Build a non-degenerate (shear-gated) fragmentation module

[awickert]

Motivation

We cannot naively "turn on fragmentation." The model has a fragmentation(g)
process (constant per-km rate; currently always off), but enabling it as-is is
both physically wrong and partly unidentifiable. The near-source component of
fragmentation is already absorbed by the fitted source size D0 (see #1). This
issue is to build the part that is identifiable and physical.

What is identifiable: distributed, shear-gated fragmentation

Fragmentation is a distinct, fittable process only if it is distributed along
the transport path — count growing / size shrinking with distance beyond what
abrasion explains (a flat-everywhere D0 cannot mimic this). The flume data say
it is:

• shear-threshold-like — ≈0 at 52 Pa (small flume), present at ~268 Pa (large
  flume); not a clean power law; and
• angularity-controlled — angular pebbles fragment, well-rounded ones do not.

So the physical parameterization is g active only where local shear
τ(x) > τ_crit (the steep, high-energy upper reaches) and off downstream as clasts
round and τ drops.

Sequencing / preconditions

This is deliberately not next. Build only after, in order:

1. Document model identifiability: near-source fragmentation is already captured by D0 #1 done — the D0 / near-source identifiability documentation.
2. Establish that distributed fragmentation actually happens. A go/no-go
   diagnostic: is there a downstream distance signature in count/size beyond what
   abrasion + D0 already explain? If the residuals show no distance-distributed
   fragmentation, there is nothing here to build — D0 (Document model identifiability: near-source fragmentation is already captured by D0 #1) is the whole story.
3. Only if (2) is positive, build it — and it must manage real hydraulics:
   a per-reach shear field τ(x) along source→site paths (slope × depth via the
   hydraulics module, over the channel network), with a τ_crit prior bracketed by
   the flume 52 Pa (off) / 268 Pa (on).

Proposed work (once unblocked)

1. Design a non-degenerate, shear-gated parameterization: g(τ(x)) switching on
   above τ_crit.
2. Flume g-prior extraction: per-lithology fragmentation magnitude from the
   "fragment-inclusive minus abrasion-only" mass difference (granite ≈ 0; angular
   vs. rounded handling documented).
3. Synthetic identifiability tests: confirm shear-gated g is recoverable
   jointly with D0 / production / abrasion given the count + size channels and a
   spatial shear field — and document the regimes where it is not.
4. Integrate with the per-reach shear field.

Dependency (blocked)

Needs the per-reach shear field τ(x) (real hydraulics over the channel
network) — the transport-path / channel-network effort. Until then, fragmentation
stays off and its near-source component lives in D0 (#1).

Relationship

• Builds on / cross-references Document model identifiability: near-source fragmentation is already captured by D0 #1 (the identifiability documentation).
• Consumes the spatial shear field from the channel-network / transport-path work.

🤖 Generated with Claude Code