Docs / What was changed in each source

Changes to openLCA LCIA methods 2.8.2

Part of What was changed in each source. What is still open about GreenDelta's package is on the issue tracker under the label greendelta.

GreenDelta publishes the LCIA methods it implements for openLCA as one archive, openLCA LCIA Methods 2.8.2 2026-07-01.zip: 45 methods — ReCiPe 2016, CML-IA, TRACI, IMPACT 2002+, EF 3.1 and its predecessors, IPCC, USEtox, EPS and others — and the 61,211 elementary flows they characterise between them. This list is those flows. The characterisation factors are not read here.

That makes it a list of what 45 methods characterise rather than of what inventories emit. Theophylline sent to agricultural soil is on it, in 11 compartments, because a toxicity method has a number for it; a substance no method scores is absent however often a process releases it.

The flows carry GreenDelta's own identifiers, except for 4,479 that reuse an identifier ecoinvent 3.8 or 3.12, or EF 3.1, already ships. Most rows have a registry number (55,630 of 61,211), written zero-padded the way SimaPro writes it — 000058-55-9 for theophylline — and read here without the padding, so one number reaches one substance whichever list it came from.

The compartments are ecoinvent 2's words — Emission to air / low population density, Resource / in ground — and 21 of the 33 are the ones GreenDelta's EF 3.1 package already uses. They are placed the same way here as for that package. Of the other twelve, two are refused (below), and four hold more than one kind of flow so the name decides:

The six noise rows in Immaterial emission go where BAFU's do.

Waste rows refused

The 186 rows in Waste / unspecified (94) and Waste / ecopoints 97, CH (92) — Packaging waste, Slags, Chemical waste and the like — are final waste leaving the product system, which Ecological Scarcity 1997 scores, and not exchanges with the environment. They are refused, as AGRIBALYSE's final waste is (#189), and each is published as an excluded source concept with the reason, so somebody counting GreenDelta's rows against this list's correspondence can see a decision rather than a gap.

Black carbon refused

GreenDelta ships 13 rows of black carbon sorted by particle size: Particulates, < 2.5 um, black carbon (5 air compartments), > 2.5 um, and < 10um, black carbon (4) and > 10 um, black carbon (4). For particulate matter, EF 3.1 and ECO-COSTS 2025 give them the same factors as the plain particle flow of the same size and compartment. IPCC 2021 also gives them a climate factor — 658.6 kg CO2-eq per kg over 100 years — that no plain particle row has.

Placing them on PM2.5 or PM10 would put that climate factor on every PM2.5 or PM10 emission. Whether black carbon is a substance of its own is not decided yet (#384), so the 13 rows are refused and published as excluded with the reason.

EF 3.1's own flows

125 rows carry an EF 3.1 identifier: GreenDelta's copy of EF's flow, renamed in ecoinvent's spelling. Particulates, < 10 um in unspecified air is EF's particles (PM10), and Copper ion under Resource / in ground is EF's copper resource. The merge sends each one to the EF 3.1 flow with that identifier, recorded as shared_identifier rather than as a name-and-registry-number match.

Particle sizes

43 of the package's particle rows are read in particulate-flow-classes.json. All but one spelling are ecoinvent's, BAFU's or Stepwise's, and are read as those lists' are. The new spelling is Particulates, > 0.2 um, and < 2.5 um, read as EF 3.1's Particles (PM0.2 - PM2.5).

Location-specific rows refused

6,206 rows write a place into the name. Every one is refused and published as excluded with the reason. They cover 372 places, from countries (ES, NO) to regions (RoW, UCTE, IAI Area, Africa):

• 4,690 water rows: Water, river, CH, Water, RoW, Water, cooling, unspecified natural origin, IAI Area, Africa;
• 949 emissions to air: Nitrogen oxides, ES, Ammonia, FR, Sulfur dioxide, AT;
• 134 emissions to water: Nitrate, DE, Phosphorus, DK, and Nitrogen, NO, which is nitrogen emitted in Norway (GreenDelta's description says NO = Norway);
• 433 land rows: Transformation, from forest, extensive, GA, Occupation, permanent crop, BR.

The unregionalised row is kept: Water, river, Nitrogen oxides, Transformation, from forest, extensive. The location-specific rows carry factors their unregionalised row does not always agree with. Ecological Scarcity 2013, for one, gives Nitrogen, NO a factor (57,000) and plain Nitrogen none. These are left out with the rows.

Factors for a place left out of BAFU's EF 3.1

The same decision applies to the factors GreenDelta states for EF 3.1, which BAFU's source reads out of this package. Occupation, agriculture has one land-use factor for no place and 400 more, one per country or region. The package states 183,654 EF 3.1 factors; the 91,304 tied to a place are left out, and the 92,350 for no place are read. Every one of the 33,045 flows has one. 69,116 of those left out are land use and 12,928 water use. The JRC's own EF 3.1 factors per country are unaffected.

