And what did we do to put it there.
Measuring and monitoring the health of coastal ecosystems.
Nothing here is deployed. Every value is modelled.
A working coast has structure: kelp and seagrass, a food web at its full length, water clear enough for light to get down.
Lose the light and you lose the meadow, whatever else you fix.
Coastal systems differ too much between places for an absolute number to mean much. Health is measured against a nearby site nobody built on.
That reference is the yardstick. Everything else is read as a departure.
Comparably impacted, different mechanisms.
One generalized coast. Five sites, six measures, each read against the reference.
Generalized archetypes, not a real place. Or open it full screen.
Which contaminant you chase decides what you sample.
Levels placed by isotope, transfer measured on every edge.
Every value modelled. Or open it full screen.
Shellfish take the particles. Macroalgae take the dissolved fraction, metals included.
Complementary rather than additive, which is what integrated multi-trophic aquaculture is.
On the generalized coast, both together take the salmon farm from 0.77 to 0.43.
An intervention improves the water by putting contaminants into tissue.
So where that tissue goes, and whether it is clean enough to sell, is part of the plan.
An intervention layer without a tissue layer is half a plan.
A farm, a current, and the band where the crop is worth taking.
Modelled, no site data. Or open it full screen.
The sampler is calculable from physics. The organism is not.
The model fuses samples, sensors and feeds into a value per cell, each carrying its own uncertainty.
Uncertainty is the output that decides where the next sample goes.
Bellingham Bay is the worked example: eight stations, measured depth, a panel protocol, and a regulator with a deadline.
Washington legislated a copper phase-out three times and postponed it three times. It reports again by 30 June 2029. WA Ecology
The apparatus, on one coast.
Drag the chart to move time. Click a hex to read it.
Every value synthetic. Or open it full screen.
A plume dilutes the same way everywhere. What it dilutes into is local.
Every new coast needs its own reference site and its own calibration, and gets the method for free.
That is what predictable means here.
Each phase is priced, and each ends on a criterion written so the answer can be no.
| Phase | What happens | Success criterion |
|---|---|---|
| 0 · desk no field cost |
Literature check, confirm whether the port asset registry exists in usable form, run a context scan against the bounding box, write the evidence ledger. | The registry, context and evidence files exist, and evidence decides whether phase 1 starts. |
| 1 · one station one season |
A single vertical panel array with all four coating treatments, staggered retrievals, imagery, and co-deployed passive samplers with both accumulator classes. Every panel set carries an uncoated control and the incumbent copper coating, so the result is a trade statement. | A calibrated two-group growth curve for one station, and a first answer on hard versus soft accumulators. |
| 2 · four stations the gradient |
The full gradient design, metabarcoding, and the chitosan sampler arm. | Transfer functions with stated uncertainty, and a pressure field that verifies against held-out panels. |
| 3 · product operator trial |
Asset attribution, the decision layer, and an operator trial on the cleaning deferral decision. | One real cleaning decision changed by the map, with the counterfactual documented. |
Two tests, and the first validates the instrument rather than the product.
Declared interest: chitosan is the anchor intervention of this design, and Island Lab's principal led product strategy at Tidal Vision from 2023 to 2026. Reference treatments, blind scoring and symmetric publication are the structural answer. Source: DESIGN.md sections 5.4, 5.5 and 13.
Nothing here is deployed and no client is committed.