Methane point-source detection and quantification from hyperspectral imagers

Hyperspectral imagers in the shortwave infrared can detect individual methane plumes from oil and gas facilities, landfills and coal mines, attributing emissions to specific sources with quantified uncertainty.

Sensors

GHGSat-D/C series: Commercial high-resolution imaging spectrometers targeting the 2.3 µm methane absorption band. Ground sampling distance approximately 25 m, enabling attribution to individual pieces of equipment. Minimum detectable emission rate roughly 100 kg/h under clear-sky conditions. Tasked on demand; revisit depends on constellation size and scheduling.
EMIT (ISS): NASA imaging spectrometer covering 380–2500 nm at 60 m ground sampling distance, mounted on the International Space Station. Designed for mineral dust mapping but demonstrated capable of detecting large methane point sources above roughly 500 kg/h. Non-sun-synchronous ISS orbit gives variable overpass times; not designed for systematic methane monitoring but useful for opportunistic detection of super-emitters.
TROPOMI / Sentinel-5P: Passive UV-SWIR spectrometer with a 7 × 5.5 km pixel (resampled to 5.5 × 3.5 km since August 2019). Daily global coverage. Detects column-averaged methane mixing ratios (XCH4) but cannot resolve individual facility-scale plumes; essential for regional context and cross-validation of point-source inventories.
MethaneSAT: EDF-funded satellite launched March 2024, targeting the 1.65 µm and 2.3 µm absorption features at roughly 100–200 m resolution with a wide 200 km swath. Designed for systematic regional and facility-scale monitoring. Quantification performance figures are being validated against aircraft campaigns; treat published pre-launch specifications as provisional until peer-reviewed results appear.

Why 2.3 µm and not something simpler

Methane absorbs solar radiation in several shortwave-infrared windows, but the 2.3 µm band is particularly useful for point-source detection because the absorption cross-section is large enough to produce a measurable column enhancement over a compact plume, while the atmospheric background is relatively stable. The 1.65 µm band is also used, notably by MethaneSAT, and offers somewhat lower interference from water vapour at the cost of a weaker methane signal.

The measurement principle is straightforward in concept: the sensor compares sunlight reflected from the surface through a plume-laden column against adjacent clear-sky pixels. Excess absorption in the methane bands, after accounting for surface albedo variation and atmospheric scattering, gives a column enhancement in parts per million metres (ppm·m). Convert that to a mass column, apply a wind-speed estimate, and you have a mass flux in kg/h. Each of those steps carries uncertainty, and the wind step is usually the largest.

The wind problem is not a footnote

Mass-flux retrieval from a satellite image requires knowing the wind speed and direction at plume height at the moment of overpass. Reanalysis products such as ERA5 report winds at hourly resolution on a roughly 31 km grid. At a single facility, the local wind can differ from the reanalysis value by 20–50% depending on terrain, stability class and time of day. That uncertainty propagates directly into the emission estimate: a 30% wind error produces a 30% emission error, before any retrieval uncertainty is added.

Aircraft campaigns that fly simultaneous transects can reduce wind uncertainty substantially, but they are expensive and cannot be sustained operationally. Some operators use meteorological masts at the facility itself; others apply ensemble wind methods that sample multiple reanalysis products and report a credible interval rather than a point estimate. Honest reporting of methane flux estimates should always state the wind source and its associated uncertainty, typically ±30–50% for reanalysis-only retrievals.

Detection limits and the super-emitter tail

Current commercial hyperspectral sensors, specifically the GHGSat constellation, have a published minimum detectable emission rate of approximately 100 kg/h under clear-sky, high-albedo conditions. That threshold rises over dark surfaces such as dense vegetation or open water, where the reflected solar signal is weak. Cloud cover renders a scene completely unusable; there is no radar analogue for this measurement, so cloudy revisits are simply lost.

The practical consequence is that sensors at this resolution are designed to catch super-emitters, the roughly 5% of facilities that account for a disproportionate share of total sector emissions, a pattern documented in aircraft survey campaigns over the Permian Basin and other major producing regions. Facilities emitting below the detection threshold are invisible to current spaceborne hyperspectral sensors. TROPOMI can detect very large aggregated regional anomalies but cannot separate individual sources at typical facility density. The gap between 100 kg/h and the true emission floor of a leaking valve remains a genuine measurement challenge.

