Power: methodology

We start from what each power plant reports to its government regulator every year: its emissions and how much electricity it generated (Australia: Clean Energy Regulator; United States: Environmental Protection Agency; India: Central Electricity Authority). We add up each company’s plants according to how much of each plant it owns, and project forward using the dates plants are due to close. The sections below describe Australia first; section 10 lists what differs for the United States and India.

Last updated 6 October 2026.

1. Data sources

NGER data for each financial year is published by 28 February of the following year. AEMO updates its closure file roughly every three months.

2. Years

All power years are Australian financial years, labelled by the year they start: 2024 means FY2024-25 (1 July 2024 to 30 June 2025). This applies to NGER actuals, projections, closure handling and the national inventory comparison.

Sources published on a calendar-year basis (Climate TRACE, IEA and IAM pathways) are matched to the financial year that starts in the same calendar year, a six-month offset that we note rather than adjust.

3. Which company a plant belongs to

We publish two bases, as for India and the US. Equity share is the default; operational control is one click away.

• Operational control. NGER assigns each facility to the company with operational control, which is not necessarily the owner. Gladstone Power Station is reported by its operator (NRG Victoria I), although Rio Tinto is the largest owner (42.1%; NRG Energy 37.5%). Company targets are drawn in this view, because companies usually set targets on the plants they operate.
• Equity share. NGER plants are linked to GEM units by AEMO unit ID (GEM’s gas tracker lists DUIDs), by reviewed aliases (e.g. Callide B and Callide C are separate units at one GEM site), or by name within the state. Matches rejected on review (e.g. two different plants called Moranbah) are listed in r_pipeline/process/gem_power_match_au.R. A plant’s share is the capacity-weighted share across its linked GEM units. Linked plants cover 97% of NGER fossil emissions since FY2013-14; in FY2024-25, 96% of emissions are split by a GEM share. Plants not in GEM (mostly renewables, small and off-grid plants) stay 100% with their NGER reporter’s group.
• Which group a company page shows. Each published company page shows the equity view for the reporter’s parent group in GEM: for example NRG Victoria → NRG Energy Inc, Alinta Energy → Sembcorp Industries, EnergyAustralia → CLP Holdings. The group’s equity view includes its share of plants operated by others (e.g. Origin’s 50% of Osborne, operated by ATCO). CK William is owned by three parent groups (CK Asset 40%, CKI 40%, Power Assets 20%), so it has no equity view.
• Ownership over time. GEM lists current owners. Equity shares are applied to the years a plant was reported by its latest NGER reporter, and to projections. Earlier years, under a previous reporter, stay 100% with that reporter. For example, Bayswater before AGL bought it in 2014 stays with Macquarie Generation, not AGL.
• Joint ventures reported twice. Some joint-venture plants are listed in full under more than one NGER reporter. Callide C appears under both CS Energy and the co-owner’s entity (IG Power) with identical figures. In the operational view each company’s series matches its NGER total, and when summing across companies the plant is counted once. In the equity view the plant is counted once and then split (CS Energy 50%). A check confirms the equity and operational national totals are equal in every year and scenario.
• Spelling changes. NGER sometimes spells a company’s name differently from year to year (for example “CS Energy Limited” vs “C S Energy Limited”, “NRG Victoria 1” vs “NRG Victoria I”). Names that differ only in spacing or punctuation are merged automatically, using the most recent spelling. Other known variants are mapped explicitly. Without this, the variants would appear as separate companies and leave gaps in the real company’s history.

4. Historical emissions

• Measure. Emissions are NGER Scope 1 for designated generation facilities, as published, in MtCO₂-e. The company total is NGER’s published corporate total, which equals the sum of the company’s facilities.

• Generation. Generation is NGER’s electricity production in MWh. The FY2013-14 file reports only GJ; we convert it (1 MWh = 3.6 GJ).

• Generation sent out. NGER electricity production is gross: the Clean Energy Regulator defines it as total generation including the plant’s own use (auxiliary load). India (CEA) and the US (eGRID) report net generation, i.e. sent out. So that intensity means the same in every country, we publish generation and intensity on a sent-out basis:

  sent out = NGER generation × (1 − auxiliary %)

  The auxiliary % is one factor per plant from AEMO’s 2025 Inputs and Assumptions Workbook, applied the same way to every company (no company-specific calibration):

  1. the plant’s AEMO generating units (DUIDs), if listed in the workbook;
  2. otherwise the station name;
  3. otherwise the workbook median for the plant’s fuel, flagged as estimated. This covers WA, NT and off-grid plants (the workbook covers the NEM only) and plants that closed before 2025.

