Namibian modern pollen and vegetation source join

We paired the published soil-pollen counts with the published vegetation-area rows for the same ten Namibian sites. The producer verifies the original TSV bytes and matches event identifiers and displayed coordinates. The paired rows preserve all 15 publisher radius labels, 100–1,500 m, and five taxon values per site and radius: 750 taxon pairs. The 100 m subset has 50 taxon pairs. The same pollen count is repeated across a site's radius rows; these are ten pollen samples, not 150 independent pollen samples. The report records source identities and counts.

The original tables are modern pollen, PANGAEA.931033 and vegetation, PANGAEA.931035, published by Tabares, Ratzmann, Kruse, Theuerkauf, Mapani and Herzschuh (2021) in PANGAEA.931037. Both original TSV headers declare CC BY 4.0. The source rows use pollen counts and calculated vegetation area in square metres. The two fraction columns divide by the five listed taxa only; they do not estimate total terrestrial pollen or total land cover. The parent PANGAEA description says mean cover was calculated by 100 m rings. We retain the publisher's ring-radius labels without assuming that the 100 m subset is the right pollen-source radius. The tables do not state exact soil-sampling or satellite-image dates.

This is a source-row join for an out-of-domain modern development cohort. We have not fitted a model, scored predictions, or validated Nordic vegetation reconstruction with these rows.

Descriptive radius sensitivity

The radius sensitivity report and site-level distances compare the two five-taxon compositions at each publisher Ring sect radius [m] label. At each site and label, the distance is the Hellinger distance sqrt(sum((sqrt(p_i) - sqrt(v_i))²)/2), where p_i is the taxon's share of the five pollen counts and v_i is its share of the five calculated vegetation areas. Zero means identical conditional compositions; one is the maximum distance. This is a descriptive composition mismatch, not a model prediction score.

There are ten independent sites at each of the 15 labels; the same site pollen sample repeats across labels. Mean distance across those ten sites is 0.3664 at the 100 m label, 0.3427 at 400 m and 0.3777 at 1,500 m. The full report gives all 15 mean, median and site-range values, and the CSV preserves each of the 150 site-label distances. The publisher describes 100 m rings; the exact paired areas contain nine decreases among 700 adjacent radius comparisons across 50 site–taxon series, affecting eight series. Such decreases cannot be cumulative nonnegative areas on the same fixed map. The JSON also reports the ratio of each site's five listed areas to a hypothetical full circular 100 m annulus for all 150 site–radius rows, treating each publisher radius label as the outer radius. These ratios are geometry diagnostics, not measured whole-cover fractions: exact ring-sector masks and taxon overlap are not established here. None of this chooses an ecological pollen-source radius. Pollen productivity and dispersal differ by taxon; exact pollen-sampling and image dates are absent from these tables. The cohort remains outside the Nordic lake target and is already exposed in development.

To reproduce from the two original PANGAEA TSV files, place PANGAEA.931033.tab and PANGAEA.931035.tab in one source directory, then run from the repository root:

python3 docs/report/vegetation-reference/namibia-pollen-vegetation/replay.py \
  --source-dir "$SOURCE_DIR" \
  --output-dir artifacts/jobs/namibia-pollen-vegetation-replay
cmp artifacts/jobs/namibia-pollen-vegetation-replay/paired.csv \
  docs/report/vegetation-reference/namibia-pollen-vegetation/paired.csv
cmp artifacts/jobs/namibia-pollen-vegetation-replay/report.json \
  docs/report/vegetation-reference/namibia-pollen-vegetation/report.json

The producer refuses any source bytes differing from the pinned SHA-256 hashes. Its output is deterministic and contains no local file paths.

The radius sensitivity producer consumes those same pinned original tables through the source-row join. Reproduce its two governed outputs from the repository root into an absent directory under artifacts/:

PYTHONDONTWRITEBYTECODE=1 python3 \
  docs/report/vegetation-reference/namibia-pollen-vegetation/radius_sensitivity.py \
  --source-dir ../bijux-pollenomics-data/vegetation/namibia-modern-pollen-vegetation \
  --output-dir artifacts/jobs/namibia-radius-sensitivity-replay
cmp artifacts/jobs/namibia-radius-sensitivity-replay/radius-sensitivity.csv \
  docs/report/vegetation-reference/namibia-pollen-vegetation/radius-sensitivity.csv
cmp artifacts/jobs/namibia-radius-sensitivity-replay/radius-sensitivity.json \
  docs/report/vegetation-reference/namibia-pollen-vegetation/radius-sensitivity.json