Cross-border coverage gap mapping for national sovereignty and security planning
Satellite-derived terrain and clutter data let regulators simulate mobile coverage on both sides of a land border, exposing zones of foreign signal dominance and domestic coverage absence before they become security or compliance problems.
Sensors
- TanDEM-X: Delivers a global digital elevation model at 12 m posting (TanDEM-X 12 m DEM) and a 90 m public release, with absolute vertical accuracy better than 10 m and relative accuracy around 2 m over moderate terrain. The fine-resolution product is the primary input for line-of-sight and diffraction calculations in border propagation models.
- Copernicus DEM GLO-30: Freely available at 30 m posting globally, derived from TanDEM-X data and edited for voids and water bodies. Suitable as a baseline terrain layer where the full 12 m commercial product is not licensed, with absolute vertical accuracy typically within 4 m over open terrain.
- GEDI (Global Ecosystem Dynamics Investigation): Spaceborne lidar on the ISS providing canopy height estimates at 25 m footprint spacing along discrete tracks between roughly 51.6° N and S latitude. Published canopy height products (L2A, L2B) give mean and maximum canopy heights used to construct vegetation clutter layers for rural propagation loss estimation.
- Sentinel-2 MSI: 10 m multispectral imagery at 5-day revisit (with both satellites). Land-cover classification derived from Sentinel-2 bands distinguishes built-up areas, forest, cropland and bare ground, providing the clutter category assignments that propagation engines require for diffraction and absorption coefficients.
Why a border is not just a line on a map
Radio frequency propagation ignores political boundaries entirely. A base station sited on a ridge two kilometres inside a neighbouring country can deliver a stronger signal into a valley town than any domestic tower, simply because terrain favours it. For a regulator, that is not an abstract concern: it means foreign operators may be providing de facto coverage in sovereign territory without a licence, without contributing to the universal service fund, and without subjecting their traffic to national lawful-intercept obligations.
The inverse problem is equally serious. A government that cannot demonstrate domestic coverage at its own borders has a weak position in spectrum coordination negotiations, and may be failing USF obligations to border communities. Quantifying both failure modes requires the same underlying analysis: a credible propagation model fed by accurate terrain and clutter data.
What the terrain data actually provides
Propagation engines such as ITM (Longley-Rice), SPLAT! or commercial variants like EDX SignalPro require two distinct inputs: a bare-earth digital terrain model for diffraction geometry, and a surface clutter model that adds height and absorption to the terrain profile. These are different things, and conflating them is a common source of error in coverage audits.
TanDEM-X at 12 m posting gives the terrain component with sufficient vertical fidelity to resolve ridge geometry that controls first Fresnel zone clearance at UHF and low-band LTE frequencies. The Copernicus DEM GLO-30, freely accessible, is adequate for rural border zones where terrain variation is gradual, though it will underestimate diffraction loss on sharp ridgelines. Neither product is a digital surface model: they require the vegetation and building layers to be added separately.
GEDI canopy height profiles fill the vegetation gap in forested border regions. Because GEDI samples along discrete orbital tracks rather than continuously, interpolation between tracks introduces uncertainty, particularly in heterogeneous forest. Published GEDI L2A products report root-mean-square height errors of roughly 2 to 5 m depending on canopy structure and slope. That range matters when a forest stand is 8 m tall and the propagation model's clutter table assigns meaningfully different loss coefficients at 5 m versus 10 m canopy height.
Sentinel-2 land-cover classification provides the spatial distribution of clutter categories across the full border zone at 10 m resolution. The classification is not the clutter height; it is the category label that a propagation engine maps to a height and loss coefficient from its internal tables. Accuracy of the classification, typically 85 to 90 per cent overall for well-trained models on Sentinel-2, propagates directly into uncertainty in the predicted coverage boundary.
Simulating the foreign signal footprint
Once terrain and clutter layers are assembled, the analysis proceeds from known or estimated tower locations on the foreign side of the border. In many cases, tower positions are partially observable from open sources: national regulatory databases, OpenCellID crowd-sourced data, or direct detection in high-resolution commercial imagery. Antenna heights, azimuths and transmitted power are often less certain, and the model should be run across a plausible parameter range rather than a single assumed configuration.
The output is a set of received-signal-level contours on the domestic side of the border. Zones where the foreign signal exceeds a minimum usable threshold, typically around minus 90 to minus 95 dBm for LTE, while domestic signal falls below that threshold, are the areas of effective foreign signal dominance. These are the zones a regulator needs to document, monitor and, if policy requires, address through domestic infill or interference coordination.
Honest limits of the method
Propagation modelling is prediction, not measurement. Even with 12 m terrain data and a well-trained clutter classification, predicted path loss at a specific point can differ from measured values by 6 to 12 dB in complex terrain, which translates to substantial uncertainty in the exact position of a coverage boundary. The analysis is most reliable for identifying broad zones of concern rather than precise metre-level boundaries.
Cloud cover does not affect the terrain or clutter inputs, since those are derived from SAR (TanDEM-X) and lidar (GEDI) rather than optical sensors. However, Sentinel-2 classification quality degrades in persistently cloudy regions, particularly tropical forest border zones, where cloud-free compositing may require aggregating imagery over months and the resulting land-cover map may not reflect current conditions after logging or agricultural clearance.
