Skip to content

LandClim author-example count preparation audit

The source-bound report independently sums every numeric raw taxon column and every prepared count cell in the retained 17-site author examples. The cohort contains nine sites in grid 512 and eight in grid 562. Across those sites, 501,713 raw counts split into 299,660 prepared counts and 202,053 raw-minus-prepared counts. A fresh replay now partitions the latter into 200,247 author zz.do not use counts, 418 unmapped counts and 1,388 outside the author time windows. No counts fall into the non-PPE category. The report retains these categories, source hashes and count totals for every site.

The producer verifies the private 193-file author-source snapshot manifest for each consumed source file. It compares fresh totals with the independent 17-site count reference, which retains the original independent report identity and numeric rows without the older challenge's plans, logs or failed jobs. The consumer replays the author's first exact lookup match and time binning for all 17 sites. WHH05 has 2,482 raw = 2,448 retained + 34 zz.do not use. The unmapped counts occur at NTWITCHEN (144) and WHH11 (274); the 1,388 outside-window counts occur at EXEBRID.

The report keeps the eight unmapped source headers verbatim with zero-based column positions and counts. NTWITCHEN has Euphorbia t. (7), Leguminoseae undiff. (1), Plantago major/minor (9) and Ranunculus t. (127); all four raw headers have a trailing space. WHH11 has Prunus spp (20), Hedera Helix (38), Filipendunla (11) and Ranunculus acris-type` (205, with a trailing grave accent in the source header). These are unresolved author-lookup names, not approved synonym or PPE assignments.

The lookup distinguishes Plantago from the separate leading-space label Plantago. Six LOBBSBOG counts under the former are excluded by the author rule; trimming leading spaces would incorrectly retain them. A focused R source check confirmed the exact Plantago mapping. The separate R formatting replay shows all 17 prepared files and 15,028 cells matching the author files.

This calculation supports the author-example raw-to-prepared relationship. It does not identify these examples as historical rows of the 1,128-site ESSD publication, resolve the ISTS PFT definition, provide measured vegetation truth, or authorize new source uses.

Replay into an absent directory under artifacts/:

PYTHONDONTWRITEBYTECODE=1 .venv/bin/python -m \
  bijux_pollenomics_dev.essd_reproduction.author_preparation_audit \
  --repository . \
  --author-source-root /path/to/private/landclim/author-reference \
  --output artifacts/jobs/landclim-author-preparation-audit/replay
cmp docs/report/source-relations/landclim/author-preparation-audit/report.json \
  artifacts/jobs/landclim-author-preparation-audit/replay/report.json