Swedish native pollen against the LandClim II author lookup

The 39-bin result applies the authors' published LandClim II taxonomy lookup to the same three native NISP datasets used in the held-out Swedish direct-pollen study. The lookup is already held in the external LandClim author-reference bundle; its SHA-256 is 3a745f9b7606e5d0b59feaba2683b930c8d3990e4810f8217b69160a29a1b507. The authors' preparation-code description calls this the synonym table for harmonizing taxonomy. The method paper, Sect. 2.2, states that pollen morphological types were harmonized into 31 RPP taxa.

Site	Native NISP	Author-assigned	Author-excluded	Unmapped	Assigned fraction
Abborrtjärnen	88,004	86,691	1,311	2	98.5080%
Bussjösjön	56,450	40,915	9,277	6,258	72.4801%
Bökesjön	42,548	41,407	1,110	31	97.3183%

The lookup has 10,788 rows, 9,731 distinct source strings and 31 assigned targets in addition to zz.do not use and an age-field target. Duplicate source strings agree on their target. Exact source strings map Alnus to Alnus glutinosa, Corylus to Corylus avellana, Quercus to Quercus deciduous, and Poaceae to Poaceae. Those assignments recover large counts that the separate literal parameter-name screen could only mark unmatched. The lookup also explicitly excludes names such as Humulus/Cannabis (4,647 BUS counts). Its listed Poaceae undiff. string does not cover BUS Poaceae undiff. (<40 µm) (6,096 counts) or Poaceae undiff. (>40 µm) (158 counts); those remain unmapped without a source-backed, size-specific decision.

The author lookup does explicitly assign its separate spellings Poaceae undiff. <40um to Poaceae and Poaceae undiff. >40um to Cerealia-t. These are source-witnessed candidates for the two BUS labels, which differ in punctuation and the micrometre symbol. If those two name variants were accepted, 6,254 BUS counts would move from unmapped mass: 6,096 to Poaceae and 158 to Cerealia-t. The BUS author-assigned fraction would rise from 72.4801% to 83.5589%. The governed bin and site results retain that scenario in separate candidate fields; the actual author lookup counts and the 6,258 unmapped BUS counts remain unchanged. No other name is normalized by this rule.

The exclusion is especially consequential in BUS bin 500–1000 cal BP: Humulus/Cannabis contributes 3,369 of 7,175 native NISP counts (46.9547%) there. In bin 0–500 cal BP it contributes 1,254 of 9,341 counts (13.4247%). These are direct native source counts in the enumerated exclusions, not vegetation percentages. A 31-target reconstruction would address the author-selected pollen subcomposition, with markedly different retained mass across these two bins; the excluded taxon must remain visible in any score or interpretation.

These are author lookup assignments, not an independently reviewed biological identification or a numerical reconstruction. The lookup target names are not identical to every published PPE/FSP parameter label; for example it emits Quercus deciduous and Picea, while the parameter table uses Quercus deciduous t. and Picea abies. The separate author parameter binding resolves the exact target-to-RPP/FSP correspondence in the retained author code profile. Applying it to Neotoma still requires a complete counting-denominator decision and sample-time basin geometry. In particular the Bussjösjön basin review records a material coordinate conflict. No Swedish LRA sample is admitted.

The report retains the 31 target count totals, leading excluded and unmapped source names for each site, source identities and exact output CSV hash. Replay from the repository root with the already retained external author-reference bundle:

PYTHONDONTWRITEBYTECODE=1 .venv/bin/python -m \
  bijux_pollenomics_dev.source_completion.landclim_taxonomy_mass \
  --repository . \
  --author-source-root ../bijux-pollenomics-data/landclim/author-reference \
  --output artifacts/jobs/landclim-taxonomy-mass/replay

Compare the replayed bins.csv and report.json byte-for-byte with this directory. The producer refuses changed lookup bytes, conflicting duplicate names, changed Swedish source scope and any native count that fails to close.