This document maps the FEM energy and emissions data — the Assessment, FacilityType, ReportedEnergyUse, EnergyUse, and Emissions records — against the GHG Protocol’s scope 1 and scope 2 inventory guidance. It is written for data consumers (brands, analysts, inventory builders) who fold facility data into their own GHG accounting and need to know, per accounting requirement:

• what data the FEM collects and what HAM computes from it,
• what assumptions stand in where data is missing,
• where the methodology aligns with the guidance, where it aligns with caveats a consumer must carry, and where there is a gap the consumer must fill from outside the FEM.

It is a companion to guidance-ghgp-fem.md (the methodology itself, cadence by cadence) and guidance-eeci.md. This document does not repeat the calculation mechanics; it judges them against the guidance.

Verdict vocabulary used throughout:

• ✅ Aligned — use the data as-is for this requirement.
• ⚠️ Aligned with caveats — the mechanism is GHGP-shaped, but there are assumptions or coverage limits you must disclose or correct for.
• ❌ Gap — the FEM does not carry this; fill it yourself or scope it out of your inventory explicitly.

Alignment at a glance

Two slices of the same alignment. The first table takes the lens of the figures and statements a GHGP report is expected to carry — the Corporate Standard’s required reporting information (chapter 9) plus the Scope 2 Guidance’s dual-reporting figures — and answers, per line item: can you produce this from HAM, and with what attached? The second slices by the accounting requirements behind those figures. Verdicts use the vocabulary above; where a line item splits (one part met, another not), both verdicts appear. Each row links to the section carrying the full reasoning — the caveats here are the headline, not the whole story.

The figures a report carries

The requirements behind the figures

Reference guidance

The alignment below is anchored to the two normative GHG Protocol documents:

1. GHG Protocol Corporate Accounting and Reporting Standard (revised edition, PDF) — the foundational standard: principles, boundaries, scopes, gases, base years.
2. GHG Protocol Scope 2 Guidance (2015 amendment, PDF) — dual reporting, the market-based method, contractual-instrument Quality Criteria, residual mix. Chapter and table numbers cited in this document (e.g. Table 6.1, §6.4.1) refer to this edition.

On sector guidance. There is no GHG Protocol sector edition for apparel, footwear, or textiles covering scope 1 and 2 inventory methodology. The apparel-specific material that exists (notably the SBTi Apparel and Footwear guidance) addresses target setting, and for a facility’s own scope 1 and 2 accounting it defers to the Corporate Standard. This document therefore judges the FEM against the two standards above directly. Where it borrows practical inventory advice that is not itself a normative requirement — the “investigate year-over-year changes above 10%” screen, for example — that advice comes from the GHG Protocol’s Scope 1 & 2 GHG Inventory Guidance (November 2019), a dairy-processing sector edition reviewed by the GHG Protocol for conformance with the Corporate Standard. It is cited as a source of practice, not as an authority for apparel facilities.

What a FEM is, in GHGP inventory terms

A FEM is an annual, single-site facility self-assessment, optionally corrected by an accredited verifier. Translated into the handbook’s inventory-design vocabulary:

• Reporting entity = one facility site. There is no corporate roll-up, no subsidiary structure, and no consolidation-approach question (equity share / financial control / operational control). The implicit model is 100% operational control of a single site.
• Reporting period = the FEM reporting year (e.g. FEM 2025 reports calendar year 2025 data). There are no explicit period-start/period-end fields on the assessment or the records; calendar-year alignment is a program convention stated in question guidance, not a data field. Partial-year operation is not captured.
• Role in a consumer inventory: a facility’s scope 1 and scope 2 data is, for most data consumers, an input to their scope 3 (purchased goods & services). The scope labels on the records are the facility’s scopes, not yours.

Everything below judges the facility-level inventory the records represent.

The five accounting principles, scored

The Corporate Standard’s five principles, applied to the FEM records:

Alignment detail

Organizational boundaries

Guidance: choose equity share, financial control, or operational control; apply it consistently to all scope 1 and 2 sources; classify leased assets accordingly.

FEM: no consolidation concept exists. One FEM = one site, implicitly 100% operational control. Vehicles are explicitly bounded as “owned and controlled … e.g., leased vehicles that are used by the facility” — operating-leased vehicles the facility uses are in boundary, which matches the operational-control treatment of leases.

Verdict: ⚠️ Aligned with caveats. For a consumer, the caveat is aggregation-side: if you consolidate facilities into a corporate view, the equity/control decision is yours, and the FEM gives you no ownership-share data to weight by. Shared sites, on-site subcontractors, and landlord/tenant splits are invisible.

Operational boundaries and scope assignment

Guidance: categorize each source as scope 1 (owned/controlled combustion, process, fugitive) or scope 2 (purchased electricity, steam, heat, cooling); report scopes separately.

FEM: every energy source carries an explicit per-cadence scope in the reference data, and every EnergyUse/Emissions row carries a scope field. Current (fem2023+) assignment:

• Scope 1 — all on-site combusted fuels (coal, natural gas, LNG/CNG, LPG/propane, diesel, petrol, fuel oil, biomass variants, biogas, biodiesel, ethanol, fabric waste, hydrogen), on-site generation (solar PV, solar thermal, wind, micro-hydro), and all vehicle fuels.
• Scope 2 — purchased electricity, purchased renewable electricity, purchased steam, district heating, purchased chilled water.

Notable movements across cadences (relevant if you consume multiple years): wind was scope 2 in fem2017–fem2022 and is scope 1 (on-site generation) from fem2023; geotherm/hydro were scope 2 options through fem2022 and are only reachable as mix components afterwards; microhydro has always been scope 1.

Why the older cadences differ: no on-site/off-site distinction

Those movements are not reclassifications for their own sake. Before fem2023 the FEM did not ask whether a renewable source was on-site or off-site. A facility reporting “wind” could equally have meant a turbine on its roof or a wind tariff from its utility, and nothing in the answer distinguished the two.

