forest canopy to capture both the ground surface and the treetop surface in a single flight. Airborne LiDAR routinely achieves vertical accuracy within a few centimeters.
Radar-based methods, particularly Interferometric Synthetic Aperture Radar (InSAR), use satellite radar signals to map elevation. The advantage is coverage: radar works in all weather conditions, day or night. The Shuttle Radar Topography Mission (SRTM) used this approach to map nearly the entire globe. Accuracy is lower than LiDAR, typically in the range of a few meters vertically, but the near-global availability makes radar DEMs invaluable.
Photogrammetry extracts elevation from overlapping aerial or satellite photographs. By analyzing the same ground point from two slightly different camera angles, software can calculate its height. This method has been used for decades and remains common for regional mapping projects.
Raw elevation points from any of these sources are rarely perfectly spaced, so interpolation algorithms fill in the gaps. Common methods include Inverse Distance Weighted (which estimates unknown points based on nearby known points, giving more weight to closer ones), Kriging (a statistical approach that accounts for spatial patterns in the data), and Spline (which fits a smooth mathematical surface through the points). The choice of interpolation method affects the final DEM’s accuracy, particularly in areas with sparse input data.
Freely Available Global DEMs
Several global DEMs are available for free download, each with different resolution and accuracy characteristics. The most widely used include SRTM (available at 30-meter and 90-meter resolution), ASTER GDEM3 (30 meters), AW3D30 (30 meters), Copernicus DEM (30 meters), and TanDEM-X (12 meters).
Among these, TanDEM-X consistently shows the highest vertical accuracy. In comparative testing across varied terrain in the Andes, TanDEM-X achieved a root mean square error of about 3 meters, followed by AW3D30 at roughly 6 meters. The 90-meter SRTM performed worst, with errors above 11 meters even after correction. ASTER GDEM3 also tends to rank near the bottom for accuracy. If you’re choosing a free DEM for a project, TanDEM-X or AW3D30 are generally the strongest options, though availability and licensing terms vary by region and use case.
Accuracy drops as terrain gets steeper. Across all datasets, error increases with slope. Flat areas produce the most reliable elevation values, while mountainous terrain and dense vegetation introduce the most uncertainty.
What You Can Do With a DEM
A DEM on its own is just a grid of numbers, but GIS software can derive a wide range of useful products from it. Slope maps calculate the steepness of the terrain at every pixel, which matters for construction planning, agriculture, erosion risk, and landslide susceptibility. Aspect maps show the compass direction each patch of ground faces, which influences sunlight exposure, snowmelt patterns, and habitat suitability.
Hydrology is one of the most common DEM applications. By analyzing which direction water would flow from each cell, GIS software can delineate watersheds, extract stream networks, and calculate flow accumulation. Indices derived from these calculations, like the stream power index (which estimates erosion potential) and the topographic wetness index (which predicts where water accumulates), are standard tools in environmental science and land management.
Beyond these core products, DEMs support viewshed analysis (determining what’s visible from a given point), volume calculations for earthwork projects, 3D visualization, flight simulation, telecommunications tower placement, and flood inundation modeling. Nearly any GIS project that involves terrain relies on a DEM at some stage.
Resolution and Accuracy Tradeoffs
Choosing the right DEM comes down to matching resolution and accuracy to your project’s needs. A 30-meter DEM is adequate for regional watershed analysis or broad terrain classification. Site-level engineering, flood modeling in flat urban areas, or precision agriculture typically demands 1 to 5-meter LiDAR-derived data. Using a coarser DEM than your analysis requires can flatten out important terrain features, while an unnecessarily fine DEM slows processing without improving results.
Vertical accuracy matters just as much as horizontal resolution. A DEM with 12.5-meter pixels but 9-meter vertical error may actually be less useful for slope analysis than a 30-meter DEM with 3-meter vertical error. Always check the published accuracy metrics for any DEM before building an analysis around it, and be especially cautious in steep terrain where errors compound.