Unlit methane venting detection at offshore oil and gas platforms
Routine unlit venting leaves no thermal signal, hiding it from flaring inventories. Combining TROPOMI methane columns, Sentinel-1 platform catalogues, and ERA5 wind fields exposes it.
Sensors
- Sentinel-5P TROPOMI: Measures tropospheric methane columns (XCH4) at 5.5 × 7 km pixel resolution (post-2019 upgrade from 7 × 3.5 km along-track), daily global coverage. Minimum detectable column enhancement above background is roughly 10–20 ppb over a single overpass, with persistent sources detectable at lower thresholds through temporal stacking.
- Sentinel-1 SAR (C-band, IW mode): Provides 10 m resolution, all-weather, day-night imagery used to build and maintain a georeferenced catalogue of platform locations, including small unmanned wellhead platforms that do not appear in public AIS or BOEM-equivalent registries. Revisit is 6–12 days at mid-latitudes depending on orbit geometry and acquisition mode.
- VIIRS DNB / Nightfire: Detects combustion events (flares) at night via thermal and near-infrared radiance. Its role here is deliberate absence: a platform showing persistent methane enhancement but no concurrent Nightfire detection is the primary detection criterion for unlit venting. VIIRS spatial resolution is approximately 375 m at nadir for the DNB band.
- GHGSat-D/C series: Commercial shortwave-infrared imaging spectrometers with roughly 25 × 25 m pixel resolution and a 12 × 12 km scene footprint. Used for targeted follow-up on platforms flagged by TROPOMI, providing facility-level emission rate estimates. Tasked on demand; not a continuous monitoring asset.
- ERA5 reanalysis (ECMWF): Hourly wind fields at 0.25-degree resolution used in Gaussian plume and Bayesian inverse models to back-calculate surface emission rates from observed TROPOMI column anomalies. Not a satellite sensor, but integral to the quantification step.
Why unlit venting is the gap in every flaring inventory
Global gas flaring inventories, including the World Bank's GGFR programme and NOAA's satellite-derived estimates, rest almost entirely on thermal detection. A flare produces a bright, hot combustion signal that VIIRS Nightfire picks up reliably even at its 375 m resolution. Unlit venting produces none of that. The gas escapes cold, and the platform looks identical to one that is simply shut in.
This matters because venting methane is, molecule for molecule, a far more potent near-term climate forcing than the CO2 produced by combusting the same gas. Operators vent rather than flare for various reasons: flare tip maintenance, low-pressure wells where ignition is unreliable, or deliberate avoidance of visible flare activity during inspection periods. None of these reasons appear in any thermal-based inventory, and the gap is not trivial. Published studies using TROPOMI have found persistent methane enhancements above offshore fields with no corresponding flare signal.
Building the platform catalogue from SAR before touching the gas data
The first analytical step is geometric, not spectroscopic. Sentinel-1's C-band SAR backscatter is exceptionally sensitive to the corner-reflector geometry of metal structures on a flat sea surface. Fixed platforms, floating production units and even small wellhead jackets appear as bright point targets well above the surrounding ocean clutter. Processing a multi-year Sentinel-1 archive over a target basin produces a persistent-scatterer catalogue that is more complete than any public registry, because it includes structures that predate mandatory reporting or fall below regulatory thresholds.
Each catalogued structure is assigned a position, a first-detection date, and a persistence score. This catalogue becomes the spatial prior for the methane attribution step: TROPOMI pixels are only interrogated for enhancement if they contain a known platform within the wind-fetch footprint at overpass time. That constraint dramatically reduces false positives from agricultural or wetland methane sources that would otherwise confound offshore attribution.
Reading the column: what TROPOMI can and cannot resolve
TROPOMI's XCH4 retrieval uses the 2.3-micrometre shortwave-infrared absorption band, comparing methane absorption against a modelled atmospheric column and surface albedo. At 5.5 × 7 km resolution, a single offshore platform's venting plume is almost never resolved as a discrete spatial feature. What TROPOMI sees is an enhancement distributed across one or several pixels downwind of the source.
The honest detection floor matters here. A single overpass over a single platform venting at rates below roughly 1,000–2,000 kg per hour will typically not produce a statistically significant pixel-level anomaly against the background variability in TROPOMI's XCH4 product. Temporal stacking, averaging 20 or more overpasses under consistent wind conditions, lowers that effective threshold considerably, but it also means the method is better suited to identifying chronic venting over weeks or months than to catching a single overnight event. Cloud cover is a hard limit: TROPOMI's shortwave-infrared retrieval fails under even thin cloud, and offshore basins in the North Sea or Gulf of Guinea can have cloud fractions above 80 percent in winter months, reducing usable overpasses significantly.
From column anomaly to emission rate: the inverse-modelling step
Once a persistent column enhancement is associated with a catalogued platform, the next step is converting the observed XCH4 anomaly into a surface flux estimate. The standard published approach applies a Gaussian plume model or a more sophisticated Bayesian inversion, using ERA5 hourly wind speed and direction at the overpass time to define the expected spatial footprint of a point source at the platform location.
