EaaA Brief v0.2

Draft v0.2 — published for review

This document outlines the principles, changes, and structure of EaaA — Everything as an Activity — a protocol and framework built on top of the existing Life Cycle Assessment standard, LCA. EaaA does not discard LCA; it builds on what works, corrects what does not, and establishes independent governance where the current standard has been captured by commercial interests.

v0.2 follows a first implementation of the framework against real datasets, which settled several of v0.1's open questions; their answers are written in. Seven founding absolutes (§2.1) sit above the principles.

The framework and the platform are two different things. This document describes the framework. The Economics. platform is built on it to validate it, and nothing here describes that platform. EaaA is a draft: developed in the open, validated against published datasets, and published here for independent review.

1. Versioning

EaaA uses a clear versioning scheme to signal continuity with existing practice rather than replacement. The name EaaA survives only as this version label: v2.0 of the LCA line is EaaA.

2. Founding Principles

EaaA is governed by the following principles, which take precedence over implementation convenience:

✦ PRINCIPLE

Keep what works. Change what doesn't. LCA professionals already know the general terms — activities, flows, inputs and outputs. That knowledge is preserved; only exchange is retired as a term (§5).

✦ PRINCIPLE

No single commercial entity may serve as the normative reference for any foundational category in EaaA. All normative references must be independently governed and openly accessible.

✦ PRINCIPLE

Factory or forest. Every activity — whether industrial or natural — is modelled by the same rules, declaring its exchanges into physical compartments. There is no privileged class of process.

✦ PRINCIPLE

Geographic boundaries, flow nomenclature, and impact categories are defined by independently governed, peer-reviewed sources — not database vendors.

2.1 Founding Absolutes

Absolutes sit above principles. Every other rule must be consistent with them, and an implementation that violates one is wrong however convenient it is. The first implementation showed how easily an EaaA application drifts back into LCA — a preloaded dictionary standing in as the end of a chain, a declared functional unit, money as a special case — and these are what stop it.

✦ ABSOLUTE A-001

Everything is an activity. Anything that has to be created, produced, grown, mined, refined, or transported is an activity. The graph is activities all the way down to the points where it touches nature. A thing produced inside the technosphere is never a terminal node — it is the output of the activity that produced it, and that activity carries its own upstream chain.

✦ ABSOLUTE A-002

No cut-offs. Ignorance is shown as low accuracy, never hidden. EaaA never truncates a chain to manufacture a usable number. Where data is missing, incomplete, or only generically known, the model represents that as a low accuracy score on the affected flows rather than cutting it off. A low score is a correct, honest output, not a failure. Early users are expected to score poorly; that is the design working.

✦ ABSOLUTE A-003

Money is a resource; currency is its unit. A monetary magnitude is a money flow — a resource of the financial type — denominated in a currency unit (ISO 4217). Money is treated like carbon, water or topsoil: one flow among the environmental flows, scored and superseded on the same gradient. Money is never classed as "economics": economics is the management of all limited resources, never one of them.

✦ ABSOLUTE A-004

An activity has no functional unit; its outputs carry the magnitude. An activity is not declared "per 1 kg" or "per 1 M.EUR" of anything. What it produces is stated as output flows, each with its own resource, unit and value, and there is no privileged reference output the rest are allocated against. Comparison happens by asking the graph a question, not by declaring one in advance.

✦ ABSOLUTE A-005

Nothing is free text. Every thing a flow names — its resource, its unit, its geography, its sphere — and every category an impact is scored into is a record in a governed reference vocabulary, anchored to an independent authority: PubChem and the IMA mineral list for substances and minerals, the UN CPC and SIEC for products and energy carriers, ISIC and ISCO for sectors and occupations, ESA 2010 for money, UCUM and ISO 4217 for units, geoBoundaries, ISO 3166 and UN M49 for places, peer-reviewed methods for characterisation factors. A name is a label; the link is the identity. A source's label that cannot be linked is an open question — a dispute, resolved by a person and recorded as a cross-reference — never a string carried into the graph.

✦ ABSOLUTE A-006

Every reference is versioned, and every use of one is pinned. A resource, an impact category and a characterisation factor carry the version of the authority that published them; a flow names the resource version it means; a link names the flow versions it joins, and through them the producing activity version; a stored result records the engine that produced it. A calculation is reproducible because nothing it read can change underneath it — a newer version is a new row beside the old, never an overwrite.

✦ ABSOLUTE A-007

EaaA asserts nothing of its own. Every row records the source it came from and the source's own identifier, and a record's identity is that source key: a rebuild from source produces the same record, and a correction at the source is a new record, not an edit. Where the framework must hold a list no authority publishes — the physical spheres, the non-administrative geographies, the curated atmosphere and hydrosphere compositions — it publishes that list under the framework's own organisation as a reviewed list with the same provenance as any other, and never under a data provider's.

A-001 is first because it is the single assumption that, when dropped, turns EaaA back into LCA: the moment a produced thing may be a terminal referent, its cradle-to-gate burden vanishes. A-002 is the reason A-001 matters — refusing terminal produced things is what forces the missing burden to show up as a low score instead of disappearing. A-003 makes the platform's founding claim, environmental cost as visible as monetary price, structural rather than rhetorical. A-004 follows from the other three: the graph is complete, nothing is cut off, and money is a flow, so an activity's outputs are simply its outputs. A-005 is what makes A-001 checkable — a produced thing can be the output of an activity only if the thing itself is an identity both sides join on. A-006 is what makes a score under A-002 mean the same thing tomorrow. A-007 is the governance principle of §3.1 made structural: separation of the data role from the standards role, enforced by provenance rather than by policy.

3. What Changes from LCA to EaaA

3.1 Governance and Standards Capture

The current LCA ecosystem suffers from what may be called standards capture: the dominant commercial database (Ecoinvent) makes internal engineering and classification decisions that become de facto industry norms through adoption at scale. This occurs without peer review, open comment periods, or democratic governance.

3.2 Geographies

LCA geographies are largely defined by Ecoinvent's internal needs, with electricity markets being a prominent example of unnecessary complexity. EaaA anchors all geographic definitions to geoBoundaries (geoboundaries.org), a peer-reviewed, openly licensed source maintained independently of any LCA agenda.

