Sensors
- VIIRS Day/Night Band (DNB) and M-band (Suomi-NPP / NOAA-20): DNB provides 750 m nominal pixel size; M-band channels M10 and M11 (1.61 µm and 2.25 µm) at 375 m are the primary flare-detection channels. Daily global revisit. Saturation at very bright flares is managed by the VIIRS low-gain stage. Detection threshold is roughly 1–3 MW radiant power, meaning small or intermittent flares are missed.
- Landsat 8/9 OLI and TIRS: OLI Band 7 (2.11–2.29 µm) at 30 m resolution resolves individual flare stacks that VIIRS cannot separate at dense fields. Revisit is 16 days per satellite, 8 days combined for Landsat 8 and 9. Useful for cross-calibrating VIIRS radiance at known large flares and for detecting sub-MW sources.
- Sentinel-3 SLSTR: Dual-view thermal and short-wave infrared channels at 500 m and 1 km. Near-daily revisit at mid-latitudes. SLSTR's S5 (1.61 µm) and S6 (2.25 µm) channels overlap spectrally with VIIRS M-band, providing an independent radiance cross-check and extending the record where NOAA-20 data are unavailable.
- VIIRS Nightfire (VIIRS NF, Colorado School of Mines / NOAA): A published algorithm that fits a Planck-curve to multi-band VIIRS radiance to retrieve flare temperature (typically 1,400–1,800 K for gas flares) and radiant heat. Temperature retrieval reduces the uncertainty in converting radiant power to combustion volume. Archive runs from 2012.
What a flare actually tells a satellite
A gas flare emits thermal radiation across a predictable spectrum. At combustion temperatures of roughly 1,400 to 1,800 K, the peak emission falls in the short-wave infrared, squarely within VIIRS M-band channels M10 and M11. The VIIRS Nightfire algorithm, developed at the Colorado School of Mines and published through NOAA/NESDIS, fits a Planck curve to the multi-band radiance to recover two quantities: flare temperature and radiant heat in megawatts. That radiant heat is the gateway to volume.
The conversion from radiant power to gas volume uses combustion physics. Published relationships, most thoroughly documented in the World Bank Global Gas Flaring Tracker methodology, express flare gas volume as a function of radiant heat, combustion efficiency (typically assumed at 98% for a well-maintained flare, lower for unlit or cold flares), and gas heating value. Gas composition varies by basin, so analysts apply field-specific priors drawn from regulatory filings or published reservoir characterisation where available. The resulting volume estimate carries an uncertainty of roughly ±20 to ±30% at field level, which is honest but still material for royalty disputes running into millions of dollars.
The World Bank tracker as public baseline
The World Bank Global Gas Flaring Tracker, produced annually in collaboration with NOAA, is the authoritative public reference for this method. It reports country-level and basin-level flaring volumes derived from VIIRS, and has been used by governments and the GGFR partnership to benchmark national reporting against satellite observation since the mid-2000s. The tracker's country totals are publicly available and frequently diverge from figures submitted to regulators, which is precisely what makes satellite-derived estimates commercially interesting.
For royalty and carbon-liability work, the tracker provides a credible, peer-reviewed methodology that analysts can cite in disputes. The limitation is that it operates at country and large-basin granularity. Field-level disaggregation, the resolution needed to audit a specific concession or price a specific operator's liability, requires additional processing: isolating individual flare pixels, applying field-specific gas-composition priors, and integrating over time series rather than annual snapshots.
Where the method breaks and what to do about it
VIIRS has a detection floor. Flares below roughly 1 to 3 MW radiant power, corresponding to very small or intermittent venting events, fall into noise. At a dense field where multiple stacks sit within a single 375 m pixel, VIIRS cannot attribute radiance to individual sources. These are not edge cases: in mature onshore basins with closely spaced wellpads, co-location ambiguity is the norm rather than the exception.
Landsat 8 and 9 partially address both problems. At 30 m resolution, OLI Band 7 can separate stacks that VIIRS merges, and the lower saturation threshold of OLI in its standard gain setting allows detection of smaller sources. The trade-off is revisit: 8 days combined versus daily for VIIRS. The practical approach is to use VIIRS for temporal frequency and total-field radiance budgets, then use Landsat to apportion that radiance across individual sources identified in the higher-resolution imagery. Cloud cover affects both sensors equally; in persistently cloudy regions such as the Niger Delta, monthly compositing may be the only viable strategy, and seasonal gaps in the time series must be flagged explicitly rather than interpolated silently.
From radiance to a royalty audit
Royalty regimes in most petroleum-producing jurisdictions calculate the state's take on produced volumes, with flared gas either deducted from the royalty base or subject to a separate flaring penalty. The audit question is whether the operator's reported flared volume matches what the satellite sees. A systematic underreport of flaring inflates the apparent production volume subject to royalty, or reduces the penalty base.
