For 5G propagation modeling, terrain data accuracy requirements scale with frequency band. At sub-1 GHz, a DTM with 5 to 10 m vertical RMSE and 25 m horizontal resolution is generally acceptable. At 3.5 GHz, vertical RMSE should be ≤2 m at 5 to 10 m horizontal resolution. At 26 GHz mmWave, vertical RMSE ≤1 m at 1 to 2 m horizontal resolution is required. Errors beyond this threshold cause misclassification of LoS/NLoS conditions that no model calibration can correct.
Terrain data accuracy has two components:
Vertical accuracy (RMSE): How close the elevation value in the dataset is to the true ground height at that location. Expressed as Root Mean Square Error in meters. Lower is better.
Horizontal resolution: How large each grid cell is. A 5 m resolution DTM has one elevation value per 5 x 5 m area. Lower resolution means terrain features narrower than one cell are not captured.
Both matter, and they interact. A DTM with excellent vertical accuracy at 25 m resolution still fails to capture a ridge that is only 10 m wide. A DTM at 1 m resolution with poor vertical accuracy has dense sampling but unreliable elevation values at each point.
Every radio propagation calculation references the terrain along the path between transmitter and receiver. The terrain height at each point determines:
An error in terrain elevation shifts the calculated path profile, which shifts the computed propagation loss. In flat terrain, this matters less because small elevation errors have limited impact on the flat path profile. In hilly or mountainous terrain, a 5 m elevation error can change a LoS path to an NLoS path or vice versa, representing a discrete error of 15 to 30 dB.
For LoS propagation at any frequency, the signal requires not just direct LoS but clearance of the First Fresnel Zone, an ellipsoidal volume around the direct path. If an obstacle (building, hill, tree) enters the First Fresnel Zone, diffraction loss begins even before full blockage occurs.
The radius of the First Fresnel Zone at the midpoint of a path scales with wavelength. At 3.5 GHz (wavelength ≈ 8.6 cm), Fresnel zones are smaller than at 1800 MHz, meaning terrain features must be accurately captured at higher resolution to correctly compute Fresnel zone clearance. At 26 GHz (wavelength ≈ 1.15 cm), Fresnel zones are very small, making terrain accuracy even more critical for borderline LoS paths.
| Frequency | Vertical RMSE | Horizontal Resolution | Notes |
|---|---|---|---|
| Sub-1 GHz (700 to 900 MHz) | ≤5 to 10 m | 25 to 50 m | SRTM-grade acceptable for rural |
| 1800 to 2100 MHz (4G bands) | ≤3 to 5 m | 10 to 25 m | SRTM acceptable for macro, not urban |
| 3.5 GHz (n78) | ≤2 m | 5 to 10 m | DTM from satellite stereo or LiDAR |
| 26 GHz mmWave | ≤0.5 to 1 m | 1 to 2 m | LiDAR-grade or high-res satellite stereo |
These thresholds apply to the DTM (Digital Terrain Model, bare earth only). If only a DSM is available, accuracy requirements are effectively higher because the DSM conflates terrain, building, and vegetation heights.
Both the DTM and DSM play different roles in the propagation chain:
DTM (bare earth): Used to establish the ground elevation profile. Errors in DTM affect path loss calculations along terrain, particularly in hilly or undulating areas.
DSM or 3D buildings (surface features): Used to model building and vegetation obstruction above the terrain surface. Errors in building height affect diffraction calculations and LoS/NLoS classification.
For flat urban terrain (for example, a city on a coastal plain), the DTM is relatively stable and may require less frequent updates. The DSM or 3D building data, which changes as the built environment changes, is often the more critical accuracy investment.
For deployments in hilly urban environments (many European, Asian, and Latin American cities), both DTM accuracy and building height accuracy are critical.
If terrain height is consistently underestimated by 5 m in a hilly area, the model will predict more LoS coverage on hillsides than actually exists. The resulting coverage map shows service where there is none.
Drive testing is used to calibrate propagation model coefficients, adjusting the model until its predictions match measured signal levels. Calibration can correct for systematic statistical biases in the model, but it cannot fix a geometry error. If the DTM says a hill is 50 m tall and the real hill is 60 m tall, the model will get the right answer for the specific drive test route but will generalize incorrectly to other nearby paths.
