Informal fuel depot and artisanal crude-oil refinery detection
Artisanal crude refining leaves overlapping signatures across thermal, radar and optical data. Satellite analytics can locate active sites within days and map historical contamination across entire creek systems.
Sensors
- Landsat 8/9 TIRS: Thermal infrared bands 10 and 11 at 100 m native resolution (resampled to 30 m). Detects persistent surface temperature anomalies from open-flame distillation pits. Revisit 8 days per satellite; combined Landsat 8 and 9 gives roughly 8-day repeat at the equator. Cloud cover over the Niger Delta severely limits usable acquisitions, particularly during the wet season (May to October).
- Sentinel-1 C-band SAR: 6-day repeat at mid-latitudes in IW mode, 10 m ground range resolution. C-band backscatter detects surface oil films on water bodies as low-return anomalies (Bragg scattering suppressed by oil). Also maps disturbed soil and compacted access tracks around storage pits. Unaffected by cloud or smoke, which makes it the primary sensor during wet-season operations.
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands; 20 m in red-edge and SWIR. NDVI and red-edge indices map hydrocarbon-induced vegetation stress with documented sensitivity to soil contamination halos. 5-day revisit at the equator with both satellites. Cloud remains a constraint; time-series compositing over 4 to 8 weeks is typically required to obtain a clean optical view.
- ICEYE X-band SAR: Commercial constellation offering sub-metre spotlight mode and roughly 1 m stripmap. X-band is more sensitive than C-band to fine surface texture changes, useful for resolving individual storage vessels or earthen pit structures. Tasked on demand; latency from order to delivery can be under 24 hours for priority acquisitions. Cost is per-scene, so it is best reserved for site confirmation after screening.
- VIIRS Day/Night Band and VIIRS Nightfire: 375 m pixel at nadir; daily global revisit. The Nightfire algorithm (Colorado School of Mines, published via EOGDATA) detects sub-pixel combustion events by fitting blackbody curves to SWIR and NIR channels. Open-flame refining fires are detectable at night even when the pit area is well below the pixel footprint, though the method cannot distinguish refinery fires from other combustion sources at this resolution without corroborating data.
What a refining pit leaves behind, and why it is readable from orbit
Artisanal crude refining in the Niger Delta typically involves stolen crude fed into improvised steel drums or earthen pits and heated by open flame. The process is inefficient and dirty. It produces a persistent thermal signature while active, a surface oil slick on adjacent water bodies, a soil contamination halo where spilled crude saturates the root zone, and a dead-vegetation ring that persists for months or years after a site is abandoned. Each of these signatures maps to a different part of the electromagnetic spectrum, which is why multi-sensor analysis is far more reliable than any single sensor alone.
UNEP's 2011 assessment of Ogoniland documented hydrocarbon contamination at depths exceeding one metre at many sites, a finding that underscores why spectral vegetation indices continue to flag old refinery locations long after the fire is gone. The soil signal outlasts the operator.
Thermal detection: what the fire gives away, and where it fails
Landsat TIRS and VIIRS Nightfire are the two primary thermal tools. Landsat's 100 m TIRS pixels are well-suited to detecting clusters of active pits; a single active refining site with multiple fires can raise the apparent brightness temperature of a pixel by several degrees above background, which is a clear anomaly in a mangrove or forest setting. VIIRS Nightfire operates at night and can flag sub-pixel combustion events by their spectral shape, which is a useful screening tool across wide areas.
The honest limits are significant. Cloud cover suppresses optical and thermal acquisitions for weeks at a time during the wet season. Smoke plumes from the fires themselves can saturate or obscure adjacent pixels. A dormant site, one that has been temporarily shut down in response to enforcement activity, produces no thermal signal at all. Thermal data confirms activity; it cannot confirm absence. Any enforcement programme that relies on thermal alone will systematically miss the intervals when operators pause.
SAR and the oil-on-water signature
Synthetic aperture radar is the workhorse sensor for this application during cloud-obscured periods. Oil films dampen the capillary waves that produce Bragg scattering in C-band SAR, creating low-backscatter patches on water surfaces that are visually and algorithmically distinct from clean water or wind-roughened water. Sentinel-1 can map these slicks at 10 m resolution across entire creek networks on a 6-day cycle.
The ambiguity problem is real and must be stated plainly. Calm water, biogenic surfactants from algae and certain rainfall patterns all produce low-backscatter patches that resemble oil films. Discrimination requires contextual analysis: slick geometry, persistence across multiple passes, proximity to known infrastructure, and co-registration with optical or thermal anomalies. A single SAR pass showing a dark patch is a lead, not a finding. ICEYE X-band spotlight imagery, at sub-metre resolution, can then resolve whether the site contains storage vessels, earthen pits or compacted access routes, providing the specificity that Sentinel-1 cannot.
The dead-vegetation halo: a signal that survives enforcement
Hydrocarbon soil saturation suppresses plant growth by displacing oxygen in the root zone and introducing toxic aromatics. The result is a characteristic dead or stressed-vegetation ring around storage pits and spill points. Sentinel-2 red-edge bands (bands 5, 6 and 7, at 20 m resolution) and NDVI computed from bands 4 and 8 are both sensitive to this stress. In the Niger Delta context, where surrounding vegetation is dense mangrove or secondary forest, the contrast is pronounced.
