Two implementations, two numbers

Ruled publish. This list publishes European Commission — JRC's number for each of the 6 factors, as derivation: ruled.

EF 3.1 ships kresoxim-methyl twice in every context -- the same name and the same CAS number, 143390-89-0, under two EC numbers, 417-880-0 and 604-351-6 -- and the two rows carry different factors. This build merges them and publishes one number; the ecoinvent Centre's implementation publishes the other, and in all 24 contested rows of this substance its number is exactly the one this build declined, never a third number. So the disagreement is not about anybody's science: it is one question about EF's own duplication, answered twice, differently.

The JRC's number is published because it is the number EF states for the row ecoinvent's own flow corresponds to -- the prepared correspondence places ecoinvent's `Kresoxim-methyl` on `08a91e70-3ddc-11dd-95c2-0050c2490048`, and 134.73 in agricultural soil is that row's factor. ecoinvent's 53,540 belongs to `8e374806-0927-45ba-bbd5-d31fa2f6cf0e`, which nothing else maps onto. That reads as their pipeline taking the factor from the other of two rows sharing a CAS rather than as a considered choice between two numbers.

What this ruling does not claim is that 134.73 is right and 53,540 is wrong as science. EF publishes both, and the 397x gap between them in soil is a question for the JRC. The list's declined number is still recorded on the factor that kept it (#375), so a reader who wants the other number can find it. (#377)

Ruled publish. This list publishes European Commission — JRC's number for each of the 6 factors, as derivation: ruled.

EF 3.1 ships kresoxim-methyl twice in every context -- the same name and the same CAS number, 143390-89-0, under two EC numbers, 417-880-0 and 604-351-6 -- and the two rows carry different factors. This build merges them and publishes one number; the ecoinvent Centre's implementation publishes the other, and in all 24 contested rows of this substance its number is exactly the one this build declined, never a third number. So the disagreement is not about anybody's science: it is one question about EF's own duplication, answered twice, differently.

The JRC's number is published because it is the number EF states for the row ecoinvent's own flow corresponds to -- the prepared correspondence places ecoinvent's `Kresoxim-methyl` on `08a91e70-3ddc-11dd-95c2-0050c2490048`, and 134.73 in agricultural soil is that row's factor. ecoinvent's 53,540 belongs to `8e374806-0927-45ba-bbd5-d31fa2f6cf0e`, which nothing else maps onto. That reads as their pipeline taking the factor from the other of two rows sharing a CAS rather than as a considered choice between two numbers.

What this ruling does not claim is that 134.73 is right and 53,540 is wrong as science. EF publishes both, and the 397x gap between them in soil is a question for the JRC. The list's declined number is still recorded on the factor that kept it (#375), so a reader who wants the other number can find it. (#377)

Ruled publish. This list publishes European Commission — JRC's number for each of the 6 factors, as derivation: ruled.

The same ruling as kresoxim-methyl's freshwater ecotoxicity, for the same reason and about the same two EF rows: this substance is published twice in every context under two EC numbers, and the ecoinvent Centre's implementation takes its factor from the row this build declined. In all six contexts here ecoinvent's number is exactly the declined one. The JRC's is published because it belongs to the row ecoinvent's own flow corresponds to.

Worth noting separately: the two soil rows differ by 38x and the four air and water rows by 4x to 8x, so unlike the ecotoxicity case the gap is not concentrated in one compartment. It is the same duplication either way. (#377)

Ruled publish. This list publishes European Commission — JRC's number for each of the 6 factors, as derivation: ruled.

The same ruling as kresoxim-methyl's freshwater ecotoxicity, for the same reason and about the same two EF rows: this substance is published twice in every context under two EC numbers, and the ecoinvent Centre's implementation takes its factor from the row this build declined. In all six contexts here ecoinvent's number is exactly the declined one. The JRC's is published because it belongs to the row ecoinvent's own flow corresponds to.

Worth noting separately: the two soil rows differ by 38x and the four air and water rows by 4x to 8x, so unlike the ecotoxicity case the gap is not concentrated in one compartment. It is the same duplication either way. (#377)

Ruled publish. This list publishes European Commission — JRC's number for each of the 4 factors, as derivation: ruled.

Sulfur hexafluoride is the insulating gas in high-voltage switchgear and the most potent greenhouse gas either implementation characterises. The JRC's EF 3.1 states 25200 kg CO2-eq per kilogram in all six air compartments; ecoinvent's implementation of the same method states 24300 in the four its own flow list has. 3.6% apart is outside `FACTOR_TOLERANCE`, so the merge published neither, and the list scored an SF6 emission to unspecified air at zero while scoring one at aircraft cruise height at 25200 -- because those are the two compartments where only the JRC speaks.

**The JRC's number is AR6 as published.** Table 7.SM.7 of the Chapter 7 supplementary material reads `Sulfur hexafluoride SF6 3200 0.567 445 18300 2250 25200 10700 34100 11.2 26200 12.1 30600` -- lifetime 3200 years, radiative efficiency 0.567 W m-2 ppb-1, GWP100 25200 -- in the report on ipcc.ch, in the final government distribution, and in the authors' machine-readable supplement, which has not changed since September 2021. The February 2022 errata amends one row of Table 7.15 (methane) and nothing else in the chapter. EF 3.1's climate category is that table, and 25200 is what it says.