For 5G FWA at 3.5 GHz or 26 GHz, terrain accuracy determines which premises are classified as qualifiable for service. A terrain error that misclassifies a borderline LoS premise as NLoS (or vice versa) produces a false fail or false pass, leading to either a missed revenue opportunity or a truck roll to a premise that cannot receive service.
Airborne LiDAR produces DTMs with vertical RMSE of 0.1 to 0.25 m at 0.5 to 2 m horizontal resolution. This is the highest-quality source available. LiDAR is cost-effective for small areas (cities, specific project zones) but prohibitively expensive for national or large-area coverage.
Sub-meter satellite imagery (e.g., WorldView, Pleiades) can be processed in stereo to produce DSMs at 0.5 to 2 m horizontal resolution with vertical RMSE of 1 to 3 m. After vegetation and building filtering, the resulting DTM has vertical accuracy in the 1 to 2 m RMSE range, sufficient for 3.5 GHz planning and near-sufficient for mmWave in many environments.
SRTM (Shuttle Radar Topography Mission) at 30 m resolution has vertical RMSE of approximately 5 to 10 m. It is the most commonly used global dataset for sub-1 GHz planning. The Copernicus DEM (GLO-30) at 30 m provides improved accuracy (vertical RMSE ≈ 4 m) and is openly available. TanDEM-X at 12 m resolution achieves vertical RMSE of ≈1.5 to 2 m and is the best radar-based option for 3.5 GHz planning.
A standard DTM derived from a DSM may contain elevation artifacts along rivers, lakes, and floodplains. Bridges get incorporated, water surfaces show irregular heights, and drainage patterns are inconsistent.
A hydro-enforced DTM (HDTM) applies hydrological correction to remove these artifacts: water bodies are flattened to consistent elevation, rivers follow downhill flow continuity, and bridge surfaces are removed from the terrain model. For RF planning, this matters because:
LuxCarta produces hydro-enforced DTM products specifically for applications where terrain consistency across water features is important.
LuxCarta delivers DTM and DSM products at multiple resolutions, from 10 m resolution DTMs for national-scale projects to sub-2 m resolution DSMs from high-resolution satellite stereo imagery. The DTM pipeline includes hydro-enforcement (HDTM) as a standard production option, ensuring clean terrain profiles across water bodies and drainage networks.
For the most demanding mmWave applications, LuxCarta can incorporate customer-provided LiDAR data as the terrain base and add LuxCarta's AI-extracted building and vegetation layers on top, combining best-in-class terrain accuracy with AI-scale building capture. All terrain products are delivered in GeoTIFF format at the target coordinate reference system for the planning project.
The BrightEarth platform allows on-demand DTM extraction for any area of interest, with immediate delivery in GeoTIFF format compatible with Forsk Atoll, InfoVista Planet, and other standard planning tools.
No. SRTM at 30 m horizontal resolution and approximately 5 to 10 m vertical RMSE is insufficient for mmWave propagation modeling. At 26 GHz, terrain errors of this magnitude translate directly into LoS/NLoS misclassification and large path loss prediction errors. Sub-2 m resolution terrain data with ≤1 m vertical RMSE is required for production mmWave network planning.
Yes, but the process requires careful filtering of buildings and vegetation from the DSM. Automated algorithms can perform this filtering, but the quality of the resulting DTM depends on the density and accuracy of building and vegetation removal. In dense urban areas, incomplete building removal leaves elevation artifacts at building footprint locations that distort terrain profiles. For critical mmWave planning, it is preferable to start with a purpose-built DTM rather than a filtered DSM.
Assess your DTM against a set of control points with known elevation (from GPS survey or high-accuracy LiDAR). Calculate RMSE across at least 50 to 100 control points distributed across the project area, including areas of varying terrain relief. Compare the resulting RMSE against the thresholds for your target frequency band.
Yes. Macro cells communicate over longer distances and across more terrain variation, so terrain accuracy affects the path profile across kilometers. Small cells at street level in flat urban areas may have nearly zero terrain variation within their coverage radius. Here, building and vegetation geometry matters far more than terrain accuracy. The DTM quality requirements for small cell planning in flat cities are therefore lower than for macro cells in hilly terrain.
LuxCarta provides AI-powered 3D geospatial data solutions for telecom, simulation, and smart city applications worldwide. Learn more at luxcarta.com or explore on-demand extraction at BrightEarth.