Border-zone ground-disturbance and incursion monitoring
SAR backscatter change and high-resolution optical differencing can detect vehicle tracks, berm construction, trench digging, and vegetation clearance along contested borders, often within hours of the event. Each method has hard limits that analysts must account for.
Sensors
- Sentinel-1 SAR (IW mode): C-band (5.405 GHz), 10 m ground range resolution in Interferometric Wide swath mode, 250 km swath, 6-day repeat at the equator (12-day for a single satellite). Bare-earth backscatter is highly sensitive to surface roughness changes of a few centimetres, making it well suited to arid border terrain. Cloud and darkness are irrelevant to the sensor.
- Planet Dove (PlanetScope): 3–4 m resolution, near-daily global revisit across the constellation of roughly 200 smallsats. Four to eight multispectral bands depending on generation. Useful for rapid optical confirmation of SAR-flagged disturbance and for detecting vegetation clearance via NDVI differencing. Blocked by cloud; no night capability.
- Airbus Pléiades Neo: 30 cm native resolution, four multispectral plus panchromatic bands, tasked on demand with same-day or next-day delivery in many regions. At this resolution, individual vehicle tyre tracks, shallow trenches, and berm scarps are visually resolvable. Swath is 14 km, so targeted tasking requires accurate prior cueing from wider-area sensors.
- Maxar WorldView-3: 31 cm panchromatic, 1.24 m multispectral, plus 16 SWIR bands. The SWIR channels can distinguish freshly disturbed mineral soil from surrounding undisturbed surface by reflectance signature, adding a spectral confirmation layer beyond visible-band differencing. Tasked commercially; latency depends on contract priority.
What a radar sees when someone drives across a desert
C-band radar backscatter is governed by surface roughness at centimetre scales. Undisturbed arid terrain, compacted by wind and time, returns a predictable, low-variance signal. A vehicle crossing that surface crushes and displaces soil aggregates, raises fine dust that resettles unevenly, and leaves tyre-pressure depressions that alter the local dielectric geometry. The result is a measurable backscatter anomaly that persists for days to weeks in dry conditions, depending on wind and subsequent traffic.
Sentinel-1's IW mode, with its 10 m pixels and 6-day repeat cycle, is the workhorse for this kind of monitoring at scale. A coherent change detection approach, comparing amplitude or phase between successive passes, can flag disturbance patches well below the nominal pixel footprint when the change is spatially coherent across multiple pixels. Published research on arid-zone SAR monitoring has demonstrated detection of disturbed areas as small as a few hundred square metres using Sentinel-1 time-series, though the exact threshold depends on soil moisture, surface composition, and the magnitude of the disturbance.
Optical differencing: what it adds and where it fails
High-resolution optical imagery from Planet, Pléiades Neo, or WorldView-3 provides a different kind of evidence. Where SAR flags a backscatter anomaly, a 30 cm optical image can resolve whether the cause is a vehicle track, a scraped berm, a cleared treeline, or an artefact. The spectral contrast between freshly exposed subsoil and the surrounding surface is often large enough to be visible even at 3 m resolution, particularly in arid terrain where vegetation is sparse and soil colour is distinctive.
The limits are not subtle. Dense forest canopy blocks optical sensors entirely, and it substantially reduces SAR sensitivity to sub-canopy activity. C-band radar penetrates only a few centimetres of dry soil and a few tens of centimetres into very dry sand; it does not see through a closed forest canopy. In forested border zones, neither method reliably detects shallow trench digging or personnel movement beneath the trees. Cloud cover is a further constraint on optical sensors: in tropical or monsoon-affected border regions, optical revisit can effectively drop to once every several weeks during wet seasons. SAR remains available through cloud, which is precisely why the two methods are used together rather than separately.
Berm construction and trench digging: the temporal signature
Earthwork construction has a distinctive temporal signature in SAR time-series. A berm being built over days or weeks produces a progressive backscatter increase along its footprint as loose, rough material accumulates. Once construction stops and the surface stabilises, backscatter gradually returns toward background levels. Trench digging creates a different pattern: a linear low-backscatter feature (the trench floor, which is smoother and shadowed) flanked by higher-backscatter spoil heaps on either side.
Detecting these features reliably requires a baseline archive. Without several months of pre-event imagery, distinguishing a new berm from a pre-existing track or field boundary is difficult. Sentinel-1's free archive, which extends back to 2014, provides that baseline for most border regions globally. Commercial SAR constellations such as ICEYE or Capella can reduce revisit to hours when rapid cueing is required, though those are separate procurement decisions beyond the open-data stack.
Vegetation clearance as an early indicator
Before heavy earthworks begin, border-zone activity often starts with vegetation clearance: scrub cut for sight lines, trees felled for access tracks, or ground stripped for vehicle staging areas. This is detectable earlier than the earthworks themselves, and at lower resolution.