**ecoinvent's number is a different atmospheric lifetime, not a corrected printing.** Their v3.12 report (§7.3.4, Table 9) moves SF6 from 18300/25200/34100 to 18200/24300/29000 for GWP20/100/500 and calls the change negligible. It is not a rounding: hold AR6's radiative efficiency and replace the 3200-year lifetime with one near 1000 years, and the three horizons come out at 18175, 24358 and 28986 -- ecoinvent's three numbers to three significant figures, and AR6's lifetime reproduces none of them. A shorter lifetime is what a decade of measurements suggests: WMO's 2022 Ozone Assessment annex lists SF6 at the 3200-year lifetime with an alternative range of 850-1280 years (Ray et al. 2017, Kovacs et al. 2017), and its own metrics, 18400/24700/29800, sit about 1.6% above ecoinvent's, which is the ratio of WMO's radiative efficiency to AR6's.

So this is not two teams transcribing one table differently. It is ecoinvent updating an input the method's own publisher has not updated. The JRC's number is published because this list publishes EF 3.1, and EF 3.1's climate change category is IPCC AR6's table: adopting a lifetime AR6 did not adopt would make our EF 3.1 something other than EF 3.1, silently, in the column a practitioner reads as EF's.

What this ruling does not claim is that 25200 is the better science. It probably is not -- the newer lifetime estimates are measurements and AR6's 3200 years is an older model result, and WMO 2022 already prints the alternative beside it. The argument is about which question this list answers. ecoinvent's 24300 stays published under ecoinvent's name, where a reader who wants the newer lifetime can find it and say so in their own assessment; and if AR7 adopts it, `ruled_about` ends this ruling rather than letting it outlive its evidence. See #724.

The aggregate category, where the factor reaches every climate score.

Ruled publish. This list publishes European Commission — JRC's number for each of the 4 factors, as derivation: ruled.

Sulfur hexafluoride is the insulating gas in high-voltage switchgear and the most potent greenhouse gas either implementation characterises. The JRC's EF 3.1 states 25200 kg CO2-eq per kilogram in all six air compartments; ecoinvent's implementation of the same method states 24300 in the four its own flow list has. 3.6% apart is outside `FACTOR_TOLERANCE`, so the merge published neither, and the list scored an SF6 emission to unspecified air at zero while scoring one at aircraft cruise height at 25200 -- because those are the two compartments where only the JRC speaks.

**The JRC's number is AR6 as published.** Table 7.SM.7 of the Chapter 7 supplementary material reads `Sulfur hexafluoride SF6 3200 0.567 445 18300 2250 25200 10700 34100 11.2 26200 12.1 30600` -- lifetime 3200 years, radiative efficiency 0.567 W m-2 ppb-1, GWP100 25200 -- in the report on ipcc.ch, in the final government distribution, and in the authors' machine-readable supplement, which has not changed since September 2021. The February 2022 errata amends one row of Table 7.15 (methane) and nothing else in the chapter. EF 3.1's climate category is that table, and 25200 is what it says.

**ecoinvent's number is a different atmospheric lifetime, not a corrected printing.** Their v3.12 report (§7.3.4, Table 9) moves SF6 from 18300/25200/34100 to 18200/24300/29000 for GWP20/100/500 and calls the change negligible. It is not a rounding: hold AR6's radiative efficiency and replace the 3200-year lifetime with one near 1000 years, and the three horizons come out at 18175, 24358 and 28986 -- ecoinvent's three numbers to three significant figures, and AR6's lifetime reproduces none of them. A shorter lifetime is what a decade of measurements suggests: WMO's 2022 Ozone Assessment annex lists SF6 at the 3200-year lifetime with an alternative range of 850-1280 years (Ray et al. 2017, Kovacs et al. 2017), and its own metrics, 18400/24700/29800, sit about 1.6% above ecoinvent's, which is the ratio of WMO's radiative efficiency to AR6's.

So this is not two teams transcribing one table differently. It is ecoinvent updating an input the method's own publisher has not updated. The JRC's number is published because this list publishes EF 3.1, and EF 3.1's climate change category is IPCC AR6's table: adopting a lifetime AR6 did not adopt would make our EF 3.1 something other than EF 3.1, silently, in the column a practitioner reads as EF's.

What this ruling does not claim is that 25200 is the better science. It probably is not -- the newer lifetime estimates are measurements and AR6's 3200 years is an older model result, and WMO 2022 already prints the alternative beside it. The argument is about which question this list answers. ecoinvent's 24300 stays published under ecoinvent's name, where a reader who wants the newer lifetime can find it and say so in their own assessment; and if AR7 adopts it, `ruled_about` ends this ruling rather than letting it outlive its evidence. See #724.

The fossil sub-category, where both implementations file this substance and neither files it as biogenic or land-use: a synthetic fluorinated gas of industrial origin. Ruled separately from the aggregate because the queue asks separately, and an answer to one is not an answer to the other.