Inversion thesis: crisp water tokens, vague water kinds

Reading findings that refine the hydrology-as-granular-partitions thesis #AHZGJ5

1. Inversion thesis: mountains have solid kinds but vague token boundaries #H9QC2R #HXHZA9 water bodies have crisp token boundaries (waterline = minimax spill level) but a messy kind taxonomy — Smith & Bittner’s own dictionary-derived water-body partition has cycles and duplicate ‘lake’ cells, so location is not functional #V48MYV #XWJ9VN contrast with well-ordered spatial partitions #27XPCS Algorithms solve token demarcation; kind assignment (lake vs pond vs tarn) reinherits lexical vagueness #PMUYXM

2. Physical grounding of crispness: water collects in concavities and forms a barrier to movement #A92K4K perception objectifies convexities #FXBZZX so concavities become object-like only when ponding converts the graded concavity into a sharp bona fide surface. A lake is a fiat parsing whose boundary placement is outsourced to gravity — the strongest form of the canonicity claim.

3. The basin merge tree doesn’t just satisfy MB3/DR3 (subcell → parthood, #UCRZGG #2BXUD4 — it saturates it: every subcell relation records a real sub-catchment parthood, unlike typical partitions that trace over parthood #U7W64E Minimal cells are atoms only relative to the partition #GDCDRU persistence-threshold refinement is governed by the logic of partition systems #4QQD4A

4. Catalogue of residual fiat choices, ontologically heterogeneous: (a) routing operator SFD/MFD #T73CNF — the only one threatening canonicity, though the filled surface z* is routing-independent; (b) persistence threshold = granularity proper; (c) flat/saddle tie-breaking = fiat demarcation in a bona fide corridor; (d) kind assignment = lexical partition problem.

5. Empirical support: mountains absent as objects from geographic databases #H9QC2R but lake polygons are standard in hydrography datasets — GIS practice already reflects the computable/non-computable boundary asymmetry.