NDVI differencing on Planet Dove imagery can flag clearance patches of a few hectares within a day or two of the event in cloud-free conditions. The signal is unambiguous in semi-arid or savanna terrain where natural vegetation loss at that rate and in that geometry is not expected. In temperate or tropical zones, seasonal senescence and agricultural clearing create false-positive noise that requires careful masking using historical phenology data. The masking is tractable but adds processing time and requires a multi-year baseline to calibrate.
Putting the methods together: a practical monitoring architecture
No single sensor solves this problem. A practical architecture uses Sentinel-1 as the persistent, all-weather change-detection layer, running automated backscatter-difference alerts on a 6-day cadence across the full border zone. Planet Dove provides near-daily optical context for cloud-free periods, allowing rapid visual triage of SAR alerts. Pléiades Neo or WorldView-3 is tasked on demand for confirmed anomalies where sub-metre resolution is needed to characterise the activity type and scale.
The alert latency from event to analyst report depends on the Sentinel-1 acquisition schedule for the specific orbit track, processing time, and whether commercial optical tasking is triggered. In practice, for a border zone with regular Sentinel-1 coverage, a significant earthwork event can be flagged within 6 to 12 days using open data alone, or within 24 to 48 hours if commercial SAR is on standing task. That latency is honest and matters for operational planning.
Satellize runs this kind of multi-source change-detection stack for government clients, combining open constellation data with commercial tasking on client licence. The same analytical pipeline that underpins the Tonga crop-estimation programme, adapted for spectral and temporal change rather than crop phenology, forms the basis of border-monitoring products. Analysts reviewing outputs via the Overhead column will recognise the methodology.
What the archive cannot tell you
Satellite imagery confirms that ground was disturbed. It does not confirm who disturbed it, or why, without corroborating intelligence. A backscatter anomaly consistent with a vehicle track is exactly that: consistent with. It could be military, smuggling, herding, or maintenance. Characterisation requires convergence of geometry, timing, context, and sometimes spectral analysis of associated features such as tyre width or equipment shadows at sub-metre resolution.
There is also a resolution floor below which ground truth is unavailable from space. A single person walking across sand leaves no detectable signature in any current commercial system. A squad on foot in scrub is similarly invisible. The methods described here are sensitive to vehicle-scale and earthwork-scale activity, not to individual personnel movement. Buyers who expect otherwise will be disappointed, and any vendor who promises otherwise is overstating the physics.
Typical figures
| SAR spatial resolution (Sentinel-1 IW) | 10 m (range) × 10 m (azimuth) ground projected |
| Optical resolution range | 3–4 m (Planet Dove) to 30–31 cm (Pléiades Neo, WorldView-3) |
| Sentinel-1 revisit (single satellite) | 12 days at equator; 6 days with both satellites; shorter at higher latitudes |
| Planet Dove revisit | Near-daily global, cloud-permitting |
| Minimum detectable disturbance (SAR, arid terrain) | Approximately a few hundred square metres for coherent earthwork; individual tracks detectable when multi-pixel coherent |
| SAR frequency | C-band, 5.405 GHz (Sentinel-1); penetrates cloud and darkness, not closed forest canopy |
| Sentinel-1 archive depth | 2014 to present (global coverage varies by year) |
| Alert latency (open data stack) | 6–12 days from event to flagged anomaly; 24–48 hours with commercial SAR on standing task |
| Coverage per acquisition (Sentinel-1 IW) | 250 km swath; suitable for full border-zone coverage in a single pass |
| Delivery formats | GeoTIFF change maps, GeoJSON alert polygons, PDF analyst reports, GIS-compatible vector layers |
Analytics Satellize can run
| SAR backscatter change alert | Amplitude change detection on Sentinel-1 IW time-series; log-ratio differencing between current and N-day baseline composite | GeoJSON polygon feed of flagged disturbance patches, updated each Sentinel-1 pass, with backscatter-change magnitude and confidence score |
| Coherent change detection map | Interferometric coherence differencing (InSAR) between successive Sentinel-1 passes; coherence loss indicates surface change | GeoTIFF coherence-difference layer with analyst annotation of high-confidence disturbance zones |
| Vegetation clearance alert | NDVI differencing on Planet Dove time-series with multi-year phenology baseline masking to suppress seasonal false positives | Weekly GeoJSON layer of newly cleared vegetation patches above configurable area threshold |
| Earthwork characterisation report | Sub-metre optical analysis (Pléiades Neo or WorldView-3) of SAR-cued anomalies; feature measurement, shadow-based height estimation, soil-colour spectral classification | PDF analyst report with annotated imagery, feature dimensions, estimated construction timeline, and activity-type assessment |
| Border-zone baseline profile | Multi-year Sentinel-1 and Landsat archive processing to establish normal backscatter and spectral envelopes for each terrain segment along the monitored border | GIS layer of baseline statistics per segment, used as the reference against which all subsequent alerts are calibrated |
| Multi-source fusion alert | Automated correlation of SAR anomaly polygons with same-period optical change detections; alerts issued only when both sensors independently flag the same location | High-confidence alert feed with reduced false-positive rate, delivered as GeoJSON with sensor-source attribution and timestamp for each contributing observation |
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.