• The lowest applicable ADM level is the geographic floor for provider declarations.
• If no provider exists at that level, municipal supply is assumed as the default.
• An activity such as 'generate electricity' is created by a provider and declares the geography where it applies.
• Electricity market geographies are simplified — grids that cross administrative boundaries are modelled at the nearest enclosing ADM level, with cross-boundary declarations where required.

✦ PRINCIPLE

ADM2 is the floor. A provider declares at the lowest administrative level that applies, and ADM2 is the lowest the framework resolves; anything finer is a sub-ADM region (below). Where geoBoundaries' ADM2 does not carry a jurisdiction's own county coding, a second authority supplies it: for the United States, the Census Bureau's FIPS county codes, joined to geoBoundaries' county shapes as synonyms — the same mechanism by which ISO 3166 and CLDR names already sit beside a country — so that county-keyed data — TRACI's eutrophication factors are keyed on FIPS — resolves to a geography the accuracy ladder can climb from.

3.2.1 Non-Administrative Geography Extension

geoBoundaries covers all recognised administrative territories. It does not cover international waters, international airspace, Antarctica, or other areas outside national jurisdiction. EaaA defines a small named extension for these geographies, governed under the same independent principles as the ADM standard.

The extension is not a fallback of last resort — it carries real geographic information that affects impact assessment. SOx emitted in the Arctic has different ecological significance than SOx emitted in a tropical shipping lane. Declaring both as 'worldwide' loses that distinction entirely.

Named non-administrative geographies in EaaA:

• Oceans: North Atlantic, South Atlantic, North Pacific, South Pacific, Indian Ocean, Southern Ocean, Arctic Ocean
• Airspace: ICAO NAT (North Atlantic), ICAO EUR, ICAO PAC, ICAO MID, ICAO AFI, ICAO SAM, ICAO CAR — relevant for aviation emissions
• Polar: Arctic region (above 66°N), Antarctic region (below 60°S)
• Global: permitted only for CO2, CH4, N2O, and other well-mixed greenhouse gases where geographic declaration carries no additional impact information

3.3 Sphere Hierarchy

LCA uses a technosphere/biosphere split to distinguish human-made from natural processes. This creates persistent edge cases (e.g. managed forests, aquaculture) and does not reflect physical reality. EaaA replaces this with a physical compartment hierarchy based on the layers of the Earth and atmosphere.

The key insight: the biosphere is not a compartment. It is a class of actors (living systems) that exchange flows between physical compartments. A tree is an activity, not a destination.

3.4 Flow Nomenclature

LCA flow names (e.g. 'Carbon, fossil', 'Air, non-urban air or from high stacks') encode substance, type, compartment, and spatial qualifiers into a single free-text string. This approach is fragile, query-hostile, and accumulates inconsistencies over time as new qualifiers are appended ad hoc.

EaaA separates these into typed, structured fields. The name becomes a human-readable label only; the identity of a flow is defined by its attributes.

EaaA calls the what of a flow a resource. A resource is a substance where it is a chemical entity, and otherwise a mineral, an energy, a phenomenon, land, a biological entity, a product, a socioeconomic quantity or money (§3.4.1.1). EaaA says resource where any kind is meant and substance only for the chemical kind.

3.4.1 Resource Identity is Contextual, Not Purely Chemical

A resource in EaaA is the identity that distinguishes one flow from another; a substance is a resource of the chemical kind. Substance identity is chemically anchored where possible — PubChem is the normative reference, with CAS recorded as a cross-reference for interoperability with regulatory and legacy systems. But identity in EaaA is finer-grained than chemistry alone.

"Water" the chemical species (H₂O, PubChem CID 962) is one substance entry. "Water, treated municipal effluent" and "Water, deionised industrial process water" are separate substance entries — they share the same chemistry but have different inventory and impact behaviour. They are different substances in EaaA even though they are the same compound in chemistry.

✦ PRINCIPLE

Substance identity in EaaA is chemically anchored where possible but may be subdivided where context affects inventory or impact behaviour. The substance vocabulary is the controlled distinction layer; flows reference substances directly without further qualification.

The test for splitting a substance:

• Different characterisation behaviour → split. If treated effluent and untreated wastewater would carry different impact factors in any EaaA-recognised method, they are separate substances.
• Different origin only, same composition and behaviour → don't split. Water from the Rhine and water from the Thames share the same substance entry; geography on the flow handles location.
• Different treatment state, same composition → judgement call, lean toward split if downstream impact differs.

A substance entry that subdivides a PubChem-anchored chemical retains the chemistry fields (CID, formula) for cross-reference, but its identity in EaaA is finer-grained than chemistry alone.

3.4.1.1 Resource Structure

Resource identity in EaaA is captured by structured attributes rather than a free-text name:

• type — substance, mineral, energy, phenomenon, land, biological, socioeconomic, product, or financial. Required. The broad category of thing being declared.
• identifier — the key the resource's authority gives it: a PubChem CID for a substance, an IMA name for a mineral, a CPC subclass or SIEC code for a product, an ISIC class or ISCO skill level for a socioeconomic quantity, an ESA 2010 transaction for money. Required.
• state — a single governed term capturing physical state or contextual subdivision: fossil, biogenic, depleted, treated_effluent, deionised, … Optional. Part of the identity, so two rows may share an identifier and differ only here.
• formula and inchikey — the chemistry, where there is one, kept for cross-reference and exact matching; never the identity.
• source — the authority that publishes the resource. A resource published by an external authority is standardised by that fact; one no external authority publishes is published by the framework itself (A-007).
• synonyms — every name, CAS number and abbreviation any vocabulary gives the resource, kept beside it and never deduplicated across vocabularies: a name two authorities agree on outranks one a single authority spells that way.

Together these resolve the cases that LCA handled by appending qualifiers to a free-text name. Hydrogen gas and hydrogen liquid are separate resources (same chemistry, different state). Depleted uranium is a separate resource (same chemistry, state = depleted). Treated effluent is a separate resource (same chemistry, state = treated_effluent). Noise is a resource with type = phenomenon and no chemistry fields; ionising radiation is not a phenomenon but a radionuclide substance.

3.4.2 The Unique-Resource Rule

An activity declares at most one flow per substance per direction. An activity that produces 100 litres of treated effluent declares one output flow of 100 litres — not one output flow per downstream consumer. Consumption is tracked separately from declaration.

