Library · Guides
Definitions
Every term the platform uses, what it means here, and — where it is commonly confused with something else — what it is not. The explanation beside a figure is this text.
- Terms
- 42
- Groups
- 7
Figures and basis
- Ensemble meanThe reported figure.
- The mean of a Monte-Carlo ensemble over every uncertain parameter. It is the reporting basis for every emission mass on the platform: it is additive across sources, years and facilities, and it carries the 95% interval. Several loss terms are multiplicative and right-skewed, so the mean of the distribution sits above the value obtained by putting every central estimate in at once.Not to be confused with: The deterministic run, which is lower. Quoting one under the other's label understates or overstates the inventory by the difference.Read the method →
- Deterministic runA diagnostic, not the reported figure.
- One evaluation of the engine with every parameter at its central value. Useful for checking that a change did what was intended, because it has no sampling noise. Sub-period totals are deterministic sums; the ensemble describes the whole record.Read the method →
- 95% intervalThe range the ensemble places the figure in.
- The 2.5th to 97.5th percentiles of the Monte-Carlo ensemble. Shown as prominently as the central figure because, on a methane estimate, the width of the interval is often the most important thing on the screen. An interval wider than 60% of the central value is marked wide.Read the method →
- Carbon intensityGreenhouse-gas emissions per unit of saleable energy.
- Grams of CO₂-equivalent per megajoule of saleable energy delivered — oil plus exported gas. Computed over a range as total carbon over total energy, never as an average of monthly ratios, which would weight a shut-in month as heavily as a full one.Read the method →
- Methane intensityMethane emitted as a share of methane produced.
- Distinct from carbon intensity and governed by different parameters, which is why the platform reports both.Not to be confused with: Methane intensity against marketed gas, which uses a smaller denominator and gives a larger figure for the same emission.
- GWP₁₀₀Global warming potential over 100 years.
- The platform uses 29.8 for fossil methane, the IPCC Sixth Assessment value, to express methane as CO₂-equivalent.
- As ofA figure as the platform stated it at a past moment.
- Every versioned row records when it was superseded, so any figure can be shown as it stood on a given date — the basis of the What changed view and of a briefing's baseline.
- SupersessionHow a figure changes without being overwritten.
- A correction or recomputation marks the prior rows superseded and inserts new ones. A figure a regulator has already acted on stays exactly as it was, and the history of how a number moved is part of the record.
Evidence and tiers
- IPCC tierHow a figure was estimated, not how good it is.
- Tier 1 applies a default factor to a facility proxy; Tier 2 uses metered activity with a published or assumed parameter; Tier 3 is facility-specific engineering estimation on measured inputs. The platform reports a tier mix weighted by methane mass, because twelve sources are not twelve equal things.Not to be confused with: A quality score. Tier 1 on a source carrying a negligible share of methane is a proportionate choice, not a shortcoming.Read the method →
- Evidence basisWhere a number came from.
- Measured, calculated, modelled or estimated. Independent of tier: a Tier 2 figure can rest on measured activity and a modelled parameter at once, and both facts travel with it.Read the method →
- Provenance lineThe rule under a headline figure.
- Every headline figure carries a line stating its basis, its tier, the period it covers and a link to the lineage record that produced it. It is the evidence trail made visible on the page.
- LineageThe inputs behind one figure.
- The returns, fluid properties, factor set, parameters, constants, engine version and random seed a result was computed from — recorded with the result and fingerprinted, so that re-running it is a test anyone can apply.Read the method →
- Factor setThe method's parameters, as a published version.
- Every uncertain parameter's distribution and every emission factor belongs to a versioned factor set. Publishing a new one recomputes every figure; a published set cannot be changed, and each figure's lineage names the set it was computed under.Read the method →
- Flare combustion efficiency (η_c)The fraction of flared methane actually destroyed.
- The dominant unmeasured parameter in the inventory. The flare term carries most of the national methane, and η_c is what converts flared volume into slipped methane — which is why measuring it ranks first.Read the method →
- Variance-reduction rankingWhat to measure first.
- Sources ordered by the uncertain mass a perfect measurement of each would remove — its expected size combined with its present uncertainty.Not to be confused with: Expected value of perfect information, which needs a decision and a loss function. The platform has neither and does not use the term.
- Screening gradeSized to direct work, not to design or decide.
- An assessment intended to size a problem and direct further work, not to support a design or an operational decision. All operability results on this platform are screening grade.
- Effective independent facilitiesHow many facilities the fleet behaves as, for access planning.
- Three FPSOs in one metocean regime lose and regain access together. The operability screening finds the fleet behaves close to a single facility, so a maintenance programme scheduled as if they were independent finds its crews idle and its facilities unreachable at the same time.Read the method →
Gas and production
- Produced gasWhat the facility brought to surface.
- The denominator for methane intensity as most upstream conventions define it.Not to be confused with: Marketed gas. The two can differ severalfold, so a threshold written against one is not a threshold against the other — which is why both are always shown together.
- Marketed gasWhat the facility sold.
