Break progression

What the five clocks are (2026-08-20). They are that row's own break event, measured rather than chosen: from the last clock at which the model's whitewater at the tracked crest is still at or below 0.02, to the first at which it reaches 0.60, at the takeoff station — the place where a crest first meets the break line, so where this wave's break starts and its peel begins. The window is found by seeding on the crest nearest the line at the set anchor and walking time backwards until that crest is unbroken and forwards until it is whitewater. Walking backwards is the whole point: the clock this sheet needs is before it breaks, and there is no forward-only way to find that from an anchor that already sits mid-break.

Two things this sheet got wrong first, both recorded because the numbers hid them. (1) The span. The columns used to run one full wave period at T/5. A crest advances exactly one crest spacing per period, so column 5 sat 0.80 of a spacing on — 0.20 from where the next wave upstream had been in column 1, and the sheet very nearly aliased back onto itself. That ratio is not site-specific: advance ÷ spacing = Δt ÷ T cancels the local wavelength, so it was 0.80 on every row, at every camera, at every H₀. It is visible in the sheet's own reads — laid out over a full T the tracked breakpoint goes 36 / 56 / 88 / 4 / 36 at day=small, back on column 1's station to the metre. (2) The anchor, which survived the first fix. The row was anchored on the argmax of crest height at the break line. A wave is tallest at the line, and at the line it is already breaking — so the anchor sat at or after break onset by construction, and the frames showed it: an established whitewater band up-line of the mark in column 1, foam 0.87 in the first cell, and the tracked wave then decaying across the row (crest 2.98 → 2.71 m) rather than breaking. A correct measurement of the wrong instant.

Why the foam is read at the crest and not at a station. At a point break the line is more or less permanently breaking somewhere, so whitewater at a fixed point in the water never goes quiet between waves — it is the bore left by the wave before. That is how a column whose own wave had not broken could report foam 0.87. Every acceptance number on this sheet is read on the tracked crest, so it is this wave's state and nobody else's.

What is on this sheet. Six wave sizes at preset=secondpeak. Four rows are the curated condition bundles from web-three/js/conditions.js: a bundle moves H₀, period, tide and Δf together, the way a real day does. Two rows are h0= only, which moves swell height against the site card's own T = 15 s, tide and Δf — so the pair isolates height from period and tide. Read the bundle rows for "what does this day look like" and the h0 rows for "what does size alone do".

What was left out. day=pulse and day=stormy (the bank has six days; the four sampled span the surf-worthy range and stormy is marked good: false). One site only — cross-site comparison is the other sheet's job. Second Peak is ξ 0.65, so it spills more than it plunges; Sewers (ξ 1.15) is the site to open if you are judging throw specifically. Tide is not swept independently: it rides inside the bundles, which is why the bundle and h0 rows are not directly comparable and are labelled as such.

The foam number can be occluded. It samples the projected attachment corridor whether or not terrain sits in front of it, so a low camera behind a bluff reads low foam for a wave that is breaking fine. The model-side pocket/brk readings in the JSON are the check; a cell is only flagged FLAT when both the pixels and the model say nothing broke.

The marker can be wrong, and it will say so. The line is the tracked crest and the ring is the point on it with the largest pocket — this wave's own breaking point. Both are model reads, not annotations, so each carries a number. ring off line in the row header is the worst distance, over the columns where this wave is actually at the line, between the ring and the break line it belongs on. And where no part of the wave has reached the line yet there is no ring at all, rather than a ring on the previous wave: an empty column 1 is what still unbroken looks like, and it is the single clearest tell that the anchor is now in the right place.

A row that does not break says so. Acceptance is measured per row and printed on it: whitewater at the tracked crest must start pre-break and end broken. Two rows cannot pass, and both are H₀ 0.70 m at a site whose measured peel floor is 1.08 m — day=small peaks at 0.193 and h0=0.7 at 0.341, against 0.858–0.890 for every row that breaks. Their headers say crest intact → crest gone instead of promising a break, their five clocks span the crest indicator's own collapse, and they are kept out of the published set. The gap between 0.35 and 0.85 is empty across the whole bank, which is why the 0.60 line sits in it.

Stills in a known order. Each cell states a position at a known clock; nothing on this page claims a direction of travel from the pictures. The row header's peel is the difference between two model reads at two model-derived clocks, not a motion measured off the frames — which is the distinction MEASUREMENT_LESSONS 1 is about. The frames are ordered by the model's own clock, so the ordering is established rather than inferred, and that is the only reason a reader may read the five states as a sequence.

Where this sheet says n/a, and why. Two measures go blank at Privates, and only at Privates, because that site's coastline defeats the contour fit (16.5 m RMS) and it runs on a synthetic stage rather than a surveyed seabed. foam_pix needs a baked break line to project a corridor onto; with no measured bed there is no baked line, so there is nothing to sample and the cell reads n/a instead of sampling an arbitrary band of pixels. ceilM, the depth-limited crest ceiling in the JSON, is null for the same root cause one step further on: the site runs u_depthMix = 0, the seabed sampler is a 1×1 stand-in, and the depth the shader reads back is the storage format's quantization floor rather than a seabed — so γh never binds and the "ceiling" would be 1.878·H₀ wearing a depth limit's name. Privates has no measured bed and therefore no ceiling to be over. An earlier sheet did print that number and produced a headline defect ("2.2× over its ceiling") that was entirely an artifact of dividing by it. Everything else on the Privates row — crest, foam_model, the set envelope — is measured the same way as every other site and is directly comparable.

Method caveat. Column 1 of every row is captured from a cold load at its own #sim=; columns 2–5 advance the clock with __pointbreak.setSim() plus two rAF ticks — the mode scripts/capture_temporal.mjs validated against per-frame reloads. Each cell's hash reproduces its frame either way.

Camera drift, recorded not assumed. The #aim cameras frame the baked line's action centroid and smooth over ~6 s of sim time, so advancing the clock could move the instrument between columns of one row (MEASUREMENT_LESSONS 11: an instrument that frames itself on the signal is not a fixed instrument). Each row header carries the largest camera displacement between any two of its five columns. A row with a non-trivial drift there is comparing slightly different windows, and its numbers should be read accordingly. The camera legitimately does differ between rows — the break line moves with H₀ — so only within-row drift is a concern.

Stills, not motion. Nothing here is a motion read off the pictures (MEASUREMENT_LESSONS 1). The positions in the cells and the peel in the row headers are model reads at clocks the model itself derived, and the frames are ordered by that clock, so the ordering is established rather than inferred — which is precisely what lesson 1's third failure ("peel direction is rightward, confirmed across two frames") lacked. Read as pictures, these frames still cannot tell you which way the peel runs.

← all QA sheets in this build · raw measurements (JSON) — the same provenance block, machine-readable · the essay · the repo · commit f3b6bde