MasterBrain — Emission Factor Database

What MasterBrain is

MasterBrain is the open, version-controlled emission factor database behind every calculator on GreenCalculus. Its point is not that it holds numbers — plenty of databases do. It is that every number comes back with its exact source, the place within it the value was read from, and the data version it belongs to, so a person can cite it and a machine can reproduce it in an audit.

That matters because of who has to use it. A sustainability consultant assembling a client inventory, an in-house ESG team preparing a disclosure, an auditor checking provenance, a researcher needing something citable, a founder building on a stable factor layer — none of them need a number. They need a number they can defend when someone asks where it came from, six months later, after the source has been revised.

One shape across 249 publishers

The work is the normalisation. Government and intergovernmental publishers do not agree on units, boundaries, gas coverage, vintage conventions or how to name anything. MasterBrain runs a single canonical schema across all of them, so a UK natural-gas factor, a Singapore grid factor and a US spend-based factor share one row shape, one unit convention and one provenance model. You stop writing per-source adapters and start comparing like with like.

Every row carries the same envelope: the value and its unit, the gas and GWP basis it was computed on, the source id, the exact table or cell reference, the date it was retrieved, and the GHG Protocol scope it belongs to. Nothing is derived silently. Where a publisher states an uncertainty range, the range travels with the factor; where none is published, the field is absent rather than invented.

What it is not

It is not the largest emission factor database, and it is not trying to be. Volume is easy to buy and easy to inflate — a few thousand real values become a million by multiplying out every unit and region combination. MasterBrain is deliberately a curated layer: every row traces to a named primary publication that a human has read the licence terms of. If you need a unit-process life-cycle inventory for product modelling, you want ecoinvent or a comparable LCI database, and the two are complements rather than substitutes.

It is also not complete, and it says so. The absence register records every known gap and classifies it — whether nobody publishes it anywhere, it is in our backlog, or we found it and declined.

What it costs

Free to read and free to cite. Browsing and searching the corpus needs no key at all; a free key covers programmatic use with no card. What you may redistribute is a separate question, and it is decided by each publisher rather than by us: a portion of the corpus is readable and citable but cannot be passed on, and the bulk feed excludes those rows rather than pretending otherwise. Every source and its terms are listed on the sources page.

Explore the database

Three pages go deeper than this one can: what every number is sourced from, how far the coverage actually reaches, and what is missing and why.

Sample factors — eight worked examples

Eight rows from the corpus, showing what a factor actually looks like in use — a fuel, two grids, a GWP, freight, spend, well-to-tank and a land-use-change row. Six are real canonical rows rendered live through the [gc_factor] shortcode, so the value stays current with the database rather than going stale in prose. Two — road freight and the FLAG land-use-change example — are version-pinned snapshots from families not yet migrated to the live V3 keyspace, and are labelled as such.

Live coverage

The table below is generated live from the MasterBrain database every time this page loads — it is not a static snapshot. The Status column distinguishes live families (rows currently in use by GreenCalculus calculators) from pending families (roadmap coverage with no published rows yet, surfaced here so the roadmap is visible rather than hidden).

Each Family name links to that family’s methodology page, where the source selection, normalisation, and worked examples for that family are documented in depth. The Geos column counts distinct ISO 3166-1 alpha-3 country codes appearing in the family’s canonical-row keypaths, so a value of 50 means the family carries factors keyed to fifty separate countries.

How to access MasterBrain

Five routes, in the order most people need them. Every one returns the same factor rows from the same versioned database — what changes is the shape of the answer, not the data behind it.

1. Through GreenCalculus calculators (recommended)

The primary access path is the calculator surface. Every page under GreenCalculus calculators reads MasterBrain at render time: you pick a calculator, enter activity data, and receive a result in which every number on the output panel cites the underlying factor row, its source, and the retrieval date — provenance is attached to the answer, not buried in a methodology appendix.

By Scope, the common entry points are the Stationary Combustion and Mobile Combustion calculators for direct emissions, the Scope 2 Electricity calculator for purchased power, the Spend-Based calculator for purchased goods, and the FLAG Emissions calculator for land-sector accounting.

Each calculator page loads only the factors it needs rather than the full database envelope — a sub-10% payload for most calculators, which keeps the surface fast without sacrificing provenance.

2. REST API

The productised API lives at api.greencalculus.com. A free key gives 1,000 calls a month with no card. Browsing and searching the corpus needs no key at all:

curl "https://api.greencalculus.com/v1/factors?section=fuels&key_prefix=fuels.gbr.natural_gas"

Fetching a single factor by key, and the bulk corpus feed, do require one. Every response carries the database version in its meta block and in X-GC-Version, so a client can log exactly which release produced a given number.

The full contract is published as an OpenAPI 3 document at /v1/openapi.json — sixteen paths covering factor lookup, search, and the calculation engines. Generate a client from it rather than hand-rolling one. Full reference: the API docs.

