What the Satellite Could Not Ignore

The satellite does not care who owns the leak. It only cares that the leak has a temperature and a shape.

For most of the industrial era, a methane plume, a cut in a forest, a flare that should have been a sale, could be a private fact. The operator knew, or did not. The regulator knew if a stack was reported, or if an inspector’s truck arrived on a day the wind was kind. The public knew a number in an inventory, averaged, delayed, and often estimated from how much fuel was sold times an emission factor written in a quieter decade. “We didn’t know” was a sentence with no timestamp. It could be used in 1970 and 2015 with the same posture.

That sentence now needs a date. Instruments in polar orbit have turned some of the largest leaks and clearings into public pixels. TROPOMI maps methane daily at the scale of kilometres. GHGSat and other point-source constellations resolve plumes to a facility. Landsat and Sentinel have been counting trees, and the absence of trees, at 10 to 30 metres for years that anyone can download. Seeing is not stopping. Clouds still win many mornings. Small sources still hide. Error bars are not a formality. But the gap between a super-emitter and a claim of ignorance has narrowed to the revisit time of a spacecraft, and that is a different civic world from the one in which Aliso Canyon had to be found by a nose and a phone call.

A spectrometer that works the night shift

Methane is invisible. It does not smell at the concentrations that warm the planet, though the mercaptans added to utility gas do. It absorbs in the shortwave infrared. Put a spectrometer in low Earth orbit, look at sunlight that has passed twice through the column of air—down to the ground and back—and the missing photons at those wavelengths are a map. The map is of a column, not of a pipe. Wind, albedo, aerosols, and the angle of the sun all enter the retrieval. The plume that looks like a brushstroke in a press image is a set of pixels that survived quality flags.

TROPOMI, the TROPOspheric Monitoring Instrument on the European Sentinel-5 Precursor satellite, launched in October 2017, is the workhorse of this public map. It covers the globe daily. Its methane pixels are on the order of 5.5 by 7 kilometres at nadir—small enough to pick out a basin and a very large point source, large enough to mix a city, a wetland, and a compressor station into one number. Independent assessments have put single-overpass detection limits for point sources in the range of hundreds of kilograms to several tonnes of methane per hour, depending on wind and surface brightness; round figures in the policy literature often cite on the order of a few tonnes to tens of tonnes per hour for a confident TROPOMI plume. Aliso Canyon, at its peak near 50 tonnes an hour, would have been a billboard. A wellhead leaking at 50 kilograms an hour would not.

That mismatch is the first honest limit. Daily and global is a kind of power. It is also a kind of blindness. TROPOMI is how researchers found that a handful of “ultra-emitters”—failed flares, unlit vents, blown pipelines—account for a disproportionate share of some oil-and-gas basins’ satellite-visible methane. It is how a pipeline rupture in Central Asia or a coal-mine ventilation shaft can become a paper in a week. It is not how you inventory a million small valves. The atmosphere does not care about the distinction. The instrument does.

The commercial zoom

GHGSat, a Montreal company, flew a demonstration satellite, Claire, in 2016 and then a constellation of small spectrometers with a different brief: stare at a site. Scenes on the order of 12 kilometres across, sampled at about 25 metres, can attribute a plume to a pad, a landfill cell, a mine vent. Detection limits under good wind and bright ground sit near 100 kilograms of methane per hour in company and third-party controlled-release tests; operational thresholds of 180–240 kilograms per hour appear in independent evaluations. In 2023 the company reported more than three million observations in 85 countries and nearly 16,000 detections above a 100 kilograms-per-hour “super-emitter” threshold. Those are tasked looks, not a census of Earth. You see what someone paid to look at, or what the operator chose to share.

Other sensors fill other niches. NASA’s EMIT on the International Space Station, the Italian PRISMA, Germany’s EnMAP, Maxar’s WorldView-3, Sentinel-2 and Landsat in their coarser methane tricks, Carbon Mapper’s Tanager: a messy, overlapping fleet. Controlled-release experiments—Stanford and others pumping known rates of methane into the air while satellites pass—have become the calibration theatre of this field. Some systems detect the release and quantify it within tens of percent. Some miss it because of clouds. Some report a plume that was not there. False positives and false negatives are not scandals. They are the error rates that a regulator has to put in the same sentence as the pretty picture.