Which kind of water each water row is

Every kind of water has the same registry number, so the merge tells river water from sea water or cooling water only by a reading in water-flow-materials.json. 123 water rows are left once the location-specific ones are refused (#389). 101 of them, under 54 names, are read:

• A name another list already ships is read the way that list reads it: Water, river is river water, Water, salt, sole brine, Water, cooling, well cooling water.
• A name that says what the water is, is read that way: Water, ground and Water, process, well are groundwater, Water, process, salt, ocean sea water, Cooling water returned to a river is cooling water.
• Four readings are ours rather than the name's. Water, thermoelectric groundwater consumption and its saline and surface siblings are cooling water: it is the water a power plant's cooling uses up, and every cooling name is read by its use. Thermally polluted water is returned cooling water, warm rather than contaminated. Water, with river silt is river water.
• Names that say nothing about the source, such as Water, Water/kg and Water, process and cooling, unspecified natural origin, are water of no stated kind. So is Water, unspecified natural origin filed in ground, for the reason BAFU's is (#89).

Where two names are read as one kind of water, the methods in the package score them alike, with one exception: ECO-COSTS 2025's Baseline Water Stress gives some of them different factors, for example −0.0001 to the thermoelectric rows and +0.0001 to the cooling rows. They are read as one kind of water regardless.

Twenty-two rows are refused for good, because they are not a kind of water:

• The 11 water-stress rows, Water, low water stress to Water, unspecified natural origin, very high water stress. A stress class stands in for the watershed the water came from, as a place in the name does. Ecological Scarcity 2006 and 2013, ILCD 2011 and ECO-COSTS 2025 score each class differently — Ecological Scarcity 2013 from 24 to 22,000 — and the other methods score every class as plain water.
• the 11 Water BWS rows, Water BWS 0% to Water BWS 100%. Only ECO-COSTS 2025 scores them, with a factor equal to the percentage in the name.

Which land class each land row is

All 490 land rows that are kept are read in land-flow-classes.json. Sixteen of them use words the other lists don't, and each was read as follows:

• Occupation, annual crop, conservation tillage, and conventional and reduced: cropland under that tillage, the class ecoinvent's arable, … tillage (obsolete) rows already have. Every method in the package gives them the factors of annual crop.
• annual crop, non-irrigated, monotone-intensive, in all three directions: cropland, non-irrigated, farmed monotone-intensive — a new intensity, beside diverse-intensive. Every method gives it the factors of annual crop, non-irrigated except Ecosystem Damage Potential, which scores it 0.74 against diverse-intensive's 0.61.
• Transformation, to unknown (used): to unspecified land in use, the class of Transformation, to unspecified, used. EF 3.1 scores it 1,914.6, which EF 3.0 gives the other spelling.
• Transformation, from agriculture + forest, and to: agroforestry, read as agriculture, mosaic. CORINE files agro-forestry areas (2.4.4) under heterogeneous agricultural areas (2.4).
• Occupation, sparsely vegetated areas, steppe, tundra, badlands: CORINE Land Cover class 3.3.3, published as a new cover, sparsely vegetated area. It is not read as grassland, because ILCD 2011, IMPACT 2002+ and EF 2.0 score it differently from openLCA's Occupation, grassland.

Not yet merged

openLCA LCIA methods 2.8.2 is registered and extracted, and merges last (priority 500) when a build names it. It cannot be merged yet (#385): in 20 land classes, two openLCA rows with different names reach one class — Occupation, annual crop and Occupation, arable land, unspecified use, for example — and the build stops on each until land-flow-groupings.json says the fold is meant. A grouping names the flow the two reach, so it is written from the first build that merges the list.

No documented build includes it, so this page records no merge outcomes: no row has been sent anywhere by hand, no registry number corrected, and no name split across two substances.

One name, two substances

None recorded yet.

An ionic charge was added or corrected

None recorded yet. The package names metal ions Cadmium II and Arsenic V and writes the element's registry number on many of them (#380).

Rows merged into one flow

Land-use rows shipped twice. 166 land-use names are in the package more than once with the same unit: Occupation, annual crop in Resource / land and again in Resource / unspecified, and so for 160 of them. Each copy is the same land class, and both map to the same flow. The row in Resource / land is kept as the valid one. The other copy, 170 rows in all, still maps but is published deprecated, pointing at the valid row (source-row-duplicates.json). Two names needed a different tie-break:

• Occupation, sea and ocean has no Resource / land copy, so the Resource / in water row is kept over Resource / unspecified.
• Occupation, inland waterbody, unspecified is in Resource / land twice, with one factor each. The row kept is the one whose identifier ecoinvent 3.12 also ships.

Water rows shipped twice. 25 water names are in the package more than once with the same unit, usually in Resource / in water and again in Resource / unspecified. Every copy reaches Resource / water, and the row in Resource / in water is kept as the valid one; the other copies, 28 rows, still map but are published deprecated. The copies are scored alike, except that EF 3.1 gives the Resource / land copy of Water, process, unspecified natural origin/kg a water-use factor of 0.0.

A deprecated copy does not weaken the valid row's match: two copies of one flow are not two flows sharing one.

Beyond land use, the package ships 504 pairs of flows with the same name, compartment and unit — usually GreenDelta's own flow beside ecoinvent's copy of it, and in 85 of them only one of the two has an EF 3.1 factor. None of those is ruled on yet.

Registry numbers corrected