Sensors
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands; 5-day revisit at mid-latitudes with both satellites. Provides NDVI time series for seasonal canopy phenology context and detects canopy-area change, but cannot resolve individual crowns narrower than roughly 10 m.
- TanDEM-X change DEM: Bistatic X-band SAR pair (0.5 m range resolution, 12 m posting in global DEM products). Height change products derived from multi-epoch acquisitions can detect canopy-top elevation shifts of roughly 1–2 m where coherence is maintained, though dense tropical canopy suppresses coherence.
- Planet SuperDove: 3 m resolution, 8-band multispectral, daily revisit globally. Resolves individual canopy crowns and detects lateral canopy spread at sub-pixel precision through change detection; commercial tasking adds on-demand acquisition.
- Airborne lidar repeat surveys: Point densities of 4–20 pts/m² in typical corridor surveys. Provides the highest-fidelity normalised canopy height model (nCHM) and direct measurement of Fresnel-zone clearance; repeat surveys at 1–5 year intervals quantify height delta with decimetric accuracy.
What the Fresnel zone actually demands of the terrain
A point-to-point microwave link does not travel as a geometric ray. Energy spreads into an ellipsoidal volume around the direct path, described by a series of Fresnel zones. For practical clearance planning, the first Fresnel zone radius at the midpoint of a 10 km link operating at 7 GHz is approximately 16 m. At 23 GHz the same geometry gives a radius of roughly 9 m. ITU-R P.530 recommends at least 60 per cent of the first Fresnel zone be kept clear of obstructions to avoid diffraction loss; full clearance is preferred.
That geometry is computed once at link commissioning, usually against a digital terrain model with no canopy component, or against a surface model that reflects conditions at a single point in time. Trees do not respect that snapshot. A species growing 0.5–1.5 m per year in a tropical or temperate climate can breach the clearance envelope of a 7 GHz link within a few seasons without triggering any obvious alarm until fade events begin accumulating.
Why the problem stays invisible for so long
Microwave fade has many causes: rain attenuation, atmospheric multipath, antenna misalignment, hardware ageing. Vegetation encroachment produces a slow, progressive increase in median path loss rather than the sharp, correlated rain-fade signature that network management systems are tuned to flag. The result is a link that drifts toward its fade margin over months or years, eventually triggering adaptive modulation step-downs or outage events that are logged as unexplained.
Ground inspection is the obvious remedy, but it is expensive on any network with hundreds or thousands of links spanning difficult terrain. A single field visit to a remote tower can cost more than several years of satellite monitoring. The economic case for remote sensing is therefore strongest precisely where the terrain is hardest to reach.
How multi-temporal imagery builds the detection chain
The analytic workflow starts with a computed Fresnel-zone clearance envelope for each link. Given the link coordinates, antenna heights, operating frequency and path profile from a baseline DSM, the envelope can be modelled as a series of cross-sectional ellipses at regular intervals along the path. Any terrain or canopy feature that intersects that envelope is a candidate obstruction.
Sentinel-2 NDVI time series then provides phenological context. Deciduous canopy that enters the Fresnel zone only in leaf-on conditions produces a seasonally varying fade signature; evergreen encroachment produces a monotonic one. Distinguishing the two matters because the mitigation responses differ: a deciduous encroachment may be tolerable for parts of the year, whereas an evergreen one demands action regardless of season.
Height change is the harder measurement. TanDEM-X change products can detect canopy-top elevation shifts at roughly 1–2 m sensitivity where X-band coherence holds, which is reasonable in temperate broadleaf or coniferous stands but degrades in dense tropical forest where volume scattering destroys interferometric coherence. For high-value links in those environments, Planet SuperDove at 3 m resolution can detect lateral canopy spread into the beam corridor even when height retrieval from radar is unreliable. Where the budget and the link criticality justify it, airborne lidar repeat surveys provide a normalised canopy height model accurate to tens of centimetres, directly comparable against the Fresnel envelope.
Honest limits of each approach
Sentinel-2 at 10 m cannot resolve individual canopy crowns narrower than one pixel. A single fast-growing tree with a crown diameter of 6–8 m will not appear as a discrete feature; it will contribute fractionally to a mixed pixel. This means Sentinel-2 is useful for detecting area-level canopy advance into a corridor but will miss isolated tall stems. Planet SuperDove at 3 m substantially reduces that blind spot but does not eliminate it for very narrow crowns.
Neither optical sensor penetrates cloud. In persistently cloudy tropical regions, months of imagery may be unusable. Compositing over a season recovers coverage but sacrifices temporal precision. SAR is cloud-independent but, as noted, loses height sensitivity in dense canopy. Airborne lidar is the most capable method and the most expensive; it is also a point-in-time measurement that requires re-flying to detect change. No single sensor solves all cases. A tiered approach, using Sentinel-2 for network-wide screening, Planet for flagged corridors, and lidar for the highest-criticality links, is more realistic than any single-sensor claim.
Turning detections into network decisions
A detection is only useful if it triggers a prioritised action. The analytic output for each link should express encroachment as a clearance deficit in metres relative to the 60 per cent first-Fresnel criterion, paired with an estimated growth rate derived from the multi-epoch height or NDVI time series. That combination supports two decisions: whether to schedule a vegetation management intervention, and whether to adjust the link's fade margin budget in the interim.
For a network operator managing several hundred links, the practical deliverable is a ranked list updated on a quarterly or annual cycle, with flagged links exceeding a defined clearance-deficit threshold. Satellize applies this workflow on open constellations including Sentinel-2, supplementing with commercial tasking where link criticality justifies the additional resolution, in the same way it structures agricultural analytics for programmes such as the Kingdom of Tonga crop-estimation work. The geometry and the physics are different; the multi-temporal change detection logic is not.
One detail worth noting for procurement: the baseline DSM quality determines everything downstream. A clearance envelope computed against a coarse terrain model with no canopy component will produce false positives wherever canopy already existed at commissioning. The first task in any serious programme is establishing a credible baseline, not simply running change detection against whatever elevation data is already in the network management system.
Typical figures
| Optical spatial resolution | 10 m (Sentinel-2), 3 m (Planet SuperDove) |
| SAR height product posting | 12 m (TanDEM-X global DEM); 0.25–1 m in airborne lidar corridor surveys |
| Revisit cadence | 5 days (Sentinel-2, dual satellite); daily (Planet SuperDove); on-demand (TanDEM-X tasking); 1–5 years typical for airborne lidar repeat |
| Minimum detectable canopy height change (SAR) | Approximately 1–2 m in temperate stands with adequate coherence; degrades significantly in dense tropical canopy |
| Minimum detectable canopy crown (optical) | ~10 m diameter (Sentinel-2); ~3–4 m diameter (Planet SuperDove) |
| Spectral bands used | Red, NIR, red-edge (NDVI, canopy phenology); X-band SAR (12.5 GHz, TanDEM-X) for height |
| Fresnel-zone frequency range addressed | Applicable to fixed links from ~2 GHz to 40 GHz; first-Fresnel radius shrinks with frequency |
| Archive depth | Sentinel-2: from 2015; Planet SuperDove: from 2021 at full constellation density; TanDEM-X: from 2010 |
| Cloud limitation | Optical sensors blind under cloud; SAR unaffected by cloud but coherence-limited in dense canopy |
| Delivery format | GeoTIFF height-delta rasters, GeoJSON link-corridor clearance reports, ranked CSV alert list |
Analytics Satellize can run
| Fresnel-zone clearance deficit map | Path profile extraction from DSM/nCHM against computed first-Fresnel ellipsoid at each link frequency and geometry | Per-link GeoJSON with clearance deficit in metres at worst-case cross-section, flagged against 60% first-Fresnel threshold |
| Canopy height change raster | Multi-epoch normalised DSM differencing (TanDEM-X or lidar nCHM); coherence-masked where SAR reliability is low | GeoTIFF height-delta layer clipped to link corridor buffers, with reliability mask |
| Canopy phenology classification | Sentinel-2 NDVI time series harmonic decomposition to distinguish evergreen, deciduous and mixed canopy within each link corridor | Per-corridor canopy-type classification table; seasonal fade-risk annotation on link report |
| Lateral canopy spread detection | Planet SuperDove multi-date change detection using normalised difference indices at 3 m resolution within Fresnel corridor buffer | Flagged corridor segments with estimated lateral encroachment rate (m/year) and confidence score |
| Prioritised intervention ranking | Composite scoring of clearance deficit, estimated growth rate and link traffic class (operator-supplied) | Ranked CSV of links by intervention urgency, updated quarterly or annually; compatible with standard network management import |
| Baseline DSM quality audit | Comparison of operator's existing path-profile data against current Sentinel-2 canopy extent and TanDEM-X surface model to identify stale baselines | Report identifying links whose commissioning-era elevation data predates significant canopy growth, with recommended re-survey priority |
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.