Skip to content

Rootsi workbook age-column audit

The replay verifies the four private workbook hashes against the 79-sequence overlap ledger, then reads only the source-native age-column structure. Its report retains exact sheet-row locations and values for chronology reversals. The workbooks remain in the private source store.

Sequence Sample rows Explicit source age column Finding
FLINK 109 Sample age (cal AD/BC) The value jumps from −3850.6025 at sheet row 55 to +3910.8433 at row 56, then returns to negative ages.
HOLK 40 Sample age (cal AD/BC) The bottom two ages rise from −1491.0767 to −1490.1805; the 40 exact numeric ages contain no duplicates.
RANV 141 None Column two is SALIX; no workbook sample-age field is present.
STAV 77 None Column two is Picea; no workbook sample-age field is present.

The FLINK jump is a candidate sign error, but the local bytes do not authorize a corrected value. HOLK has a small source-order reversal; the claim of duplicated ages in the inherited review is not supported by exact equality of these 40 workbook cells. RANV and STAV have site coverage metadata elsewhere, which are not sample dates. No source age uncertainty, model identity, historical publication-input identity, or per-sequence analytical rights is established. All four remain refused for dated-event and LRA use on this evidence.

A separate Neotoma source witness depth-matches all FLINK and STAV workbook rows. It identifies the FLINK row-56 sign disagreement and supplies a candidate STAV calibrated-BP point-age vector from the shared EPD lineage. It does not correct the workbook or authenticate either sequence as the historical LandClim input.

An arithmetic check rounds the extrema of 1950 − source age to integer BP and compares them with the sequence metadata bounds. HOLK's full 40-cell extrema produce −50 to 3441 BP, matching its metadata; its last sheet row instead produces 3440 BP, consistent with the bottom reversal. FLINK's full 109-cell extrema produce −1961 to 12011 BP, while its metadata say −50 to 12011 BP. Omitting the single cell immediately after the interior reversal yields the metadata bounds. This is a diagnostic of the discrepancy, not evidence that the omitted cell should be changed or that the 1950 arithmetic is an approved calibrated-age conversion. The report retains the diagnostic omission separately from the unchanged source values.

Source-holder decision needed: Does the LandClim II FLINK input use this exact 109-row workbook, and what is the authoritative cal AD/BC value at sheet row 56? For HOLK, which age-depth model and ordering govern the last two samples, and how are their uncertainty bounds represented? For both sequences, which conversion convention produced the published BP extrema? The request is for source records and model identifiers, not permission to replace the cells with a visually plausible value. RANV needs a sample-age source linked to its 141 workbook rows; STAV needs historical author-input and model-version confirmation for the Neotoma candidate age vector.

Reproduce from the repository root:

PYTHONDONTWRITEBYTECODE=1 .venv/bin/python \
  docs/report/source-relations/landclim/workbook-chronology/replay.py \
  --source-root ../bijux-pollenomics-data/landclim/rootsi-workbooks \
  --output artifacts/jobs/rootsi-workbook-chronology/report.json
cmp artifacts/jobs/rootsi-workbook-chronology/report.json \
  docs/report/source-relations/landclim/workbook-chronology/report.json