Informal border-crossing footpath and vehicle track mapping

Repeated foot and vehicle movement across ungoverned border segments leaves compaction and vegetation loss detectable in VHR optical imagery and SAR coherence products. Distinguishing smuggling routes from livestock paths requires endpoint analysis, not just track geometry.

Sensors

PlanetScope SuperDove: 3 m ground sample distance, 8 spectral bands including red-edge and NIR, daily global revisit. Sufficient to resolve individual vehicle tracks wider than roughly 2–3 m; narrower footpaths require stacking multiple acquisitions to confirm persistence.
Sentinel-1 C-band SAR: 10 m resolution in Interferometric Wide Swath mode, 6-day repeat at mid-latitudes, 12-day at the equator. Coherence differencing between successive passes detects surface disturbance caused by repeated vehicle movement, cloud-independent and night-capable.
Maxar WorldView-3: 30 cm panchromatic, 1.24 m multispectral, tasked on demand. Confirms track morphology, width, tyre-rut spacing and associated infrastructure (parked vehicles, staging areas) that lower-resolution data cannot resolve.
Airbus Pléiades Neo: 30 cm panchromatic, 1.2 m multispectral, stereo-capable for 3D terrain models. Useful for identifying track gradients and crossing-point engineering such as filled drainage ditches or cleared embankments.

What repeated movement does to ground that satellites can read

Every wheeled or foot crossing compacts the upper soil horizon, displacing loose material and reducing surface roughness. Vegetation is either trampled or deliberately cleared. Both changes alter how a surface reflects sunlight and scatters radar pulses. The signal is not dramatic, but it is consistent, and consistency is what multi-temporal analysis exploits.

In optical imagery, compacted bare soil has a higher reflectance in the red and SWIR bands relative to surrounding vegetated ground. The Normalised Difference Vegetation Index (NDVI) drops along active tracks, and the drop deepens with use frequency. PlanetScope SuperDove's red-edge band (approximately 705 nm) is particularly sensitive to early vegetation stress before full clearance occurs, giving a few days of lead time over a simple NDVI threshold. In SAR, C-band coherence between two Sentinel-1 passes is high over stable bare ground and low where the surface has been disturbed. A track used daily will show coherence loss relative to the surrounding undisturbed terrain, a pattern that persists as long as movement continues.

Where the published smuggling-route evidence points

UNODC's West Africa cocaine and cannabis trafficking assessments and IOM's Displacement Tracking Matrix data for the central Sahara both document corridor-level movement patterns across the Mali-Niger-Libya triangle, the Darién Gap in Central America, and the Thai-Myanmar border highlands. These published routes provide geographic priors: analysts do not scan entire borders blindly but concentrate processing on documented transit corridors and their likely variants.

In the Sahel, tracks typically converge on dry riverbeds (wadis) that offer firm crossings and natural concealment from ground observation. In Central America's Darién, movement is constrained by topography into a small number of ridge-line and river-valley corridors that IOM field teams have mapped on the ground. In South-East Asia, the Golden Triangle border zones show track networks radiating from known market towns. Each geography has a different spectral and SAR signature because soil type, vegetation cover and rainfall seasonality differ, so detection thresholds must be calibrated locally rather than applied universally.

The livestock ambiguity: why track geometry alone is not enough

This is the central analytical problem. Informal livestock tracks and smuggling paths are morphologically indistinguishable from above. Both show linear NDVI depression, soil compaction and SAR coherence loss. Both can be several kilometres long. A purely geometric classifier will produce a high false-positive rate in pastoral regions.

Disambiguation requires endpoint and context analysis. Livestock tracks terminate at water points, corrals, markets or seasonal grazing areas, all of which have their own satellite signatures: circular bare-ground patches around boreholes, rectangular enclosures, clustered settlement structures. Smuggling tracks tend to terminate at road junctions, river crossings, or locations that make no sense for animal husbandry but do make sense for vehicle transfer. Staging areas sometimes show parked vehicle shadows in WorldView-3 imagery. Temporal rhythm also helps: livestock movement is diurnal and seasonal; smuggling movement may show different temporal clustering, though this distinction is harder to confirm from optical data alone given cloud and acquisition gaps.

No automated classifier yet published achieves reliable separation without human review of endpoint context. Honest assessment: the satellite layer narrows the search area and ranks candidate tracks by activity level; it does not replace ground-truth verification.

SAR coherence differencing in practice: what the method actually resolves

Sentinel-1 coherence differencing compares two complex SAR images acquired 6 or 12 days apart. Where the surface is unchanged, phase relationships between pixels remain stable and coherence is high (approaching 1.0 over dry bare ground). Where vehicles have passed, soil roughness changes and coherence drops, sometimes to values below 0.3 over active tracks. The technique was demonstrated in published literature for road extraction in agricultural and semi-arid environments and has been applied to informal track detection in conflict-affected Sahel zones.