The method also cannot detect informal or temporary transmitters, nor can it account for ducting conditions that occasionally propagate signals far beyond their nominal coverage footprints. Atmospheric ducting is a separate analysis. For border zones where ducting is a documented phenomenon, propagation modelling from terrain data alone will underestimate the extent of foreign signal intrusion during anomalous propagation events.
From model to regulatory instrument
The practical deliverable for a regulator is not a raw propagation raster but a set of structured outputs: a coverage gap inventory with geographic coordinates and estimated population affected, a foreign-dominance zone map with signal-level confidence intervals, and a comparison layer showing the domestic coverage commitment from USF licence conditions against the modelled reality.
These outputs can be refreshed periodically as land cover changes, as new towers are built on either side of the border, or as spectrum assignments shift. The satellite-derived terrain and clutter inputs are stable on a multi-year timescale for terrain and a seasonal timescale for vegetation, so the main update trigger is changes to the tower inventory rather than the underlying geodata.
Satellize builds and maintains the terrain and clutter model stack as a reusable analytical asset, combining open Copernicus and GEDI data with licensed TanDEM-X products where resolution demands it. The same model architecture used in the Kingdom of Tonga crop-estimation programme, where terrain strongly conditions observable conditions on the ground, applies directly to border propagation work.
What a government should ask before commissioning this analysis
The quality of the output is bounded by the quality of the tower inventory on both sides of the border. If the foreign tower database is incomplete or out of date, the foreign-signal simulation will be incomplete. A realistic scope of work should state explicitly which tower data sources are being used and how gaps in the foreign inventory are handled, whether by conservative worst-case assumptions or by sensitivity analysis across a range of plausible configurations.
Governments should also specify the frequency bands of interest before the analysis begins. Propagation behaviour differs substantially between 700 MHz rural LTE, 1800 MHz urban LTE and 2600 MHz, and a clutter model calibrated for one band is not automatically valid for another. The deliverable should state the bands modelled and the propagation model variant used, so that the analysis can be independently audited or extended.
Typical figures
| Terrain model resolution (primary) | TanDEM-X 12 m posting; Copernicus DEM GLO-30 at 30 m (free tier) |
| Vertical accuracy (terrain) | Relative ~2 m, absolute <10 m (TanDEM-X 12 m); absolute typically <4 m (GLO-30) over open terrain |
| Clutter classification resolution | 10 m from Sentinel-2 MSI; overall accuracy typically 85–90% for trained land-cover models |
| Canopy height sampling (GEDI) | 25 m footprint diameter; discrete tracks; coverage between approximately 51.6° N and S latitude |
| GEDI canopy height accuracy | RMSE approximately 2–5 m depending on canopy structure and terrain slope (published L2A product) |
| Sentinel-2 revisit | 5 days at equator (two-satellite constellation); cloud-free composite may require 1–3 months in humid tropics |
| Frequency bands supported by propagation model | Model-dependent; typically 100 MHz to 6 GHz for ITM/Longley-Rice class engines; must be specified per engagement |
| Predicted path-loss uncertainty | Typically 6–12 dB at a specific point in complex terrain; coverage zone boundaries reliable to within ~500 m to 2 km |
| Archive depth (terrain/clutter inputs) | TanDEM-X DEM: single-epoch global product; Sentinel-2 archive from 2015; GEDI from 2019 |
| Deliverable formats | GeoTIFF coverage rasters, GeoPackage/Shapefile gap and dominance zone polygons, PDF regulatory report with confidence intervals |
Analytics Satellize can run
| Terrain and clutter model stack | Fusion of TanDEM-X or GLO-30 bare-earth DEM with Sentinel-2 land-cover classification and GEDI canopy heights; standard ITU-R P.452 or ITM clutter category mapping | GeoTIFF terrain and clutter rasters, ready for import into propagation engines; metadata report on input accuracy |
| Domestic coverage footprint simulation | Point-to-area propagation modelling (ITM/Longley-Rice or equivalent) from domestic tower inventory; signal-level raster at specified frequency bands | Received-signal-level GeoTIFF and coverage polygon layer with confidence-interval bands; USF compliance comparison table |
| Foreign signal dominance zone map | Propagation simulation from foreign tower inventory (open-source or licensed data); overlay with domestic signal raster to identify zones where foreign signal exceeds domestic at minimum usable threshold | Foreign-dominance polygon GIS layer with estimated population affected (gridded population overlay); PDF regulatory summary |
| Coverage gap inventory | Boolean intersection of domestic coverage absence zones with populated-place and road-network datasets; ranked by estimated affected population | Prioritised gap inventory table with coordinates, estimated population, and distance to nearest domestic tower |
| Sensitivity analysis across tower parameter uncertainty | Monte Carlo or scenario-based propagation runs varying foreign tower EIRP, antenna height and azimuth within plausible ranges; envelope of worst-case dominance extent | Worst-case and best-case dominance zone polygons; written assessment of key uncertainty drivers |
| Periodic update and change-detection report | Sentinel-2 land-cover change detection between epochs; re-run of propagation model where clutter or tower inventory has changed | Change-flagged update report on defined refresh cycle (quarterly or annual); updated GIS layers |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.