A legacy unauthenticated route remains at /wp-json/greencalculus/v1/factors for read-only browsing, alongside /wp-json/greencalculus/v1/gwp-values, which returns the IPCC AR4/AR5/AR6 100-year GWP table as self-documenting JSON.

3. MCP server, for AI agents

MasterBrain is available to any MCP-capable agent at mcp.greencalculus.com. The agent looks a factor up mid-conversation and hands back the source, the version and the DOI with its answer, instead of guessing a number from training data.

{ "mcpServers": { "greencalculus": {
    "url": "https://mcp.greencalculus.com",
    "headers": { "Authorization": "Bearer YOUR_KEY" } } } }

Twelve tools are exposed: lookup_factor, lookup_factors, search_factors and resolve_factor for retrieval, calculate_activity, calculate_embodied, calculate_pcaf, calculate_electricity, calculate_freight, calculate_spend and calculate_business_travel for computation, and explain_absence for the gaps. resolve_factor takes a plain-English description and returns ranked candidate keys, which is usually what an agent actually wants; lookup_factors batches a list of keys into one call instead of fifty.

4. Client libraries

JavaScript and Python clients, plus a Postman collection, are published at github.com/jeremiahsay/greencalculus-sdk. Install them with npm install greencalculus or pip install greencalculus, or generate your own client from the OpenAPI document above.

5. Bulk data and the citable dataset

The consolidated feed at /v1/corpus returns the whole redistributable corpus in one paginated stream, and states in its own metadata how many rows it excluded and why. It is licence-gated rather than volume-gated: rows whose upstream terms forbid redistribution are never in it, so what you receive is what you may pass on.

Frozen, citable extracts live under the data directory, and the dataset carries a DOI (10.5281/zenodo.17202747) for academic citation. Cite the version you used, not the live database — see Citation and licensing.

Direct PHP, for plugin extensions only

If you are extending the plugin itself, three memoised accessors are available: gc_mb_raw_factors(), gc_mb_canonical_taxonomy_families() and gc_mb_canonical_sources_registry(). Each caches per request, so they are safe to call repeatedly in one page load. This is an internal path — third-party sites should use the REST API above, not call PHP directly.

Methodology and provenance

Source selection criteria

For any taxonomy family, MasterBrain selects a source by a four-tier hierarchy, always preferring the highest applicable tier for the geography in question.

First, primary government and regulatory publishers: the US EPA, UK DESNZ / DEFRA, the EU Commission, Eurostat, the ONS, the IEA, and the IPCC. When an official factor exists for the geography, it wins.

Second, intergovernmental and multilateral sources where no single government owns the factor: IPCC working-group reports, ISO standards, the GHG Protocol, the GLEC Framework, and IATA RP 1726.

Third, peer-reviewed academic datasets — the ICE Database v3.0, EXIOBASE, and the Cambridge CBECI for crypto — used where government and intergovernmental coverage runs out.

Fourth, industry consortia such as PCAF, the Pathfinder Framework, and ASHRAE refrigerant standards, used only when no government factor exists.

Every source resolves to a registered entry in the canonical sources registry. A row cannot cite a source that is not registered — the validator blocks the release if it tries.

Normalisation rules

Source data arrives in incompatible shapes: different units, different GWP bases, different country codings, different file structures. MasterBrain normalises all of it into one canonical form before a row is accepted.

Units resolve to SI base where possible — kg, kWh, MJ, m³, tonne-km. Geographic codes become lowercase ISO 3166-1 alpha-3 inside keypaths (gbr, usa, chn, deu, ind). The global-warming-potential basis defaults to IPCC AR6 100-year in engineering mode, with AR5 100-year preserved alongside for DEFRA regulatory-mode compatibility. And every row uses the nested factor/source/scope v3 canonical shape — a single structure across all families, which keeps downstream consumption simple no matter which family a consumer is reading.

{
  "key": "fuels.gbr.natural_gas.kwh_gcv",
  "factor": {
    "value": 0.18231,
    "unit": "kg CO2e per kWh",
    "gwp_set": "AR5_100"
  },
  "source": {
    "id": "DEFRA_2026",
    "cell_ref": "'Fuels'!D42"
  },
  "scope": { "ghg_protocol": "scope1", "category": "stationary_combustion" }
}

The validator pipeline

Every MasterBrain release runs through scripts/check-mb-shape.sh before merge. A release that fails any check does not ship. Three of those checks are worth understanding because they map directly to classes of silent data corruption.

CHECK 1 — outer-vs-inner key equality

Each canonical row is keyed by a dotted string — for example fuels.gbr.natural_gas.kwh_gcv — and also carries that string in an internal 'key' field. CHECK 1 enforces that the two match exactly, for every row. This catches the “row exists but lookup misses” class of bug, where a row is present in the array but unreachable because its declared key disagrees with its position. Backwards-compatibility aliases (_alias_of rows) are exempt, because their outer key intentionally differs from the canonical key they redirect to.

