Gas flaring volume estimation from nocturnal radiance
VIIRS DNB and SWIR channels can estimate gas flaring volumes at oil fields and refineries globally, providing regulators and investors with independently derived production signals. The World Bank GGFR publishes annual benchmarks using this method.
Sensors
- VIIRS Day-Night Band (DNB): 375 m native resolution, daily global coverage from Suomi-NPP and NOAA-20. Detects low-radiance flares but saturates above roughly 2.6 × 10⁻⁷ W cm⁻² sr⁻¹ in the panchromatic visible band, limiting it to smaller or more distant flares without supplementary channels.
- VIIRS M-band SWIR (M10, M11): 375 m resolution at 1.61 µm and 2.25 µm. These channels saturate at much higher radiance than DNB, capturing the thermal peak of intense flares that overwhelm the visible band. Used together with DNB, they extend the dynamic range across the full global population of flare intensities.
- Landsat 8/9 OLI and TIRS: 30 m multispectral resolution with a 16-day revisit per satellite. OLI Band 7 (2.2 µm SWIR) resolves individual stacks at dense facilities where VIIRS cannot distinguish sources. Archive extends to 1972 across the broader Landsat programme, enabling long-run trend analysis.
- Sentinel-2 MSI: 20 m resolution in SWIR bands (Band 11 at 1.61 µm, Band 12 at 2.19 µm), 5-day revisit at the equator with both Sentinel-2A and 2B. Useful for attributing flares to individual stacks or process units at complex facilities, though the sensor is not optimised for high-radiance thermal sources and can saturate on very intense flares.
What the radiance signal actually tells you
Combustion in a gas flare produces a blackbody-like emission spectrum peaking in the near-infrared and SWIR, with a tail extending into the visible. The VIIRS Day-Night Band detects that visible tail at night. The brightness of the DNB signal is proportional to the flare's radiant heat flux, which in turn correlates with the volume of gas being combusted, once combustion efficiency and flare temperature are accounted for.
The relationship is not perfectly linear. Flare temperature varies with gas composition, stack design, and wind. Combustion efficiency in real operations ranges from roughly 85 to 98 percent, introducing uncertainty in the conversion from radiance to volume. Published methods, including those underpinning the World Bank GGFR annual estimates, apply empirical calibration against ground-truth metering data to constrain this uncertainty. The GGFR reported approximately 139 billion cubic metres of gas flared globally in 2023 using this approach, giving analysts a validated reference point against which field-level estimates can be cross-checked.
Why two sensors are better than one
The DNB saturates on intense flares. A major production field in the Permian Basin or the Niger Delta can push the DNB pixel into saturation, at which point you know a flare is large but cannot quantify it. The VIIRS M-band SWIR channels saturate at a much higher radiance threshold, so they capture the signal that the DNB cannot. Running both channels together spans a dynamic range that covers everything from a small wellpad flare detectable only at night to a supergiant offshore platform burning several million cubic metres per day.
For facilities where individual stack attribution matters, Landsat 8/9 OLI or Sentinel-2 SWIR imagery at 20 to 30 m resolution can separate stacks that sit within the same 375 m VIIRS pixel. The trade-off is revisit: Landsat revisits a given point every 8 days with both satellites combined, Sentinel-2 every 5 days. Cloud cover reduces effective revisit further, and neither sensor is tasked nocturnally by default, so daytime SWIR observations capture the thermal anomaly rather than the visible flame. Analysts typically use VIIRS for high-frequency volume tracking and Landsat or Sentinel-2 for spatial attribution.
Reading flaring as a production signal
Associated gas is a by-product of oil production. When a field lacks pipeline infrastructure or gas processing capacity, operators flare rather than capture. The volume of gas flared therefore tracks oil production closely at fields where the gas-to-oil ratio is relatively stable. This makes flaring radiance a useful independent proxy for production activity, particularly in jurisdictions where official output figures are delayed, revised, or simply not published.
The signal has real limits as a production proxy. A field that installs gas capture infrastructure will show declining flare radiance even as oil output rises. Planned maintenance shutdowns, regulatory enforcement actions, and seasonal demand changes all affect flaring independently of production. Analysts who treat flaring radiance as a single-variable production indicator without accounting for these factors will draw incorrect conclusions. The honest use is as a corroborating signal alongside other indicators, not a standalone production meter.
For compliance monitoring, the picture is somewhat cleaner. A facility that reports zero flaring but shows persistent nocturnal radiance at its location has a discrepancy that is difficult to explain away. Regulators in several jurisdictions have begun incorporating satellite flaring data into inspection targeting, precisely because the signal is independent of operator self-reporting.