  Factors are about 5.5% for NSW black coal, 6.7% for Queensland black coal, 8.4% for Victorian brown coal, 1.8% for CCGT, about 0.7-1% for OCGT and reciprocating engines, and under 0.1% for wind and solar. In FY2024-25, 168 TWh of NGER generation used a plant-specific factor (157 TWh by unit, 11 TWh by station name) and 56 TWh a fuel median. NGER gross generation is kept in the downloads (generation_gross_twh). Historical company sent-out generation is the NGER corporate total × the sent-out share of the company’s plants that year. Projections carry each plant’s sent-out baseline forward, so actuals and projections are on the same basis.

• Intensity is NGER Scope 1 emissions per MWh sent out.

• Gross option (Australia). Australian company pages have a Sent out / Gross switch on the intensity and generation charts, and show gross intensity under the headline figure. Gross is NGER generation before own use. Sent out stays the default because it is what customers receive and it matches the other countries. To compare a company with a climate pathway, gross is usually used, because the pathways (for example the IEA’s and the IPCC AR6 models) count electricity generated gross. The Carbon Budget Tracker uses gross generation for this reason (generation_twh in cbt_power_series.csv). Emissions are the same on both bases. India (CEA) and the US (eGRID) publish net generation only, so they have no gross option.

• How others report it. Company pages show company-reported figures and TPI next to ours, for context only. They are not directly comparable: companies use their own boundaries and methods. For example, AGL’s “operated and contracted generation intensity” (used by TPI) includes electricity AGL buys under contract, and AGL’s reported emissions are Scope 1 and 2. For AGL in FY2024-25, our sent-out generation is 33.6 TWh, against 31.3 TWh AGL reports as sent out from its operated plants. The 7% gap is not explained by own-use factors alone (it implies 12% own use and losses, against 5.8% from AEMO’s factors) and may reflect which plants AGL counts as operated. We do not adjust our figures to match any company. Each company-reported value is recorded with document, page, quote and link in data/raw/CompanyReports/power_company_reported.csv. TPI labels financial years by their start year, as we do.

• Climate TRACE. Company pages (Australia, US) also show Climate TRACE v5.2.0, which models plant emissions rather than reporting them. Climate TRACE plants are linked to ours (Australia by name, US by location within 2 km) and given our GEM equity shares, so the comparison shows Climate TRACE’s modelling, not differences in ownership. Each page lists the plants behind the largest gaps, split into generation and emissions per MWh. Across the 39 published companies the median gap is 7%; single plants differ much more (e.g. Yallourn −25%, Collie +114%). Tests against NGER, eGRID and CEA: QA item G4.

• Rounding. Published values are rounded to 0.1 Mt.

• Ownership-change years. Years in which a company’s total moves by more than 25% are flagged as possible part-year ownership changes. For example, AGL FY2014-15 includes about 10 months of Macquarie Generation (Bayswater, Liddell), and Origin FY2013-14 includes part of a year of Eraring.

• Latest data. The most recent NGER year is always used as an actual. Projections start the following year.

5. Plant closures

• Source. Closure dates come only from AEMO’s Generating Unit Expected Closure Year file, at the level of individual generating units.

• Matching. Each NGER facility is matched to its AEMO station or stations by name, with a reviewed list of explicit matches where names differ. Examples:

  • NGER’s “Yallourn Power Station” is AEMO’s “Yallourn W”.
  • “Tarong Power Stations” covers AEMO’s Tarong and Tarong North.
  • AGL’s Loy Yang is Loy Yang A only.

  Matches are only allowed between the same kind of plant, so a power station cannot be matched to a solar farm or battery with a similar name.

• Closure rule. A unit closing in calendar year Y emits up to the end of the financial year ending 30 June Y, and nothing after. Where AEMO gives an exact closure date, Y is the year of that date. Otherwise Y is AEMO’s Expected Closure Year, which we read as a calendar year (AEMO does not specify). Examples:

  Plant	AEMO closure	Last year with emissions
  Eraring	30 Apr 2029	FY2028-29
  Yallourn W	1 Jul 2028	FY2027-28
  Bayswater	2033	FY2032-33

  The rule is exact for closures between January and 1 July. A closure later in the year would lose up to six months of emissions; none of the companies we publish currently has one.

