Industries / Industrial equipment and services / Industrial services / Leak detection
Reporting
Reporting is how leak findings become numbers: the tonnes a plant or company says it released, worked out from readings, counts of parts and, more and more, direct measurement.
6
Milestones achieved, from 1995 to 2024.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1995
From readings to kilograms
A Method 21 analyzer, the portable instrument used in leak surveys, reports the concentration of vapor at a leak in parts per million, not how much is escaping. EPA's Protocol for Equipment Leak Emission Estimates (EPA-453/R-95-017, November 1995), an update of a 1993 version, set out four ways to turn surveys into mass emissions, from an average factor for each type of component to equations that convert each screening reading into a leak rate, fitted to leaks that had been enclosed in bags and measured. It added equations for the petroleum industry, fixed "default-zero" rates for components that read zero, and "pegged" rates for readings above the analyzer's limit of 10,000 or 100,000 parts per million by volume.
Credit
EPA's Office of Air Quality Planning and Standards
2010
Leaks enter federal reporting
EPA's rule of 30 November 2010 added petroleum and natural gas systems (subpart W) to its Greenhouse Gas Reporting Program, for facilities emitting 25,000 tonnes of carbon dioxide equivalent a year or more; it required reporting, not controls. Gas processing plants, transmission compressor stations, underground storage, LNG facilities and gas distribution gate stations must survey for leaks with an OGI camera, an infrared laser instrument or Method 21, then multiply the number of each type of leaking component by a fixed "leaker" emission factor. Onshore production sites report leaks from counts of equipment times average factors. A 2016 revision added survey methods matching EPA's new leak standards for well sites and compressor stations, with leaker factors to use with them.
Credit
US EPA
2018
The inventory falls short
Ramón Alvarez of the Environmental Defense Fund and 23 co-authors estimated US oil and gas methane from measurements at facilities, checked against aircraft observations over areas with about 30% of US gas production. They put 2015 emissions from the supply chain at 13 ± 2 teragrams (million tonnes) a year, 2.3% of gross US gas production and about 60% more than EPA's inventory, likely, they wrote, because inventory methods miss emissions released during abnormal operating conditions.
LessonReports built from factors for normally working equipment leave out the rare large releases, so they need checking against measurement.
Credit
Ramón A. Alvarez, Daniel Zavala-Araiza and David R. Lyon (Environmental Defense Fund) and co-authors, in Science
2018
California logs every leak
California's Air Resources Board (CARB) adopted its oil and gas methane rule on 23 March 2017, and from 1 January 2018 it required quarterly leak surveys and a yearly report listing each leak: the inspection date, the Method 21 instrument and its calibration date, the component type and tag, the reading, the repair date and the reading after repair. CARB's summary of 2022, published in May 2025, covers 100 operators at 451 facilities: on average 2,105,729 components were surveyed each quarter, 10,082 leaks of 1,000 parts per million by volume (ppmv) or more were found (0.48% of components), and 9,918 were repaired or the part replaced. CARB estimated, with correlation equations, that about 10% of the leaks gave 51% of the emissions and that the repairs cut emissions from the covered components by 18%, about 1,600 tonnes of methane. CARB notes that the data are reported by the operators; its checks found errors needing follow-up in 0.26% of the leaks.
Credit
California Air Resources Board
2020
Reports tied to measurement
On 23 November 2020, 62 companies with 30% of the world's oil and gas production joined OGMP 2.0, a methane reporting framework of the Oil and Gas Methane Partnership led by the UN Environment Programme (UNEP), the European Commission and EDF. Its top level, Level 5, requires a company to reconcile its source-by-source estimates with measurements of whole sites, and OGMP checks the reported data and plans independently. By UNEP's October 2025 report, membership had grown to 153 companies with 42% of world production, and 65 companies with 17% of world production had reached the "Gold Standard", Level 5 reporting for their material assets.
Credit
UNEP's International Methane Emissions Observatory, the European Commission and EDF
2024
Large releases must be reported
EPA's revision of the oil and gas reporting rule, published 14 May 2024 and effective 1 January 2025, added "other large release events": abnormal events releasing methane at an instantaneous rate of 100 kg an hour or more, the same threshold as EPA's Super-Emitter Program, must now be reported. EPA said the change answers studies showing that super-emitters, the few sources releasing very large amounts, make up much of total emissions, and the revision moved more of the rule's methods to measured data.
