BRENNEN

Industries / Industrial equipment and services / Industrial services / Leak detection

Aircraft, drones and satellites

This work looks for leaks from above: instruments on aircraft, drones and satellites either see methane by the sunlight it absorbs or measure it in the air they fly through, so that wide areas can be checked in days.

7

Milestones achieved, from 2015 to 2026.

3

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 2015

    A few sources dominate

    In April 2015 NASA's Jet Propulsion Laboratory (JPL) flew its next-generation Airborne Visible/Infrared Imaging Spectrometer (AVIRIS-NG), which records sunlight reflected from the ground in hundreds of narrow bands of color, over the Four Corners region of the US Southwest, where satellites had shown high methane. The team mapped more than 250 methane plumes from well pads, storage tanks, pipelines, gas processing plants and a coal mine vent at 1 to 3 m resolution, from a detection limit of about 2 to 5 kg an hour up to about 5,000 kg an hour. The largest 10% of sources gave 49 to 66% of the total, and ground crews confirmed and repaired two pipeline leaks found during the flights.

    LessonBecause a few sources carry most of the methane, a fast survey of many sites can find most of it.

    pmc.ncbi.nlm.nih.gov

    Credit

    Christian Frankenberg, Andrew Thorpe, Robert Green and colleagues at JPL and partner institutions; Proceedings of the National Academy of Sciences, 2016

  2. 2016

    A satellite for single sites

    GHGSat-D, called Claire, launched on 21 June 2016 on an Indian Polar Satellite Launch Vehicle, was the first of a planned fleet of small satellites that GHGSat Inc. designed to measure greenhouse gases from individual facilities. It imaged 12 by 12 km scenes with an effective resolution of 50 by 50 m. GHGSat's own engineers put its detection threshold at 1,000 to 3,000 kg of methane an hour, and a 2022 quality assessment done by Serco for the European Space Agency found it could map regional emissions but had limited ability to resolve individual point sources. GHGSat cut stray light about a hundredfold in its first commercial satellite, launched on 2 September 2020.

    LessonA first satellite proved the idea, but only for very large leaks; its errors set the design of the next ones.

    amt.copernicus.orgearth.esa.int

    Credit

    GHGSat Inc.; assessment by Serco for the European Space Agency

  3. 2017

    Refineries measured from above

    Tegan Lavoie, Paul Shepson and colleagues at Purdue University, with the Environmental Defense Fund, flew a small aircraft carrying an atmospheric chemistry laboratory around three oil refineries and three gas-fired power plants to measure the methane coming from each. They estimated that the refineries' yearly methane emissions may be 11 to 90 times the figures calculated from operators' reports to EPA's Greenhouse Gas Reporting Program. Shepson pointed to "leaky compressors, valves, and industrial hardware" and said a gas imaging camera could find the leaks.

    purdue.edu

    Credit

    Lavoie, Shepson and colleagues, Purdue University and the Environmental Defense Fund; Environmental Science & Technology, 2017

  4. 2021

    A free satellite joins in

    Sentinel-2 is a pair of European Union satellites built to map land, which take pictures with 20 m pixels in shortwave infrared bands that methane partly absorbs and revisit a place every 2 to 5 days. Daniel Varon of Harvard University, with Daniel Jacob of Harvard and colleagues at GHGSat, showed that it can detect and size methane sources of about 3 tonnes an hour or more over even ground. They followed a well-pad device in Algeria's Hassi Messaoud oil field, seen in 93% of cloud-free scenes at 2.6 to 51.9 tonnes an hour until a flare lit in August 2020 shut it down, and a compressor station in Turkmenistan's Korpezhe field whose rates closely matched GHGSat-D's.

