Ivory and bushmeat stockpile facility detection near park boundaries
High-resolution optical imagery can identify informal storage and processing compounds near park boundaries by detecting construction patterns, vehicle activity and cold-storage infrastructure inconsistent with declared land use.
Sensors
- Maxar WorldView-3: 0.31 m panchromatic, 1.24 m multispectral (8 bands including SWIR). Sufficient to resolve individual vehicle types, roof materials and shadow geometry of structures under 5 m tall. Revisit approximately 1 day at mid-latitudes with off-nadir tasking, though cloud cover in humid forest corridors regularly degrades usable acquisition frequency.
- Airbus Pléiades Neo: 0.30 m panchromatic, 1.2 m multispectral (4 + 4 bands). Stereo and tri-stereo modes allow height estimation of structures, which helps distinguish cold-storage units from agricultural sheds. Tasking latency from order to first acquisition typically 24–72 hours.
- Planet SuperDove (PlanetScope): 3 m multispectral, 8 spectral bands, near-daily global revisit. Resolution is too coarse to identify individual structures reliably, but the daily cadence is well suited to detecting new compound clearance events, vehicle presence changes and access-track formation over weeks. Acts as a screening layer before costly tasking of sub-metre sensors.
- Sentinel-2 MSI: 10 m (visible and NIR), 20 m (red-edge and SWIR), 5-day revisit at the equator with both satellites. Free and openly archived since 2015. Useful for corridor-scale change detection and for flagging anomalous land-cover patches in buffer zones, but cannot resolve individual structures at compound scale.
What a consolidation compound actually looks like from above
Wildlife trafficking in East and Central Africa does not move product directly from kill site to export port. Published UNODC and TRAFFIC casework documents an intermediate step: consolidation points, typically informal compounds within 10–50 km of park boundaries, where ivory and bushmeat are sorted, processed and held before onward movement by road or river. These sites are not purpose-built. They occupy existing agricultural plots, small-scale timber yards or family compounds, which is precisely what makes ground-level enforcement difficult.
From above, the signatures are legible. Consolidation compounds tend to show compacted bare earth in patterns inconsistent with crop cultivation: no row structure, no seasonal tillage marks, no irrigation geometry. Structures often include one or more metal-roofed buildings larger than a typical rural dwelling, positioned away from the main access track. Cold storage, where present, appears as insulated box units or modified shipping containers. WorldView-3 at 0.31 m panchromatic can resolve these features; Pléiades Neo stereo pairs can add height estimates that distinguish a 2.5 m refrigeration unit from a flat-roofed store.
Vehicle activity is a secondary indicator. Repeated imagery over weeks can show whether trucks are present intermittently at irregular hours, which is inconsistent with declared farming activity but consistent with consolidation logistics. Planet's daily cadence is well matched to this kind of temporal pattern analysis, even though individual vehicles are only marginally resolved at 3 m.
The corridors that published casework has already named
UNODC's World Wildlife Crime Reports and TRAFFIC's published seizure analyses identify specific geographic corridors where ivory and bushmeat move from source to port. The Selous-Niassa corridor in Tanzania and Mozambique, the Garamba-South Sudan axis, and the forest-edge zones of Cameroon and the Central African Republic appear repeatedly. These are not speculative geography: they are documented in publicly available enforcement and NGO literature.
Knowing a corridor exists does not tell you where within it a facility sits. That is the analytical problem satellite imagery addresses. A defined search zone of, say, 5,000 km² along a known corridor can be screened at Sentinel-2 resolution to identify candidate patches of anomalous bare earth or new structure development, then prioritised for sub-metre tasking. The two-stage approach keeps commercial imagery costs proportionate to the intelligence value.
Spectral and structural indicators: what each band actually tells you
Bare compacted earth has a distinct spectral signature in the red and SWIR bands. In Sentinel-2 Band 11 (1.6 µm SWIR) and Band 12 (2.2 µm SWIR), disturbed soil and compacted surfaces separate clearly from surrounding vegetation and from undisturbed agricultural land. WorldView-3 adds eight SWIR bands between 1.19 µm and 2.37 µm, giving finer discrimination of surface materials including metal roofing, tarpaulins and concrete, all common in informal storage compounds.
Vegetation indices are a useful negative indicator. A plot declared as active farmland should show NDVI values cycling seasonally. A plot used for storage will show suppressed, static NDVI over the compound footprint, even when surrounding fields are productive. This mismatch between declared use and observed vegetation behaviour is one of the more reliable screening signals available from freely archived Sentinel-2 time series.
Thermal infrared would be the obvious tool for detecting refrigeration units, but Landsat 8/9 TIRS at 100 m resolution and ASTER at 90 m are too coarse to isolate a single cold-storage unit in a mixed compound. Airborne thermal or drone-based sensing would be required for that level of detail, which is outside the scope of satellite-only analysis.
