Special operations forward operating base and austere airstrip detection
Austere forward bases leave faint but readable marks: unpaved strips, radial vehicle tracks, sparse shelters. Multi-temporal optical analysis turns those surface scars into intelligence.
Sensors
- Planet SuperDove: 3 m ground sample distance, 8 spectral bands (including red-edge and NIR), daily revisit at mid-latitudes. The temporal cadence is the primary tool: comparing sequential daily composites reveals when track scars first appear and how activity pulses over time.
- Maxar WorldView-3: 0.31 m panchromatic, 1.24 m multispectral, 16 SWIR bands. At this resolution individual tyre-track ruts, fuel-drum shadows and aircraft wheel marks become visible. Tasked on demand; latency to first collect is typically 24–72 hours depending on orbital geometry and cloud.
- Airbus Pléiades Neo: 0.30 m panchromatic, 1.2 m multispectral, twice-daily revisit capacity per satellite (four-satellite constellation). Useful when Maxar tasking is unavailable or a second sub-metre source is needed for cross-confirmation.
- Sentinel-2 MSI: 10 m in visible and NIR bands, 5-day revisit (2–3 days with both satellites). Free and openly archived since 2015. Too coarse to confirm structures, but adequate for detecting the spectral signature of compacted bare earth against surrounding vegetation and for cueing higher-resolution tasking to a candidate location.
What a dirt strip gives away before anyone lands on it
A paved runway requires months of construction and leaves unmistakable signatures. An austere airstrip does not. A flat, compacted stretch of laterite or hardpan, 600–900 metres long and 15–20 metres wide, can be prepared in days with graders and hand tools. That is precisely what makes it hard to find through conventional means and, paradoxically, what makes it detectable by satellite.
Compaction changes soil reflectance. In visible and NIR bands, freshly compacted bare earth in a vegetated or semi-arid landscape appears as a spectrally distinct pale scar. At 10 m resolution Sentinel-2 can flag the anomaly. At 3 m Planet SuperDove resolves the strip's geometry. At 0.31 m WorldView-3 shows the surface texture: the absence of stones larger than fist-size, the smoothed centreline, the slight berming at the edges left by a grader blade. These are not ambiguous features once you know what to look for.
Radial tracks as a structural signature
Formal military installations follow grid layouts because grids are efficient for logistics and force protection. Forward operating bases in denied terrain follow no such convention. Structures cluster near water, shade or existing cover, and vehicle routes radiate outward to observation posts, fuel caches and perimeter positions. The result is a spider-web pattern of wheel tracks that is geometrically distinctive and largely absent from agricultural or pastoral land use.
Open-source investigators examining imagery of Sahel and Syrian conflict zones have documented this pattern repeatedly. Tracks in undisturbed terrain compact and bleach the surface, remaining visible in high-resolution optical imagery for months after activity ceases. Multi-temporal differencing, comparing a baseline image from before any activity against the current scene, isolates new linear features with high specificity. Planet's daily archive makes it possible to date the first appearance of each track segment to within one or two days, which is operationally significant.
The honest limit here is spatial resolution. At 3 m, individual tracks are detectable but not always interpretable. A single vehicle pass may not produce a scar wide enough to distinguish from a footpath at that resolution. Confirmation requires sub-metre tasking, and sub-metre tasking requires knowing where to look. The workflow is therefore two-stage: Planet for cuing, WorldView-3 or Pléiades Neo for confirmation.
Structural confirmation at sub-metre resolution
Once a candidate location is identified, a sub-metre collect resolves the question of what is actually there. At 0.31 m panchromatic resolution, WorldView-3 can distinguish a canvas shelter from a parked vehicle, a fuel berm from a natural ridge, and an aircraft parking revetment from a dry streambed. Shadow geometry at known solar elevation angles provides height estimates for structures too small to interpret from plan view alone.
Aircraft presence is the most definitive indicator. Light fixed-wing aircraft used in special operations contexts, such as the Pilatus PC-12 or various STOL types, have wingspans of 16–20 metres. At sub-metre resolution the aircraft itself is resolvable. Parking areas, tyre marks on the strip threshold and fuel containers confirm operational use. The absence of aircraft does not indicate inactivity: austere strips are typically used intermittently, which is another reason the temporal baseline from Planet matters more than any single high-resolution snapshot.
Cloud, shadow and the limits of optical methods
Every method described above is optical. Cloud cover breaks the temporal chain. In the Sahel, cloud is seasonal and manageable. In equatorial forest zones or during monsoon periods, persistent cloud can deny optical access for weeks. SAR (synthetic aperture radar) penetrates cloud and can detect surface roughness changes consistent with track formation, but SAR coherence analysis for subtle ground disturbance requires careful baseline selection and is less interpretable than optical imagery for non-specialist consumers. It is a complement, not a replacement.