MethaneSAT was supposed to occupy the middle: a wide swath, on the order of 200 kilometres, at roughly 100-by-400-metre pixels, precise enough to total a basin and still see large pads. The Environmental Defense Fund and partners launched it on 4 March 2024. Early images of the Permian, Appalachian, Uinta, and other basins were used to argue that observed oil-and-gas methane exceeded some official inventories by large factors. Contact was lost on 20 June 2025. After weeks of attempts the spacecraft was judged unrecoverable, a power failure. Fifteen months of data remain to be mined. The basin-scale niche is, for the moment, a gap. A satellite that cannot be ignored when it is working can still die. That, too, belongs in an honest account.

Aliso Canyon, and the week the basin doubled

On 23 October 2015, Southern California Gas reported a leak at well SS-25 in the Aliso Canyon underground storage field above Porter Ranch. The well was not sealed until 11 February 2016, with a permanent plug announced a week later. For 112 days a single point in the San Fernando Valley emitted natural gas at rates that, at their peak, roughly doubled the methane output of the entire Los Angeles basin. Aircraft mass-balance flights led by Stephen Conley, published in Science, put the total at about 97,100 metric tons of methane. The California Air Resources Board’s later synthesis, blending those flights with reservoir-pressure data, arrived at 99,650 ± 9,300 metric tons; the Board used 109,000 as the figure to fully mitigate. About 8,000 households were relocated. The smell, the headaches, and the fear were local. The climate pulse was state-sized: on the order of a fifth of California’s methane for those sixteen weeks.

TROPOMI was not yet flying. The leak was found the old way, then measured the new way. Scientific Aviation flew transects. NASA JPL’s AVIRIS and HyTES imaged the plume from aircraft. The Hyperion instrument on EO-1 and Japan’s GOSAT saw enhancements from orbit. A later remote-sensing study, using Landsat-8 among other sensors, argued that a plume was already visible on 26 October—three days after the wellhead failure and well before the first mass-balance flight on 7 November—and that the earliest orbital snapshot implied a rate even higher than the first aircraft estimate. Whether that particular retrieval becomes the canonical peak is a specialist argument. The civic point does not depend on it. A leak that large had a temperature and a shape. Orbiting instruments could see the shape. Seeing it in October did not close the well in October. Relief-well crews did that in February. The satellite is a witness. It is not a valve.

Aliso is now the case that satellite people use to say the era of surprise is over for events of that size. It is also the case that Porter Ranch residents use to say that a witness is not a remedy. Both uses are correct. The next Aliso, if it is large enough and the sky is clear, will have a timestamp on a public map. The next Aliso, if it is a thousand small leaks, may not.

Forests, 30 metres at a time

Methane is one invisible. A missing tree is another, and it photographs more easily, which is why it was inventoried sooner and still argued over more.

Landsat, a NASA–USGS program, has been imaging the land at about 30 metres since the 1970s; the archive became free in 2008, which is as important as any sensor improvement. Sentinel-2, from the European Copernicus program, added 10-metre optical bands and a faster revisit. Brazil’s National Institute for Space Research had not waited for a free global product. PRODES, the Projeto de Monitoramento do Desmatamento na Amazônia Legal, has used Landsat, with heavy human interpretation, to produce official annual Amazon deforestation figures since 1988. DETER, a coarser, faster MODIS-based alert, was built so that enforcement would not have to wait for the annual map. Those systems are why Brazil could claim, in the late 2000s, a documented drop in Amazon clearing, and why later rebounds could be argued with numbers instead of anecdotes.

In 2013, Matthew Hansen and colleagues published in Science the first globally consistent, 30-metre maps of forest loss and gain for 2000–2012, processed on Google Earth Engine from more than 650,000 Landsat scenes. They reported 2.3 million square kilometres of loss and 0.8 million of gain over the twelve years. The tropics were the only climate domain with a statistically significant trend of increasing loss. Brazil’s reduction was visible; Indonesia’s increase was visible; the southeastern United States appeared as a checkerboard of harvest and regrowth that national pride does not always file under “deforestation.” Global Forest Watch put the maps in a browser. A journalist in Jakarta and a ranger in Acre could look at the same pixel.