Honest limits of the method
Cloud cover is the dominant operational constraint. The forest-edge corridors of Central and East Africa sit in regions where persistent cloud can reduce usable optical acquisitions to fewer than 30 per year in the wettest months. SAR sensors such as Sentinel-1 penetrate cloud but offer only 10 m resolution in Interferometric Wide mode, which is insufficient for compound-level analysis. SAR change detection can flag new structure development, but confirmation requires a clear optical acquisition.
Camouflage is a genuine countermeasure. Operators who are aware of aerial surveillance can cover structures with vegetation or shade netting, park vehicles under tree canopy and limit compound activity to periods of known cloud cover. These measures reduce detection probability significantly. The method is most effective against operators who do not yet know they are being watched, or who underestimate the resolution of commercial satellites.
Attribution is also limited. Imagery can show that a compound exists and that its activity pattern is anomalous. It cannot, by itself, prove what is stored inside. Satellite analysis generates leads for ground-based enforcement, not prosecutable evidence on its own. That is an important boundary to state clearly to any client commissioning this work.
From screening to actionable intelligence
A practical workflow runs in three stages. First, corridor-scale screening using Sentinel-2 time series identifies candidate sites by flagging bare-earth anomalies, static vegetation patches and new access tracks within a defined buffer zone around park boundaries. This can cover hundreds of thousands of hectares at low cost using openly archived data.
Second, candidate sites are ranked by a composite score: proximity to known trafficking corridors, structure count, vehicle activity frequency and spectral anomaly strength. The top-ranked sites receive commercial tasking from WorldView-3 or Pléiades Neo for sub-metre confirmation imagery.
Third, confirmed sites feed into a GIS layer delivered to enforcement partners, with change-detection alerts triggered when subsequent imagery shows significant activity changes. Satellize runs this kind of multi-source analytical pipeline for government clients; the approach is structurally similar to the crop-estimation programme the company operates for the Kingdom of Tonga, adapted here for enforcement rather than agricultural monitoring. Outputs are formatted for direct import into standard GIS environments used by national park authorities and wildlife crime units.
Typical figures
| Best spatial resolution (optical) | 0.30–0.31 m panchromatic (Pléiades Neo, WorldView-3); 1.2–1.24 m multispectral |
| Screening resolution | 3 m (Planet SuperDove); 10 m (Sentinel-2 visible/NIR) |
| Revisit (sub-metre tasking) | Approximately 1 day with off-nadir tasking; cloud-limited to fewer than 30 clear acquisitions per year in high-humidity corridors |
| Revisit (screening layer) | Near-daily (Planet); 5-day (Sentinel-2, equatorial) |
| Key spectral bands | Visible, NIR, red-edge, SWIR (1.19–2.37 µm for WorldView-3); SWIR Bands 11 and 12 for Sentinel-2 soil discrimination |
| Minimum detectable structure | Structures approximately 3 m × 3 m at sub-metre resolution; compound clearings from approximately 200 m² at 3 m resolution |
| Archive depth | Sentinel-2: from 2015 (open archive); Planet: from approximately 2016 (licensed); Maxar: from approximately 2008 for some areas |
| Thermal detection of refrigeration | Not achievable at satellite resolution; Landsat TIRS at 100 m and ASTER at 90 m are too coarse for single-unit detection |
| Delivery formats | GeoTIFF change-detection rasters, GeoJSON site polygons, PDF intelligence summary, GIS-ready vector layers (Shapefile / GeoPackage) |
| Typical tasking latency | 24–72 hours from order to first acquisition attempt; cloud delays variable |
Analytics Satellize can run
| Corridor screening map | Sentinel-2 SWIR bare-earth index and NDVI anomaly detection across defined buffer zones around park boundaries | GIS layer of candidate anomaly patches, ranked by composite score, covering the defined search corridor |
| Compound confirmation report | Sub-metre optical analysis of structure count, roof material, vehicle presence and shadow-derived height estimation using WorldView-3 or Pléiades Neo stereo | PDF site intelligence report with annotated imagery, structure inventory and activity assessment for each confirmed compound |
| Vehicle activity time series | Change detection on Planet daily imagery stack to extract vehicle presence frequency and access-track usage patterns | Tabular activity log per site with presence/absence flags and trend chart, updated on a defined schedule |
| New-clearance alert | Automated NDVI suppression and bare-earth emergence detection on Sentinel-2 time series within the monitored buffer zone | Email or API alert with coordinates and preliminary imagery when a new candidate clearance event exceeds detection threshold |
| Declared versus observed land-use comparison | Overlay of national land-use registry or agricultural permit data against observed spectral land-cover classification | Discrepancy report identifying plots where observed activity is inconsistent with declared use, formatted for enforcement referral |
| Multi-temporal compound progression archive | Historical imagery stack from Sentinel-2 and available commercial archive to reconstruct compound development timeline | Annotated time-lapse GIF and accompanying GIS layer showing compound growth stages and activity peaks |
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.