Layer model — two passes and a ledger
Generated by build_layer_model.py. Pass 1 is the defensible model. Pass 2 is everything admitted on an assumption, and every assumption is in the ledger.
Generated by build_layer_model.py. Pass 1 is the defensible model. Pass 2 is everything admitted on an assumption, and every assumption is in the ledger. Subtract pass 2 and pass 1 is recovered unchanged.
Pass 1 — the confident model
Built from 10 of 13 records: complete reads, unambiguous, target unit correctly identified.
| contribution | n | records |
|---|---|---|
| top of Sherwood known | 2 | ST13SE1 (28.6 m), ST12NE6 (83.0 m) |
| top bounded below | 4 | ST22NE88 (>50 m), ST22NW7 (>78 m), ST22NW5 (>59 m), ST32SW1 (>32 m) |
| Sherwood absent | 4 | ST24SW1, ST23NW8, ST22NE81, ST23NW6 |
A top surface cannot be fitted from this. It needs points where the top is known and there are 2. What pass 1 does establish is a presence/absence domain: proven present in the west, proven absent over the Cannington basement high, and everywhere near the store only bounded from above.
Pass 2 — records set aside, re-examined
ST34SW98 — ADMIT-WEAK
Held back from pass 1 because: partial read - the pages read do not cover the depth its bound would imply.
Contributes ‘Mercia Mudstone cover present at this location’ and nothing about depth. The bound implied by its total depth is NOT used.
ST01SE18 — ADMIT-WEAK
Held back from pass 1 because: top_kind ‘ambiguous’ - not a clean observation of the target unit.
Annotations run in correct stratigraphic order (Pebble Beds > Lower Marls > Lower Sandstone), so they are one coherent reading rather than two competing ones - each is queried, but they do not contradict each other. Nearest pass-1 control is ST12NE6 at 16 km, which does not conflict. Admitted as a THIN-PRESENCE point at low confidence and lower resolution.
ST12NW18 — ADMIT-INFERRED
Held back from pass 1 because: top_kind ‘wrong_unit’ - not a clean observation of the target unit.
The logged section is Aylesbeare Mudstone / Exeter Group - both stratigraphically BELOW the Sherwood Sandstone. For a lower unit to sit at outcrop the Sherwood must be absent here, eroded rather than buried. That is an absence observation, and it agrees with the three basement points in pass 1 rather than contradicting anything.
Ledger
Every assumption applied in pass 2. Nothing else was assumed.
| # | record | assumption | effect on the model | conf | reversible by |
|---|---|---|---|---|---|
| 1 | ST34SW98 |
Treat a partial read as evidence of lithology present on the pages read only; discard any depth bound derived from total depth. | Adds a presence-of-cover point. No depth constraint added. | low | Remove the point; pass 1 unchanged. |
| 2 | ST01SE18 |
Where every formation call in a log is queried but the calls form a correctly ordered sequence, accept the sequence and not the individual identifications. | Adds a second point where the target is PRESENT, thin (~17 m of basal Pebble Beds). Top depth carried at low confidence. | low | Remove the point; pass 1 returns to one penetrating observation. |
| 3 | ST12NW18 |
A formation stratigraphically BELOW the target, logged at or near outcrop, implies the target is absent at that point. | Adds one ABSENCE point. Does not add any depth to the top surface. | medium | Remove the absence point; pass 1 unchanged. |
What pass 2 changed
- 3 of 3 held-back records admitted, all at reduced confidence.
- Points where the target is PRESENT: 2 → 3 (the added one is thin and low-confidence).
- Points where the target is ABSENT: 4 → 5.
- Depth constraints added: none. Every pass-2 admission is a presence/absence statement, not a depth. The top surface is no better constrained than it was after pass 1.
That last line is the important one. Pass 2 widened the domain the model covers without improving the resolution of the surface, which is exactly what should be expected from records that were set aside for being doubtful.