Attribution: knowing which stack, not just which county

At GHGSat's 25 m ground sampling distance, the plume origin can often be attributed to a specific tank battery, compressor station or wellhead rather than simply to a facility. That attribution matters for operators who need to dispatch maintenance crews and for regulators who need to enforce permit conditions. EMIT's 60 m pixels allow attribution to a facility complex but rarely to individual equipment.

Attribution confidence depends on wind direction stability during the overpass. A steady wind produces a coherent plume that can be traced back to a point. A turbulent or shifting wind produces a diffuse, difficult-to-attribute cloud. Gaussian plume fitting and more sophisticated atmospheric transport inversions both require some minimum degree of plume coherence to return a credible source location. When the wind is light and variable, the honest answer is often a polygon rather than a point.

Integrating point-source detections into a monitoring programme

A single overpass is a snapshot. Repeated tasking over weeks and months reveals whether an emission event was a transient equipment fault or a persistent structural leak, a distinction with very different regulatory and commercial implications. Systematic monitoring also allows operators to verify that repairs were effective, closing the loop between detection and remediation.

TROPOMI's daily global coverage provides the regional methane background against which point-source detections can be contextualised. A facility-scale GHGSat detection that coincides with an elevated TROPOMI column in the same basin on the same day is more credible than one that appears in isolation. Combining sensors at different scales is standard practice in published methane attribution studies. Satellize runs TROPOMI column-anomaly screening as a first-pass layer to prioritise commercial tasking requests, reducing the number of costly high-resolution acquisitions needed to characterise a basin. The analytics approach draws on the same published inversion methods used in academic methane attribution work, applied operationally rather than retrospectively.

Honest limits before you commission a survey

Cloud cover is the principal operational constraint. In humid tropical regions or during winter at high latitudes, clear-sky overpass probability can fall below 30% in a given month, making consistent monitoring difficult without a large constellation and careful scheduling.

Surface albedo matters more than most buyers expect. Retrievals over bright desert surfaces or snow are generally more reliable than those over dark boreal forest or open water. Published detection-limit figures typically assume moderate-to-high albedo; ask for the albedo-adjusted threshold relevant to your facilities before interpreting a negative detection as confirmation of compliance. Finally, MethaneSAT's performance figures are still being validated; the wide-swath regional capability is real, but treat quantitative claims with appropriate caution until independent validation studies are published.

Typical figures

Spatial resolution (GHGSat)	~25 m ground sampling distance
Spatial resolution (EMIT / ISS)	60 m ground sampling distance
Spatial resolution (TROPOMI)	5.5 × 3.5 km per pixel (post-August 2019)
Revisit (TROPOMI)	Daily global coverage
Revisit (GHGSat)	On-demand tasking; frequency depends on constellation scheduling
Key spectral bands	2.3 µm methane absorption window (primary); 1.65 µm (MethaneSAT secondary)
Minimum detectable emission rate	~100 kg/h (GHGSat, clear sky, moderate-to-high albedo); higher over dark surfaces
Wind uncertainty contribution to flux error	±30–50% typical for reanalysis-only wind input
Cloud tolerance	None; optical/SWIR measurement requires clear sky
TROPOMI archive depth	From May 2018 (Sentinel-5P launch)

Analytics Satellize can run

Regional methane column anomaly screening	TROPOMI XCH4 column retrieval, anomaly detection against seasonal baseline	Monthly basin-level anomaly map (GeoTIFF or GIS layer) with flagged areas for priority tasking
Facility-scale plume detection report	Matched-filter or IMAP-DOAS retrieval on GHGSat or EMIT imagery	Per-facility PDF report with plume image, detected column enhancement and detection/non-detection verdict
Emission rate estimate with uncertainty bounds	Integrated mass enhancement method using ERA5 wind ensemble	Point-source emission estimate in kg/h with ±1σ credible interval, delivered as structured data feed
Persistent super-emitter identification	Time-series aggregation of repeat detections across multiple overpasses	Ranked facility list with detection frequency, median emission rate and first/last detection dates
Repair verification overpass	Targeted re-tasking after reported remediation, compared against pre-repair baseline	Before/after detection report confirming emission reduction or flagging continued leakage
Multi-sensor cross-validation layer	Spatial and temporal co-location of TROPOMI regional signal with facility-scale detections	Attribution confidence score per detection, GIS layer with sensor provenance metadata

Sources

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.