A satellite-derived time series, aggregated to monthly or quarterly intervals to match reporting periods, can be compared directly against operator submissions. Discrepancies above a threshold, say two standard deviations from the expected radiance-volume relationship calibrated on fields with metered data, flag candidates for regulatory inquiry. This is not a definitive proof of misreporting; flare composition uncertainty, combustion efficiency variation, and sensor noise all contribute to the spread. But it is an independent, non-falsifiable signal. An operator cannot retroactively alter what a satellite recorded at 1:30 a.m. local time in 2021.
Carbon-liability pricing follows the same logic. Under emerging carbon-border adjustment mechanisms and voluntary carbon markets, the flared volume estimate feeds directly into a Scope 1 emissions calculation. Methane slip from incomplete combustion, typically 1 to 5% of flare gas by volume depending on combustion efficiency, adds a further warming contribution that some buyers now price separately.
Delivery format and what Satellize runs
The analytic outputs for this use case are field-level time series rather than snapshots. A useful deliverable is a monthly CSV or GIS layer giving, for each identified flare cluster: centroid coordinates, radiant heat in MW, estimated gas volume in million standard cubic feet or thousand standard cubic metres, temperature where Nightfire retrieval converges, and a quality flag indicating whether the estimate is cloud-free, partially composited, or flagged for co-location ambiguity. Confidence intervals should accompany every volume figure.
Satellize runs this pipeline on open constellations, primarily VIIRS and Landsat, and can add commercial tasking for targeted high-resolution confirmation. The same analytical infrastructure that underpins the Tonga crop-estimation programme handles multi-source time-series ingestion and field-boundary attribution. Clients in royalty auditing typically want a standing monthly feed covering a defined concession list, delivered ahead of the operator's reporting deadline so that discrepancies can be raised before payments are settled rather than after.
Typical figures
| Primary detection pixel size | 375 m (VIIRS M-band); 750 m (VIIRS DNB); 30 m (Landsat OLI Band 7) |
| Revisit frequency | Daily (VIIRS, Suomi-NPP + NOAA-20 combined); 8 days (Landsat 8 + 9 combined); near-daily (Sentinel-3 SLSTR) |
| Spectral bands used | VIIRS M10 (1.61 µm), M11 (2.25 µm), DNB (0.5–0.9 µm); Landsat OLI Band 7 (2.11–2.29 µm); SLSTR S5/S6 (1.61/2.25 µm) |
| Minimum detectable flare | Approximately 1–3 MW radiant power (VIIRS); sub-MW sources detectable with Landsat OLI at 30 m |
| Flare temperature retrieval range | Approximately 1,400–1,800 K for gas flares via VIIRS Nightfire Planck-curve fitting |
| Volume estimation uncertainty | ±20–30% at field level, depending on gas composition prior and combustion efficiency assumption |
| Archive depth | VIIRS from 2012; Landsat from 1972 (TM/ETM+/OLI); Sentinel-3 from 2016 |
| Latency | VIIRS near-real-time products available within 3–6 hours of overpass; monthly composites typically produced within 5 days of month end |
| Coverage | Global daily; polar and high-latitude fields covered by multiple daily VIIRS overpasses |
| Delivery formats | Monthly GeoTIFF radiance composites, field-level CSV time series, GeoJSON flare-cluster layer with volume and quality attributes |
Analytics Satellize can run
| Field-level monthly flaring volume estimate | VIIRS Nightfire radiant-heat retrieval calibrated against published combustion-physics conversion factors and field-specific gas-composition priors | Monthly CSV and GeoJSON layer per concession, with volume in Mscf or km³, temperature, radiant heat, and confidence interval |
| Operator-reported versus satellite-derived flaring discrepancy flag | Statistical comparison of satellite volume estimates against regulatory submissions; z-score flagging of outliers beyond two standard deviations | Quarterly discrepancy report with field-level flags, magnitude of divergence, and supporting radiance evidence for regulatory or legal use |
| Scope 1 flaring emissions estimate (CO₂ and methane slip) | Volume-to-emissions conversion using published combustion efficiency ranges and IPCC Tier 2 emission factors; methane slip calculated at 1–5% of flare gas volume | Annual emissions report per asset or portfolio, formatted for carbon-market disclosure or carbon-border adjustment mechanism submission |
| Co-located flare source apportionment | Landsat OLI Band 7 at 30 m used to identify individual stacks within VIIRS pixels; radiance apportioned by Landsat-derived source fraction | Stack-level attribution layer overlaid on field boundary map, delivered as GeoTIFF and GIS-ready shapefile |
| Multi-year flaring trend and intensity baseline | VIIRS archive compositing from 2012 to present; field-level trend decomposition separating production-driven variation from efficiency improvement or regulatory response | Time-series chart and data table per field, suitable for ESG due-diligence packs or pre-acquisition asset assessment |
| Cloud-gap-flagged monthly composite | Cloud masking using VIIRS quality flags; partial-month compositing with explicit coverage fraction reported per field; no silent gap-filling | Monthly raster composite with per-pixel cloud-fraction metadata, ensuring audit trail integrity for regulatory submissions |
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.