✦ PRINCIPLE

A flow describes a substance crossing an activity boundary. The producing activity declares what it produces; consumption is tracked through links between flows, not through additional flow rows on the producer.

This rule rests on §3.4.1: same-substance-different-kind cases are resolved by substance subdivision, not by multiple flow rows of the same substance. The wastewater plant's "treated effluent" and "process water" are different substances, so they are different flow rows with no rule violation.

Consumer linkage is tracked separately from the producer's declaration. A producer's output flow declares the total quantity it produces; each downstream consumer declares its own input flow for the same substance, and the two are linked. Many-to-many between activities is mediated by these links, not by duplicate flow rows on the producer. The unique-resource rule is preserved cleanly.

4. What Stays the Same

EaaA retains the core conceptual architecture of LCA. The following terms and concepts are preserved with the same meaning:

• Activities (unit processes)
• Flows (elementary and intermediate)
• Functional unit
• System boundary
• Life cycle inventory (LCI)
• Life cycle impact assessment (LCIA) — method structure preserved; the normative method per category is to be established (§6.3)

The goal is that an LCA practitioner reads EaaA and recognises the landscape, even where specific elements have changed.

4.1 EaaA is Attributional by Design

EaaA is an attributional framework. This is not a limitation — it is a deliberate architectural choice that follows directly from the activity graph model.

✦ PRINCIPLE

EaaA describes what is declared to happen. The activity graph contains real declared flows between real declared activities. Consequential questions — what does this activity displace? what would have happened otherwise? — are study-level interpretations performed by comparing two different declared systems. They are analysis on top of EaaA, not a feature within it.

This resolves the recycling and end-of-life problem that causes so much controversy in LCA. There is no need for avoided burden calculations, system expansion assumptions, or allocation choices around recycled content. The model simply describes what happens:

• Waste plastic flows to a collection activity → to a transport activity → to a processing activity → outputs recovered pellets
• If those pellets are used by a downstream activity, they flow there as an input — carrying the full upstream burden of collection, transport, and processing
• If those pellets are not used, they route to a landfill output activity in whatever geography they end up in
• The model contains what happened. What it does not contain did not happen in this declared system

A practitioner who wants to ask 'what is the benefit of recycling this plastic versus landfilling it?' builds two EaaA models — one with the recycling chain, one without — and compares them. That comparison is consequential analysis. It is performed by the practitioner on top of EaaA models, not by the framework itself.

5. Inputs and Outputs (Replacing Exchanges and Flows)

LCA uses the terms 'exchange' and 'flow' to describe what moves between activities and the environment. While 'exchange' carries useful connotations in academic contexts, it introduces unnecessary abstraction for practitioners outside that tradition. EaaA replaces this with plain language:

5.1 Intermediate Flows

When one activity's output becomes another activity's input, the flow does not enter a physical compartment. Physical compartments only apply when a flow leaves the activity chain entirely and enters the environment.

Example: Pump Water (A) outputs 1l of water. Bake Bread (B) inputs 0.4l of water from A. The 0.4l is an intermediate flow with no compartment. The remaining 0.6l is not a loss — it is unused capacity from activity A, available to other activities.

✦ PRINCIPLE

Intermediate flows between activities do not belong to a physical compartment. A flow only enters the compartment hierarchy when it leaves the activity chain entirely.

5.2 The Activity Graph — Inputs, Outputs, and References

Activities in EaaA are nodes in a directed graph. They can reference other activities as both inputs and outputs — there is no structural distinction between consuming an activity and producing into one. Both are declared references with a direction.

✦ PRINCIPLE

An activity may declare any other activity as an input or an output. The framework does not require activities to be owned, governed, or created by the same database. A reference is sufficient — just as a software application declares a dependency without owning the dependency.

This means complex or bundled outputs do not need to be decomposed inside the producing activity. A wheat farm does not need to declare dozens of trace soil emissions directly. It declares an output reference to a wheat compost activity — which itself handles the elementary flow declarations to physical compartments. The farming activity stays clean. The compost activity handles the complexity.

Different farming practices, geographies, or seasons can reference different compost activities. The wheat farming activity just points to whichever is appropriate. If a better compost activity becomes available, the reference is updated — the farming activity itself is unchanged.

The complete EaaA model is therefore: activities exchanging flows, with physical compartments at the edges. Every flow either moves between two activities (intermediate) or connects an activity to a physical compartment (elementary). Nothing else exists in the model.

5.2.1 A Resource Labels a Flow; It Never Ends the Chain

The first implementation drifted exactly where A-001 predicts. A flow referenced a resource, a resource was an entry in a preloaded dictionary, and so a produced thing — gold, flour, a manufactured chemical — became a terminal referent: the chain stopped at the dictionary entry and the cost of producing it never entered the calculation. That is the LCA "background dataset as opaque endpoint" behaviour EaaA exists to remove.

✦ PRINCIPLE

A resource identity labels a flow for validation and matching; it is never the endpoint of a chain. Every flow terminates at exactly one of two places: a physical sphere (elementary) or another activity (intermediate). A flow that consumes a produced thing must point at the producing activity, not at a standalone identity. The same identity — gold, CO₂, water — may be drawn from nature as an elementary flow or produced by an activity as an intermediate one; consumers join to whichever supplies them, through the shared identity.

An intermediate input that names a produced resource and has no producing activity behind it is an honest-but-incomplete state: permitted, scored low, and surfaced as an improvement target (A-002). It is never refused and never silently complete.

5.3 Leaf Activities — An Emergent Property

A leaf activity is not a declared attribute in EaaA — it is an emergent property of the activity graph. An activity is a leaf when it has no input flows from other activities. This is discovered by traversing the graph, not declared upfront.

✦ PRINCIPLE

Leaf is a graph property, not a flag. An activity with no upstream activity inputs is a leaf at that point in the model. This may change as the model grows — a previously terminal activity may gain upstream inputs when new datasets are added.

This distinction matters for a key reason. Consider solar energy:

• Solar radiation arriving at a wheat field is a terminal input — there is no meaningful upstream activity that produced the sun's energy in an LCA sense. It is declared as an elementary input flow drawn from the atmosphere compartment, referenced to an independent survey source (e.g. NASA POWER) scoped to the activity's geography.
• The solar panel activity, however, is not a leaf. It has inputs: panel manufacturing, land use, inverter production, installation. It has outputs: electricity, waste heat, end-of-life panel flows. It is a full activity with a complete upstream chain.