- Produced gas less what was flared, used as fuel and reinjected. A smaller denominator, so a larger methane intensity for the same emitted mass.
- Gas-oil ratioGas produced per barrel of oil.
- Standard cubic feet of gas per barrel of oil. It rises as a reservoir depletes, which is why gas rates can rise while oil falls.
- Arps declineHow a field's rate falls over time.
- The standard petroleum-engineering description of production decline, used to project oil rate forward.Read the method →
- History matchA forecast anchored on the latest reported rate.
- The trajectory's shape comes from the fitted model; its starting level is set by the most recent reported rate.Read the method →
- Water cutThe share of produced liquid that is water.
- It rises as a field matures, and with it the produced-water degassing term.
- Mscf, MMscf, Bcf, bopdVolume units used in returns.
- Thousand, million and billion standard cubic feet of gas; barrels of oil per day.
Policy and standards
- Screening testA comparison against a benchmark, with no legal force.
- The 2030 tests compare a projected pathway against a policy benchmark or a market comparator, so that a pathway can be discussed against something.Not to be confused with: A compliance determination. Only a competent authority makes one, and nothing this platform produces is one.
- Zero Routine FlaringThe World Bank initiative to end routine flaring by 2030.
- Non-routine flaring — safety, start-up, emergency — continues, which is why the screening floor is a small share of produced gas rather than zero.
- Global Methane PledgeA 30% cut in methane from 2020 levels by 2030.
- Applied here at facility level as a test, not as an assumption.
- OGMP 2.0The Oil and Gas Methane Partnership reporting framework.
- Its highest level, Level 5, is source-level measurement reconciled against site-level measurement.
- Standards crosswalkHow each part of the method maps to each framework.
- A cell holds a clause reference only once someone has entered it from the authoritative text, a second person has reviewed it against that text, and a third — never whoever entered it — has approved it. Until then the cell says to be populated, and briefings mark the crosswalk as a draft.
Filing and review
- Monthly returnOne facility-month of activity, filed by the operator.
- Days on production, oil, gas produced, flared, used as fuel, exported and reinjected, and produced water — the fields of the Petroleum Commission's return. Filed as a CSV or through the form.Read the method →
- QC findingA rule in the published catalogue that a return did not meet.
- An error blocks acceptance and returns the submission to the operator; a warning is shown to the reviewer and does not. The catalogue is public, and an operator can pre-check a file against it before filing.Read the method →
- Submission statusWhere a filing stands.
- Submitted, then validating; then validated (ready for a reviewer) or corrections requested (returned with findings); then accepted (promoted into the inventory) or returned for revision (with a note). A corrected file supersedes the original, which keeps its findings on the record.Read the method →
- Review queueRule findings and AI questions, in priority order.
- Each item is decided by a person with a note of at least ten characters: confirmed as an issue, explained, dismissed or sent for investigation. The order's basis is stated — a fitted triage model, or the documented fallback.Read the method →
Assurance
- Reproducibility checkA stored figure re-run from its lineage.
- A sample of lineage records is re-run every night with the recorded inputs, factor set, engine version and seed; each result — reproduced or not, and by how much — is recorded and public in summary.Read the method →
- Audit trailEvery state-changing action, before and after.
- Who acted, when, from where, on what, and the state before and after. The database refuses any change to or deletion of an entry.
- Briefing and evidence packA published PDF and the files that prove it.
- Every number in a briefing is recorded as a figure with its lineage; the evidence pack carries the snapshot, the figures and a manifest of hashes. Published briefings are frozen by the database and listed with their SHA-256 on the transparency record.Read the method →
- SHA-256A fingerprint of a file's exact bytes.
- Recorded when a file is stored and checked before it is served. The verifier page also hashes what it receives in the browser, so the check does not rest on the server's word.Read the method →
- Demonstration dataAnonymised facilities and scaled volumes.
- A demonstration deployment shows Facility A, B and C with every volume scaled by an undisclosed factor, under a standing banner. Its figures are internally consistent and are not the real ones.
AI and the assistant
- AI questionA question, not a finding.
- The anomaly screening flags a month that departs from its own seasonal pattern. A shutdown, a compressor trip or a well coming on would look like this — so would a mis-keyed column. Only a person can say which, and nothing an AI module raises changes a figure.Read the method →
- Triage priorityThe order of the review queue.
- A fitted model's probability that an item is a real issue — once at least 200 labelled reviews with both outcomes exist — otherwise the documented fallback: severity, then anomaly score, then age.Read the method →
- Satellite reconciliationTop-down retrievals checked against the bottom-up inventory.
- A divergence becomes a question in the review queue. A retrieval cannot attribute methane to a process and sees only large emitters under favourable conditions, so it never replaces a source-resolved figure.Read the method →
- Grounded assistantAnswers only from recorded figures, with citations.
- It retrieves figures under the asker's own access, cites each one, and checks every number in its answer against what it retrieved. It refuses forecasts, compliance verdicts, enforcement advice, calculations and questions about people.Read the method →