A digital surface model (DSM) is a 3D representation of the Earth’s surface that includes everything sitting on top of the ground: buildings, trees, vehicles, bridges, and any other elevated features. It captures the world as you’d see it from above, recording the height of the tallest object at every point across a landscape. This makes it fundamentally different from terrain models, which strip away all those features to show only bare ground.
How a DSM Differs From a DEM or DTM
Three terms come up constantly in mapping and geospatial work, and they’re easy to confuse. A digital surface model records the top of whatever is present at each location. In a forested area, that means the treetops. In a city, that means rooftops and overpasses. In an open field, it’s just the ground itself.
A digital terrain model (DTM) or digital elevation model (DEM) represents only the bare earth. To create one, aboveground features like trees, buildings, and vehicles are filtered out of the raw data, leaving a clean picture of the ground surface underneath. Think of it as what the landscape would look like if you removed every structure and cleared every tree.
The practical difference matters enormously. If you subtracted the terrain height from the surface height at any given point, the result tells you how tall the object at that spot is. Forestry researchers use exactly this method to measure tree height: they take the highest DSM value at a tree’s location and subtract the corresponding terrain elevation. Urban planners use the same math to map building heights across a city.
How DSMs Are Created
Two main technologies produce digital surface models: LiDAR and photogrammetry.
LiDAR (light detection and ranging) works by firing rapid laser pulses from an aircraft or drone toward the ground and measuring how long each pulse takes to bounce back. Each returning pulse becomes a point in a massive “point cloud,” a collection of millions of individual height measurements scattered across the landscape. A DSM is built from the first returns of this data. The first return is whatever the laser hits first on its way down, whether that’s a rooftop, the top leaf of a tree, or bare ground in an open area. These raw point clouds are stored in .las files, which can contain billions of individual points for a single survey area.
Photogrammetry takes a different approach, using overlapping photographs shot from slightly different angles to calculate depth, much like human eyes use two slightly offset viewpoints to perceive distance. Stereo pairs of satellite or aerial images are matched and processed to extract surface heights. Satellite-based photogrammetry can produce DSMs at roughly 1-meter resolution using high-resolution imagery with a ground sampling distance of 0.5 meters.
Common File Formats
If you’re downloading DSM data, the file format tells you what kind of processing has already been done. A .las file means you’re getting raw point cloud data that still needs to be converted into a usable surface. A .tif file (GeoTIFF) means someone has already processed the points into a raster, a grid of pixels where each pixel stores a single elevation value. Most DSM products distributed for general use come in raster format, since they’re far easier to work with in GIS software and don’t require specialized point cloud processing tools.
Accuracy and Resolution
The quality of a DSM depends heavily on how the data was collected. Aircraft-mounted LiDAR systems produce some of the most accurate results, with vertical accuracy around 5 to 7 centimeters and horizontal accuracy near 1 meter in high-quality surveys. Satellite-derived DSMs are less precise but cover much larger areas at lower cost. In comparative studies, airborne LiDAR DSMs achieved vertical accuracy around 0.67 meters against reference models, while satellite-based DSMs came in closer to 1.07 meters. That gap is significant for detailed engineering work but perfectly acceptable for regional planning or large-scale environmental monitoring.
Resolution, the size of each grid cell in the final raster, typically ranges from 0.5 to 2 meters for high-quality products. A 1-meter resolution DSM means each pixel covers a 1-by-1-meter square on the ground, which is detailed enough to distinguish individual buildings and large trees but won’t pick up a fire hydrant.
Flood Risk and Urban Planning
Flood modeling illustrates why DSMs exist alongside terrain models rather than replacing them. Bare-earth DEMs work well for mapping natural drainage patterns and river channels, but they completely miss the structures that redirect water in real life: levees, embankments, bridge decks, and retaining walls. A DEM-based flood simulation will miss the fact that an embankment raises local terrain by 1 to 2 meters, which can be the difference between a neighborhood flooding or staying dry.
Using a raw DSM in flood models creates the opposite problem. Trees and dense vegetation show up as solid barriers that block water flow in the simulation, even though real floodwater passes through forest and brush with some resistance. In one study, DSM-based simulations inflated surface roughness in areas with tree canopy, artificially reducing the modeled flow of water through wooded corridors along rivers.
The most effective approach combines both models. Researchers have developed hybrid methods that overlay structural features like embankments and bridges from the DSM onto the bare-earth DEM while excluding vegetation. This preserved the hydraulic influence of built infrastructure while eliminating false barriers caused by trees. The hybrid method captured localized flood depth increases of 1.5 to 2.0 meters behind embankments and reduced false flood predictions in vegetated areas by 12 to 18 percent compared to using a raw DSM alone.
Forestry and Vegetation Analysis
Measuring millions of trees by hand is impossible. DSMs make it practical. By calculating the difference between the DSM (which records the treetop canopy) and the DTM (which records the ground beneath), researchers produce what’s called a canopy height model. Every pixel in this derived product represents the height of the vegetation at that spot. Forest managers use canopy height models to estimate timber volume, monitor growth rates, track the spread of disease or pest damage, and assess wildfire fuel loads.
The same principle applies to agriculture. Comparing DSM-derived crop heights over time reveals growth patterns, identifies areas of stress where plants are shorter than expected, and helps optimize irrigation and fertilizer application.
Where to Find DSM Data
Many countries make DSM data freely available through government agencies. In the United States, the USGS 3D Elevation Program distributes LiDAR-derived surface and terrain models for much of the country. The National Ecological Observatory Network (NEON) provides high-resolution DSM, DTM, and canopy height data for ecological research sites. For global coverage, the Copernicus programme offers a 30-meter resolution DSM derived from satellite radar data, and several commercial providers sell sub-meter products for areas where public data isn’t detailed enough.
Before downloading, check what you’re actually getting. Some providers label bare-earth data as a “digital elevation model” when it’s technically a terrain product with surface features removed. Others bundle DSM and DTM together. Knowing the distinction helps you grab the right dataset for your project without wasting time processing data you don’t need.