CHECK 2 — no integer outer keys

Canonical rows are dictionary-keyed by their dotted string. CHECK 2 blocks any row that has ended up under an integer index instead — which happens silently when a row literal is pasted as an anonymous list element rather than as a keyed dictionary entry. Such a row looks fine on inspection but is unreachable to factor lookups, so the validator refuses the release until the row is re-keyed.

Vocabulary strict-blocks

Each activity_type term carries its own strict-block defining required fields, forbidden fields, and value constraints. A gwp_value row, for instance, must have a non-null factor.gas, a factor.gwp_set matching AR4_100|AR5_100|AR6_100, and an empty factor.unit. Because vocabulary proliferation is far harder to roll back than vocabulary consolidation, existing terms are reused wherever a new term would overlap.

Versioning model

Three version tracks move in coordination.

The MasterBrain data version (GC_MB_VERSION) is date-anchored — currently v2026.262 — and bumps on row additions, value updates, and vocabulary changes. The validator version is a comment at the top of scripts/check-mb-shape.sh — semantic (e.g. v1.x) — and bumps when the vocabulary extends or a strict-block changes. The plugin version (GC_VERSION) is semantic (e.g. v6.x) and bumps on UI, schema, or shortcode changes.

A value-changing MasterBrain bump cascades: it triggers a per-page Major bump on every Calculator and Methodology page that renders the affected value, and each of those is recorded as its own entry in the changelog, one per affected page.

Update cadence

MasterBrain re-publishes on a realistic, upstream-driven schedule: weekly during active development phases (current), easing to roughly bi-weekly during stability phases. Releases are triggered by upstream publications — DEFRA’s annual June release, EPA’s annual eGRID update, IPCC working-group reports per cycle — plus GreenCalculus’s own coverage-expansion phases.

For context on cadence: ecoinvent typically issues one major release per year, DEFRA / DESNZ updates annually in June, and IPCC working-group reports cycle on six-to-eight-year intervals. MasterBrain re-publishes within days of each upstream release. Every release is recorded as an mb_release entry, visible in the changelog.

Why these sources

A factor database is only as defensible as its source-selection rule. This is ours, stated so you can disagree with it. The full registry — every publisher, licence, vintage and update cadence — is on the sources page.

Geography decides first

United States — US EPA. The EPA is the regulatory authority under the Clean Air Act and Subpart C of the Greenhouse Gas Reporting Programme, and US reporting entities are expected, often required, to use its factors for SEC climate disclosure, CDP and state rules such as California SB 253. MasterBrain inherits the EPA’s own internal hierarchy rather than second-guessing it: AP-42 for stationary combustion, MOVES for mobile, eGRID for electricity.

United Kingdom — DESNZ. Formerly DEFRA and BEIS, DESNZ is the UK Government’s official publisher, and its factors are what SECR, ESOS and CDP submissions are built on. The dataset is released each June and MasterBrain re-publishes within days.

European Union — Eurostat. Eurostat AEA and GVA factors cover EU geographies. Within a geography we prefer AEA over EPA’s USEEIO where the American industry classification does not map cleanly onto the European one — the NACE and NAICS codes diverge, and forcing one onto the other silently reassigns emissions between sectors. Currency basis is preserved per row, so a EUR 2019 figure says so rather than being quietly converted.

Global-warming potentials: both, not one

AR6 (2021) is the most recent IPCC assessment and the best current atmospheric science. DEFRA still uses AR5 for regulatory continuity. Rather than pick a side and force it on everyone, MasterBrain ships AR4, AR5 and AR6 100-year values together, with an engineering-mode default of AR6 and a regulatory-mode default of AR5 — so a disclosure that must be AR5 stays AR5, and an analysis that should use current science can.

The values come from IPCC AR6 Working Group I, Chapter 7 Supplementary Material, Table 7.SM.7, which carries authoritative 100-year and 20-year GWPs for roughly 200 gases. Each row cites that table through source.cell_ref, so a GWP traces to a specific cell rather than to “the IPCC”.

Sector methodologies, where a sector has one

Freight — GLEC. The Global Logistics Emissions Council Framework is the industry-standard freight methodology, endorsed by the Smart Freight Centre and the European Commission, covering road, rail, sea, air and inland waterway on a tonne-kilometre basis. It is what MasterBrain uses for Scope 3 Categories 4 and 9.

Aviation — IATA RP 1726. The airline industry’s standardised passenger-CO₂ method, used by offset programmes and by corporate Scope 3.6 reporting. Aviation rows reference it where applicable, alongside the ICAO calculator methodology.