Sentinel-2, once it was operational, made the same logic faster. Two satellites, 10-metre bands, a revisit of five days at the equator in combination, and an open archive meant that a clearing could be dated to a week rather than a year. Alert systems that use Landsat and Sentinel together—GLAD, RADD, and their national cousins—do not replace a court’s definition of illegal. They replace the months in which a cut could be a rumor. A ministry that still publishes an annual statistical yearbook as its first deforestation number is now choosing a tempo, not living in a data vacuum.

The asterisks are the same family as the methane asterisks. “Forest” in Hansen’s product is tree cover above a height and canopy threshold, not a land-use class. A palm plantation can be gain. A burned woodlot can be loss. PRODES and Global Forest Watch disagree in places because they are not answering identical questions: one is tuned to primary-forest conversion in Legal Amazonia with interpreters in the loop; the other is an automated global change product. Selective logging and understory fire degrade a forest that still looks like canopy from 30 metres. Clouds, especially in the west of Amazonia and in Central Africa, hide months. A 2017 comparison against finer commercial imagery found that both GFW and PRODES underestimated loss at two Amazon sites, GFW less so. The pixel is public. The definition is still a fight. That fight is an improvement on having no pixel.

Public pixels, institutional lag

The lag is now the story. A TROPOMI plume can be processed in days. A GHGSat tasking can return a facility name. A GLAD alert can flag a 10-metre clearing weekly. Ministries, companies, and courts still run on annual inventories, self-reports, and the pace of a site visit. The mismatch produces a new genre of journalism: here is the leak, here is the owner, here is the year-old factor that was supposed to represent it. Sometimes the owner already knew and was waiting for a part. Sometimes the owner is a state company in a country that did not invite the satellite. Sometimes the plume is a wetland, a feedlot, or a retrieval artifact, and the first tweet is wrong.

Attribution is the hard step the press image skips. A 7-kilometre TROPOMI pixel over a crowded basin is not a court exhibit. A 25-metre GHGSat plume over a named pad is closer, and still wants a wind field and a second overpass. Companies have begun to buy the data on themselves, which is both compliance and control. Civil-society groups have begun to publish the same scenes. The politics of who gets to see the leak are not solved by the physics of absorption.

There is also a justice lag. Porter Ranch had the resources to leave. A landfill plume over a poorer census tract may be just as visible from orbit and less visible in a hearing. Deforestation alerts fire fastest where the canopy was intact enough to make a crisp edge; gradual degradation of a community forest may never trip the algorithm. Public pixels are not the same as public power. They are a precondition that did not used to exist.

What the instrument cannot do

Clouds. Shortwave-infrared methane retrievals fail through water cloud. A week of weather can hide a week of venting. Operators know this; so do the people who schedule illegal clearing. Night, high latitudes in winter, snow, and dark forests all degrade some products. A leak that is large and brief—a blowdown, a maintenance vent—can fall between overpasses. A leak that is small and chronic can sit forever below the detection limit and still add up, in a thousand copies, to a national total.

Quantification error is not a rounding problem. Emission rate is retrieved from a column enhancement and a wind. Winds at 10 metres are not winds at 2 metres. A factor-of-two mistake is not rare in hard scenes. Inventories that mix satellite plumes with emission factors can double-count or miss the diffuse remainder. Wetlands and ruminants produce methane that is not a “leak” and will light up a basin map if you do not mask them. Oil-and-gas advocates will say so, sometimes fairly.

None of that returns us to 2014. The super-emitter, the unlit flare, the pipeline rupture, the square kilometre of new clearing in a park: these are classes of event for which “we didn’t know” has become a claim that can be checked. The check will sometimes exonerate. A pixel can be wrong. The check is still a check.

A date on not knowing

Aliso Canyon was reported on 23 October 2015. TROPOMI reached orbit in October 2017. Hansen’s global map was in Science in November 2013. PRODES is a 1988 institution. Claire flew in 2016. MethaneSAT lived from March 2024 to June 2025. Those dates are not trivia. They are the points after which particular kinds of ignorance are optional.