Rather than leave those volumes unattributed, the model resolves each pre-fem2023 renewable to its most common real-world case. Geothermal, hydro, and wind are treated as off-site — purchased generation whose attributes arrive with the electricity — so under the location-based method they are re-attributed to purchased grid electricity. The attribution is a fixed list applied at record-generation time:

# assessments/fem2022/extras/ghgp.stl — location-method attribution
offsite_renewables = [
    "geotherm",
    "hydro",
    "wind",
    "purchrenew",
]
source = this.row["_source"]
if this.row["_source"] in offsite_renewables:
    source = "electricpurch"
else:
    source = EMISSIONS[this.row["_source"]].constant(
        "attributed source", this.row["_source"]
    )

microhydro, solarphoto, and solarthermal are absent from that list because they are unambiguously on-site by nature, and stay at their own zero emission factor under both methods.

Two consequences for a consumer of pre-fem2023 records:

• A facility that genuinely had on-site wind is overstated under the location-based method — its generation is priced at the grid factor. There is no field that lets you identify or correct those facilities; the information was never collected.
• The market-based method is where the assumption shows. These sources keep their own zero factor under the market method (the modeled case is off-site generation with retired certificates), so geothermal, hydro, and wind are the only sources whose market and location rows differ before fem2023. See “Cadence applicability”.

From fem2023 the distinction is a reported fact rather than an assumption: on-site generation is its own question set, and the facility’s own answer about selling certificates drives the attribution.

Verdict: ⚠️ Aligned with caveats. Two caveats:

1. Scope-level totals are on the Assessment record, in both methods: total_market_scope_one_emissions, total_market_scope_two_emissions, and their location counterparts. Like total_*_emissions they exclude biogenic, which the memo totals carry, so scope 1 + scope 2 accounts for the whole non-biogenic figure with nothing left over.

   FacilityType does not carry a scope split — GHGP reports scopes at the entity level, and a facility type is a sub-facility breakdown. If you need one, sum the Emissions rows filtered to one scope_two_method and grouped by scope.

2. Sold-certificate volumes are split across scopes. On-site generation whose certificates were sold is re-priced at the grid factor under the market method, and those rows carry scope: "two"; the retained share keeps scope: "one". This follows Scope 2 Guidance Table 6.1 (owned/operated generation → certificates sold to 3rd party) and §6.3, which puts consumption from owned generation whose attributes were sold into scope 2 activity data — §6.4.1 requires it to be treated “as though it were purchased from the grid”. The standard’s carve-out (“scope 1 reporting shall still reflect any emissions from the generator”) does not apply, because none of the affected sources — solar PV, solar thermal, wind, micro-hydro — combusts anything.

   Note this is the one place where the scope label does not follow the reported source. Elsewhere it always does: purchased steam decomposed into its coal fraction stays scope 2, because the supplier burned the coal, not the facility.

Gases and global warming potentials

Guidance: account for all seven Kyoto gases; use 100-year GWPs from the most recent IPCC assessment report, consistently across years; report the GWP source.

FEM:

• Fuel emission factors are single blended kg CO₂e/MJ values — derived from IPCC stationary-combustion defaults with AR6 GWPs (fem2025 citations; ethanol from US EPA 2025). There is no per-gas (CO₂/CH₄/N₂O) breakdown anywhere in the data.
• Refrigerant GWPs (in-assessment only, see Fugitives) come from mixed sources and vintages — IPCC AR6 for a subset, plus vendor tables, EU F-gas documents, EPA tools, and older Kyoto-era values; some entries have no value at all and compute as zero.

Verdict: ⚠️ for combustion CO₂e (blended factors are standard practice for corporate inventories; the vintage is documented), ❌ for per-gas reporting (if your program requires gas-by-gas disclosure — e.g. CH₄ and N₂O reported separately — you cannot derive it from FEM data), and ❌ for refrigerant GWP consistency (no single AR vintage; treat any refrigerant-derived number as indicative).

The refrigerant GWP caveat bites on the fugitive rows described under “Scope 1 — fugitive emissions”: those values are only as good as the GWP table behind them. Refrigerants with no authoritative GWP produce no row at all rather than a zero, so the effect of a gap is an understatement you can see, not a wrong number you cannot.

Scope 1 — stationary combustion

Guidance: collect fuel quantities per source from meters or purchase records; convert physical units to energy; apply fuel-appropriate emission factors; document heating-value basis.

FEM: facilities report annual quantity + unit per fuel source (the purchase-records method, in the handbook’s terms). Unit conversion to MJ is systematic (GNU units-derived conversion tables, with volume/mass/energy paths per fuel). Fuel EFs are IPCC-derived per source. Blends are split: diesel/biodiesel by reported percentage; petrol/ethanol by volume-weighted share. Biomass with certificates and generated biomass can be decomposed into sub-source mixes (agricultural residue, crops, forest residue, mill residue, solid wood) by reported percentages.

Verdict: ✅ Aligned, with two disclosures to carry:

• The records apply the latest fuel EF vintage to all cadences (methodological-consistency choice, documented in the methodology guide). If you need vintage-of-year factors, recompute from the reported MJ (EnergyUse.value) with your own factors.
• Heating-value basis (HHV vs LHV) is not declared per fuel in the data; the factors follow their IPCC/EPA sources’ conventions.

Scope 1 — mobile combustion

Guidance: include owned/leased vehicles under your control; collect fuel consumed (minimum), ideally distance and vehicle characteristics.

FEM: a dedicated vehicle-energy table (15 fuel options including electric, hydrogen, and renewable purchase) collects annual fuel quantity per source for “owned and controlled vehicles” (leases included). Rows carry coverage: ["vehicle"] and flow through the same EF machinery. Vehicle blends (diesel/biodiesel, petrol/ethanol) are split like stationary blends.