The practical value for enforcement is that this signature persists. A site abandoned six months ago still shows reduced NDVI. Time-series analysis, comparing monthly or quarterly composites over a multi-year archive, can reconstruct the operational history of a site: when it became active, how it expanded, and whether it has been relocated rather than abandoned. Sentinel-2 archive depth extends to 2015, giving roughly a decade of coverage for retrospective analysis.
Turning signatures into an operational detection workflow
An effective workflow runs in three tiers. First, VIIRS Nightfire and Sentinel-1 SAR provide wide-area screening on a near-daily and 6-day cadence respectively, flagging candidate locations across an entire region. Second, Sentinel-2 time-series composites confirm vegetation stress halos and map the spatial extent of contamination around flagged sites. Third, commercial SAR (ICEYE spotlight) or very-high-resolution optical tasking provides site-level confirmation before any enforcement action is taken.
Satellize runs this multi-sensor screening workflow on open constellations with commercial tasking added on client licence. The approach is similar in structure to the analytics developed for the Tonga crop-estimation programme, adapted here for anomaly detection rather than crop-area estimation. Output is typically a site-level alert layer in GIS format, updated on a defined cadence, with confidence scores derived from how many independent sensor signatures corroborate each candidate.
One honest caveat about the whole approach: detection is not attribution. Satellite data can locate a site and characterise its activity level. Establishing who operates it, and building a case that meets evidentiary standards, requires ground truth. The satellite layer narrows the search area; it does not replace the investigator.
What the published record says about detection limits
UNODC and UNEP field reports on Niger Delta artisanal refining describe sites ranging from single-pit operations of a few metres diameter to multi-pit complexes covering several hectares. Single small pits are at or below the reliable detection threshold for Sentinel-1 and Landsat TIRS when operating in isolation; the thermal and backscatter anomaly may not exceed noise. Clusters of pits, or sites with associated storage tanks and access tracks, are reliably detectable at Sentinel-1 and Sentinel-2 resolutions. ICEYE sub-metre spotlight can resolve individual drums and pit structures, but at per-scene cost that makes it impractical for area-wide screening.
Published remote-sensing studies in journals such as Remote Sensing (MDPI) have demonstrated detection rates for oil-spill slicks in the Niger Delta exceeding 80 percent under favourable SAR acquisition conditions, with false-positive rates that require optical or multi-temporal corroboration to manage. No published study claims reliable detection of single-operator, single-pit sites from space alone.
Typical figures
| Thermal spatial resolution (Landsat TIRS) | 100 m native, 30 m resampled product |
| SAR spatial resolution (Sentinel-1 IW) | 10 m ground range |
| SAR spatial resolution (ICEYE spotlight) | ~1 m |
| Optical/multispectral resolution (Sentinel-2) | 10 m (VIS/NIR), 20 m (red-edge, SWIR) |
| Revisit cadence (SAR screening) | 6 days (Sentinel-1 pair); <24 h on-demand (ICEYE) |
| Revisit cadence (thermal) | 8 days per Landsat satellite; ~8-day combined Landsat 8+9; daily (VIIRS) |
| Cloud-cover impact | Severe for optical/thermal in Niger Delta wet season (May–Oct); SAR unaffected |
| Minimum reliably detectable site | Multi-pit complex >0.5 ha; single small pit at or below reliable threshold |
| Archive depth | Sentinel-1 from 2014; Sentinel-2 from 2015; Landsat from 1972; VIIRS from 2012 |
| Typical delivery format | GeoTIFF anomaly layers, GeoJSON site-alert feed, PDF site report |
Analytics Satellize can run
| Active-site thermal anomaly alert | Brightness-temperature differencing against seasonal baseline (Landsat TIRS) combined with VIIRS Nightfire sub-pixel combustion detection | GeoJSON alert feed updated on each clear-sky Landsat pass and nightly VIIRS pass, with anomaly magnitude and persistence score |
| Oil-on-water slick map | Sentinel-1 SAR backscatter thresholding with contextual filtering to suppress biogenic and wind-shadow false positives; multi-pass persistence check | GIS polygon layer of confirmed and candidate slick extents, updated every 6 days |
| Vegetation-stress halo map | Sentinel-2 NDVI and red-edge index time-series compositing; change detection against pre-activity baseline | Quarterly raster layer showing contamination halo extent and severity class, with historical progression from 2015 archive |
| Site-confirmation dossier | ICEYE X-band spotlight tasking analysed for pit structures, storage vessels and access tracks; fused with thermal and SAR screening results | PDF site report with annotated imagery, coordinates, activity timeline and confidence assessment |
| Regional screening dashboard | Automated multi-sensor fusion scoring candidate locations by number of corroborating signatures (thermal + SAR + NDVI anomaly) | Web-accessible ranked candidate list with sensor evidence links, refreshed on each new acquisition cycle |
| Historical activity reconstruction | Multi-year Sentinel-1 and Sentinel-2 archive stack analysis to date site activation, expansion and apparent abandonment or relocation | Site-level timeline report suitable for legal or regulatory documentation |
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.