Imports — EU CBAM. The Carbon Border Adjustment Mechanism defines default embedded-emissions values for goods entering the EU when supplier-specific data is unavailable. Importers need those defaults to report, so they live under the cbam section and are read from the same database as everything else.

When sources disagree

They do, and the disagreement is kept rather than averaged away. Where two credible publishers give different values for the same activity, MasterBrain stores both under distinct keys with their own provenance, and the selection rule above decides which one a calculator reaches for by default. Averaging them would produce a number no publisher stands behind and no auditor can trace.

Standards and frameworks alignment

How MasterBrain maps onto the major accounting and disclosure frameworks: what it supplies for each, which factor families carry it, and whether that coverage is live today. Every standard here has its own reference page — follow the name for the requirement itself, the clause numbering and the reporting boundaries.

Authorship, expertise and verification

Author

Every factor in MasterBrain is authored and maintained by Jeremiah Say, Lead Systems Architect and Methodology Curator at GreenCalculus, working from Singapore. His full profile, prior work, and contact details are on the author page.

The credentials that matter here are structural rather than personal: the methodology and its source code are open and inspectable, and every factor traces to a primary citation a reader can verify independently. Jeremiah also contributes the underlying methodology to the public record, including the Wikipedia articles on the GHG Protocol, IPCC AR6, life-cycle assessment, and CSRD. The release-verification process is detailed in the next section.

Verification standard

Verification is in-repo and reproducible, not a claim on a page. Two parity gates must pass before any merge, every release hands off through a recorded audit trail, and the public changelog documents the result. You can read the data-layer releases via the data-layer changelog and the plugin releases via the plugin changelog. The verification approach is consistent with ISO 14064-3, the standard for the validation and verification of GHG assertions.

Feedback and corrections

To report a correction or request a new factor or geography, use the contact channel on the author page — primary-source integration for new geographies is a roadmap item driven directly by these requests.

Frequently asked questions

Figures on this page are read live from MasterBrain v2026.262 (2026-10-15). The surrounding prose was last reviewed on October 9, 2026.

When these numbers change

MasterBrain is re-published continuously — DEFRA each June, EPA’s eGRID annually, IPCC per assessment cycle, and our own coverage phases in between. A factor you cited last quarter may have been revised since.

Factor Watch is a short email when a value you rely on moves: what changed, from what to what, which release it landed in, and the source that revised it. No digest, no marketing.

Citation and licensing

How to cite MasterBrain

Cite both the version and the retrieval date — they are not optional, because the database is updated weekly and a citation without them will not reproduce.

Prose form: Copy

Say, J. (2026). MasterBrain — The GreenCalculus Emission Factor Database (v2026.262, retrieved 2026-10-15). GreenCalculus. https://greencalculus.com/masterbrain/

BibTeX: Copy

@misc{say2026masterbrain,
  author       = {Say, Jeremiah},
  title        = {{MasterBrain --- The GreenCalculus Emission Factor Database}},
  year         = {2026},
  version      = {2026.262},
  url          = {https://greencalculus.com/masterbrain/},
  note         = {Retrieved 2026-10-15}
}

The version and retrieved date are the load-bearing fields here. Because MasterBrain re-publishes weekly, a factor value cited today may differ from the same keypath next month — recording the version is what lets a reader reconstruct your number.

License terms

MasterBrain is released under the CC-BY-4.0 (Creative Commons Attribution 4.0 International) license. It permits commercial use, redistribution, modification, and embedding in proprietary tools, subject to attribution. Attribution means citing MasterBrain per the format above, linking back to https://greencalculus.com/masterbrain/, and indicating any modifications you made.

One transitivity point: when MasterBrain attributes a factor to an upstream source — DEFRA, EPA, IPCC, and the rest — that upstream source’s own license terms also apply. Most government sources are public-domain or open-data licensed, but verify per source for any commercial-redistribution edge case.

Historical version retention

Every released MasterBrain version is retained indefinitely. Historical versions are accessible three ways: through the mb_release records (admin-side, and exposable at /mb-release/<slug>/ URLs if published publicly in future), through the plugin’s git history in the greencalculus-mb repository, and through the deploy log (scripts/deploy-log.txt), which records every shipped version with its timestamp and commit SHA.

For reproducibility, recompute against the version cited in your inventory — do not silently re-cite an old figure against the current version, because the value may have moved.

Permitted and prohibited uses

Permitted: corporate sustainability inventories, academic research, regulatory disclosures, internal tooling, and downstream commercial products, all with attribution.

Prohibited: misrepresenting MasterBrain as your own dataset, stripping attribution, and using MasterBrain values in any way that contradicts an upstream source’s own attribution requirements. The dataset is free and open; the one thing it asks in return is honest provenance. The full author and contact details are on the author page, and the source is in the open-source plugin repository on GitHub.