Routine gas flaring volume quantification
Satellites detect gas flares by their infrared signature and convert radiant power to estimated gas volumes. The method is independent of operator self-reporting, works globally every night, and is honest about its uncertainty floor.
Sensors
- VIIRS Nightfire (Suomi NPP / NOAA-20): The workhorse for routine flare monitoring. VIIRS bands M7, M8, M10, M11 and I4 span 0.7–3.7 µm; the Nightfire algorithm fits a dual-Planck curve to sub-pixel combustion sources, resolving temperatures from roughly 700 K to above 2,000 K. Nominal pixel size 375–750 m. Daily global night-time coverage; daytime retrievals are possible but noisier. Archive from 2012 (NPP) and 2018 (NOAA-20).
- Landsat 8 / 9 OLI + TIRS: 30 m multispectral plus 100 m thermal (resampled to 30 m). OLI band 7 (2.11–2.29 µm SWIR) saturates on large flares but resolves spatial structure on smaller ones. Useful for geolocating individual stacks within a facility that VIIRS sees as a single hotspot. 16-day revisit per satellite; combined 8-day revisit with both. Daytime only.
- Sentinel-3 SLSTR: Dual-view radiometer with 1 km pixels in SWIR (1.37, 1.61, 2.25 µm) and mid-IR (3.74, 10.85, 12.0 µm). Fire Radiative Power products are operationally generated by EUMETSAT. Roughly daily global revisit at mid-latitudes. Less spatially precise than VIIRS Nightfire for individual stack attribution, but provides an independent radiometric cross-check.
- MODIS (Terra / Aqua, MYD14 / MOD14): 1 km thermal anomaly product. Useful for long time-series analysis back to 2000, and for contextualising VIIRS retrievals against a two-decade baseline. Sensitivity is lower than VIIRS for small flares; the instrument is ageing and VIIRS is its operational successor.
What a flare's colour temperature gives away
A gas flare is a blackbody radiator. Its spectral emission follows Planck's law, so the ratio of radiance in two or more shortwave-infrared bands encodes the combustion temperature. The VIIRS Nightfire algorithm, developed at the Colorado School of Mines Earth Observation Group and published in the peer-reviewed literature, fits a dual-Planck model to the sub-pixel signal: one component for the hot combustion core, one for the cooler background. The output is Fire Radiative Power (FRP) in megawatts.
FRP is then converted to gas volume using empirical combustion-efficiency factors. The World Bank's Global Gas Flaring Reduction Partnership (GGFR) has published the conversion coefficients most widely used in regulatory contexts. The key assumption is that flare combustion efficiency falls in a known range, typically cited as 95–98 % for a well-operated flare. Poorly maintained or wind-affected flares can fall below 90 %, and that uncertainty propagates directly into the volume estimate. No satellite retrieval can measure combustion efficiency directly.
The size-temperature ambiguity and why it matters
Here is the fundamental problem. With only two or three spectral channels, a Planck-curve fit cannot uniquely separate a small, very hot flare from a large, cooler one that produces the same integrated radiance. VIIRS Nightfire addresses this by fitting both temperature and sub-pixel area simultaneously, but the solution is underdetermined when the flare occupies a very small fraction of the pixel. Published uncertainty estimates for individual VIIRS flare detections run to ±50 % or more on volume for small flares, narrowing to roughly ±20–30 % for large, persistent sources.
Landsat's 30 m resolution helps resolve the geometry. A facility that appears as a single VIIRS hotspot may contain three or four separate stacks at Landsat scale, each with a different radiometric signature. Fusing the two datasets, using Landsat to constrain spatial structure and VIIRS for nightly temporal sampling, reduces the ambiguity considerably. It does not eliminate it. Any volume figure derived from radiance alone carries an irreducible uncertainty that honest reporting must state.
From nightly detection to annual volume: the aggregation chain
A single VIIRS overpass gives one FRP measurement per night per facility. Converting that to an annual flared-gas volume requires integrating across all valid observations, gap-filling for cloud-obscured nights, and applying a gas-composition assumption (the calorific value of the gas affects the FRP-to-volume coefficient). Cloud cover is a genuine constraint: in equatorial regions during monsoon season, valid night-time observations may be available fewer than half the nights in a month.
The standard approach, used by the World Bank's Global Flaring Monitor and by national regulators in several jurisdictions, is to compute monthly FRP sums from all cloud-free observations and then scale to a monthly total using the observed fraction of cloud-free nights. This introduces additional uncertainty in highly cloudy environments. Annual volumes are then reported with explicit confidence intervals rather than point estimates. Regulators who accept satellite-derived volumes as compliance evidence, including some US state-level programmes and the EU's forthcoming methane regulation framework, are increasingly requiring those confidence intervals to be stated.