The concept of a leaf is still useful — it identifies the terminal nodes in any given model, and those nodes are typically elementary flows from the environment rather than manufactured activities. But it does not need to be stored as a property on the activity itself.

✦ PRINCIPLE

Input leaves are rare. A terminal input — one with no upstream activity — is legitimate only when it is a genuine elementary draw from a physical sphere: ore in the ground, water from a river, gold from a stream, CO₂ from air, solar radiation. Every other input is sourced from a producing activity. The legitimate input-leaf set is small and largely enumerable; an input leaf outside it is the exception that draws scrutiny, not the default.

Output leaves are not a concern: an elementary output to a sphere is always terminal and always complete. Leaf remains a derived property of the graph, never a stored flag — but the rarity of input leaves is a stated expectation the model is checked against.

5.4 Scaling — The Output Flow as the Atom

In LCA, datasets are sized to a study functional unit. EaaA has no functional unit (A-004). An activity is declared at the scale its flows describe: its output flows state what it produces and how much, its input flows state what that takes, and the activity is linear at that scale. Scaling happens at the point of use. When activity B draws 40% of one of activity A's output flows, it inherits 40% of every input and output of A.

✦ PRINCIPLE

Every activity in EaaA is declared at a scale at which its inputs and outputs are linear. Scaling behaviour is handled by the scale the flows are stated at, not by the framework. There are no non-linear activities in EaaA — only activities whose flows have not yet been stated at the right scale.

Non-linear real-world behaviour (e.g. exponential CO2 costs at high pumping volumes) is handled by stating the activity at the appropriate granularity: one activity stated per 10 ml for low-volume behaviour, another per 50 l for high-volume behaviour. These are separate activities producing the same resource at different scales.

5.5 Time Factors — Degradation, Amortisation, and Run-In

Some activities change efficiency over time. EaaA defines a single mechanism for all time-dependent behaviour: the time factor. A time factor is an optional declared rate of change on an activity's output ratio, with a direction, a starting value, and an optional settling point.

✦ PRINCIPLE

The time factor is the only permitted form of non-linearity in EaaA. It must be explicitly declared on the activity. An activity without a declared time factor is assumed to be stable at its declared unit efficiency across its lifetime.

Three real-world patterns are covered by a single time factor declaration:

• Degradation — efficiency declines over time toward a lower baseline. Example: turbine wear and silt accumulation in a hydroelectric dam reduce electricity output per unit of water flow.
• Amortisation curve — output ratio starts above baseline (carrying construction or setup burden) and declines toward baseline as the initial cost is recovered. Example: a new dam in its first years of operation carries the environmental cost of construction; by year 20 that burden is fully amortised.
• Run-in — output ratio starts below baseline and improves as the system optimises. Example: a new industrial process running below designed efficiency during commissioning.

A time factor declaration requires: a starting value (multiplier on the baseline output ratio), a rate of change (per day, per year, or per unit of output), a direction (improving or degrading), and an optional settling point (the value at which change stops). Once the settling point is reached, the activity operates at its baseline unit efficiency indefinitely.

A time factor that degrades to zero is the standard pattern for post-closure and post-use temporal outputs. The producing activity is complete and closed. The output activity it declared continues running, its emissions declining along the declared curve, until the time factor reaches its settling point of zero.

Example — acid mine drainage: a copper mine declares an output reference to an acid drainage activity at closure. That activity outputs to Hydrosphere at a peak rate, with a time factor degrading to zero over 50 years. The mine is closed; the drainage activity runs on independently.

Example — geological CO2 storage: a CCS injection activity declares an output to a geological storage activity. That activity holds CO2 in Geosphere → Deep subsurface and declares a small leakage output to Atmosphere → Troposphere with a very slow time factor over centuries. The permanence of storage is expressed by the shape of the leakage curve — near-zero rate for high-integrity geology, higher rate for uncertain formations.

5.6 The Accuracy Model

EaaA introduces a first-class accuracy concept that measures how well an assembled model reflects the actual system being studied. Unlike LCA pedigree scoring — which scores each dataset in isolation — the EaaA accuracy model scores the fit between each input and the activity it serves, and that score propagates upstream through the activity graph.

✦ PRINCIPLE

Accuracy in EaaA is not about how good a dataset is in isolation. It is about how well each declared input matches the activity being modelled. A perfectly measured dataset from the wrong geography still degrades accuracy when used in a mismatched context.

Component scores

The accuracy model has three independent component scores. Each is meaningful on its own and queryable separately. Future components — technological correlation, measurement basis — follow the same pattern and can be added without changing the architecture.

Source credibility is read off the organisation that publishes the producing activity, not off a flag on the activity: an organisation whose dataset is peer-reviewed carries that standing, and every activity it publishes inherits it. Nobody enters it, so nobody can inflate it. The EaaA-verified tier is granted by the framework's own review and never claimed by a provider.

Unknown is 0.5, never 0 and never 1

Where a component cannot be measured — the consumer or the producer declares no geography or no reference year, the input names no producer at all, or the activity has no product inputs — the component scores 0.5. Zero would assert a mismatch nothing measured; one would assert a fit nothing measured. A hollow sector and a genuinely terminal process are indistinguishable from the graph, and 0.5 is what indistinguishable scores. An activity with no product inputs therefore scores 0.5 on all three components rather than scoring nothing.

Final score — no grades

The final score is the mean of the three component scores.

There are no letter grades. Bucketing a continuous score into five boxes invites comparison of the boxes rather than the numbers, at thresholds that are provisional. The score is the whole of the claim, shown as a number from 0 to 100.

Reason codes

Each component score carries the reasons behind it — the rung of the ladder each input landed on (parent geography, a five-year gap, an unreviewed source, an unknown consumer geography) with the count of inputs on that rung. A reason names what the scoring did; it is not a grade and it moves no score. Several distinct causes can return the same number, and the codes are what tells them apart — which is what an auditor needs and a headline number cannot give. The consumer's own declared geographic level rides on the result as a fact, not a score: a country-wide activity whose inputs are all country-wide is a perfectly matched model of a vague question.