• Multi-unit stations. When a station’s units close in different years, the station’s emissions fall in proportion to the share of units still open, each unit counting equally because AEMO does not publish unit capacities. Stanwell’s four units close in 2043, 2044, 2045 and 2046, so its emissions step down by a quarter each year.

• How closure dates have moved. For every plant we show each change AEMO has recorded since July 2021, from the change logs in its Expected Closure Year releases, for example Eraring: 2031/32 → 2025 (Feb 2022) → 2027 (May 2024) → 30 April 2029 (Jan 2026). When we measure a change, a year-only value is placed at mid-year. The Power overview page lists every change for coal and gas plants.

• No AEMO date. AEMO’s file covers the National Electricity Market (NSW, Vic, Qld, SA, Tas) only. Plants in Western Australia and the Northern Territory therefore have no closure date and are projected to keep operating. Each one is flagged “No AEMO closure date (WA/NT)”. The same applies to units for which AEMO records no closure year (shown as 2200 in AEMO’s file, e.g. Callide C).

6. Projections

• Starting point. Each plant starts from its average over the latest three NGER years (currently FY2022-23 to FY2024-25), or fewer years for a newer plant. The window moves forward with each NGER release.

• Which plants are projected. Only plants that report in the latest NGER year. Plants that have closed (e.g. Liddell, April 2023) or now report under a different company drop out automatically.

• Projections. Each company has separate, labelled projections that are compared, never blended:

  Projection	What it assumes
  AEMO closures	Each plant continues at its starting level until its units close on AEMO’s dates (section 5). Generation from closed units is assumed to be replaced by zero-emission supply.
  Closures delayed as in the past	The same, with each closure moved later by the typical delay AEMO has recorded when a date was moved later since July 2021, by fuel, across all NEM units: currently coal +1.9 years (4 delays: Callide B, Eraring twice, Tarong) and gas +3.4 years (13 delays), rounded to whole years. Recalculated with each AEMO release. Note: AEMO has also moved many coal closures earlier since 2021 (e.g. Loy Yang A 2048 to 2035, Gladstone 2035 to 2029).
  Closures 5 years late	Stress test: every closure moved 5 years later.
  Company targets	Emissions follow the company’s stated target line (section 7). Only for companies with a target in tonnes.
  No closures	Each plant continues at its starting level to 2050.

• Why a three-year average. It smooths one-off outages and unusual weather years. The trade-offs:

  • it carries older outages forward (Callide C was largely offline after unit failures in 2021 and 2022, so its average of 1.8 Mt is below its FY2024-25 level of 3.7 Mt);
  • it lags when coal plants are steadily running less.

• Sanity rule (checked automatically on every data update). No projected year may exceed the highest of the company’s last three actual years, unless the company has a plant that is new to it within that window, from construction or acquisition. Companies that rely on this exception are listed in the check output. For example, APA’s projection is 0.04 Mt above its recent high because it took over Newman and Port Hedland from Alinta in FY2023-24.

• Independent check. AGL and Origin projections were compared with a separate, independently coded calculation. Cumulative 2025-2050 emissions agree within 3% (AGL 282 vs 279 Mt; Origin 84 vs 86 Mt), and every difference comes from a documented choice: the three-year starting average, per-unit closures, and the newer AEMO release, in which Origin moved all its gas plants’ closure years five years later.

7. Company targets

• Source. Targets come only from the Oxford Net Zero Tracker snapshot, and every target line is labelled with the snapshot date (currently 20 January 2026).
• Baseline. Percentage reductions apply to the company’s NGER Scope 1 emissions in the target’s baseline financial year. For example, AGL’s FY2019 baseline is 42.7 Mt.
• Years. A target “for FY2035” or “by 2035” is read as the financial year ending June 2035.
• The line. It starts at the baseline year and runs in straight lines between the target milestones. After the last milestone it stays flat.
• Milestones.
  • Most companies: the interim and end targets in the tracker’s structured fields.
  • Where the tracker’s notes transcribe a company’s full year-by-year schedule, we use it and label the line “detailed schedule”. This is currently AGL only (Climate Transition Action Plan 2025: −17% in FY26, −19% in FY27-FY34, −52% in FY35, ≥90% and net zero in FY36).
  • End point: the tracker’s end-target %. If there is none, we use a gross % stated in the tracker’s text (AGL: 90% gross by 2050). If neither exists, the line goes to zero, labelled “net zero (may rely on offsets)”.
• Scope. Company targets often cover Scope 1+2 or Scope 1-3, but are compared here with NGER Scope 1 for power generation. These lines are flagged as approximations.
• Intensity targets. Some targets are in emissions per unit of output, for example Origin’s −40% Scope 1-3 equity emissions intensity (t CO₂-e/TJ) by FY2030. These are shown with the company’s intensity data in their own units, not converted to tonnes. Where the tracker’s structured field and its text disagree about the target type, we follow the text and flag it.
• Parent-group targets. A target set by a parent group covering far more than the Australian business (CLP for EnergyAustralia, Engie for International Power) is listed but not applied.
• Example finding. AGL’s stated targets sit above its own AEMO closure projection in every year. Following its target line would mean 401 Mt over 2025-2050, against 282 Mt on AEMO closure dates, which is roughly the same as its coal closing 4-5 years late.

