Sämbosjön source age bounds against fixed pollen bins¶
The source-bound report, 111 selected sample rows, 22-bin coverage ledger and boundary ledger compare the selected Sämbosjön calibrated sample ages with their own reported younger and older bounds. This is an age-only diagnostic for the existing direct observed-pollen result. It does not open pollen observation parts, change the frozen predictions or recalculate scores.
58 of 111 sample intervals have positive-width overlap with two nominal 500-cal-BP bins; the other 53 lie wholly within their point-age bin. Each of the 22 scored nominal bins contains at least one boundary-crossing sample. All 21 internal boundaries, 1,000–11,000 cal BP, are crossed by at least one reported interval. No interval overlaps more than two bins or extends beyond the existing 1–22 bin range. Nominal bin 1 has three samples and none has its entire reported interval inside that bin. The bin ledger retains the point-age sample count, the number of crossing and wholly contained nominal samples, and the count of intervals that could overlap each bin. These overlap counts are separate marginal possibilities; they are not a simultaneous reallocation of the 111 samples.
For an interval [younger, older], a bin counts as overlapped only when their
intersection has positive width. Touching an older bin boundary at one point
does not count as crossing it. The source's point age defines the nominal bin;
the younger and older values define overlap. Every selected point age lies
within its reported bounds. The retained chronology is the source-default
calibrated Giesecke et al. 2014 (MADCAP) smooth spline based on IntCal09.
Its sample claims and chronology notes give no confidence level for these
bounds, and no joint age posterior was retained. This calculation therefore
does not assign probabilities to neighboring bins or invent new age draws.
The fixed pollen score used point-age bins: 22 bins were scored, with IDW MAE 0.022991 on the declared 16-label subcomposition. Because all 22 bins contain at least one boundary-crossing sample, their membership is conditional on the selected point ages. The interval-overlap screen alone cannot tell how a different age assignment would change each bin's pollen composition, positive count support or MAE. The machine report binds the selected sample and chronology identities to the retained Neotoma manifest, age-claim part, direct-pollen input receipt, prediction freeze and unchanged score files.
Replay from the repository root into an absent directory under artifacts/: