Docs / What we have found

Units and the conditions of measurement

Part of What we have found, the class where the unit on a row does not say enough: at what conditions, of what substance, standing for what.

A cubic metre of gas is not a quantity

ecoinvent 3.8 ships coal-mine off-gas in m3. From 3.9 onwards, the same flow under the same identifier is in Sm3 — standard cubic metres. The substance did not change; the older export simply did not say at what conditions its cubic metre was measured. A gas volume varies by roughly a fifth between 0 °C and 60 °C at constant pressure, so a number carried across without the qualifier is a number with an unstated error bar. EF 3.1's copy of the same flow still says m3.

What a unit can fail to say

A number without its unit is not a measurement, and for some quantities the unit alone is not enough either — a volume of gas needs the temperature and pressure it was measured at. Source lists are inconsistent about both, in ways that do not announce themselves.

A kilogram of ore is not a kilogram of metal

The other way a unit can be insufficient is that it does not say what is being weighed. ecoinvent measures some resources as the mineral that comes out of the ground; EF 3.1 measures the element the mineral is made of. Both are in kilograms, so nothing in the units says a conversion is needed, and mapping the two at parity credits an inventory with a kilogram of metal for every kilogram of ore.

The correspondence table knows this, and applies it to one pair and not the other. GLAD's file has a ConversionFactor column and uses it: rows 4856 to 4861 send ecoinvent's three TiO2, … in ground flows to EF's elemental titanium with a factor of 0.6, which is titanium's share of TiO₂ by mass (47.867 / 79.866 = 0.599). Rows 5364 and 5365 send zirconia, as baddeleyite, in ground to elemental zirconium with no factor at all — and baddeleyite is ZrO₂, of which zirconium is 91.224 / 123.22 = 0.740 by mass. The same substitution, the same table, stated once and omitted once.

ecoinvent's own characterisation numbers show the size of it, because ecoinvent states its resource-depletion factor on both weighing bases:

4.0273e-06 / 0.74033 = 5.4399e-06, which is ecoinvent's own zirconium number to four figures. It is one number stated twice, on two weighing bases, and read as a 35% disagreement by anything that compares them as though both meant the same kilogram. Gypsum is the same story with water in place of oxygen: ecoinvent gives Calcium sulfate 4.5462e-05 and Gypsum 3.5949e-05, and the dry salt is 136.14 / 172.17 = 79.1% of the hydrate's weight — 4.5462e-05 × 0.79073 = 3.5946e-05.

What it cost was this project's own, and is recorded here because the source fact is what set it up: both ecoinvent flows of each pair were mapped onto EF's single flow, the model holds one factor per flow and category, so two numbers arrived where one could be kept and neither was published. Calcium sulfate carried no resource-depletion factor at all, and zirconium's was published as the JRC's alone although ecoinvent had stated it twice and agreed. The decision is to publish each substance with the number its own publisher stated about it — baddeleyite is published as the mineral it is — and, where a conversion genuinely is the answer, to state it on the mapping row with both units named and apply it even when both units are kg (#596).

The name says what the kilogram stands for, and the unit does not

A third way a unit can be insufficient: the row is measured in one quantity and named for another. Stepwise 2006 ships uranium as a raw-material resource three times, and every one of them in kilograms:

All three carry 7440-61-1 and the formula U: one substance, three rows, and what separates them is an assumption about the fuel cycle — 451 GJ/kg is the net once-through figure of the ETH-era data, 560 GJ/kg the gross content, and 2,291 GJ/kg a cycle with reprocessing, which extracts about four times what a once-through cycle does and is still under 3% of complete fission. The conversion is stated in the name and nowhere a machine reads, and the method's own factor is that same number: the row is a kilogram of uranium standing for an amount of energy.

Read as shipped, a kilogram in one row and a kilogram in another differ by a factor of five in what they mean, and none of them is the kilogram this list publishes uranium in — the resource is published in megajoules, EF 3.1's basis. So the three rows are rebased onto MJ, each by the factor its own name states rather than by a single project default, which would restate the 451 row by 24%, and renamed to Uranium with the vendor's spelling kept as a synonym. The correction is written once for the whole SimaPro lineage rather than per list, because the habit came with ecoinvent 2's flows and arrives in every list that inherited them (#713).

The rest of the class, per source list

BAFU's water rows are the clearest case of a unit being an accident rather than a decision. Four Water discharge rows are in kilograms while 110 rows of the same four compartments are in cubic metres; standing timber is shipped /kg in two compartments and /m3 in a third, and across the 11,947 datasets of the 2026 v1 archive the volume rows are used 134 times and the two mass rows once each, for 0.0499 kg and 1.9 × 10⁻¹¹ kg. Publishing a substance whose unit depends on which vendor row happened to create it is what the corrections here prevent; each states its conversion factor, and for liquid water that factor is the density, 1000 kg/m³ (#409, #384).