Unauthorised oil and gas well drilling detection
Illegal well drilling leaves a predictable sequence of surface scars: cleared pads, access tracks, flare light and tank batteries. Sentinel-1 SAR and PlanetScope optical time series can flag new pad events within days, though distinguishing unlicensed from licensed activity still requires concession-boundary cross-referencing.
Sensors
- Sentinel-1 C-band SAR (ESA): 10 m ground range resolution in Interferometric Wide Swath mode, 6-day repeat at mid-latitudes (12-day from a single satellite). Backscatter change from bare soil clearance is detectable regardless of cloud cover or night. Coherence change detection between pre- and post-event image pairs can flag pad-scale disturbance of roughly 0.1 ha and above, though dense vegetation suppresses coherence even before human activity.
- PlanetScope SuperDove: 3 m optical resolution, daily revisit globally. Eight spectral bands including red-edge and near-infrared allow NDVI differencing to confirm vegetation removal. Useful for visual confirmation and track mapping after SAR flags a candidate site, but cloud cover in equatorial basins such as the Niger Delta can interrupt the time series for days or weeks at a stretch.
- Landsat 8/9 OLI: 30 m multispectral resolution, 16-day repeat per satellite (8-day combined). Free archive back to 1972 provides a long baseline for detecting pad proliferation over time. Thermal Infrared Sensor (TIRS) band at 100 m resolution can detect persistent flare heat signatures, though it struggles to resolve individual small flares against background soil temperature in daytime acquisitions.
- Maxar WorldView-3: 31 cm panchromatic, 1.24 m multispectral. Tasked on demand; not a systematic monitor. Used to resolve ambiguous SAR detections: a WorldView-3 image can confirm whether a cleared pad holds a rig, a tank battery or simply a legitimate agricultural structure. Revisit depends on tasking priority and cloud; typically 1-4.5 days over a target at mid-latitudes.
- VIIRS Day/Night Band and VIIRS Nightfire (NOAA/Mines): 750 m resolution, nightly global coverage. VIIRS Nightfire uses multi-band fitting to detect and characterise sub-pixel combustion sources. Persistent new flare points in a basin, especially ones absent from licensed-facility registers, are a useful screening signal. Detection threshold is roughly 2-5 MW radiant power for a single flare; small or intermittent flares may be missed.
What a drill pad looks like before the rig shows up
Pad construction follows a predictable sequence that satellite sensors can track step by step. First comes vegetation clearance: typically a rectangle of 0.5 to 2 hectares bulldozed to bare soil. That clearance produces an immediate, strong increase in C-band SAR backscatter over vegetated terrain, and an equally sharp NDVI drop in optical imagery. The access track, usually 4 to 8 metres wide, extends from the nearest road and appears as a linear backscatter anomaly that can run for several kilometres through otherwise intact land.
The rig itself arrives later and is transient. What persists after drilling are the tank battery, the wellhead structure and, frequently, a flare stack. In basins where gas is vented or flared rather than captured, the VIIRS Nightfire algorithm can detect the combustion signature as a new persistent point source. In the Permian Basin, published work using Landsat and VIIRS has tracked the proliferation of pad sites at basin scale, linking nighttime light growth to production records. The Nigerian Delta presents a harder problem: cloud cover exceeds 80% of days over much of the year, making optical time series unreliable as a primary detection layer and pushing the workflow toward SAR.
Detection is the easy part. Attribution is the hard part.
Identifying a new pad is a change-detection problem. Attributing it as unlicensed is a geospatial overlay problem, and the two are often conflated in procurement conversations. A detected pad event becomes evidence of potential illegality only when it falls outside a published concession boundary, inside a protected area, or within a licensed block assigned to a different operator than the one apparently active on the ground.
Concession boundary data quality varies enormously by jurisdiction. In Iraq, the Ministry of Oil publishes block boundaries, and cross-referencing against them is feasible. In Nigeria, the Department of Petroleum Resources maintains block data, though boundary precision and update frequency are imperfect. In some jurisdictions, concession data is not publicly available at all, which means attribution requires the client to supply the authoritative boundary layer. Without that layer, the analytic product is a change-detection alert, not an enforcement-grade attribution. Honest workflow design keeps those two outputs clearly separate.
SAR coherence change detection: what the method actually measures
Interferometric coherence between two Sentinel-1 acquisitions measures how similar the radar scattering from a patch of ground is between two dates. Intact vegetation depresses coherence because leaves move between passes. Bare soil and man-made structures maintain high coherence. A pad clearance event therefore produces a coherence increase over the cleared area, which is counterintuitive but reliable: the newly exposed soil is more temporally stable to radar than the vegetation it replaced.
The practical detection floor for this approach is roughly pad sizes above 0.1 to 0.2 ha in open terrain. Dense tropical forest complicates matters because canopy volume scattering already suppresses coherence before any disturbance occurs; in that environment, intensity change detection on the backscatter itself is more reliable than coherence differencing. False positives include seasonal flooding, agricultural ploughing and legitimate construction. A well-designed workflow applies a land-cover mask to exclude agricultural zones and runs a minimum persistence filter: a single-date anomaly is a candidate, but two or three consecutive acquisitions showing the same footprint raise confidence substantially.