The method was formalised for TROPOMI by Varon et al. (2018, 2019) and has been applied to both onshore and offshore facilities in peer-reviewed literature. Emission rate uncertainties from a single overpass are large, typically a factor of two or more, because ERA5 wind fields at 0.25-degree resolution do not capture local turbulence and boundary-layer structure accurately at the platform scale. Averaging across multiple overpasses reduces random error but introduces bias if venting is episodic rather than continuous. The output is best reported as a range, not a point estimate, and any regulatory use of these figures should treat them as screening-level evidence rather than metered measurements.
The null-flare test: using VIIRS as a negative control
The analytical signature of unlit venting is not just a methane enhancement. It is a methane enhancement coincident with a VIIRS Nightfire null. For each flagged platform and each relevant time window, the workflow checks whether any Nightfire thermal anomaly was recorded at the platform location. A platform with persistent XCH4 enhancement and zero Nightfire detections over the same period is the primary detection class.
A secondary class covers platforms where Nightfire records occasional flaring but the methane enhancement is disproportionately large relative to what the flare volume would produce. This suggests a combination of flared and vented gas, with the vented fraction invisible to thermal sensors. Separating these two classes requires care: VIIRS Nightfire has a detection floor for small or low-temperature flares, and a very small flare may be present but undetected. GHGSat tasking on flagged platforms can resolve ambiguity by providing a spatially resolved emission map at roughly 25 m resolution, confirming whether the methane source is co-located with the flare stack or elsewhere on the deck.
What this looks like in practice, and where Satellize fits
The workflow produces three deliverables: a georeferenced platform catalogue with persistence and activity metadata, a ranked list of platforms showing anomalous methane-to-flare ratios over a user-defined monitoring window, and per-platform emission rate estimates with uncertainty bounds derived from the inverse model. These can be delivered as GIS layers, structured reports, or API feeds depending on the client's regulatory or investment workflow.
Regulators in licensing jurisdictions use this kind of analysis to prioritise inspection visits, since physically boarding every platform in a large offshore basin is impractical. Investors and insurers use it to assess whether an operator's self-reported venting volumes are consistent with what the atmosphere shows. Satellize runs this stack on open constellations, adding commercial GHGSat tasking on client licence where facility-level confirmation is needed. The Tonga crop-estimation programme is a different domain entirely, but it reflects the same principle: that open satellite data, properly processed, can answer questions that ground-based reporting leaves open. If you are assessing a specific basin or licence block, the right starting point is a scoping call to define the monitoring window and the acceptable uncertainty threshold for your regulatory context.
Typical figures
| TROPOMI XCH4 spatial resolution | 5.5 × 7 km per pixel (post-August 2019 upgrade) |
| TROPOMI revisit | Daily global coverage; usable overpasses reduced by cloud (shortwave-infrared retrieval fails under cloud) |
| TROPOMI single-overpass detection floor (offshore point source) | Approximately 1,000–2,000 kg CH4/hour; lower with temporal stacking over 20+ overpasses |
| Sentinel-1 SAR platform detection resolution | 10 m (IW mode); revisit 6–12 days at mid-latitudes |
| VIIRS Nightfire flare detection resolution | ~375 m at nadir; detection floor for small/low-temperature flares not published as a fixed threshold |
| GHGSat follow-up resolution | ~25 × 25 m pixel, 12 × 12 km scene; tasked on demand, not continuous |
| ERA5 wind field resolution | 0.25 degrees (~28 km), hourly; introduces emission rate uncertainty of factor 2 or more per overpass |
| Sentinel-1 SAR archive depth | From 2014 (Sentinel-1A launch); TROPOMI archive from October 2017 |
| Emission rate output uncertainty | Factor of 2 or more (single overpass); improves with multi-overpass averaging but remains screening-level |
| Deliverable formats | GeoJSON/Shapefile platform catalogue, CSV emission ranking, PDF regulatory report, API feed |
Analytics Satellize can run
| Georeferenced offshore platform catalogue | Sentinel-1 SAR persistent-scatterer detection and multi-temporal bright-target extraction over user-defined basin | GeoJSON or Shapefile layer with platform position, first-detection date, and persistence score; updated on new Sentinel-1 acquisitions |
| Methane column anomaly map per platform | TROPOMI XCH4 pixel extraction within ERA5 wind-fetch footprint of each catalogued platform; background subtraction using upwind reference pixels | Time-series CSV of per-platform XCH4 anomaly (ppb above background) for each cloud-free overpass in the monitoring window |
| Null-flare coincidence flag | VIIRS Nightfire event database queried against platform catalogue for concurrent detections; platforms ranked by methane-enhancement-to-flare-activity ratio | Ranked alert list of platforms with persistent XCH4 enhancement and low or zero Nightfire detections; updated monthly |
| Inverse-modelled emission rate estimate | Gaussian plume or Bayesian inversion (Varon et al. published method) using ERA5 hourly wind at overpass time; multi-overpass averaging to reduce random error | Per-platform emission rate range (kg CH4/hour) with stated uncertainty bounds; suitable for regulatory screening, not metered compliance |
| GHGSat targeted confirmation | Commercial tasking of GHGSat-C imaging spectrometer on flagged platforms; facility-scale methane map at ~25 m resolution | Scene-level emission map confirming source location on platform (flare stack versus other deck equipment); PDF report with spatial overlay |
| Basin-level venting inventory | Aggregation of per-platform estimates across all catalogued structures in a licence block or national EEZ; comparison against operator self-reported figures where available | Structured basin report with total estimated unlit venting flux, confidence classification per platform, and recommended inspection priority ranking |
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.