    LessonSatellites built for other jobs can watch the largest leaks often, but only above a few tonnes an hour.

    amt.copernicus.org

    Credit

    Varon, Jervis, McKeever, Spence, Gains and Jacob; Atmospheric Measurement Techniques, April 2021

  5. 2024

    Aircraft pass a blind test

    Sahar El Abbadi, Adam Brandt and colleagues at Stanford University released metered gas for more than 700 blind measurements over six weeks while five aircraft systems flew over without knowing the rates: Carbon Mapper, GHGSat-AV, Insight M, MethaneAIR and Scientific Aviation. The aircraft consistently quantified releases above 10 kg of methane an hour, and GHGSat-AV and Insight M detected releases below 5 kg an hour. For the systems that image methane looking down, the ratio of estimated to true rates (the parity slope) ran from 0.76 to 1.13; for the one that samples air as it flies, it was 0.5.

    LessonThe authors judged that aircraft sensing had matured since earlier tests and was ready for "an increasingly important role" in policy and regulation.

    pmc.ncbi.nlm.nih.gov

    Credit

    El Abbadi, Chen, Burdeau, Rutherford, Chen, Zhang, Sherwin and Brandt, Stanford University; Environmental Science & Technology, June 2024

  6. 2025

    Seven super-emitter notices

    EPA's Super Emitter Program, created in its 2024 methane rule, lets certified third parties report methane releases of 100 kg an hour or more, found by remote sensing at or near oil and gas sites; EPA passes each notice to the operator, who must investigate and respond. EPA approved Carbon Mapper's airborne system for the program on 2 January 2025. EPA's public data hold seven notices, all from Carbon Mapper, for releases seen on 9 and 10 January 2025 in California's southern San Joaquin Valley at 149 to 812 kg an hour, one of them at a renewable fuel refinery near Bakersfield. In their responses, operators named a compressor (twice), an oil well, a storage tank, and a maintenance job that drained fluid from a gas line; one found no source; and one operator stated in all four of its responses that EPA had never made it aware of the event. EPA then put the program off until 22 January 2027, and on 29 October 2026 proposed to rescind it, citing "insufficient evidence of the nature and duration of super-emitter events".

    LessonA notice system needs both a reliable sensor and a working channel to the operator; here the operator's first complaint was about the channel.

    epa.govservices.arcgis.comhklaw.com

    Credit

    US EPA; Carbon Mapper; the operators' responses

  7. 2026

    Drones tested in France

    Over four weeks in June and September 2024, Audrey McManemin and colleagues at Stanford University ran a blind test of eight commercial systems at TADI, TotalEnergies' test site near Pau, France: four drone systems, GHGSat's satellites, a network of 12 fixed point sensors, a laser camera and an infrared camera. All four drone teams detected releases below 0.7 kg an hour, and each detected 90% of releases below 6 kg an hour, but their estimates of leak size varied widely. After the true rates were revealed, one drone team traced its errors to a faulty wind sensor, and another to transcription errors, sensor saturation and its handling of wind data.

    LessonThe authors found low and variable wind a key cause of errors in leak size, so wind has to be measured as carefully as methane.

    amt.copernicus.org

    Credit

    McManemin and colleagues, Stanford University, with TotalEnergies; Atmospheric Measurement Techniques, February 2026

Watch for

  1. Smaller leaks from orbit

    GHGSat says it is designing a second-generation satellite that will detect methane leaks from 50 kg an hour, half the roughly 100 kg an hour its current satellites detect, and cover about 20,000 square kilometers a day, with launch planned for early 2028.

    Achieved whenthe satellite is in orbit and an independent blind test measures it detecting 90% of releases at or near 50 kg an hour.

    spaceq.ca

    Claimed by

    GHGSat, September 2026

  2. Aircraft sensing near 3.5 kg

    GHGSat says its aircraft-mounted sensor, DATA.AIR, detects leaks approaching 3.5 kg of methane an hour in tests at Colorado State University's METEC. EPA approved the second-generation DATA.AIR for periodic screening at under 10 kg an hour on 20 October 2026.