The bread example

The bread activity in Bavaria illustrates how the three components interact:

• Water input: source = a peer-reviewed organisation (0.85), geography = same ADM2 (1.0), temporal = same year (1.0) → 0.95
• Wheat input: source = ecoinvent (0.85), geography = Bavaria ADM1, one level up (0.8), temporal = 2-year difference (0.9) → 0.85
• Yeast input: source = industry reported (0.4), geography = Germany ADM0, two levels up (0.6), temporal = 5-year difference (0.7) → 0.57
• Energy input: source = ecoinvent (0.85), geography = global average (0.1), temporal = 3-year difference (0.7) → 0.55

The yeast and energy inputs are immediately visible as the weakest links. A practitioner can prioritise those for data improvement.

Accuracy propagates upstream

Activity-level accuracy is the mean, per component, over the activity's product inputs. Each input is scored for its own fit and then blended with the producer's own stored score for that component — the mean of the two — so a well-matched input from a badly-sourced upstream cannot score as if the upstream were sound. A low-accuracy dataset anywhere in the chain degrades every activity that uses it, measurably and traceably.

✦ PRINCIPLE

Propagation is a mean, not a minimum. A minimum would make the result insensitive to everything except the single worst node; a mean keeps every input's fit visible in the number.

Three rules fell out of building it:

• A mixed input is one input, blended by share. Where several producers feed one input flow (§3.4.2), each link is scored against its own producer and the results collapse into a single entry weighted by each link's share — so 95% from a good producer and 5% from a poor one scores essentially as the good one, and splitting a supplier in two does not move the mean.
• Unknown does not propagate. A producer whose own score is the 0.5 placeholder — a terminal process, an activity with no inputs declared — contributes nothing to its consumer's blend; the consumer takes its own fit alone. Blending it in would drag every consumer of every leaf toward 0.5 as though something had been observed.
• Accuracy does not scale. A consumer scales a producer's inventory and impacts by the amount it draws. A score must not triple because three kilograms were consumed, which is why accuracy is carried apart from the quantities.

✦ PRINCIPLE

Accuracy makes gaps visible rather than hiding them. A flow with no accuracy scores is still declared — the activity is complete — but the missing scores tell any downstream user exactly where the model needs improvement.

5.7 LCA Dataset Import Protocol

The unit linearity principle provides a formal pathway for importing existing LCA datasets. LCA datasets were often built at arbitrary or study-specific scales. The import protocol converts them to valid EaaA activities by identifying the unit at which their inputs and outputs are internally linear.

• Step 1: Take the LCA dataset as defined.
• Step 2: Test inputs and outputs at multiple scales.
• Step 3: Identify the unit at which the ratio of inputs to outputs is stable — this is the linear unit.
• Step 4: Declare that as the canonical EaaA unit for this activity.
• Step 5: The dataset is now a valid EaaA activity, usable at any percentage of that unit.

Where no stable linear unit can be found, the dataset was likely aggregated across incompatible conditions or is internally inconsistent. Such datasets fail the import protocol and are flagged for review rather than silently carried over.

✦ PRINCIPLE

The LCA import protocol is also a quality filter. Datasets that cannot be linearised reveal problems in their original construction. The protocol surfaces these problems rather than papering over them.

5.8 Calculation Semantics — What a Result Is Per

Removing the functional unit (A-004) raises a question LCA never has to answer, because the functional unit answers it in advance: with no declared basis and possibly several outputs, what is an activity's result per, and which output does it belong to?

✦ PRINCIPLE

A result is per the activity as its flows describe it. An activity's inventory is its own elementary flows plus its scaled upstreams, and its magnitude is set by the flow values themselves. Nothing is declared and nothing is normalised. A consumer's share is the exchange ratio — what its input draws divided by the producer's output flow it draws from — taken against the specific output flow the link points at. That ratio is the whole of "per", chosen per link by the consumer, never declared by the producer.

Three consequences follow, and the third is the one an LCA practitioner will read as a defect:

• Outputs the consumer does not draw are out of its scope. An activity producing x and y gives a consumer of x exactly what it drew. It inherits no share of y; y's fate — recycled, landfilled, sold on — belongs to whatever activity draws y. No allocation, no system expansion, no co-product carried down the link.
• A monetary-only link is a value share. Where a source states only what one activity paid another (EXIOBASE), the ratio is monetary on both sides and cancels, so burden propagates correctly without any physical quantity. The link keeps the product direction — producer output to consumer input — because burden moves with goods, not with money. A product flow carried in a currency unit is a stated low-accuracy state, never a claim that N kilograms moved.
• Shares across consumers may sum past the producer's own burden, and that is intended. You cannot have x without running the activity, so every consumer that needs x is honestly answerable for what running it costs. The invariant this breaks — that summing every product's footprint across the economy equals total emissions — is the attributional accounting claim allocation exists to preserve. EaaA does not make it. It answers what does this chain cost, per chain. Platform totals will not reconcile against a national inventory, by construction.

5.8.1 Direction of Traversal Is the Attribution

✦ PRINCIPLE

Upstream answers what this cost to make; downstream answers where it went. The same directed links, walked either way. An unused output's fate is a different root, not a different rule: nobody declares at production what happens to scrap, the model records it when some activity draws it and records nothing when none does. A downstream traversal is navigational, never additive — walking downstream and accumulating would push an end-of-life burden back onto the producer, the mirror image of the allocation error above.

A whole-life number is a root choice, not a sum. Cradle-to-gate is the production activity's result. Cradle-to-grave is the end-of-life activity's result, whose chain already contains production, which already contains its materials. Chain results are never added together: "production + destruction" counts production twice, because destruction contains it. The correct instruction to a practitioner is to pick the downstream-most root.

6. Impact Assessment — Starting Framework

Impact categories in EaaA are grounded in the physical compartment hierarchy established in Section 3.3. Every elementary flow already declares its compartment and geography — impact methods can therefore apply geographic weighting automatically, without the retrofitting that LCA requires. LCA categories are the starting reference but are not imported wholesale. Each is reviewed against the compartment structure and re-anchored to independent peer-reviewed characterisation factors.

✦ PRINCIPLE

Characterisation factors in EaaA come from peer-reviewed science updated on a defined cycle — not from any commercial database. The governance principle that applies to geographies and leaf activities applies equally to impact assessment: no single vendor holds normative authority over characterisation factors.

6.1 Category Structure — Mapping from LCA

The following table maps LCA impact categories to EaaA compartments. This is a starting framework — the final category list, characterisation factors, and geographic weighting rules are subject to peer review and independent governance.