8. Coverage

We publish the ten largest generators: ranked by FY2020-21 emissions among companies still emitting in the latest NGER year. In FY2024-25 they emitted 112.5 Mt, 83% of NGER power-generation emissions (135.8 Mt). That total counts joint-venture plants listed under two reporters once, and treats blanks in the NGER files as zero.

9. Known limitations

• Equity shares are GEM’s current owners, applied to the years under the plant’s latest NGER reporter; ownership changes within that period are not tracked. GEM’s parent names are used as published, except where GEM itself uses two names for one group (Verve Energy and Synergy; GDF Suez, Engie Services and ENGIE SA).
• Sent-out generation is estimated from AEMO’s assumed own-use factors, not metered. AEMO’s public unit data (and OpenElectricity) is measured as generated, so it cannot replace this. Factors are AEMO’s current values, applied to all years.
• WA and NT plants have no closure dates (see section 5).
• Unit closures are weighted equally within a station, not by capacity.
• AEMO’s Expected Closure Year is read as a calendar year.

10. United States and India

The same method as Australia, with each country’s official sources. Sections 1-9 describe Australia; this section lists what differs.

• Companies are parent groups. EIA and CEA name operating subsidiaries (e.g. Georgia Power, NTECL), so companies are GEM parent groups. GEM parents are rolled up to their ultimate parent company with GEM’s Global Energy Ownership Tracker (only wholly owned subsidiaries, and never into governments, so state utilities stay separate), plus a short reviewed list (Union Electric → Ameren, Entergy Louisiana → Entergy).
• Equity share (default): each plant split between parent groups by GEM equity shares; plants not linked to GEM go to their operator’s group. Operational control: 100% of each plant to its operator’s group. An operator belongs to the group that owns at least half of what it operates (by emissions); otherwise it is its own group. GEM’s owners and the national operator are both current, so shares apply to all years.
• Linking to GEM: US gas and oil plants by the EIA IDs in GEM (IDs that GEM gives to more than one plant are left to name matching); coal plants by name within state, checked against capacity; reviewed aliases for CEA abbreviations and renamed stations (e.g. Derang = Angul, Sembcorp Gayatri = Painampuram). Linked plants cover 94% of US and 92% of Indian emissions.
• Company targets (US): from the Oxford Net Zero Tracker, on the same rules as Australia, for the 19 published groups it covers (listed companies). Absolute targets are drawn on the operational-control view from the group’s emissions in the baseline year; for baselines before 2018 (e.g. 2005) this is the group’s current plants in that year, which can differ from the portfolio the company reports its own baseline on. Intensity targets and net-zero-only targets are listed, not drawn.
• Not yet added: targets for Indian companies; company-reported figures and TPI comparisons for the US and India.

11. Europe

The same method, with European sources (EU 27, Norway, Iceland, Liechtenstein and the United Kingdom):

• Emissions: verified emissions per installation under the EU emissions trading system (European Commission Union Registry, 2013–2025) and, for the UK from 2021, the UK emissions trading system. These include heat from combined heat and power plants. Installations count as power if they are combustion installations linked to a power plant in Global Energy Monitor’s trackers, or registered as electricity and heat producers. Steelworks power plants that burn blast-furnace gas are counted in steel, not power.
• Companies: the 30 largest parent groups (Global Energy Monitor), equity share by default and operational control (the installation’s operator in the registry) alongside. Today’s owners are applied to all years.
• Generation and intensity: net generation per unit from ENTSO-E (units of 100 MW or more), shown for a company only where its plants with complete generation data carry at least 80% of its emissions; otherwise left blank. Plant-level generation is not shown.
• Closures: planned retirement years from Global Energy Monitor. Few gas units have announced dates, so the planned-retirements projection is mostly coal closures.
• Not yet added: company targets and company-reported figures for European groups.