Findings about the BAFU/UVEK data — the open ledger
A finding is a measured observation about a BAFU dataset or release: an amount that cannot be right, a supplier declared with a location that is not registered, a flow the published flow list does not carry, a dataset flagged as cumulative that carries no flows. It is not a contribution layer (those are contribution records); a finding may lead to a layer, or to a recommendation, or to nothing but a note that saves the next person a day.
Every row below was re-measured on BAFU's own release files (the ecoSpold v1 archive, the LCIA results workbook, the change log) before it was written up; two earlier write-ups elsewhere were wrong because they measured a local import instead. The derivations, scripts and discussion threads live on the linked issues. Dataset values and names are quoted freely; text from the LCI reports is never pasted (Terms of Use clause 4, author copyright).
Status vocabulary
| Status | Meaning |
|---|---|
| open | observed and measured; not yet sent to BAFU |
| upstream-reported | sent to lca@bafu.admin.ch (date on the issue) |
| upstream-fixed | BAFU corrected it in a named release |
| informational | a method or scope difference, not a defect; recorded so nobody re-derives it |
| handled-in-layer | a contribution layer applies it in a candidate build; the upstream question stays open |
| withdrawn | did not survive re-measurement against BAFU's own files; kept so nobody re-finds it |
Ledger (seeded 2026-10-07 from the lci-bafu-catalog ledger, BAFU:2025 v2 and BAFU:2026 v1)
| # | Dataset(s) | Observation | Release | Status | Thread here · derivation |
|---|---|---|---|---|---|
| F1 | 188725 (ES), 178351 (CH), 242090 (CH) strawberries, plastic tunnel | 126.22 / 18.085 / 18.079 kg Disposal, polyethylene, 0.4% water, to municipal incineration per kg fruit, against 0.0096 / 0.042 / 0.036 kg of polyethylene granulate entering; the tunnel film is already disposed of in the tunnel dataset (per m²·a). The LCIA workbook scores them at 383 / 55 / 55 kg CO₂-eq per kg; the heated greenhouse in the same group scores 2.6. Identical in both releases; not in the 2026 change log. |
2025 v2, 2026 v1 | open | #6 · lci-bafu-catalog #1 |
| F2 | 191385 Refrigerant R134a, at plant (RER) | HFC-134a 0.019, HCFC-124 0.010, CFC-113 0.010 kg per kg — each exactly 10× the corresponding ecoinvent 3.6 dataset (0.0019 / 0.001 / 0.001); GWP100 108.6 against roughly 19. A documented process difference or a unit slip? | 2025 v2, 2026 v1 | open | #7 · #2 |
| F3 | hard-coal supply chains + 2 others | 23 exchanges declare the supplier location ENTSO; no dataset is registered as ENTSO, 25 as ENTSO-E, and every one of the 23 suppliers exists as ENTSO-E. Two more exchanges declare CH for suppliers that exist only as GLO / RER (Disposal, concrete, 5% water, to inert material landfill; Direct air capture system, sorbent-based, 100ktCO2). An importer linking by name + location drops all 25 silently; one linking by dataset number does not. |
2026 v1 | open | #8 · #3 |
| F4 | whole release | Withdrawn. The claim that ~7.6 % of elementary exchanges were absent from the flow list was an artefact of an earlier donor-seeded import; re-measured, one flow name is absent. | — | withdrawn | #4 |
| F5 | flow list, all releases | The XML declares chloride to water with seven sub-compartments (ocean, river, groundwater …); the published flow list carries one Chloride without sub-compartment. A structure question, not a value defect. |
2026 v1 | informational | #5 |
| F6 | CH concrete family | Treatment, concrete production effluent (0.0143 m³/m³) is carried by 4 of the 63 CH concrete datasets and by none of the civil-engineering concretes. Inconsistent modelling or a scope choice? |
2026 v1 | informational | #6 |
| F7 | published LCIA workbook 2026 v1 | 33 rows where the ecotoxicity total differs from inorganics + organics by more than 5 % (worst 23.4 %); 110 datasets score zero in every category. | 2026 v1 | open | #7 |