Verdict: ⚠️ Aligned with caveats. Fuel-quantity-only: no vehicle counts, classes, or distances, so no fuel-economy cross-checks and no CH₄/N₂O refinement by vehicle technology (subsumed in the blended CO₂e factor). Vehicle electricity is re-attributed like facility electricity (grid factor), which is correct. The coverage split (production / domestic / vehicle) only exists from fem2023; earlier cadences cannot isolate vehicle energy.

An assessment with no vehicle rows is ambiguous on its own; vehicle_energy_status on Assessment (fem2023+) tells the cases apart: not_applicable (the facility reports no company owned and controlled vehicles), untracked (vehicles are reported but the tracking opt-in is anything short of a full “yes”, no fuel sources are reported for them, or a reported source is missing the quantity or unit its row needs), or tracked (every reported fuel source carries what its row needs). Like ghgp_completion_status, tracked means completely tracked — one gap drops the signal. Undefined before fem2023 and while the vehicle questions are unanswered. The domestic-coverage rows carry the analogous domestic_energy_status, whose not_applicable also covers the facility that reports domestic energy combined with production — there the data is in the combined-coverage rows, not missing.

Scope 1 — fugitive emissions (refrigerants)

Guidance: quantify refrigerant losses via screening, sales-based, or mass-balance methods; account per gas because GWPs differ; small releases matter.

FEM: collects, per refrigerant type, the quantity added to existing equipment during the reporting year (Q6.1.2) — a top-up proxy roughly equivalent to the service component of the sales-based method. In-assessment calculations multiply by per-refrigerant GWPs into totalRefrigerantEmissions, which joins the in-product totalGHGemissions figure.

Verdict: ⚠️ Aligned with caveats. Refrigerant releases are on the records from fem2023, as Emissions rows carrying emission_type: "fugitive" and scope: "one", identified by refrigerant through reported_source_key. They are emitted under both accounting methods, so a scope 1 total filtered to either sees the same figure, and they roll up into total_<method>_scope_one_emissions.

They are deliberately not in total_*_emissions or EECI, which stay pinned to energy-derived rows so their meaning did not change when fugitives were added. Filter on emission_type when you sum this table yourself, and note that a refrigerant key will not join against the energy source reference data.

Carry these caveats:

• Top-up-only is a floor: no installation/disposal losses, no beginning/ending inventory, no equipment capacity. Note this is a methodological floor, not an error — annual top-up is the service-component proxy the sales-based method recognises. What it cannot capture is the year in which equipment is installed or decommissioned.
• Top-ups are lumpy in a way annual reporting does not smooth. Equipment is often recharged every few years rather than annually, so a single reporting year can show a large top-up against neighbouring years showing none. The reported figure is a faithful record of that year’s service, but it is a poor estimate of that year’s leakage rate. Averaging across consecutive assessments is the right correction and is not something the FEM can do for you today — each assessment is evaluated in isolation.
• The GWP table has mixed vintages, and 15 of the 127 refrigerants offered have no GWP value. Those produce no row at all rather than a zero-valued one — a missing row is a visible gap, whereas a zero would assert that the release had no warming effect. For four of them (r717 ammonia, r718 water, r728 nitrogen, r740 argon) zero would have been approximately right anyway; for the rest it would not, r400 being a CFC blend whose true GWP is in the thousands. Populating these is in progress; until then a facility reporting one of the fifteen contributes nothing to the fugitive total — but not silently: refrigerants_status reports it (see below).
• No refrigerant data exists before fem2023.
• Implausible top-ups fail the data-quality gate (HAM 0.3). A refrigerant total of at least 10,000 t CO₂e that also exceeds the assessment’s entire energy-related scope 1 + 2 raises Assessment: Refrigerant emissions of … kg CO2e are implausibly high against energy-related emissions of … kg CO2e and fails pass_data_quality. The ceiling is the most a single very large industrial system can release in a year (IPCC 2006 Vol. 3 Ch. 7 Table 7.9: ~10 t charge at 7–25 %/yr, at the highest GWPs), and a facility running several has an energy footprint far above it; below the ceiling, reported refrigerant emissions run 1–10 % of a facility’s energy emissions, the flagged ones 15× to 20,000× — a kilogram entered as a tonne, or a charge entered as a top-up. The fugitive rows are still published, so the entry can be inspected; the assessment simply drops out of pass_data_quality-filtered analysis. In FEM 2024 the flag removes ~11 Mt of spurious scope 1 (two self-assessed filings hold 10 Mt of it); other cadences lose under 1 Mt.

An assessment with no fugitive rows is ambiguous on its own; refrigerants_status on Assessment (fem2023+) tells the cases apart:

• not_applicable — no refrigeration, cooling or air-conditioning emission sources; no rows are expected.
• unmodeled — a positive refill was reported for a refrigerant with no authoritative GWP. The rows understate the facility’s releases, and no answer the facility could give would fill the gap — reported ahead of the other states because it taints whatever rows do exist.
• untracked — refrigerant use applies but the facility does not know which refrigerants it uses, its tracking answer is anything short of a full “yes”, or a declared refill is missing the quantity or unit its row needs. Missing rows here mean missing data, not absent releases.
• tracked — every declared refill carries what its row needs. A year with no refills is tracked with no rows — zero releases, not missing data (though see the lumpiness caveat above).

The field is undefined before fem2023 and while the air-emission source questions are unanswered — deliberately not not_applicable, since the facility never said either way.

Treat the figure as a screening estimate that is understated by construction, not as a complete fugitive inventory.

Scope 1 — process emissions

Guidance: include process emissions (chemical/physical transformations) in scope 1 where they occur.