Credit
US EPA
Watch for
Federal reporting may pause
On 16 September 2025 EPA proposed to remove reporting duties for most industries in its Greenhouse Gas Reporting Program, including gas distribution, and to suspend reporting by the rest of the oil and gas category until reporting year 2034. It received more than 50,000 comments and, by an interim final rule of 14 October 2026, moved the deadline for 2025 data to 10 March 2027 while it considers them.
Achieved whenEPA publishes a final rule that removes or suspends subpart W reporting and it takes effect.
Claimed by
US EPA, September 2025
Measured reports for a quarter
OGMP 2.0 says that 7% of world oil and gas production, 238 assets, reached Level 5 in 2024, and that under current company plans production reported at Level 5 will reach about 25% of world output by 2027 and about 30% by 2030. It also reports that Level 5 assets had a methane intensity of about 0.1% in energy terms, a figure from the companies' own reports.
Achieved whenOGMP's assurance of 2027 data confirms Level 5 reporting for at least 25% of world production.
Claimed by
UNEP's Oil and Gas Methane Partnership 2.0, March 2026
Open
01
Alberta's alternatives, scored
One public table of every alternative leak program the Alberta Energy Regulator has approved, giving for each the sites, the detection methods, the modeled cut against the default OGI program and, from its performance report, the leaks found, the share repaired, repairs delayed, leaks recurring, days from detection to repair, the estimated cut and that cut as a share of the modeled one.
The briefCredit · unclaimed02
An open check for logs
Free software that reads an export from a plant's leak detection and repair database (component tag, date and time, technician, instrument, reading, leak definition, repair dates, recheck reading) and flags the faults that audits look for: readings closer together than the instrument can respond, components on the list never monitored, leaks without a first repair attempt in 5 days or a repair in 15, repairs with no recheck, delay-of-repair entries past the next shutdown, and drifting calibrations.
The briefCredit · unclaimed
Open 01 · the brief
Alberta's alternatives, scored
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
One public table of every alternative leak program the Alberta Energy Regulator has approved, giving for each the sites, the detection methods, the modeled cut against the default OGI program and, from its performance report, the leaks found, the share repaired, repairs delayed, leaks recurring, days from detection to repair, the estimated cut and that cut as a share of the modeled one.
Why it is still open
The AER's approvals page links 55 PDFs, executive summaries with the modeled cut and performance reports with the results, but its reporting guide makes the emission reduction summary a voluntary item. The Sundre group's report shows how far results can fall from the model: 14.4% against 50.4%. Searches in November 2026 found the individual reports and no comparison across them.
Built with what exists
pdfplumber to pull the tables; a model that answers fixed questions about each report, such as Gemini Decision 1 (see the Note "A thousand questions, one pass"); pandas; the open-source LDAR-Sim model (github.com/LDAR-Sim/LDAR_Sim) to rerun a program's design with its measured repair rates.
First step
Code the Sundre, Ember and Canadian Natural reports by hand to fix the fields.
Done looks like
Every approved program with a performance report coded, with a second person checking a fifth of them; the table and the median ratio of estimated to modeled cut published, with each figure linked to its page in the source PDF.
Open 02 · the brief
An open check for logs
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
Free software that reads an export from a plant's leak detection and repair database (component tag, date and time, technician, instrument, reading, leak definition, repair dates, recheck reading) and flags the faults that audits look for: readings closer together than the instrument can respond, components on the list never monitored, leaks without a first repair attempt in 5 days or a repair in 15, repairs with no recheck, delay-of-repair entries past the next shutdown, and drifting calibrations.
Why it is still open
EPA's guide sets out these checks as QA/QC elements of a good program, and audits of LDAR records are sold by consultants and run inside commercial databases. A US patent granted in 2019 to Inspectionlogic covers timing each component reading by the instrument's response time. A GitHub search in November 2026 for "LDAR", "Method 21" and "leak detection repair" found the LDAR-Sim model and an Android app for one analyzer, but no open checker for records.
Built with what exists
Python with pandas and a data validation library such as pandera; a generator of synthetic plant records with planted faults; de-identified exports from a plant or contractor willing to share one.
First step
Write the rule list from EPA's guide and generate one synthetic year for a 5,000-component process unit.
Done looks like
On 50 synthetic plant-years holding 1,000 planted faults of eight kinds, at least 95% found with no more than 2% of clean records flagged; then a run on one real de-identified export, with the plant's own LDAR auditor reviewing every flag and the agreement published.