    Achieved whenMETEC or another independent tester publishes the detection curve behind the 3.5 kg figure.

    spaceq.caepa.gov

    Claimed by

    GHGSat, September 2026

  3. A French methane satellite

    Absolut Sensing of Toulouse reports that its GESat GEN1 microsatellite, a 16U spacecraft (built from 16 standard 10 cm units) that measures methane with a compact interferometer at 50 m resolution, reached an average precision of 111 parts per billion and an average detection sensitivity of 450 kg an hour over 160 processed scenes.

    Achieved whenthe paper passes peer review and an independent controlled-release test confirms the detection limit.

    egusphere.copernicus.org

    Claimed by

    Absolut Sensing, with researchers at France's Laboratoire de Météorologie Dynamique, in a preprint of May 2026

Open

  1. 01

    One table of blind tests

    An open table of every published blind controlled-release result for methane leak detection, one row per release per system, giving the platform, release rate, wind, whether the release was detected, the estimated rate and the test campaign. From it, fit for each system and campaign the release rate it detects 90% of the time, and set that beside the threshold EPA approved for the same system, where there is one.

    The brief
    Credit · unclaimed
  2. 02

    Every refinery, from the catalog

    A public table of every methane plume in Carbon Mapper's open catalog within a set distance of each US refinery, with its date, estimated rate and instrument; the number of times each refinery was flown over or imaged; and the methane the refinery reported to EPA for the same year.

    The brief
    Credit · unclaimed

Open 01 · the brief

One table of blind tests

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open table of every published blind controlled-release result for methane leak detection, one row per release per system, giving the platform, release rate, wind, whether the release was detected, the estimated rate and the test campaign. From it, fit for each system and campaign the release rate it detects 90% of the time, and set that beside the threshold EPA approved for the same system, where there is one.

Why it is still open

The results sit in separate papers and repositories, such as Stanford's aircraft releases (data linked from the 2024 paper on GitHub), the 2024 TADI test (data on Zenodo) and METEC's 2022 round of continuous monitors, while EPA's approvals list thresholds from under 1 to under 15 kg an hour with no shared scoreboard. A search in November 2026 found compilations for a single satellite fleet and detection models for single campaigns, but no table across technologies.

epa.govamt.copernicus.orgpmc.ncbi.nlm.nih.gov

Built with what exists

The published supplements and data repositories; pdfplumber for tables in supplementary PDFs; Python with pandas or DuckDB; logistic regression in statsmodels for detection curves.

  1. First step

    Load the Stanford aircraft data and the TADI Zenodo record and fit one detection curve for each platform.

  2. Done looks like

    At least eight campaigns and 3,000 release rows coded, with 200 rows checked by a second person against the sources; a detection curve with uncertainty for each system in each campaign; table and code published under an open license.

Open 02 · the brief

Every refinery, from the catalog

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public table of every methane plume in Carbon Mapper's open catalog within a set distance of each US refinery, with its date, estimated rate and instrument; the number of times each refinery was flown over or imaged; and the methane the refinery reported to EPA for the same year.

Why it is still open

Aircraft measurements at three refineries in 2017 estimated methane 11 to 90 times the reported figures, a small sample. Carbon Mapper's catalog is open: a query on 19 November 2026 of a box from Houston to Port Arthur, Texas, returned 100 methane plumes seen from April 2021 to September 2026, 19 of them labeled sector 1A1 (energy industries), at about 200 to 3,100 kg an hour. No comparison refinery by refinery across the country was found.

purdue.eduapi.carbonmapper.org

Built with what exists

Carbon Mapper's public plume API; refinery locations from the US Energy Information Administration; methane reported under EPA's Greenhouse Gas Reporting Program; NASA's Common Metadata Repository to count EMIT satellite overpasses; Python with GeoPandas.

  1. First step

    Pull the catalog's plumes around ten Gulf Coast refineries and count how often each was imaged.

  2. Done looks like

    All US refineries listed with overpass counts and plume detections; for each refinery with at least three detections, an implied yearly rate, with its assumptions stated, beside the reported figure; 30 plume attributions checked by hand on aerial imagery; code and table published.