Confusion sources and detection limits
Not every persistent nocturnal light source is a flare. Fishing vessels using bright lights to attract squid, industrial facilities with continuous outdoor lighting, and gas-fired power stations all produce radiance that can be confused with flaring in automated detection pipelines. The VIIRS Nightfire algorithm, published by the Colorado School of Mines Earth Observation Group and distributed via the EOGDATA platform, applies temperature-based discrimination to separate high-temperature combustion sources from low-temperature artificial lighting. Even so, urban industrial zones require manual review.
The 375 m spatial resolution of VIIRS means that a cluster of wellpads within a few hundred metres of each other appears as a single source. Attribution of flaring to a specific operator within a dense field requires the higher-resolution Landsat or Sentinel-2 imagery described above. Below roughly 10 MW of radiant heat flux, VIIRS detection becomes unreliable, meaning small routine flaring events at individual wellheads may go undetected. Cloud cover is a further constraint: persistent cloud over equatorial producing regions such as parts of West Africa can reduce monthly detection rates significantly.
From pixel to deliverable
A practical flaring monitoring programme typically produces three things: a time series of estimated flaring volumes per facility or field, an anomaly alert when a facility's flaring deviates materially from its historical baseline, and a spatial map of active flare locations updated on a monthly or near-real-time cadence. The volume estimates carry an uncertainty range, honestly stated, that reflects the calibration assumptions and the sensor's dynamic range limitations.
Satellize runs this class of analytics on open VIIRS and Landsat archives, with commercial high-resolution tasking added where stack-level attribution is required under a client licence. The Kingdom of Tonga crop-estimation programme demonstrates the same underlying workflow: open-constellation data processed through calibrated empirical models to produce an actionable number, not just a map. For energy clients, the deliverable is a structured data feed or GIS layer that slots into existing compliance or investment workflows, with methodology documentation sufficient for regulatory submission.
Typical figures
| Primary spatial resolution | 375 m (VIIRS DNB and M-band); 30 m (Landsat 8/9 OLI); 20 m (Sentinel-2 SWIR) |
| Revisit frequency | Daily global (VIIRS, Suomi-NPP + NOAA-20 combined); 8 days (Landsat 8+9 combined); 5 days (Sentinel-2A+2B combined at equator) |
| Spectral bands used | VIIRS DNB (~0.5–0.9 µm panchromatic); VIIRS M10 (1.61 µm), M11 (2.25 µm); Landsat OLI Band 7 (2.11–2.29 µm); Sentinel-2 Band 11 (1.57–1.66 µm), Band 12 (2.10–2.28 µm) |
| DNB saturation threshold | Approximately 2.6 × 10⁻⁷ W cm⁻² sr⁻¹; intense flares require SWIR channels to avoid saturation |
| Minimum detectable flare (VIIRS) | Approximately 10 MW radiant heat flux under clear-sky conditions; smaller events may be missed |
| Volume estimation uncertainty | Typically ±20–30% at field level using published calibration methods; higher for individual stacks |
| Archive depth | VIIRS from 2012 (Suomi-NPP launch); Landsat from 1972 (Landsat 1); Sentinel-2 from 2015 |
| Latency (VIIRS open archive) | Near-real-time products available within hours via NASA FIRMS and EOGDATA Nightfire; monthly composites within days of month-end |
| Coverage | Global, including offshore platforms; cloud cover reduces effective detection in tropical regions |
| Delivery formats | GeoTIFF, GeoJSON, CSV time series, API feed; methodology report for regulatory use |
Analytics Satellize can run
| Monthly flaring volume estimate per facility | VIIRS DNB + SWIR radiance-to-volume conversion using published GGFR/Nightfire calibration coefficients | Structured CSV or JSON time series with uncertainty bounds, updated monthly |
| Flare anomaly alert | Statistical deviation from rolling 12-month baseline radiance; threshold set per facility type | Email or API alert with facility ID, deviation magnitude, and supporting imagery thumbnail |
| Active flare location map | VIIRS Nightfire temperature-discriminated detection, filtered against known non-flare light sources | Monthly GeoJSON or GIS layer of confirmed flare centroids with radiant heat flux estimates |
| Stack-level flare attribution | Landsat 8/9 OLI or Sentinel-2 SWIR hotspot detection at 20–30 m resolution, co-registered to facility footprint | Facility-level report identifying which stacks or process units are active, with imagery evidence |
| Multi-year flaring trend analysis | Annual VIIRS composite time series from 2012 to present, normalised for cloud cover and satellite geometry | Trend report with field-level charts suitable for ESG disclosure or regulatory filing |
| Compliance discrepancy flag | Cross-referencing satellite-derived flaring estimates against operator self-reported volumes where public data exists | Discrepancy summary table with confidence rating, for use in regulatory inspection targeting or investor due diligence |
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.