FEM: the air-emissions section collects a qualitative inventory of emission sources (boilers, generators, incinerators, ovens, production processes: dyeing, printing, solvents, ODS use, etc.) and compliance status — but no quantities except refrigerant top-ups. Nothing from it enters the records.

Verdict: ⚠️ Aligned with caveats for wastewater, ❌ Gap otherwise.

On-site wastewater treatment is on the records from fem2023, as Emissions rows with emission_type: "process". Each wastewater stream is priced on the organic load it sheds — volume × the drop in BOD5 across treatment — times a per-process yield from the US EPA’s biogenic-emissions estimation methodology, and reported as two rows:

• CH₄ under wastewater_ch4, scope: "one". Methane is scope 1 even though its carbon is biogenic; the Corporate Standard’s biogenic carve-out is for CO₂ only.
• CO₂ under wastewater_co2, flagged biogenic: true. It is therefore excluded from the scope rollups, as biogenic CO₂ should be.

Both of the FEM’s BOD5-monitoring shapes are covered: a facility reporting one before/after pair per stream produces one row pair per stream, and a facility monitoring at sub-process level produces one row pair per treatment stage. The full derivation — equations, parameters, and the assumption each parameter choice carries — is in Appendix A.

Because rows only materialize when they carry a positive value, an assessment with no wastewater rows is ambiguous on its own. The Assessment record disambiguates it with wastewater_emissions_status:

• not_applicable — no on-site (or zero-liquid-discharge) treatment with a secondary (biological) stage; no rows are expected.
• untracked — treatment is applicable but the facility does not monitor BOD5 in a usable shape, or reported no stream with the volume and before/after readings the calculation needs. Missing rows here mean missing data, not absent emissions — fill from outside the FEM or treat as a coverage gap.
• tracked — the rows reflect reported treatment data.

The field is undefined before fem2023 and when the wastewater routing questions are unanswered — deliberately not not_applicable, since the facility never said either way.

Three caveats:

• Two of the yield parameters stand in for conditions the FEM does not ask about: aerobic units are priced as well-managed (the FEM carries no overloaded-aeration signal) and facultative lagoons as shallow (no depth question). Both are the lower-emitting of their published alternatives; Appendix A quantifies the sensitivity.
• N₂O is not modelled. The IPCC method derives it from nitrogen load, which the FEM does not collect.
• The wastewater CO₂ does not appear in total_*_biogenic_emissions, which remains an energy-combustion total. Sum the rows on biogenic if you want a complete biogenic memo across all emission types.

Beyond wastewater, process emissions remain a gap. The air-emissions section gives presence/absence signals only, so chemical-process emissions must be filled from outside the FEM.

Biogenic emissions

Guidance: report biogenic CO₂ from biomass combustion separately from the scopes (memo item); biogenic CH₄/N₂O belong in scope 1 proper.

FEM: Emissions rows carry a biogenic boolean (from the attributed source’s category), and the assessment totals implement the memo-item structure: total_market_emissions / total_location_emissions exclude biogenic; total_market_biogenic_emissions / total_location_biogenic_emissions carry it separately.

Verdict: ⚠️ Aligned with caveats.

• The split is per-source, not per-gas: a biogenic fuel’s entire blended CO₂e (including its CH₄/N₂O component, which GHGP would keep in scope 1) lands in the biogenic memo. The scope-1 line you derive from FEM data is therefore slightly under the GHGP definition for facilities burning biomass, and the memo slightly over. The misallocated portion is the CH₄/N₂O share of a biogenic fuel’s blended factor, which is a low-single-digit fraction of its CO₂e — so this is very unlikely to be material at facility level. It is disclosed for completeness rather than because it needs correcting; if your programme reports biogenic memo items to a materiality threshold, check it against your own rather than assuming.

• The two partly-biomass waste fuels are split. fabricwaste and muniwaste carry blended emission factors covering both a biomass and a fossil fraction. Rather than leave the whole figure in the scopes, the records attribute the biomass share to a <source>_biogenic counterpart — fabricwaste_biogenic, muniwaste_biogenic — which is flagged biogenic and therefore lands in the memo totals, while the remainder stays in scope 1 (or scope 2 for muniwaste, which only reaches a facility inside a purchased steam mix, where the supplier did the burning).

  The split re-allocates and does not re-price: both halves carry the parent’s emission factor, so a facility’s total CO₂e is unchanged and only its distribution between the scope totals and the biogenic memo moves.

  Fractions used: muniwaste 50%, which its factor citation states outright. fabricwaste 33%, which is provisional — that factor blends a biological component modelled as biomass municipal waste with a synthetic component modelled as industrial waste, but the ratio was never recorded. Back-solving the published factor against the IPCC defaults those components come from puts the biological share between 0.27 and 0.34, and roughly a third is also the global natural-fibre share. Treat the fabricwaste biogenic/fossil split as indicative until the derivation is confirmed; the totals it feeds into are not affected either way.

• Biogenic totals cover fuel combustion only. Wastewater CO₂ is biogenic and on the records, but is not in these totals — sum the Emissions rows on biogenic for a complete memo.

Scope 2 — location-based method

Guidance: grid-average factors appropriate to the facility’s location; totals for all purchased electricity, steam, heat, cooling.

