Sensors
- Sentinel-2 MSI: 10 m visible and near-infrared bands, 20 m SWIR bands (B11 at 1610 nm, B12 at 2190 nm). Revisit roughly 5 days at the equator with both satellites. SWIR bands distinguish active burn scars and exposed heavy-metal-contaminated soil from background land cover. Free archive from 2015.
- Landsat 8/9 TIRS: Thermal infrared at 100 m resolution (resampled to 30 m in products), two thermal bands centred at 10.9 µm and 12.0 µm. Revisit 8 days per satellite, 16 days per sensor alone. Detects persistent heat anomalies from smouldering burn piles, though spatial resolution limits detection to larger or clustered sites. Free archive from 1972 (Landsat 1) through present.
- Planet SuperDove (PlanetScope): 3 m resolution, 8 spectral bands including red-edge and two NIR bands, daily revisit over most land areas. Useful for confirming site boundaries and monitoring activity changes between coarser-resolution passes. Commercial tasking required; no free archive.
- Maxar WorldView-3: 31 cm panchromatic, 1.24 m multispectral, 3.7 m SWIR (8 bands, 1195–2365 nm). The SWIR suite is particularly relevant: published research has used similar high-resolution SWIR data to map soil contamination extent at sub-parcel scale. Tasked commercially; not suitable for routine revisit monitoring but valuable for site characterisation and legal documentation.
- VIIRS (Suomi NPP / NOAA-20): 375 m active-fire detection product (VNP14) identifies open-burning events with sub-daily revisit. At this resolution individual e-waste sites are rarely resolved, but persistent fire detections co-located with known informal settlements provide a useful screening layer. Data available via NASA FIRMS.
What burned cable insulation leaves in the soil
Informal e-waste processing concentrates heavy metals, including lead, cadmium, chromium and copper, into the topsoil at levels that suppress or eliminate vegetation. At Agbogbloshie in Accra, published soil studies have recorded lead concentrations orders of magnitude above WHO reference values across an area of several hectares. That contamination does not move quickly. Even after burning activity ceases, the spectral signature of bare, metal-laden soil persists for years, which is why satellite archives are useful not just for detecting active sites but for mapping the cumulative damage and tracking whether remediation is occurring.
The vegetation response is equally diagnostic. Chlorophyll stress and outright die-off reduce the normalised difference vegetation index (NDVI) in contaminated zones relative to surrounding land. In tropical and subtropical settings where vegetation recovers quickly after disturbance, a persistently low-NDVI patch adjacent to an informal settlement is anomalous and worth investigating. The signal is not unique to e-waste, which is why spectral analysis must be combined with contextual cues: proximity to urban fringe areas, road access, and the presence of thermal anomalies.
How SWIR bands distinguish an active burn from ordinary bare ground
Sentinel-2 bands B11 (1610 nm) and B12 (2190 nm) are sensitive to moisture content and mineral composition of exposed soil. Freshly burned areas and acid-stripped soil show characteristic reflectance profiles in these bands that differ from dry agricultural soil, construction spoil or naturally sparse savanna. Published research on Guiyu in China's Guangdong province, one of the world's largest e-waste processing regions, has used Landsat SWIR composites to delineate contaminated zones and track their expansion over multi-year periods.
Active burning adds a thermal component. Smouldering cable fires and acid-bath residue burning maintain temperatures well above ambient for hours. Landsat TIRS detects these as localised thermal anomalies, though at 100 m resolution a single small burn pile may be below the detection threshold unless multiple piles cluster together. WorldView-3's SWIR suite, when tasked specifically, can resolve individual burn areas at sub-5 m scale, which is useful for enforcement documentation. The honest caveat: cloud cover over equatorial West Africa and monsoon-season South-East Asia can interrupt optical and thermal observations for days or weeks at a stretch.
Building a detection workflow from screening to confirmation
A practical detection programme runs in three tiers. First, VIIRS active-fire detections and Sentinel-2 NDVI anomaly mapping screen large areas at low cost to flag candidate sites. This is computationally cheap and can cover entire national territories on a monthly cycle. Second, Sentinel-2 SWIR time-series analysis at flagged locations tests whether the spectral signature matches the contamination profile rather than, say, a brick kiln or agricultural burn. Third, commercial tasking of Planet SuperDove or WorldView-3 provides the spatial resolution needed to characterise a confirmed site, count infrastructure (cable-stripping tables, acid drums, vehicle access), and produce imagery suitable for regulatory or legal use.
Change detection between archive epochs is often more informative than any single image. A site that appears as a modest bare patch in 2018 Sentinel-2 imagery but has expanded to three times the area by 2024 tells a story that a single current image cannot. Landsat's archive stretching back to the early 1970s, and Sentinel-2's free archive from 2015, make multi-year trajectories straightforward to construct.