6.2 What EaaA Adds to Impact Assessment

Because every elementary flow in EaaA declares a compartment and geography at the point of declaration — not retrospectively — impact methods can apply geographic weighting automatically and consistently. This is the most significant structural improvement over LCA impact assessment.

• Water scarcity: 1 litre extracted in Swartland (severe water stress) and 1 litre extracted in Scotland carry the same physical flow declaration but different impact weights — applied automatically from the geography declaration, no manual adjustment needed.
• Acidification: SOx from a ship in EU regulated waters and SOx in the South Atlantic carry different geographic weights — the named ocean region declaration enables this without any additional practitioner decision.
• Eutrophication: nitrogen flows declared to a specific watershed geography enable watershed-level loading assessment, not just national averages.

The accuracy model also interacts with impact assessment. A flow declared with low geographic accuracy — global average where a specific geography is needed — carries a degraded impact result. The accuracy score propagates through to the impact assessment output, making geographic data gaps visible in the final results rather than hidden in averages.

6.3 Characterisation Factor Governance

Characterisation factors are the multipliers that convert declared flow quantities into impact scores. In LCA these are bundled inside impact methods (ReCiPe, CML, TRACI) that are updated infrequently and governed inconsistently. EaaA separates the category structure from the factors, allowing each to be updated independently.

• Climate change: IPCC Assessment Reports — AR6 currently. Updated ~7 years. Global warming potentials are the most stable and widely agreed factors in LCA.
• Ozone depletion: WMO Scientific Assessment of Ozone Depletion. Updated ~4 years.
• Water scarcity: WRI Aqueduct or equivalent peer-reviewed water stress dataset. Updated annually.
• Land use: GBIF biodiversity data or equivalent. Geography-specific characterisation is an active research area — factors are less settled than climate change.
• Human toxicity / ecotoxicity: USEtox database — peer-reviewed, independently governed. Updated as new substance data becomes available.

6.4 Social and Financial Indicators

The framework's impact categories characterise environmental burden. EaaA also carries two themes that do not: social (employment and hours worked, each per ISCO skill level) and financial (labour income, tax revenue, gross value added, cost, and an environmental price). Every factor in them is 1.0, pinned to its unit, because what the indicator contributes is a definition — which resources belong to "tax revenue" — and those definitions are ISCO's and ESA 2010's, not EaaA's. They roll up a supply chain exactly as an emission does.

✦ PRINCIPLE

Social and financial indicators are not comparable with the environmental categories and never additive with them. Gross value added is money the chain generated; cost is money the buyer paid; an environmental price is money the chain externalised onto everyone else — three different euros under one theme. Impact categories are indicators, not a partition: reading them is choosing a question, not spending a budget.

7. Key Terms

8. Working Examples — Logic Stress Tests

The following examples are not finished LCA studies. They are deliberate stress tests of the EaaA framework — designed to find holes in the logic before the standard is formalised. Each example states how EaaA handles the situation.

Example 1: Baking Bread in Germany (Simple Linear Chain)

A bakery in Bavaria produces one loaf of sourdough bread. This is the baseline example — a simple, mostly linear chain within a single geography. It should work cleanly under EaaA if the fundamentals are correct.

Activity chain

• Extract groundwater (Geosphere → Subsurface) — ADM2: Landkreis München
• Pump and treat water (intermediate flow → bread activity)
• Grow wheat (inputs: water, soil nutrients; outputs: grain + straw as co-product)
• Mill wheat (input: grain; output: flour + bran as co-product)
• Bake bread (inputs: flour, water, energy; outputs: bread + CO2 to troposphere + waste heat)

EaaA behaviour

• Water scales linearly: bakery uses 40% of pump activity — inherits 40% of all pump inputs and outputs.
• Straw and bran are co-products with their own output declarations — they do not disappear.
• CO2 from baking is declared as output to troposphere (atmosphere → troposphere compartment).
• Waste heat: declared to the atmosphere as a flow of an energy resource, or drawn by a downstream activity as an intermediate flow (below).

Example 2: Copper Mining in the DRC (Cross-Sphere, Multi-Geography)

An open-pit copper mine in Katanga, Democratic Republic of Congo, produces copper concentrate for export to a smelter in Belgium. This example crosses geospheres, hydrospheres, and atmospheres, spans two ADM geographies, and involves multiple co-products and waste flows.

Activity chain

• Extract ore (Geosphere → Deep subsurface, ADM1: Katanga)
• Crush and float ore (intermediate flow; outputs: copper concentrate + tailings)
• Tailings disposal (solid waste output → Geosphere → Soil, Katanga)
• Acid mine drainage (liquid waste output → Hydrosphere → Freshwater, Katanga)
• Transport concentrate (intermediate flow, Katanga → Antwerp, crosses ADM1 boundaries)
• Smelt concentrate in Belgium (inputs: concentrate + energy; outputs: copper + SO2 to atmosphere + slag)

EaaA behaviour

• Tailings and acid drainage are declared outputs to physical compartments in the correct geography — Katanga ADM1, not Belgium.
• SO2 from smelting is declared to troposphere in Belgium ADM1.
• Transport crosses multiple ADM geographies — each leg declared separately with its own geography.
• Slag is a co-product of smelting — must be declared, not discarded.

Example 3: Wheat Farming in South Africa (Natural Activity, Water Stress Geography)

A wheat farm in the Western Cape, South Africa. This example tests how EaaA handles natural activities, water scarcity geography, and the 'factory or forest' principle in an African context where Ecoinvent data is sparse.

Activity chain

• Extract groundwater (Geosphere → Subsurface, ADM2: Swartland)
• Grow wheat (inputs: groundwater, solar radiation, soil nutrients, fertiliser; outputs: grain + straw + CO2 uptake from troposphere + N2O to troposphere)
• Harvest grain (intermediate output to milling activity)

EaaA behaviour

• The wheat farm is an activity like any other — factory or forest.
• CO2 uptake from troposphere is declared as an input from atmosphere — the farm draws CO2 in.
• N2O from fertiliser application is declared as an output to troposphere.
• Groundwater extraction is declared as an input from Geosphere → Subsurface at the correct ADM2 geography.
• Solar radiation: an elementary input from the atmosphere (below).