What the archive reveals that self-reporting does not
The VIIRS archive runs back to October 2011 on Suomi NPP, giving over a decade of nightly global flare observations independent of any operator or government report. Studies comparing satellite-derived flaring volumes against national statistics have found systematic under-reporting in several major producing regions, with discrepancies documented in peer-reviewed analysis published in journals including Remote Sensing of Environment. The satellite record does not prove intent; it does provide an objective baseline against which reported figures can be tested.
For enforcement agencies, the most useful product is not the absolute volume estimate but the anomaly: a facility that reports zero flaring on a night when VIIRS records a persistent 50 MW thermal source has a credibility problem. Trend analysis across quarters is more reliable than any single-night figure, and the long archive makes trend analysis straightforward.
Practical limits regulators should understand before commissioning analysis
Minimum detectable flare size for VIIRS Nightfire is approximately 1–2 MW FRP under good conditions. Smaller flares, common at marginal wells and gathering stations, fall below that threshold and are simply invisible to the method. Landsat can detect smaller sources in its SWIR bands, but its 16-day revisit means it will miss intermittent flaring events entirely.
Daytime flaring is harder. Solar reflected radiance overwhelms the SWIR signal in daylight, so VIIRS Nightfire is a night-time product by design. Facilities that flare predominantly during daylight hours, which is uncommon but not unknown, will be underrepresented. Sentinel-3 SLSTR and MODIS carry daytime fire products, but their sensitivity to small industrial flares is lower than the night-time VIIRS retrieval.
Satellize runs VIIRS Nightfire analytics on open-data streams and can cross-validate against Landsat and Sentinel-3 for facilities where spatial disambiguation or daytime coverage matters. The Overhead column has covered flaring methodology in detail for readers who want the technical background before commissioning a formal assessment.
What a delivered flaring report actually contains
A credible satellite-derived flaring assessment names every detected source by geographic coordinate, reports FRP in megawatts for each valid observation, states the cloud-free observation fraction, gives monthly and annual volume estimates in standard cubic metres or million standard cubic feet with explicit uncertainty ranges, and flags any nights where the thermal signature is anomalous relative to the facility's own historical baseline.
It does not claim sub-10 % accuracy on individual events. It does provide an independent, auditable, time-stamped record that is far harder to dispute than a gap in a paper trail. For regulators designing a monitoring programme, that combination of transparency and independence is the actual value of the method.
Typical figures
| Spatial resolution (VIIRS Nightfire) | 375 m (I-band) / 750 m (M-band) at nadir; degrades toward swath edge |
| Spatial resolution (Landsat 8/9 SWIR) | 30 m; useful for within-facility stack attribution |
| Revisit (VIIRS, single satellite) | Daily global night-time coverage; NPP + NOAA-20 together provide two passes per night at mid-latitudes |
| Revisit (Landsat 8 + 9 combined) | ~8 days at equator; daytime only |
| Minimum detectable FRP (VIIRS Nightfire) | ~1–2 MW under clear-sky night conditions; smaller sources are below detection |
| Key spectral bands | SWIR 1.6, 2.1–2.3 µm; mid-IR 3.7–4.0 µm; thermal 10–12 µm |
| Volume estimate uncertainty | ±20–30 % for large persistent flares; ±50 % or more for small or intermittent sources |
| Archive depth | VIIRS from October 2011 (NPP); MODIS from 2000; Landsat from 1972 (thermal from Landsat 4, 1982) |
| Cloud impact | Cloud-obscured nights excluded; gap-filling by scaling to observed cloud-free fraction introduces additional uncertainty in tropical regions |
| Latency (operational VIIRS products) | Near-real-time products available within ~3 hours of overpass via FIRMS; Nightfire refined products typically 24–48 hours |
Analytics Satellize can run
| Facility-level FRP time series | VIIRS Nightfire dual-Planck retrieval applied to all cloud-free night-time overpasses for a defined asset list | Monthly CSV or GIS layer with FRP in MW per observation, cloud-free fraction, and flagged anomalies |
| Annual flared-gas volume estimate | FRP integration with GGFR combustion-efficiency coefficients and gas-composition assumptions; cloud-gap scaling | Annual report per facility with volume in Mm³ or MMscf, explicit uncertainty intervals, and methodology annex |
| Spatial disambiguation of stacks within a facility | Landsat 8/9 SWIR band analysis fused with VIIRS hotspot centroids to attribute FRP to individual flare points | GIS layer of individual stack locations with estimated FRP contribution per stack |
| Anomaly detection against facility baseline | Statistical comparison of current-period FRP against the facility's own multi-year VIIRS archive; z-score flagging | Alert feed (JSON or email) when a facility's thermal signature deviates significantly from its historical pattern |
| Cross-validation against self-reported volumes | Comparison of satellite-derived annual estimates against operator or national statistics submissions; discrepancy scoring | Regulatory briefing document with discrepancy table and confidence-weighted interpretation |
| Basin-level flaring trend analysis | Aggregated VIIRS Nightfire retrievals across all detected sources within a defined basin polygon, trended quarterly over the full archive | Interactive dashboard or static report showing basin-level flared volumes from 2012 to present |
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.