FEM:

• Electricity: per-country grid factors, IEA-derived, locked per cadence (updated with each release; the fem2025 factors are based on IEA 2022 data / provisional 2023 where available). Factors are licensed and not republished in the data — records carry computed emissions only.
• Steam: energy density from facility-reported temperature and pressure per steam source (defaults 200 °C / 1 unit-converted pressure), fuel mix reported by the facility, remainder modeled as coal at 80% boiler efficiency — a deliberately conservative unknown-source assumption.
• District heating: modeled as coal at 80% efficiency (fem2023+); volume-reported heat converted via inlet/outlet temperature delta.
• Chilled water: modeled as grid electricity at COP 3.0. Note that chilled-water rows carry energy_form: "thermal" even though they are attributed and priced as grid electricity. This is deliberate and not an inconsistency: energy_form describes how the facility consumes the energy, not what was burned to produce it. Chilled water is generated off-site and piped in, so like purchased steam and district heating it is a thermal delivery. The electricity appears in the attribution (attributed_source_key), which is where a consumer building a generation-side split should look. If you are building consumption-side electric/thermal splits or EECI-style intensity metrics, energy_form is the correct field and needs no adjustment.
• The facility’s own utility emission factor (Q1.5.1) is collected but deliberately not used from FEM 2024 on (unverifiable self-reports moved the program to uniform grid factors). Supplier- specific EF questions for chilled water, district heating, and purchased renewables are likewise collected-but-unused.

Verdict: ✅ Aligned for electricity (grid-average by location is exactly the location-based method), ⚠️ for steam/heat/cooling (single-assumption models; the coal-remainder and COP defaults are conservative and can materially overstate clean district systems — if you have verified supplier factors, the collected-but-unused questions tell you which facilities have them, and you can recompute from EnergyUse.value).

Scope 2 — market-based method

Guidance (Scope 2 Guidance): report market-based alongside location-based (dual reporting); use supplier/contract-specific factors where instruments meet the Quality Criteria; use the residual mix for untracked consumption where available, else the grid average; instruments must be retired, uniquely claimed, and plausibly matched in market and vintage.

FEM: the dual structure is native — every reported energy row generates both a scope_two_method: "market" and a "location" row, and all totals exist in both flavors. Market-method mechanics:

• EACs (Q3): facilities report instrument type (REC, GO, REGO, I-REC, TIGR, Green-e, EKO, other), retired MWh (explicitly retired-not-purchased, registered to the facility’s own legal entity and address), a certificate upload, and the generation-source mix. Covered volume is re-attributed to the certified sources; uncovered volume stays on the grid factor.
• Purchased renewables without EACs: count as renewable under the market method only if the facility attests it owns the attributes (PPA-ownership question); otherwise the full volume is re-attributed to grid electricity. Unknown-source renewables (renewunknown) always fall to the grid factor — the can’t-substantiate-it-don’t-credit-it rule.
• On-site generation: the share of certificates sold is re-priced at the grid factor (the attributes left the facility) and carries scope: "two"; the retained share stays zero-emission at scope: "one". This applies under both methods, per Table 6.1 and §6.4.1 — a sold certificate means the electricity is accounted for “as though it were purchased from the grid” in the location-based total as well as the market-based one. The two methods differ in what the grid share costs: the market side uses the residual mix where one is published, the location side always the grid average.
• Residual mix: per-country AIB residual-mix factors price market-method grid electricity from fem2025, with a fall-back to the location factor wherever none is published.

Verdict: ⚠️ Aligned with caveats — structurally one of the strongest areas, with two material caveats:

1. Quality Criteria are only partially evidenced. Retirement, ownership, registry family, and a certificate file are collected; vintage (generation year), market/grid boundary, and registry serial IDs are not. You can trust the arithmetic; you cannot audit Scope 2 Quality Criteria conformance from FEM data alone. If your program requires criteria-conformant instruments, request certificates (the upload exists) or treat EAC-adjusted figures as facility-attested.

2. Residual mix reaches Europe only, from fem2025. Untracked consumption is priced at the AIB European Residual Mix where one is published and at the grid average everywhere else — which the Scope 2 Guidance permits where no residual mix exists. Four boundaries to carry:

   • European coverage only. AIB has ~34 member countries; the rest of the world keeps the grid average, so for untracked consumption the two methods agree there.
   • fem2025 onward. AIB began publishing per-country CO₂ factors with its 2024 results; earlier workbooks carry the generation mix but no emission factors. fem2023 and fem2024 keep the grid average.
   • A one-year lag. AIB publishes a year’s mix in May of the following year, after the cadence is built, so fem2025 carries the 2024 mix.
   • Non-participating members fall back. A member that did not report shows a residual mix of zero at source; those countries keep the grid average rather than being credited as carbon-free. Austria and Switzerland are both in that position for 2024.

   Expect market-based emissions for European facilities without certificates to be materially higher than location-based: a residual mix excludes certificate-backed generation, so it exceeds the grid average by construction. For Germany’s 2024 mix the gap is roughly 120%.

Carbon offsets are collected (type, quantity, reference) and — correctly — never netted against scope 1 or 2 in any record or total.

Emission factor provenance

Guidance: use context-appropriate factors; document sources.

FEM: fem2020+ fuel EFs carry citations (IPCC Stationary Combustion defaults + AR6 GWPs; US EPA for ethanol; modeling notes for derived fuels like fabric waste, coal-water slurry, steam). Pre-fem2020 factors have empty citation fields. Grid factors are IEA-licensed (values not redistributable — consume the computed emissions, not the factors). Known factor-level modeling notes worth carrying into your disclosures: CNG/LNG factors exclude compression/liquefaction energy (scope 3 in GHGP terms); hydrogen is zero at point of use with upstream explicitly deferred to scope 3; biomass certification does not change the EF.

Verdict: ✅ for fem2020+, ⚠️ for older cadences (undocumented citations; factors were carried forward from earlier program years).

Activity data collection and quality control

Guidance (handbook): standardize collection; investigate >10% year-over-year changes; verify unit conversions; cross-check against reference sources.