Example 4: Aluminium Can — Recycling Loop

A beer can is produced from primary aluminium, filled, consumed, and then recycled back into aluminium. This is a classic LCA recycling loop — one of the most contested modelling problems in LCA, particularly around avoided burden and allocation.

Activity chain

• Mine bauxite (Geosphere → Subsurface, Australia)
• Refine alumina (intermediate flow)
• Smelt aluminium (inputs: alumina + electricity; outputs: aluminium + CO2 + PFCs to atmosphere)
• Roll and form can (intermediate flow)
• Fill can — beer production (separate system)
• Consumer use and disposal
• Collect and sort used cans (intermediate flow)
• Re-smelt scrap (inputs: scrap + electricity; outputs: secondary aluminium — same quality as primary)

The LCA problem

In LCA, the recycling loop creates a multifunctionality problem: the re-smelting activity produces secondary aluminium that displaces primary aluminium. Who gets the credit — the can producer or the recycler? Three competing methods exist (allocation, system expansion, avoided burden) and give materially different results.

EaaA behaviour

• The re-smelting activity is declared normally: inputs (scrap + energy) and outputs (secondary aluminium to whoever uses it).
• Secondary aluminium is an intermediate output — it flows to the next activity that declares it as an input.
• If the secondary aluminium displaces primary aluminium, that displacement is a consequential question — it is not inherent to the activity declaration.

Example 5: Hydroelectric Dam — Congo Basin (Large Infrastructure, Long Life, Ecosystem Interaction)

A large hydroelectric dam on the Congo River. This example involves massive cross-sphere interactions, long asset life, ecosystem disruption as an output, and electricity as a co-product shared across a grid that crosses national boundaries.

Activity chain

• Construct dam (inputs: concrete, steel, labour; outputs: physical structure — a one-time activity with 60-year life)
• Reservoir flooding (output: land occupation → Geosphere → Soil; methane from decomposition → Atmosphere → Troposphere)
• Generate electricity (inputs: water flow; outputs: electricity to grid + water downstream)
• Electricity transmitted across grid (crosses DRC, Republic of Congo, potentially export)

EaaA behaviour

• Reservoir methane is declared as output to troposphere — same as any other emission.
• Land occupation (flooded area) is declared as output to Geosphere → Soil at the correct ADM geography.
• Electricity output goes to the grid — an intermediate flow to downstream activities.
• Water downstream is an output to Hydrosphere → Freshwater at the downstream geography.

Example 6: Electric Vehicle Battery — End of Life (Second Life and Cascade Recycling)

A lithium-ion battery from an electric vehicle reaches end of first life at 80% capacity. It is repurposed as grid storage (second life) before eventual recycling. This tests second-life modelling and cascade systems where the functional unit changes between lives.

Activity chain

• Manufacture battery (inputs: lithium, cobalt, nickel, energy; outputs: battery pack)
• First life — EV use (input: battery; output: mobility service over ~10 years)
• Second life — grid storage (input: used battery at 80% capacity; output: energy storage service over ~5 years)
• End of life — shred and recover materials (outputs: recovered lithium, cobalt, nickel as intermediate flows; black mass waste to Geosphere → Soil)

Worked example — recycled phone battery

• Battery activity declared: 2025, 4000mAh, degradation time factor 0.02% capacity loss per day
• Phone manufactured 2025 — inputs battery activity at year 0 (100% capacity)
• Phone used for 3 years — battery time factor running, 2028 extraction point
• At extraction: battery activity is at day 1095 of time factor → capacity ~79%
• Recycled phone declares output: battery activity reference, time factor position day 1095
• Refurbisher inputs that battery at day 1095 — receives an activity whose current output ratio is already declared
• If refurbisher uses battery in a second device: new device inputs battery at day 1095, time factor continues running
• If battery goes to materials recovery: a processing activity inputs the battery, outputs recovered lithium and cobalt as intermediate flows — each carrying their own upstream burden

Example 7: Fast Fashion T-Shirt — Globalised Supply Chain

A cotton t-shirt: cotton grown in Uzbekistan, spun in Bangladesh, dyed in Cambodia, assembled in Vietnam, sold in the UK, worn 10 times, and landfilled. Tests multi-geography chains, use-phase modelling, and the treatment of consumer behaviour.

Example 8: Natural Gas Power Plant with CCS (Negative Emissions and Cross-Sphere Injection)

A gas power plant with carbon capture and storage (CCS) captures CO2 from combustion and injects it into a saline aquifer 2km underground. This is a cross-sphere flow — atmosphere to geosphere — and produces what may be a net negative emission.

• Combust natural gas (inputs: gas from Geosphere; outputs: electricity + CO2 to atmosphere)
• Capture CO2 (input: CO2 from atmosphere; output: compressed CO2 — intermediate flow)
• Inject CO2 (output: CO2 to Geosphere → Deep subsurface)

Example 9: Urban Wastewater Treatment (Multi-Output, Receiving Water Body)

A wastewater treatment plant in Cape Town treats domestic sewage and discharges treated effluent into the Atlantic Ocean. Produces biogas (used for energy) and biosolids (applied to agricultural land) as co-products.

• Collect sewage (inputs: domestic wastewater; geography: City of Cape Town ADM2)
• Treat wastewater (inputs: electricity, chemicals; outputs: treated effluent + biogas + biosolids)
• Discharge effluent (output to Hydrosphere → Ocean, South Atlantic)
• Use biogas for electricity generation (intermediate flow — offsets grid electricity)
• Apply biosolids to farmland (output to Geosphere → Soil, Western Cape ADM1)

Example 10: Digital Service — Cloud Computing (Invisible Infrastructure, Shared Resources)

A single Google search query. The activity consumes a fraction of a server, a fraction of a data centre, a fraction of a network, and produces a result on a user device. Everything is shared and fractional. This is the hardest scaling example.

• User device (input: electricity; output: query transmitted)
• Network transmission (input: electricity across multiple geographies; output: data packets)
• Data centre processing (input: electricity + cooling water; output: query result + waste heat)
• Data centre cooling (inputs: water, electricity; outputs: evaporated water to atmosphere + waste heat)

Resolved chain — AWS EC2

• Data centre construction activity — unit: per facility, lifetime output declared (e.g. 10 billion compute-minutes). Amortised fraction consumed per compute-minute downstream.
• EC2 compute minute activity — inputs: electricity per compute-minute (AWS EU-West grid mix, geographic accuracy: region-level), cooling water per compute-minute, amortised fraction of construction activity. Output: 1 compute-minute.
• Website activity — inputs: N compute-minutes from EC2 activity (accuracy: Estimated if based on traffic averages, Measured if from CloudWatch logs). Output: website service.
• User query — inputs: fraction of website activity scaled to query volume. Output: query result.

