Cross-border smuggling infrastructure and track mapping
Compacted soil, disturbed vegetation and metal infrastructure create spectral and radar signatures that persist long after a vehicle has passed. Sentinel-2 time-series and VHR optical imagery map informal tracks and border-crossing infrastructure used by overland smuggling networks.
Sensors
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands, 5-day revisit at mid-latitudes with both satellites. Detects soil disturbance and compaction as a spectral contrast against surrounding terrain; the SWIR bands (1610 nm and 2190 nm) are particularly sensitive to moisture and soil texture changes that accompany repeated vehicle passage. Free archive from 2015.
- Sentinel-1 SAR (C-band): 20–40 m ground range resolution in Interferometric Wide Swath mode, 6-day repeat. Metal culverts, fencing and corrugated-iron structures produce a characteristic linear double-bounce return in VV polarisation. SAR penetrates cloud and works at night, which matters in the Sahel wet season when optical data is routinely blocked for weeks at a time.
- Planet SkySat: 50 cm native resolution, tasked on demand. Resolves track width to within roughly 1 m, sufficient to distinguish a single-axle footpath (under 1 m) from a vehicle route (typically 2–4 m). Colour imagery distinguishes bare compacted soil from gravel surfacing or tyre-pressed vegetation. Revisit is task-dependent, not systematic.
- Airbus Pléiades Neo: 30 cm resolution in panchromatic, 1.2 m multispectral. At this scale, tyre-track impressions, spoil heaps beside excavated tunnel vents and surface-mounted ventilation pipes become directly interpretable. Stereo tasking produces a digital surface model accurate to roughly 50 cm vertically, useful for detecting low-relief earthworks.
What disturbed soil tells a spectrometer
Compacted vehicle tracks differ from surrounding terrain in three measurable ways. They are brighter in the visible bands because surface aggregates are crushed and the albedo rises. They are drier in the SWIR bands because compaction reduces infiltration and the soil moisture signal drops. And they are spectrally flatter: the textural variation of natural scrub or crusted desert pavement is replaced by a uniform, abraded surface.
Sentinel-2's Band 11 (1610 nm) and Band 12 (2190 nm) are well-suited to picking up this contrast. Published work on the US-Mexico border has used band-ratio indices combining SWIR and red-edge channels to isolate track pixels from background soil with reasonable reliability in arid and semi-arid terrain. The method degrades in vegetated areas where canopy closes over the track, and it struggles in sandy ergs where wind reworking erases the surface signature within days. Honest assessment: Sentinel-2 is a detection and change tool, not a precision mapping instrument. It tells you a track exists and roughly when it formed. It does not tell you how wide it is or what surface type it has.
SAR double-bounce and the metal infrastructure problem
Border smuggling infrastructure often includes metal: culverts beneath informal crossings, corrugated roofing on waypoint shelters, chain-link or wire fencing used to funnel movement. In C-band SAR, a horizontal metal surface adjacent to a vertical metal face produces a strong double-bounce return that is disproportionate to the physical size of the target. A 2 m culvert can return a signal that occupies several pixels in Sentinel-1 imagery.
This is useful. A culvert crossing a dry wadi that does not appear on any road database is an anomaly worth investigating. SAR-based change detection can flag new metal installations by comparing backscatter intensity across successive passes. The limit is spatial resolution: at 20 m, Sentinel-1 cannot resolve whether the return is a culvert, a water tank or a parked vehicle. Confirmation requires optical tasking. The combination of SAR alert followed by VHR optical task is the standard workflow, not a shortcut.
Time-series logic: when a track forms, not just where it is
A single image shows a track. A time-series shows when it appeared. That distinction matters for enforcement because a track that formed in the past six weeks is operationally active; one that formed four years ago may be abandoned or may have been absorbed into legitimate use.
Sentinel-2's five-day revisit, combined with the archive running back to 2015, allows analysts to date track formation events to within a fortnight in cloud-free conditions. The method is straightforward: compute a soil-exposure or brightness index for each available scene, then flag pixels that cross a threshold and remain above it across subsequent scenes. Transient disturbances (a single vehicle pass, a flood deposit) drop back below threshold. Persistent tracks do not. In the Sahel, cloud cover during the June-to-September wet season creates gaps of four to eight weeks that can obscure the precise formation date. SAR fills part of that gap but cannot replicate the spectral discrimination of optical data.
Resolving track type at VHR: foot path, motorcycle or truck
Track width is the primary discriminant between foot traffic and vehicle traffic. Foot paths in arid terrain are typically under 0.5 m wide. Motorcycle routes, common on the Balkans route and in the Sahel, run 0.5–1.2 m. Light-vehicle routes are 1.8–3 m. Heavy goods vehicles leave tracks above 3 m, often with visible tyre-rut separation.
