Scientific concepts

Use these definitions when reading the results, the source prerequisites, and an atlas record. A term names the quantity or evidence role; the linked report supplies its actual inputs and permitted interpretation.

Counts, fractions and time

Term	Meaning in these reports
Native pollen count	The source's recorded count for a sample and taxon. Retained counts can include excluded or unmapped taxa. A percentage alone does not establish an integer count.
Denominator	The total against which a fraction is calculated. All native pollen and an assigned taxon subtotal are different denominators; the report names which one it uses.
Closed taxon response	Fractions normalized over one declared taxon set. A closed response may omit substantial native count mass and cannot be read as all vegetation.
Taxon target	A declared destination concept for native labels. Assignment, exclusion and unmapped mass stay explicit; a text-label match alone does not qualify pollen productivity.
Age model	A source-specific mapping from depth or sample identity to age. An alternate model is retained separately until a model is selected for the intended analysis.
cal BP	Calendar years before 1950. A source window labeled BP still needs its own chronology qualification; its label alone does not establish a calibrated sample age.
Analysis bin	A declared age interval used to aggregate observations. A point age can select a bin while the source age interval crosses its boundary. Contained-age sensitivity retains only samples whose support fits the declared bin.

Read coverage and naming for observation units, unresolved denominators and country dimensions. Read results and model use for the separate point-age and contained-age comparisons. Marginal age bounds do not supply a joint chronology posterior.

Predictions and reconstructions

Term	Meaning in these reports
Direct pollen prediction	Prediction of the declared observed-pollen response from training pollen. Its target is pollen composition, so its score does not measure vegetation accuracy.
LRA	Landscape Reconstruction Algorithm, comprising regional REVEALS and local LOVE reconstruction. Each run needs qualified pollen, chronology, basin, taxon, productivity and transport inputs.
REVEALS	The regional vegetation reconstruction component. A retained LandClim grid cell is an existing modeled result; displaying it does not rerun the reconstruction.
LOVE	The local vegetation reconstruction component, conditional on its regional contribution and declared model inputs. The repository's numerical profile and optimizer settings belong to the named run.
PPE	Pollen productivity estimate for a declared taxon and parameter source. It is a model input, rather than a measured vegetation outcome for the target lake.
Fall speed and transport	Inputs describing pollen settling and dispersal in the chosen model. Agreement on one implementation's arithmetic does not select universally suitable parameters.

The papers and evidence roles connect these method names to their source accounts. The model-use results state the settings and refusal conditions of the existing development calculations.

Evidence and uncertainty

Term	Meaning in these reports
MAE	Mean absolute error on a named response and aggregation unit. An equal-bin mean weights supported bins equally; it does not weight every sample or country equally.
Hellinger distance	A distance between the declared pollen compositions. It describes a different discrepancy from MAE, so two methods can rank differently under the two metrics.
Standard error	A source or model uncertainty summary for the named estimate. Individual-cell standard errors are not a pooled country uncertainty or a confidence interval.
Source correspondence	Agreement on specified source inputs, arithmetic or output bytes. Its claim is limited to that comparison.
Withheld target	A target excluded from the named training set. Prior source or outcome inspection is reported separately and can limit an independence claim.
Measured vegetation reference	A separate measured outcome with matching site, time, taxa and spatial support. A modeled grid or nearby archaeological record is a different evidence role.
Refusal	An explicit unsupported or missing-input outcome retained with its reason. A refused calculation is not a zero measurement.

Start the atlas walkthrough to inspect one modeled-context cell, trace its source and reproduce its bounded summary. The walkthrough keeps software behavior, source custody and scientific fitness as separate claims.