Published cases and what they established
The Permian Basin in West Texas has been studied extensively using Landsat time series and VIIRS nighttime light data. Researchers have tracked pad density growth at basin scale, correlating new clearances with production data and identifying sites where drilling activity preceded regulatory filings. The basin's arid, cloud-free climate makes it close to ideal for optical change detection; revisit gaps are the main constraint rather than cloud.
The Niger Delta is the opposite extreme. Published remote-sensing studies have used Sentinel-1 to map pipeline-tap sites and illegal bunkering infrastructure, and the same SAR-first approach applies to pad detection. Cloud cover makes PlanetScope and Landsat useful only for confirmation during dry-season windows. Iraqi Kurdistan presents an intermediate case: the semi-arid terrain is optically favourable, but the political complexity of overlapping Kurdish Regional Government and federal Iraqi concession boundaries means the attribution layer is particularly difficult to assemble from public sources alone.
Honest limits of the workflow
Sentinel-1's 10 m resolution means small wellhead structures on an existing pad are below reliable detection. The workflow is designed to catch pad construction events, not individual wellbore additions to an already-cleared site. A licensed operator drilling a second or third well on an existing pad produces almost no detectable surface change beyond what is already there.
Revisit latency matters for enforcement. A six-day Sentinel-1 repeat means a pad can be fully constructed, drilled and the rig moved before the second acquisition arrives. Commercial SAR constellations such as ICEYE or Capella offer sub-daily revisit but at higher cost and with narrower swaths, making systematic basin-wide monitoring expensive. PlanetScope's daily optical revisit partially compensates, but only in cloud-free conditions. Any enforcement programme should plan for a detection latency of days to a few weeks in practice, not hours. Satellize's analytics stack on open constellations is well-suited to the systematic monitoring layer; high-resolution commercial tasking is reserved for confirmation of high-confidence alerts to manage cost.
Finally, the method cannot determine whether drilling is productive or whether hydrocarbons are actually being extracted. A cleared pad with a wellhead present is a surface indicator, not a production measurement. Subsurface attribution requires regulatory data that satellite imagery cannot supply.
Designing a monitoring programme that holds up in an enforcement context
Regulators and concession holders who have used satellite change detection in enforcement proceedings have learned that the evidentiary chain matters as much as the detection itself. Image provenance, processing reproducibility and the separation between automated detection and human analyst review all affect whether a finding can be used in legal or administrative proceedings.
A practical programme architecture runs a SAR-based change-detection layer on every new Sentinel-1 acquisition over the area of interest, flags candidates above a size and confidence threshold, queues those candidates for optical confirmation within the next available cloud-free PlanetScope or Landsat pass, and overlays confirmed detections against the client-supplied concession boundary layer before generating an alert. The alert record should include the detection date, the image identifiers, the coordinates, the distance from the nearest concession boundary and a confidence tier. That structure is auditable. A simple shapefile of red dots is not.
Typical figures
| Primary SAR resolution | 10 m (Sentinel-1 IW mode, ground range) |
| Primary SAR revisit | 6 days at mid-latitudes (single satellite); 12 days near equator on some tracks |
| Optical confirmation resolution | 3 m (PlanetScope SuperDove); 30 m (Landsat 8/9 OLI) |
| High-resolution tasking resolution | 31 cm pan / 1.24 m MS (Maxar WorldView-3, on-demand) |
| Minimum detectable pad size (SAR coherence) | Approximately 0.1 to 0.2 ha in open terrain; larger threshold in dense vegetation |
| Flare detection threshold (VIIRS Nightfire) | Approximately 2 to 5 MW radiant power; sub-pixel, 750 m pixel size |
| SAR archive depth | Sentinel-1A from April 2014; Sentinel-1B from April 2016 (B decommissioned 2021) |
| Optical archive depth | Landsat back to 1972; PlanetScope from 2016 |
| Alert latency (operational) | Days to weeks depending on SAR acquisition schedule and cloud cover at confirmation step |
| Delivery formats | GeoJSON alert feed, GeoTIFF change layers, PDF audit report with image provenance |
Analytics Satellize can run
| New pad construction alert | Sentinel-1 backscatter intensity change detection and coherence differencing between sequential image pairs, with minimum-persistence filter across three acquisitions | Timestamped GeoJSON alert with coordinates, detection date, image identifiers and size estimate |
| Access track mapping | Linear feature extraction on SAR intensity and PlanetScope NDVI difference imagery using morphological filtering | Vector polyline layer of new tracks with length and bearing attributes |
| Concession boundary attribution | Spatial overlay of confirmed pad detections against client-supplied or publicly available concession boundary polygons | Alert records tagged as inside licensed area, outside licensed area or within buffer zone, with distance-to-boundary field |
| Flare onset detection | VIIRS Nightfire multi-band sub-pixel combustion fitting applied to nightly composites; new persistent point sources compared against licensed-facility register | Monthly flare-onset report with radiant power estimates and facility-match status |
| Basin-scale pad proliferation trend | Landsat OLI NDVI time-series differencing at 30 m over multi-year archive, aggregated to concession block level | Annual trend GeoTIFF and tabular summary of pad count and area by block |
| High-confidence site confirmation package | On-demand WorldView-3 tasking for SAR-flagged candidates above threshold; analyst-reviewed image interpretation | PDF site report with annotated imagery, structure identification and confidence rating for enforcement use |
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.