Open-source datasets like OpenStreetMap (OSM) and SRTM can support preliminary 5G planning and feasibility analysis, but they are generally insufficient for production network planning at 3.5 GHz and above. Key limitations include incomplete building height data in OSM (only 15 to 20% of OSM buildings have height attributes), SRTM's vertical RMSE of 5 to 10 m (too coarse for urban mmWave), inconsistent global coverage quality, and absence of the 3D vegetation and population data required for modern 5G workflows.
Several publicly available datasets are commonly evaluated for telecom use:
| Dataset | Type | Resolution | Key Limitation |
|---|---|---|---|
| SRTM (NASA) | DTM/DSM | 30 m | 5 to 10 m vertical RMSE; DSM not DTM |
| Copernicus DEM (GLO-30) | DTM | 30 m | 4 m RMSE; good globally, still coarse for urban |
| TanDEM-X 12 m | DSM | 12 m | Not freely available in full-res; DSM, not DTM |
| OpenStreetMap (OSM) | Road, building vectors | Varies | Missing building heights; inconsistent coverage |
| ESA WorldCover | LULC | 10 m | 10 classes only; no height; accuracy varies by class |
| Copernicus Land Cover | LULC | 100 m | Too coarse for 5G; 23 classes but coarse resolution |
| Global Human Settlement Layer (GHSL) | Population + building density | 100 m | Good for macro-scale demand analysis |
Before investing in commercial data, open-source datasets can support rough feasibility analysis: estimating how many macro sites might be needed, evaluating spectrum choices, or comparing deployment scenarios. SRTM plus Copernicus Land Cover provides a usable starting point for sub-1 GHz macro planning in rural areas.
In flat rural environments on 700 to 900 MHz, Copernicus DEM at 30 m resolution with ESA WorldCover LULC can support network planning. The propagation environment is relatively forgiving at these frequencies, and the coarser data does not introduce unacceptable errors for initial site placement.
Open-source data is appropriate for academic propagation research, algorithm development, and benchmarking, where data cost is a constraint and results are intended for comparative analysis rather than production deployment.
OSM road networks and building footprints are excellent for visualizing coverage maps and presenting results to stakeholders. Even where OSM building data lacks height attributes, the footprints provide useful geographic context.
OpenStreetMap contains building footprints contributed by community mappers. However, only approximately 15 to 20% of OSM buildings globally include height tags (height=, building:levels=). The vast majority are bare footprints without any vertical information.
For 5G RF planning, building data without height attributes is nearly as limited as no building data at all. The model cannot compute diffraction, LoS/NLoS classification, or building shadow zones without height.
Coverage quality also varies enormously by geography. OSM is well-mapped in Western Europe, North America, and Japan; coverage in sub-Saharan Africa, South and Southeast Asia, and parts of Latin America ranges from sparse to absent in specific cities.
SRTM (Shuttle Radar Topography Mission) is a DSM, not a DTM. It captures the top surface including vegetation and buildings, not the bare earth. For telecom DTM applications, SRTM requires building and vegetation filtering, which degrades its accuracy further.
More critically:
ESA WorldCover at 10 m resolution provides 11 land cover classes. Copernicus Land Cover at 100 m resolution provides 23 classes but at far too coarse a resolution for 5G planning. Neither dataset provides the 18 to 19 class telecom-optimized LULC classification that RF planning tools expect, nor the granularity to distinguish between adjacent urban sub-types (dense urban vs. urban vs. residential) that carry meaningfully different attenuation coefficients.
No publicly available open-source dataset provides individual tree polygons with height and canopy measurements globally. This is a significant gap for 5G planning because vegetation attenuation at 3.5 GHz and 26 GHz is not a minor correction. It is a primary propagation factor. A 12.8% improvement in 3.6 GHz prediction accuracy was observed simply by adding 3D tree data in a Barcelona study.
Population demand for 5G capacity planning requires time-of-day population density maps, daytime commercial zones versus nighttime residential distributions. This data does not exist in any open-source form at useful (10 m) resolution.
| Requirement | OSM + SRTM + WorldCover | Commercial (e.g., LuxCarta) |
|---|---|---|
| Building height data | Less than 20% coverage globally | 93%+ capture rate globally |
| Building height accuracy | N/A (mostly absent) | ≤2 m RMSE |
| Terrain vertical RMSE | 5 to 10 m (SRTM) | ≤1 to 2 m |
| LULC classification | 10 to 11 classes (open) | 18 to 19 telecom-optimized classes |
| LULC resolution | 10 to 100 m | 50 cm to 10 m |
| Vegetation height data | Not available | Individual tree heights |
| Population demand maps | Not available | 10 m, time-of-day |
| Data freshness | 2000 (SRTM) to 2021+ | Updated, on-demand |
| Global consistency | Highly variable | Consistent pipeline globally |
| Telecom tool compatibility | Requires reformatting | Delivered ready for Atoll, Planet |
For operators balancing cost and accuracy, a tiered approach can work:
This approach uses open-source data where it is adequate and commercial data where accuracy is commercially critical.
LuxCarta operates a globally scalable AI production pipeline that addresses precisely the gaps in open-source data: complete building extraction with 93%+ capture rate, accurate building heights (≤2 m RMSE), telecom-optimized LULC at 18 to 19 classes and 50 cm resolution, and individual 3D tree data with canopy and trunk separation. All of this is produced from satellite imagery, available globally, without dependency on community mapping coverage or government data release schedules.
For operators who have already used open-source data for initial planning and need to upgrade for production deployment, the BrightEarth platform enables targeted extraction for specific urban cores, transit corridors, or campus areas without ordering a full national dataset. Data is delivered in formats that plug directly into existing Atoll or Planet workflows without reformatting.
Yes, and this is a practical cost-saving option for sub-6 GHz urban projects where terrain variation is limited. Use SRTM (or Copernicus DEM) for the terrain base layer and supplement with commercial 3D building vectors and LULC for the urban areas where propagation accuracy matters most. For mmWave or hilly terrain, commercial DTM data at ≤2 m RMSE is preferable.
Yes. The Copernicus DEM (GLO-30) at 30 m resolution achieves approximately 4 m vertical RMSE, an improvement over SRTM's 5 to 10 m. It also uses more recent source data and has better coastal coverage. However, at 30 m resolution it still cannot support urban 3.5 GHz planning reliably. TanDEM-X 12 m is a better option if commercial purchase is acceptable; otherwise, Copernicus DEM is the best freely available global terrain source.
OSM building coverage is improving in some regions through community mapping campaigns and AI-assisted import projects. However, the fundamental challenge is not coverage but height data. Adding building heights to OSM at scale requires either structured surveys or automated extraction from imagery, which commercial operators do systematically. OSM height data coverage is unlikely to reach production-grade completeness in commercially important markets within a near-term planning horizon.
Many countries provide national geodata at no cost: Ordnance Survey in the UK, IGN in France, USGS in the US, and so on. Quality varies considerably. Where government datasets include accurate building heights and high-resolution DTMs, they can be excellent sources. The practical constraints are: they often require format conversion for telecom tool compatibility, coverage may stop at national borders requiring separate datasets for cross-border projects, and update schedules may not match deployment timelines.
AI-based automated extraction from satellite imagery is exactly how commercial providers like LuxCarta produce their data. The distinction between AI-generated and commercial data is increasingly minimal. The value of commercial data lies in the quality control, accuracy benchmarking, consistent classification, and telecom-ready formatting that accompanies the AI extraction pipeline, not in the AI itself.
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.