The framework provides the tools. AWS decides whether to declare per compute-minute, per vCPU-hour, or per GB of memory — just as a pump manufacturer decides per litre or per 10 litres. EaaA may publish sector guidance on preferred units for digital infrastructure, but does not mandate them.

Example 11: Petrol Generator — Co-Product vs Waste

A petrol generator burns refined petroleum to produce electricity. This example illustrates how EaaA distinguishes co-products from waste outputs, and how the same physical output can be either depending on what the model contains.

Activity chain

• Extract crude oil (input from Geosphere → Deep subsurface)
• Refine petroleum (intermediate flow — crude oil in, refined fuel out)
• Petroleum combustion activity (intermediate flow — fuel in, combustion outputs out)
• Petrol generator (inputs: petroleum combustion activity; outputs: electricity, CO2, NOx, water vapour, waste heat, spent residue)

Output classification

• Electricity → intermediate flow to whatever activity declares it as an input
• CO2 → output to Atmosphere → Troposphere
• NOx → output to Atmosphere → Troposphere
• Water vapour → output to Atmosphere → Troposphere
• Waste heat → output to Atmosphere → Troposphere as a flow of an energy resource (resolved, Example 1), unless a downstream activity draws it (below)
• Spent residue → reference to output activity (handles soil/water compartment declarations)

Co-product vs waste — the heat example

The generator also produces heat. Whether that heat is a co-product or waste is not decided by the generator activity — it is decided by the graph:

• If a building heating system declares the heat as an input → it is a co-product. This is a combined heat and power (CHP) configuration.
• If nothing declares it as an input → it is waste, and routes to an output activity that declares it to the atmosphere.

✦ PRINCIPLE

An output is a co-product if at least one downstream activity declares it as an input. It is waste if nothing claims it — at which point it must resolve to a physical compartment, directly or via an output activity. The producing activity declares the output either way. The graph determines the classification.

Example 12: Cargo Ship — Durban to Antwerp (Multi-Geography Transport)

A bulk carrier transports copper concentrate from Durban, South Africa to the port of Antwerp, Belgium. This example shows how EaaA models multi-geography transport without a special transport schema — using multiple fuel inputs and multiple geographic outputs on a single activity.

The key principle

Combustion is geography-agnostic. Burning bunker fuel produces the same CO2, NOx, and SOx per tonne of fuel regardless of where the ship is. The geographic burden lives in the fuel production activity — where the fuel was refined and what its upstream costs were — not in the combustion itself. The outputs are declared to the geography where the combustion happens.

Activity: Ship copper concentrate, Durban to Antwerp

Inputs

• Bunker fuel — 800 tonnes, loaded in Durban (input references South African refinery activity, ADM1: KwaZulu-Natal)
• Bunker fuel — 400 tonnes, loaded in Las Palmas for top-up (input references Canary Islands refinery activity)

Outputs

• CO2 → Atmosphere → Troposphere, South Atlantic (international waters) — proportional to fuel burned in this zone
• CO2 → Atmosphere → Troposphere, EU regulated waters — proportional to fuel burned inside EU ECA zone
• CO2 → Atmosphere → Troposphere, ADM1: Western Cape (port departure emissions)
• CO2 → Atmosphere → Troposphere, ADM1: Antwerpen (port arrival emissions)
• NOx → same geographic split as CO2
• SOx → same geographic split as CO2 — note: different regulatory limits inside EU ECA vs international waters, relevant for impact assessment
• Ballast water discharge → Hydrosphere → Ocean, port geography (potential invasive species vector — output activity)

What this achieves

• The geographic split of emissions is explicit and auditable — EU-regulated emissions are separated from international waters emissions.
• Each fuel input carries its own upstream geographic burden — South African refinery costs are distinct from Canary Islands refinery costs.
• No special transport activity type needed — multiple inputs and outputs on a standard activity handles the full complexity.
• The proportions of fuel burned per zone come from the route log or estimated averages — declared as part of the dataset.

Example 13: Waste Plastic — Recycling Chain (UK to Malaysia)

A plastic bottle produced in the UK is collected, sorted, shipped to Malaysia, processed into recycled pellets, and used to manufacture a new product. This example shows how EaaA handles recycling, cross-geography waste flows, and end-of-life scenarios without avoided burden calculations or system expansion.

Activity chain

• Manufacture plastic bottle (inputs: virgin PET resin, energy; outputs: bottle)
• Consumer use (input: bottle; output: empty waste bottle)
• Collect and sort plastic waste (input: waste bottles; output: sorted PET waste bales, ADM1: England)
• Ship PET waste bales to Malaysia (inputs: bales + bunker fuel; outputs: CO2 to South Atlantic + Indian Ocean geographies, bales arrive in Malaysia)
• Process plastic waste — Malaysia (inputs: PET bales, energy, water; outputs: recovered PET pellets + process wastewater + residue to landfill output activity)
• Manufacture new product — Malaysia (input: recovered PET pellets; outputs: product)

What EaaA declares

• Every step is a declared activity with declared inputs and outputs. No credit is claimed for displacing virgin PET — the downstream manufacturer simply declares recovered pellets as its input rather than virgin resin, and its upstream burden reflects that chain.
• If the pellets are not used — if the Malaysian processor cannot sell them — they route to a landfill output activity in Malaysia. The full burden of shipping and processing waste that ends in landfill is visible in the model.
• The shipping emissions are declared geographically — South Atlantic, Indian Ocean, port geographies — not as a global average.

Consequential analysis on top of EaaA

A practitioner who wants to quantify the benefit of this recycling chain builds two models: (1) the chain above, and (2) an alternative where the waste plastic goes directly to a UK landfill output activity. The difference between the two models is the consequential benefit of recycling. That comparison is performed by the practitioner — it is not a feature of the framework.

Where the Questions the Examples Raised Stand

EaaA Brief v0.2 — a draft framework, developed in the open and published for review.