A storage well that vents at 50 tonnes an hour in a clear-sky Los Angeles winter is no longer an invisible accident. A 100-hectare cut in eastern Amazonia is no longer a fact that has to travel by rumor. A landfill that breathes 500 kilograms an hour may still be invisible to TROPOMI and obvious to GHGSat if someone looks. A million household furnaces are still a model, not a plume. The art is to match the instrument to the lie. The lie that a basin is in line with its inventory can be tested. The lie that a particular 2-centimetre valve is tight cannot, from 500 kilometres, yet.

Flying River’s question is the one the press image does not ask. What is happening here that most people cannot easily see? A gas that has a spectrum. A forest that has a 30-metre memory. A delay between a public pixel and a closed valve. What happens if we follow the connection far enough? We stop accepting “we didn’t know” as a weather. We ask what year the speaker means, which sensor was flying, whether the sky was clear, and whether the thing not known was below the detection limit or merely below the level of effort. The satellite will not care who owns the leak. The rest of us have to. Seeing is the start of an argument, not the end of one. The argument is better than the silence it replaced. It is not the same as a repair.

Sources

• Veefkind, J. P., et al. “TROPOMI on the ESA Sentinel-5 Precursor: A GMES Mission for Global Observations of the Atmospheric Composition for Climate, Air Quality and Ozone Layer Applications.” Remote Sensing of Environment 120 (2012): 70–83. (Instrument and 2017 launch context.)
• Lauvaux, Thomas, et al., and related TROPOMI ultra-emitter studies (e.g. Science 2022) on large oil-and-gas methane plumes.
• Jacob, Daniel J., et al. “Quantifying Methane Emissions from the Global Scale Down to Point Sources Using Satellite Observations of Atmospheric Methane.” Atmospheric Chemistry and Physics 22 (2022): 9617–9646. (Detection thresholds by instrument class.)
• GHGSat. Constellation specifications; 2023–2025 observation and plume-count releases; controlled-release detection limits ~100 kg h⁻¹.
• Sherwin, Evan D., et al. Single-blind satellite methane controlled-release testing. Atmospheric Measurement Techniques 17 (2024): 765 and related papers (quantification accuracy; role of clouds).
• Environmental Defense Fund / MethaneSAT LLC. Launch 4 March 2024; early basin images (Permian and others); loss of contact 20 June 2025 and subsequent end-of-mission statements.
• Conley, Stephen, et al. “Methane Emissions from the 2015 Aliso Canyon Blowout in Los Angeles, CA.” Science 351, no. 6279 (2016): 1317–1320. (~97,100 metric tons; peak rates to ~60 t h⁻¹; doubling of LA-basin methane.)
• California Air Resources Board. “Determination of Total Methane Emissions from the Aliso Canyon Natural Gas Leak Incident” and 21 October 2016 news release (99,650 ± 9,300 metric tons; 109,000 metric tons for mitigation; ~20 percent of statewide methane during the event).
• U.S. EPA. Inventory update incorporating Aliso Canyon (ARB Phase I/II split; ~99,650 metric tons total).
• Thompson, David R., et al., and subsequent remote-sensing reanalyses of Aliso (Hyperion, AVIRIS; Landsat-8 early-plume estimates, including 2025 Environmental Research Communications work on 26 October 2015).
• Hansen, M. C., et al. “High-Resolution Global Maps of 21st-Century Forest Cover Change.” Science 342, no. 6160 (2013): 850–853. (2.3 million km² loss, 0.8 million km² gain, 2000–2012; 30 m Landsat.)
• INPE. PRODES and DETER Amazon monitoring program documentation (annual Landsat-based deforestation mapping since 1988; alert system).
• Global Forest Watch / World Resources Institute. Public Landsat-based loss-and-gain maps and GLAD alerts.
• Milodowski, D. T., et al. “Forest Loss Maps from Regional Satellite Monitoring Systematically Underestimate Deforestation in Two Rapidly Changing Parts of the Amazon.” Environmental Research Letters 12 (2017).
• NASA Earth Observatory. Coverage of the Hansen et al. 2013 map and Landsat’s role in forest monitoring.