FEM/HAM: in-form validations catch structural errors at entry time (mix percentages, implausible totals, unit constraints), but those never reach the records. On the records themselves there are exactly six per-assessment data-quality checks: assessment total energy implausibly high (>10¹⁰ MJ) or low (<38,574 MJ, including missing-entirely), an energy source answered “not tracked” (ghgp_completion_status: incomplete), negative reported values, purchased-renewables exactly equal to EAC MWh×3.6 (a common data-entry confusion), implausibly high LNG (>10⁹ MJ, the LNG/m³-vs-MJ confusion), and — from HAM 0.3 — refrigerant plausibility: fugitive scope 1 of at least 10,000 t CO₂e that also exceeds the assessment’s energy-related scope 1 + 2 (see the fugitive-emissions chapter). Cross-assessment rules (the section after this one in the methodology guide) add flags from a facility’s other cadences. pass_data_quality is all-or-nothing: one flag fails the whole assessment, and it is stamped on every record row.

The two flag fields answer different questions, and it is worth keeping them apart when you build screens:

• pass_data_quality — did the whole assessment pass? Carried identically on every record row, so it is the field to filter on.
• data_quality_flags — what went wrong, at the grain of the row you are reading. On Assessment this is every flag raised anywhere. On FacilityType it is the field-level flags raised by rows attributed to that facility type, which is what lets you tell which facility type a bad figure came from. Assessment-wide flags are not repeated onto facility-type rows, so a facility type can have an empty data_quality_flags while pass_data_quality is false — the failure belongs to the assessment as a whole, or to a sibling facility type.

Verdict: ⚠️ Aligned with caveats. No year-over-year discontinuity checks, no cross-facility benchmarks, no unit-conversion audits reach the records. The flags are a floor, not an assurance mechanism. Consumers doing serious QA should add their own screens (energy intensity vs production volume, YoY deltas, unit-magnitude checks) — aggregate FEM experience says unit confusion (kWh/MWh, LNG volume) is the dominant error class.

Two structural limits are worth planning around:

• Completeness is signalled on every cadence. The FEM asks, per energy source, “Does your facility track its energy use from this source?”. Where a facility says it uses a source and then says it does not track it, the assessment is reported incomplete on ghgp_completion_status, and a data-quality flag naming that source is raised — so the assessment fails pass_data_quality and drops out of any filtered analysis by default, without you having to know this check exists. Use the flags to see which sources are missing.

  A blank answer is not treated as a gap: only an explicit “I use this and don’t track it” counts, so the check never manufactures failures for facilities that left irrelevant rows alone.

  The question changes shape across cadences and the check follows it. From fem2023 each track question is paired with a “do you use this source?” column, so a gap is unambiguous. In fem2017-fem2022 there is no such pairing and the answer list includes unknown, which counts as a gap in its own right — a facility saying it does not know whether it tracks a source is exactly the uncertainty this check exists to surface. Because every cadence carries the check, pass_data_quality means the same thing across the whole record set.

• There are no year-over-year checks, because each assessment is evaluated in isolation — the model has no access to a facility’s other assessments at evaluation time. This is why the handbook’s “investigate changes above 10%” screen has no counterpart here, and why lumpy refrigerant top-ups cannot be smoothed. Cross-year screens remain consumer-side work.

Verification

Guidance: third-party verification recommended; document assurance.

FEM: a verification program exists (accredited verifiers, per-question verification tags, verifier-corrected data). When an assessment is verifier-posted, the records are computed from verifier-corrected data.

Verdict: ⚠️ Aligned with caveats. Posting status is on the Assessment record — posted_by_verifier, posted_by_facility and assessment_status — on every surface, so you can distinguish a self-assessed FEM from a verifier-corrected one without joining platform metadata. posted_by_verifier is the assurance signal to filter on: when it is true, the records were computed from verifier-corrected answers.

Read an undefined value as unknown, not as unverified. The fields are populated by the host that supplies the assessment, so they are present on the analytics tables and the platform API; the raw evaluate endpoint takes a caller-supplied context verbatim and will only carry them if the caller does. (The analytics tables additionally carry status, facility_posted and verifier_posted as platform columns; those predate these fields and say the same thing.)

What does not exist anywhere, on any surface, is assurance detail: verifier identity, verification body, accreditation, and assurance level are not present in the data the model receives, so they cannot be derived. Exposing them requires a change to the upstream platform contract and is being pursued separately. If your programme needs to evidence who verified an assessment and to what standard, that remains an out-of-band request rather than something the records can answer.

Base year and recalculation

Guidance: set a base year; recalculate it for structural changes, methodology changes, and significant errors; define a significance threshold.

FEM: none of this exists at the record level. Facilities declare an energy baseline year for target-setting questions, but it is a self-reported performance-target input, not a recalculable GHGP base year. Posted FEMs are immutable by design; methodology does evolve between cadences (and the records deliberately re-derive historical cadences under the latest fuel EFs).

Verdict: ❌ Gap — by design. Base-year policy is the consumer’s: you own the base year, the recalculation triggers, and the significance threshold. What HAM gives you is the ability to recompute any historical cadence consistently (the whole model is re-evaluated from source data on every release), which is the right substrate for consumer-side restatement.

Uncertainty

Guidance: assess and document measurement/estimation uncertainty.

FEM: no quantified uncertainty. The structural signal is provenance on every EnergyUse/Emissions row: "reported" (attribution follows facility-specific information) vs "modeled" (an assumption was applied — unknown steam source → coal, chilled water → grid electricity, unowned renewable attributes → grid, …).

Verdict: ⚠️. Use the modeled-share of a facility’s emissions as a first-order uncertainty proxy; the conservative defaults mean modeled rows skew high.