Planet SkySat at 50 cm can resolve the 1.2 m threshold with reasonable confidence. Pléiades Neo at 30 cm can measure rut separation directly. Both systems require tasking, which costs money and introduces latency of one to several days depending on orbital geometry and cloud. Neither system offers the systematic revisit of Sentinel-2. The practical workflow is to use Sentinel-2 to prioritise areas of interest, then task VHR only on the pixels that matter. Attempting to survey an entire border corridor at VHR resolution is neither cost-effective nor necessary.
Tunnel-vent markers: the surface signature of subsurface infrastructure
Cross-border tunnels, documented extensively on the US-Mexico border by US law-enforcement agencies, require ventilation. Vent shafts produce surface markers: small spoil mounds, circular or rectangular disturbed-soil patches, and occasionally metal grates or pipes visible at VHR. Pléiades Neo stereo imagery can detect the low-relief earthwork of a spoil mound as small as 0.5 m in height if the sun angle is favourable and the surrounding terrain is flat.
The limit is that spoil mounds are easy to conceal with vegetation or tarpaulins, and a vent pipe of 15 cm diameter is below the resolution of any current commercial satellite. What satellite imagery can do is flag anomalous surface disturbance in a location where no registered construction activity is recorded. Verification requires ground investigation. Satellize treats this class of detection as a cueing product, not a confirmation product, and documents that distinction explicitly in every delivery.
Satellize's analytics pipeline applies the same change-detection and anomaly-flagging logic it developed for the Kingdom of Tonga crop-estimation programme to border-corridor monitoring: systematic, scene-by-scene comparison against a baseline, with human review before any alert is escalated to a client.
Coverage, latency and what the method cannot see
Sentinel-2 covers the entire land surface every five days. Processing a 100 km border segment to a track-change layer takes a few hours of compute once the pipeline is configured. Latency from satellite pass to delivered alert can be under 24 hours for Sentinel-2 data, which is published to the Copernicus Data Space within hours of acquisition.
VHR tasking latency is longer. Cloud, orbital geometry and competing task queues mean a requested Pléiades Neo image may arrive in one day or five. For time-sensitive operations, this is a real constraint. SAR fills the gap for metal-infrastructure detection but not for soil-spectral analysis.
The method has no purchase on smuggling that leaves no surface trace: goods moved on existing paved roads, through legitimate crossings, or by air. It is a tool for detecting informal infrastructure in terrain where that infrastructure creates a physical mark. Applied to the right geography and the right question, it is specific and auditable. Applied outside those conditions, it produces noise.
Typical figures
| Optical spatial resolution (Sentinel-2) | 10 m (visible/NIR), 20 m (SWIR) |
| Optical spatial resolution (VHR) | 30 cm (Pléiades Neo), 50 cm (Planet SkySat) |
| SAR spatial resolution (Sentinel-1 IW) | 20 × 22 m (range × azimuth) |
| Revisit (Sentinel-2, mid-latitudes) | 5 days (both satellites combined) |
| Revisit (Sentinel-1) | 6 days |
| VHR tasking latency | 1–5 days depending on cloud and queue |
| Minimum detectable track width (VHR) | ~0.5 m at 30 cm resolution; ~1.2 m at 50 cm resolution |
| Archive depth (Sentinel-2) | From 2015; Sentinel-1 from 2014 |
| Key spectral bands for soil disturbance | SWIR 1610 nm (B11) and 2190 nm (B12); Red-edge 740 nm (B6) |
| Delivery formats | GeoTIFF change layers, vector track centrelines (GeoJSON/Shapefile), dated alert reports (PDF) |
Analytics Satellize can run
| Track formation date map | Sentinel-2 SWIR band-ratio time-series; threshold-crossing persistence filter to exclude transient disturbance | GeoTIFF layer with per-pixel first-detection date; accompanying PDF summary for each border segment |
| Track width and surface-type classification | VHR optical object-based image analysis; width measured from edge-detected centreline buffers; surface type from colour and texture | Vector polyline file with width and surface-type attributes; confidence score per segment |
| Metal-infrastructure anomaly alerts | Sentinel-1 VV backscatter change detection; double-bounce signature flagging against road-database mask | Point alert feed (GeoJSON) with backscatter delta and acquisition date; updated on each Sentinel-1 pass |
| Tunnel-vent surface-marker candidates | Pléiades Neo stereo DSM differencing against baseline; low-relief mound detection combined with circular soil-disturbance flag in multispectral | Prioritised point list with stereo-derived height anomaly and optical thumbnail; explicitly labelled as cueing product |
| Corridor activity index | Rolling 30-day count of new track-pixel detections per 10 km border segment; normalised against cloud-free observation frequency | Weekly time-series chart and GIS heat-map layer; suitable for operational briefing |
| Historical network reconstruction | Full Sentinel-2 archive analysis from 2015; track persistence and abandonment events mapped per season | Multi-year GeoPackage with temporal attributes; analytical report on route-shift patterns |
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.