| F8 | 102 flagged and 37 structurally aggregated candidates | 102 datasets are cumulative LCIs flagged ecoSpold dataSetInformation@type=2 (no supplier links); 37 more PlasticsEurope-era eco-profiles are aggregated by structure (no supply input, > 50 elementary flows) but not flagged. Matters for every disaggregation or copied-process layer. |
2026 v1 | informational | #3 · #11 |
| F9 | 286952 CEM II, B-LL cement, at plant (CH) | lorry transport 4.345E-4 tkm/kg in the dataset; the cited 2020 cement/concrete report's table prints 2.00E-2 tkm for CEM II/B-LL (rail 1.50E-4 in both). Found by comparing the dataset with its report. | 2026 v1 | open | #9 · #12 |
| F10 | 196588 Disposal, rectangular straw bale | flagged type=2 but carries only its production exchange — a cumulative LCI with no flows; the two straw-bale siblings carry 1,656 flows each. | 2026 v1 | open | #3 · #11 |
| F11 | 1,289 datasets named xx … / xxx … |
retired by name prefix and/or category words (obsolete: 803, notMaintained: 441, current category: 45; 57 obsolete datasets carry no prefix). The convention is undocumented; consumers that match by name anchor on retired datasets unknowingly. |
2026 v1 | open | #4 · #15 |
F8: the list behind the 37 unflagged candidates
The hackathon's extended dataset list
contains 37 distinct dataset UUIDs with detected_by=structure. Each row
provides its name, location, unit, input count and elementary-flow count. This
subset includes four names prefixed xx; excluding those gives 33, not 22.
The detection script
selects datasets outside the flagged list with more than 50 elementary exchanges
and no production-supply inputs after excluding waste-treatment services and
infrastructure inputs measured in unit. It also excludes emission-only,
disposal, discharge, provision and resource-correction activities by name.
These are structural candidates; the rule and source generation matter when
comparing another search.
The table contains 101 flagged rows plus these 37 structural rows, for 138 total.
The separate 102nd flagged dataset, 196588 Disposal, rectangular straw bale,
is omitted from that survey because it has no elementary-flow vector (F10).
The hackathon's rebuild documentation reports 22 of the 37 rebuilt, with 15 remaining. That 22 describes rebuild coverage. An independently identified set of 22 still needs a UUID comparison; equal counts do not establish that they are the same subset.
Count provenance checked on 13 October 2026 against the published table at
commit f4784965c12b82f2731e7cdcd696c11afaf434a6. This check verifies the list
behind the reported count; it is not a new census of raw ecoSpold files, a
validation of the reconstructed inventories, or an acceptance threshold for a
current import. Continue the source comparison in
#3.
Mapping import work with Chris
Mapping integration status links the concrete importer fixes from this shared work. #14 tracks conversion factors and unit compatibility; #15 tracks CF fallback instructions the current consumer cannot apply. C002 keeps the candidate method integration and reproduction requirements. Importer issues are distinct from the raw-source findings above; code fixes do not establish a new method's score parity.
How a finding moves
- Open an issue with the BAFU data finding template: dataset code, release,
what you measured and how, why it cannot be right, exposure if known. It gets
bafu-finding+needs-triage. - A second person re-measures on the release files (never on an import) and
drops
needs-triage, or setswithdrawnwith the reason — a withdrawn finding stays in the table. - A row is added here with the status and the issue link.
- Batches go to BAFU (lca@bafu.admin.ch) through the recommendations tracker;
the issue gets
upstream-reportedand the date, thenupstream-fixedwith the release that fixed it.
A finding that a contribution layer works around gets handled-in-layer; the
upstream question stays open until BAFU answers it.