Working with the records: practical rules

Hard-won rules for anyone assembling an inventory from the tables:

1. Always filter to one scope_two_method. Every reported row exists twice (market + location); summing across both double-counts everything, including scope 1.
2. Scope totals are on Assessment — use total_<method>_scope_one_emissions / ..._scope_two_emissions rather than deriving them. They exclude biogenic, matching the other total_*_emissions fields. Only FacilityType needs a hand-rolled split, from the scope field on the Emissions rows.
3. Do not treat emissions as a complete mirror of energy_use. Zero-emission rows (renewables, zero-MJ entries) are dropped from emissions; join back to energy_use for energy accounting.
4. ReportedEnergyUse is not for GHGP reporting — it is raw, pre-attribution, pre-modeling (steam un-grossed, chilled water un-divided). Use it for data-quality work only.
5. Domestic/vehicle breakouts are exact-coverage sums: facilities that do not track domestic separately report combined production+domestic rows, which the *_domestic_* fields do not include (grand totals always include everything). All domestic/vehicle breakouts are undefined before fem2023. domestic_energy_status and vehicle_energy_status disambiguate an empty breakout: not_applicable means no such rows are expected (combined reporting / no company vehicles), untracked means the data is missing, tracked means the breakout is complete.
6. FacilityType allocation is approximate: domestic and vehicle energy is split evenly across the assessment’s facility types, and per-type rows only materialize when the facility can report per-type (fem2023+) or has exactly one type (earlier).
7. Historic-cadence records ≠ what the facility saw. The records apply the latest fuel EFs and current methodology to all cadences; the in-product FEM numbers used cadence-vintage factors. This is deliberate (comparability) — cite the release, not the FEM year, as your factor vintage.

Summary

Aligned — consume as-is:

• Facility-level activity data collection for stationary and mobile combustion (purchase-records method, systematic unit conversion)
• Location-based scope 2 for purchased electricity (IEA grid averages by country and cadence)
• Dual market/location reporting structure and totals
• Scope 1 / scope 2 rollups on Assessment, in both methods
• Refrigerant fugitives as scope 1 Emissions rows (fem2023+)
• On-site wastewater CH₄ (scope 1) and CO₂ (biogenic) as Emissions rows (fem2023+)
• Tracking-status signals on Assessment (fem2023+): wastewater_emissions_status, domestic_energy_status, vehicle_energy_status and refrigerants_status, separating not-applicable from untracked data — with refrigerants_status additionally flagging refills the GWP table cannot price as unmodeled
• Posting status on Assessment, so verifier-corrected records are distinguishable from self-assessed ones
• Per-source completeness signalling on every cadence, failing pass_data_quality so incomplete assessments self-exclude
• Offsets excluded from scopes; unknown-provenance renewables not credited
• Biogenic combustion CO₂e as a separate memo total, including the biomass fraction of the partly-biomass waste fuels
• Modeled-vs-reported provenance labeling

Aligned with caveats — disclose or correct:

• Blended CO₂e factors; no per-gas breakdown
• Latest-EF re-derivation of historic cadences
• Steam/district-heat/chilled-water single-assumption models (coal-remainder, 80% efficiency, COP 3.0)
• EAC Quality Criteria only partially evidenced (no vintage, market boundary, or serial IDs)
• Residual mix covers European AIB members from fem2025 only; elsewhere and in earlier cadences, market-based grid electricity equals location-based
• Sparse data-quality checks beyond completeness; an absent source with no tracking answer still means “unknown”, not “zero”
• No year-over-year checks — assessments are evaluated in isolation
• Assurance detail (verifier identity, body, accreditation, level) is absent everywhere; only posting status is available
• Biogenic split is per-source, not per-gas (immaterial)
• fabricwaste’s biogenic fraction (33%) is provisional pending confirmation of how its blended factor was derived

Gaps — fill from outside the FEM:

• Consolidation approach / ownership shares (single-site 100% operational control is assumed)
• Reporting-period fields (calendar-year is convention, not data)
• Process emissions other than wastewater (qualitative presence only)
• Wastewater N₂O (no nitrogen data collected)
• Per-gas (CH₄/N₂O) reporting
• Base year, recalculation policy, significance threshold
• Quantified uncertainty
• Electricity T&D losses and other scope 3 adjacencies (explicitly out of scope)

Known implementation issues (as of August 2026)

None outstanding. Current-behavior notes a consumer should account for appear here as they are found; the deviations previously listed — sold-certificate scope labelling and the location-method treatment of certificate sales — are both resolved and described in place above.

Cadence applicability

The alignment above describes the current methodology (fem2024 through fem2027, which are calculation-identical). Earlier cadences differ materially:

For consumers, the practical line is: full market-based accounting and domestic/vehicle segmentation are meaningful from fem2023 onward.

That does not make the earlier market-based rows empty, and it is worth being precise about what they do carry. Because geothermal, hydro, and wind are modeled as off-site generation with retired certificates (see “Why the older cadences differ” above), those three sources are zero-rated under the market method and priced at the country grid factor under the location method in fem2017–fem2022. That divergence is real and is the pre-fem2023 market/location signal — for a facility reporting meaningful geothermal, hydro, or wind volume, the two methods give genuinely different answers and the market figure reflects the modeled certificate claim.

What the earlier cadences lack is everything else the market method normally involves: no EAC reporting, no PPA-ownership attestation, no certificate-sale adjustment on on-site generation, and no residual mix. Purchased renewables (purchrenew, fem2020+) look like they should diverge but do not — they re-attribute to grid electricity under both methods, so they contribute nothing to the market/location difference.

So: treat pre-fem2023 market-based totals as meaningful only to the extent the facility reported geothermal, hydro, or wind, and as equivalent to location-based otherwise. And carry the modeling assumption with them — the market figure is only as good as the “these renewables were off-site” premise, which the data cannot confirm.

Appendix A: the wastewater treatment calculation

This appendix documents the calculation behind the wastewater_co2 and wastewater_ch4 rows — what is computed, where every number comes from, and which assumptions a consumer inherits by using them.

Source methodology

The calculation implements the wastewater-treatment method of:

RTI International for the US EPA, Greenhouse Gas Emissions Estimation Methodologies for Biogenic Emissions from Selected Source Categories: Solid Waste Disposal, Wastewater Treatment, Ethanol Fermentation (draft, December 14, 2010), PDF — Section 3.2.1, Equations 3-1 and 3-2, with the Table 3-1 default parameters.

