Calibration
Chapple’s catalogue is calibrated against IntCal13 and stores a single contiguous range per
date — so any multimodal structure is already lost. radio-rian recalibrates from raw BP against
IntCal20 (Reimer et al. 2020), but only after deciding which curve each sample belongs to.
That decision — the reservoir class — is the point of the exercise.
Reservoir classification comes first
You cannot soundly calibrate a marine shell with a terrestrial curve. So each determination is classified from its recovered material and δ¹³C:
| Class | Count | Handling |
|---|---|---|
| terrestrial | 9,280 | IntCal20 ✓ (material terrestrial, or δ¹³C in the C3 range) |
| terrestrial_assumed | 1,864 | calibrated, but flagged — no material and no δ¹³C to confirm |
| marine / marine_suspect | 200 | not calibrated — needs Marine20 + a local ΔR |
| hardwater / problematic / dietary | 290 | flagged for review |
The metadata that makes this possible was itself recovered: material from Chapple’s notes (~80%, where his own column held 1.2%), and δ¹³C from Chapple, the TII monograph appendices, and XRONOS.
The recalibration
For terrestrial samples the age is calibrated against IntCal20 and emitted as highest posterior density (HPD) intervals — one row per interval, at 1σ and 2σ. A single date routinely yields several disjoint intervals across the curve plateaux; collapsing them to a start/end pair would destroy exactly the structure a chronologist needs. 54,287 intervals across 10,257 calibrated determinations.
Validation: UB-4082 (4447 ± 45 BP) calibrates to 3335–2930 cal BC — matching Chapple’s own
published IntCal13 range (3338–2929), the small IntCal13→20 shift, and the multimodal structure
captured as three intervals.
decided_by=rule — it leans on
the machine-inferred material draft. The marine/dietary flags that gate soundness need human
review before analytical use. The calibration arithmetic is correct and reproducible; the
curve-choice flags are drafts.