The practical resolution limit is Sentinel-1's 10 m pixel. A single-lane vehicle track of 3–4 m width will affect perhaps one pixel reliably; detection depends on the track being used repeatedly over the coherence interval, not just once. Narrow footpaths below roughly 5 m effective disturbed width are at the margin of reliable detection with Sentinel-1 alone. For those, PlanetScope stacking or WorldView-3 tasking is necessary. SAR also struggles in areas of high wind-driven sand movement, which creates coherence loss unrelated to human activity, a common problem in the central Sahara.

Building a monitoring workflow: cadence, archive and alert logic

A practical workflow for a border-zone client combines three layers. First, a Sentinel-1 coherence change product run on every available 6-day pair flags areas of new surface disturbance. Second, PlanetScope daily mosaics over flagged areas confirm optical signature and allow track geometry to be traced. Third, WorldView-3 or Pléiades Neo is tasked on the highest-priority detections for sub-metre confirmation and endpoint characterisation.

Archive depth matters. Sentinel-1 data runs back to 2014 for most regions, PlanetScope to approximately 2016 for many areas, and Landsat (at 30 m, useful only for established tracks) to the 1970s. Comparing current detections against the historical baseline distinguishes genuinely new tracks from long-established paths that have simply not been mapped before. New tracks appearing over weeks are operationally more significant than old ones.

Satellize structures this kind of layered workflow for government border-security clients, applying the same change-detection pipeline architecture used in the Tonga crop-estimation programme to a very different physical signal. Alerts can be configured as GIS polygon feeds or structured reports at weekly or fortnightly cadence, depending on the border length under surveillance and the client's analytical capacity to act on detections.

Honest limits and what they mean for operational use

Cloud cover is the most persistent constraint in Central America and South-East Asia. During monsoon months, optical imagery over the Darién or the Thai-Myanmar highlands may be unusable for weeks at a stretch. SAR fills much of this gap but cannot resolve the endpoint-context question that disambiguates livestock from smuggling routes. Clients in high-cloud regions should plan for a SAR-primary, optical-confirmatory workflow rather than the reverse.

Very lightly used tracks, perhaps crossed only a few times per month, may not produce detectable coherence loss or consistent NDVI depression within a single revisit cycle. Detection sensitivity improves with use frequency. A route used nightly by vehicles will be visible within one or two Sentinel-1 pairs; a route used weekly may take a month of stacking to confirm. This means the method is better suited to monitoring established corridors at scale than to catching a single crossing event.

Typical figures

Optical spatial resolution	3 m (PlanetScope SuperDove); 30 cm panchromatic (WorldView-3, Pléiades Neo)
SAR spatial resolution	10 m (Sentinel-1 IW mode)
Optical revisit	Daily global mosaic (PlanetScope); tasked within 1–3 days depending on latitude (WorldView-3, Pléiades Neo)
SAR revisit	6 days at mid-latitudes, 12 days near equator (Sentinel-1 single-satellite)
Minimum detectable track width	Approximately 3–4 m in PlanetScope; approximately 5 m effective disturbed width in Sentinel-1 coherence products
Spectral bands used	Red-edge (~705 nm), NIR, SWIR for vegetation stress and soil exposure; C-band (5.4 GHz) for SAR coherence
Cloud limitation	Optical unusable under dense cloud; SAR unaffected. High-cloud seasons in Darién and South-East Asia can suppress optical coverage for weeks
Archive depth	Sentinel-1 from 2014; PlanetScope from approximately 2016; Landsat 30 m from 1972 for baseline context
Latency from acquisition to alert	Sentinel-1 coherence product: 24–48 hours post-acquisition; PlanetScope optical: same-day to next-day mosaic availability
Delivery formats	GeoTIFF change layers, GeoJSON track polygons, structured PDF/HTML reports, API feed to client GIS

Analytics Satellize can run

SAR coherence change map	Sentinel-1 interferometric coherence differencing between successive 6- or 12-day pairs; coherence loss threshold calibrated per land-cover class	Fortnightly GeoTIFF layer flagging areas of new surface disturbance along defined border segments
NDVI depression track extraction	Time-series NDVI differencing on PlanetScope daily mosaics; linear feature extraction using morphological filtering to isolate track-width objects	GeoJSON polyline layer of candidate tracks ranked by persistence score and NDVI depression magnitude
Track activity timeline	Per-track NDVI and coherence time series extracted at fixed intervals; activity index computed as fraction of acquisitions showing depression above threshold	CSV or dashboard chart showing activity level per track segment over rolling 90-day window
Endpoint context classification	Object-based image analysis on WorldView-3 or Pléiades Neo imagery at track termini; classification of endpoint type (water point, corral, road junction, staging area) using published OBIA methods	Annotated imagery report with endpoint classification and confidence rating for each high-priority track
New-track emergence alert	Baseline comparison of current coherence and NDVI products against multi-year archive median; tracks absent from baseline flagged as newly emerged	Weekly alert feed (GeoJSON or email digest) listing newly detected track segments with coordinates and first-detection date
Cross-border corridor risk ranking	Spatial overlay of detected tracks against published UNODC and IOM corridor documentation; density and activity scoring per 10 km border segment	Quarterly corridor risk map in PDF and GIS format for border-management planning

Sources

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.