LandClim II modeled Nordic panels

The panel inventory contains 400 source-window × country × metric rows: four Nordic countries, 25 published PANGAEA 937075 windows, and four separately labeled model metrics. The receipt pins the normalized grid and source classification table by byte count and SHA-256. The producer validates the complete 1,875-row PANGAEA II temporal grid through the production modeled-context manifest before summarizing any row.

ET, ST, and OL are the published evergreen-tree, summer-green-tree, and open-land land-cover types. AL is the published agricultural-land/cereals plant functional type within OL. AL is not a fourth disjoint land-cover category. All values are source-model percentage-cover estimates. A source window is not a physical pollen sample chronology, and none of these values is measured vegetation truth.

Each CSV row gives the number of source grid cells, the high/low/no-pollen quality split, and the minimum and maximum individual-cell estimates and source standard errors among high- and low-quality rows. no_pollen_data placeholders are excluded from numeric ranges. Blank numeric bounds mean no usable modeled cell. Standard errors are not pooled, converted to confidence intervals, or treated as uncertainty for a country average. The table makes no area-weighted or country-mean estimate.

The 25 windows are discrete published intervals from 11,200–11,700 BP to 0–100 BP. Adjacent windows do not authorize interpolation or propagation. Country groups here use the governed source-grid country labels and the production manifest's country/window quality inventory; they are descriptive model coverage, not a boundary-release decision.

Reproduce

From the repository root, write to a new artifact directory and compare both files to this directory:

PYTHONDONTWRITEBYTECODE=1 .venv/bin/python -m \
  bijux_pollenomics_dev.source_completion.landclim_modeled_panels \
  --repository . --output artifacts/jobs/landclim-modeled-panels/replay

The producer refuses changed source labels or an incomplete model manifest. The focused checks exercise the four-panel inventory, no-pollen denominators, paired standard-error fields, and label drift. Qualified scientific interpretation and comparison to independently measured vegetation are separate decisions.