Neotoma dataset decisions for the Nordic pollen study¶
The dataset ledger joins the captured Neotoma relational dataset part, the sample-level pollen prerequisite screen, and the retained four-country direct observed-pollen study. It binds all 209 datasets and 12,388 samples to the exact source manifests and study files by SHA-256. The source part is also checked against its relational manifest. Each dataset row retains its country, site and canonical row digest.
The selected direct-pollen study used 85 datasets: Sweden 3, Denmark 2, Norway 52, Finland 28. The other 124 are explicitly outside that selected study, including seven with unassigned country. This is a study inclusion decision, not a claim that the other datasets are unusable. The sample screen finds 10,637 count-and-comparable-age prerequisite candidates across all datasets; it refuses 1,751 samples. Those candidates are not all on the study's closed 16-label axis and are not LRA inputs.
Every dataset retains unresolved LRA source requirements and has no matched independent measured-vegetation observation in this ledger. The source screen already refuses all 12,388 samples for LRA admission. Basin history, taxonomy, parameter binding and qualified source decisions remain separate. This one family ledger does not complete the all-family entity matrix or approve source use beyond the retained direct-pollen study.
Replay from the repository root:
mkdir -p artifacts/jobs/neotoma-entity-decisions
PYTHONDONTWRITEBYTECODE=1 PYTHONPATH=packages/bijux-pollenomics-dev/src \
.venv/bin/python -m bijux_pollenomics_dev.source_completion.neotoma_entity_decisions \
--repository . --output artifacts/jobs/neotoma-entity-decisions/report.json
The destination must be absent. The producer reads each of its six declared
source files once, then rechecks their byte identities before publishing the
report atomically. Compare the replay file byte for byte with this directory's
report.json.