The EPA method prices the CO₂ and CH₄ from a biological treatment unit on the oxygen demand it removes: influent flow × influent oxygen demand × removal efficiency, times per-gas conversion factors. The FEM collects exactly the inputs that product needs — annual wastewater volume per stream (m³) and BOD5 concentrations before and after treatment (mg/L) — so the method transfers without gap-filling on the activity side.

The calculation

Because 1 mg/L = 1 g/m³, the organic load removed is a mass:

BOD removed (g) = volume (m³) × (BOD5_before − BOD5_after) (mg/L)

Each gram of removed BOD is converted at (Equations 3-1/3-2):

CO₂ (g/g) = 44/32 × (1 − MCF × BG) × (1 − λ)
CH₄ (g/g) = 16/32 × (MCF × BG) × (1 − λ)

• 44/32 and 16/32 are stoichiometric ceilings: the mass of CO₂ (or CH₄) obtainable per unit of oxygen demand.
• MCF (methane correction factor) is the fraction of the removed oxygen demand converted under anaerobic conditions.
• BG = 0.65 is the document’s default fraction of anaerobically converted carbon leaving the biogas as CH₄ (the rest leaves as CO₂).
• λ (biomass yield) is the fraction of removed carbon incorporated into new biomass — sludge — rather than emitted as gas.

With the Table 3-1 defaults, per FEM process choice:

The CH₄ mass is converted to CO₂e at 27.2 kg CO₂e/kg CH₄ — the IPCC AR6 100-year GWP for non-fossil methane, the correct variant here because the carbon is biogenic, and the same AR6 vintage the fuel emission factors use. The CO₂ row carries its mass directly (GWP 1) and is flagged biogenic: true.

Worked example

1,000 m³/yr treated anaerobically from 300 mg/L down to 30 mg/L removes 1,000 × 270 = 270,000 g of BOD. The records carry:

• wastewater_co2: 270,000 × 0.594 / 1000 = 160.38 kg CO₂ (biogenic memo)
• wastewater_ch4: 270,000 × 0.234 / 1000 = 63.18 kg CH₄ × 27.2 = 1,718.50 kg CO₂e (scope 1)

How the FEM’s two reporting shapes are handled

The FEM asks how BOD5 is monitored, and the answer routes the facility to one of two mutually exclusive question sets:

1. One before/after pair per stream (combined, domestic, and/or industrial wastewater). Each stream is priced on its own volume and its own BOD drop — streams are never pooled, because summing concentrations across streams of different volumes is meaningless. A stream reporting several process kinds against its single before/after pair has its removed load split equally between them, each share priced at its own yield: one pair carries no information about how much each stage removed, and an equal split is the assumption-minimal reading. (The split matters: a worst-case-wins reading would price the whole load anaerobically and roughly double the methane of an aerobic/anaerobic train, while ignoring the anaerobic stage would eliminate it.)
2. One before/after pair per treatment sub-process (up to ten stages). Each stage is a unit in series: its own removed load, priced by its own declared process kind, against the annual volume of the wastewater stream it declares. Per-stage removals telescope to the whole-train removal, so the stage rows sum without double counting — and unlike the single-pair shape, this one needs no apportionment assumption at all.

Assumptions a consumer inherits

Two Table 3-1 parameters answer questions the FEM does not ask, and both default to the lower-emitting published row:

• Aerobic units are priced as well-managed (MCF 0, λ 0.65). The overloaded row (MCF 0.3, λ 0.45) would raise the aerobic CO₂ yield by about a quarter and add a methane term worth over twice the CO₂ on a CO₂e basis. The FEM collects no aeration-loading signal to distinguish the two.
• Facultative lagoons are priced as shallow (MCF 0.2). The deep (≥ 2 m) row carries MCF 0.8, which on a CO₂e basis is roughly 2.6× the shallow total once the methane is weighted. The FEM has no lagoon-depth question.

If your program needs conservative (high-side) wastewater figures, treat facilities reporting facultative treatment as the sensitive population; the underlying volumes and BOD values are on the assessment should you need to re-price them.

What is deliberately not modelled

• Sludge digestion (source document Equations 3-4 through 3-7). The λ term already keeps sludge-routed carbon out of the treatment rows; pricing its later digestion needs sludge flow, solids concentration, and digester-type data the FEM does not collect.
• Biogas capture and combustion (Section 3.2.2). The FEM does not ask whether methane is captured or flared, so all computed CH₄ is assumed vented. For a facility that captures and burns its biogas, the methane row overstates — conservatively — what actually reached the atmosphere.
• N₂O (Section 3.3). The method derives it from influent nitrogen (TKN), which the FEM does not collect; no defensible default exists.

Guards

Rows materialize only when they carry a positive value: a stream or stage with no tracked volume, no BOD answers, or a post-treatment reading at or above its pre-treatment reading produces no row — never a zero-valued or negative one. This is the same absent-not-asserted convention the refrigerant rows use.

Reading absence: wastewater_emissions_status

Because of the no-zero-rows convention, “no wastewater rows” has two very different readings, and the Assessment field wastewater_emissions_status tells them apart:

Applicability mirrors the FEM’s own gate on the BOD5-monitoring questions — the same conditions that decide whether the facility is ever asked for treatment data: an on-site, on-site-and-offsite, zero-liquid-discharge, or septic-combination discharge route, plus “secondary treatment” among the facility’s water pretreatment types. The field is undefined for cadences without the treatment tables (pre-fem2023) and when the routing questions are unanswered — deliberately not reported as not_applicable, since the facility never said either way.

Note that tracked asserts the inputs exist, not that emissions were positive: a facility whose effluent BOD5 equals its influent produces no rows while still reporting tracked, which is the honest reading — it measured, and the measurement showed nothing removed.