Known limits and honest ambiguities
Spectral confusion is the central methodological problem. Bare soil in arid or semi-arid settings can resemble metal-contaminated ground in SWIR bands without any e-waste present. Urban construction sites, informal brick kilns and charcoal production areas all generate thermal anomalies and NDVI suppression. No single spectral index is diagnostic on its own. Confidence rises when multiple independent signals, SWIR reflectance anomaly, thermal excess, persistent NDVI suppression and settlement context, co-occur at the same location.
Spatial resolution sets a hard floor on what is detectable. A single household-scale operation burning cable in a courtyard will not register in Sentinel-2 or Landsat data. Detection is practical for sites covering roughly 0.5 hectares or more, or for clusters of smaller operations that together produce a measurable aggregate signature. For enforcement purposes, satellite detection identifies where to look; ground inspection or very-high-resolution commercial imagery is required to confirm what is happening and to whom the site belongs.
Revisit frequency matters for monitoring active operations. A site that burns only at night or only on certain days may not coincide with a daytime optical pass. VIIRS provides near-daily fire detection but at coarse resolution. Combining VIIRS temporal coverage with Sentinel-2 spatial detail, accepting that they rarely observe simultaneously, is the practical compromise available on open data budgets.
From detection layer to enforcement-grade evidence
Regulatory agencies and development-finance institutions increasingly require spatial evidence for environmental compliance assessments. A satellite-derived site inventory, with coordinates, estimated area, activity status and multi-year change record, gives an investigator a prioritised list rather than an undifferentiated complaint. It also establishes a documented baseline: if a site is ordered to close, subsequent imagery can confirm whether activity has ceased or merely shifted to a nearby location.
Satellize runs this kind of multi-sensor screening and change-detection workflow on open constellations, with commercial tasking added where a client holds the appropriate licence. The approach is similar in structure to the crop-estimation programme Satellize operates for the Kingdom of Tonga: open-data screening at national scale, commercial imagery for site-level characterisation, and a structured analytic output rather than raw imagery. For e-waste enforcement, the deliverable is typically a GIS layer of confirmed and candidate sites, each with a confidence rating, a spectral evidence summary and an archive of dated imagery. Investigators can then allocate field resources to the highest-confidence detections first.
Typical figures
| Spatial resolution (screening) | 10–20 m (Sentinel-2 MSI); 30 m (Landsat 8/9 OLI); 375 m (VIIRS active fire) |
| Spatial resolution (confirmation) | 3 m (Planet SuperDove); 1.24 m multispectral / 3.7 m SWIR (WorldView-3) |
| Thermal resolution | 100 m native (Landsat TIRS, resampled to 30 m in delivered products) |
| Revisit cadence | ~5 days (Sentinel-2, both satellites); 16 days per Landsat sensor; daily (Planet SuperDove, VIIRS) |
| Key spectral bands | SWIR B11 1610 nm, B12 2190 nm (Sentinel-2); TIR 10.9 µm and 12.0 µm (Landsat TIRS); 8-band SWIR 1195–2365 nm (WorldView-3) |
| Minimum detectable site area (screening) | Approximately 0.5 ha for persistent soil anomaly in Sentinel-2; smaller sites require commercial imagery |
| Cloud-cover limitation | Optical and thermal sensors fully blocked by cloud; persistent cloud in tropical regions can cause gaps of 2–6 weeks in usable imagery |
| Archive depth | Sentinel-2 from 2015; Landsat from 1972; Planet from approximately 2016 (commercial) |
| Latency (open data) | Sentinel-2 and Landsat products typically available within 1–3 hours of acquisition via Copernicus Data Space and USGS EarthExplorer |
| Delivery formats | GeoTIFF, GeoPackage, GIS-ready vector layers, PDF site reports with dated imagery mosaics |
Analytics Satellize can run
| National or regional candidate-site inventory | NDVI anomaly mapping and SWIR reflectance screening over Sentinel-2 time series; VIIRS active-fire co-location filter | GIS polygon layer of candidate sites with anomaly scores, coordinates and first-detection date |
| Site-level spectral contamination map | Multi-date SWIR composite analysis and spectral mixture modelling to delineate soil contamination extent | Per-site GeoTIFF with contamination probability surface and estimated affected area in hectares |
| Thermal anomaly time series | Landsat TIRS band-ratio analysis for persistent heat signatures; VIIRS VNP14 active-fire event log | Tabular fire-event log per site with dates, estimated intensity class and Landsat thermal imagery thumbnails |
| Multi-year site expansion record | Change detection across Sentinel-2 and Landsat archive epochs using bitemporal SWIR and NDVI differencing | Annotated time-lapse image series and area-change chart per site, suitable for regulatory submissions |
| Confidence-rated site classification | Multi-signal fusion: SWIR anomaly, thermal excess, NDVI suppression, settlement-proximity scoring | Ranked site list with high/medium/low confidence ratings and supporting spectral evidence summary per site |
| Post-intervention monitoring report | Repeat Sentinel-2 and commercial-imagery acquisition at confirmed sites following enforcement action | Quarterly PDF report with before/after imagery, NDVI recovery index and activity-status assessment |
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.