Night activity is a deliberate tactic at austere bases. Optical sensors are blind to it. VIIRS Nightfire and VIIRS Day/Night Band can detect persistent artificial lighting at roughly 375 m resolution, but a single vehicle headlight or a shielded camp lamp will not register. Thermal infrared can detect engine heat signatures, but the commercially available thermal bands on WorldView-3 (eight SWIR bands at 3.7 m) are not thermal infrared in the heat-detection sense. Genuine thermal anomaly detection requires different sensors and is covered separately in this library.
Resolution also sets a hard floor on minimum detectable structure size. At 3 m Planet resolution, a single small shelter of 3 m × 3 m occupies roughly one pixel. It is not reliably detectable. At 0.31 m WorldView-3 resolution the same shelter occupies roughly 100 pixels and is unambiguous. The practical implication: sub-metre tasking is not optional for structural confirmation, it is the only method that works.
From detection to a finished intelligence product
Raw imagery is not intelligence. The analytical steps between a Planet daily composite and a finished site assessment include change-detection processing, candidate triage, sub-metre tasking coordination, feature extraction and contextual interpretation. Each step introduces latency. A detection-to-report cycle of 48–72 hours is achievable with pre-positioned tasking agreements; 24 hours is possible for high-priority requests when a WorldView-3 or Pléiades Neo pass is already scheduled over the area of interest.
Satellize structures this workflow as a monitored-area service: a defined geographic bounding box is processed against each new Planet pass, candidates above a change-detection threshold are automatically flagged, and an analyst reviews flagged scenes before any alert is issued. The same architecture that supports the Kingdom of Tonga crop-estimation programme, where field-condition changes must be detected quickly across dispersed island terrain, applies here with different feature classifiers and a different alert threshold.
Governments operating their own sovereign imagery programmes can ingest the same Planet and Sentinel-2 streams and run equivalent analytics on national infrastructure. The detection logic is not proprietary; the value is in the operational integration, the trained classifiers and the analyst layer that prevents false-positive fatigue from degrading the programme over time.
Typical figures
| Spatial resolution (cuing layer) | 3 m (Planet SuperDove), 10 m (Sentinel-2 MSI) |
| Spatial resolution (confirmation layer) | 0.30–0.31 m panchromatic (WorldView-3, Pléiades Neo) |
| Revisit rate (cuing) | Daily at most latitudes (Planet); 2–5 days (Sentinel-2) |
| Revisit rate (confirmation) | On-demand tasking; first collect typically 24–72 hours |
| Minimum detectable airstrip length | Approximately 400 m at 3 m resolution; shorter strips require sub-metre confirmation |
| Minimum detectable track width | ~3–5 m at Planet resolution (single-vehicle track marginal); reliably detectable at WorldView-3 |
| Archive depth | Planet daily archive from 2016; Sentinel-2 from 2015; WorldView archive from 2009 (tasked scenes only) |
| Cloud limitation | Optical methods fully blocked by cloud; SAR required for persistent-cloud environments |
| Spectral bands used | Visible (RGB), NIR, red-edge for land-cover discrimination; panchromatic for texture analysis |
| Delivery formats | GeoTIFF change layers, GeoJSON site polygons, PDF site assessments, alert feeds via API |
Analytics Satellize can run
| New bare-earth linear feature detection | Multi-temporal spectral differencing on Planet SuperDove NIR and red-edge bands; threshold-based change mask | Daily GeoJSON alert layer flagging candidate strips and track networks within a defined area of interest |
| Track-scar age estimation | First-detection dating from Planet daily archive; spectral fade curve fitting against known compaction timelines | Site report with estimated date of first ground disturbance and activity-pulse timeline |
| Sub-metre structural feature extraction | Object-based image analysis on WorldView-3 or Pléiades Neo panchromatic; shadow-height estimation at known solar geometry | Annotated GeoTIFF with classified features: strip, shelters, vehicle park, fuel point, perimeter tracks |
| Aircraft presence and type assessment | Manual and semi-automated object detection at 0.31 m resolution; wingspan measurement against known aircraft type dimensions | Aircraft identification table with confidence rating and imagery timestamp |
| Activity-tempo scoring | Time-series analysis of vehicle-track density and structural footprint change across Planet archive | Weekly activity-index chart per monitored site; PDF report with trend narrative |
| Candidate-site triage from Sentinel-2 cuing | Automated spectral anomaly detection (bare-earth index) across wide-area Sentinel-2 tiles; ranked candidate list for sub-metre follow-up | Prioritised tasking recommendation list with coordinates and anomaly confidence scores |
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.