A fisherman, helped by a director, runs a survey, which makes a local map, which becomes a sovereign model, which enables monitoring, which enables stewardship. This page is the parts of that problem, the people it is for, the tools each of them actually touches, and how that changes across six stages.
v0.1
Every step in the chain has to be worth doing on its own, because nobody will run step one for the sake of step five. So the test this map applies is narrow and unkind: for each actor, name the surface they open and the one thing they get. Where an actor touches nothing, they are an audience, not a user. Where only Island Lab touches a surface, it is our tool and not theirs.
00 Read one stage at a time
Selecting a stage dims the actors and the tools that stage does not touch. The whole arc is the default, and nothing on this page is hidden behind the buttons.
01 The problem, stated as problems
Age
The known Posidonia map for Pollica is EMODnet Seabed Habitats, sourced from ISPRA, surveyed 2017-11-29. Every decision about the meadow rests on a picture that is nine years old.
Cost
Re-surveying professionally is expensive and happens rarely, which is the reason the picture is nine years old rather than an accident of it.
Custody
The data that does exist is not held locally. Nobody has packaged the tool, the method and the ownership into something a comune can run and keep.
Cadence
There is one snapshot and no schedule after it, so change in the meadow is invisible. One survey is a picture; it takes two to see a direction.
Standing
A credit or policy claim rests on measurement that is independent, repeatable and not marking its own homework. An assertion by the party that benefits is not that.
02 Who it is for, and which surface they open
Actor → the one thing they get
Periplus
Portolan
Sonar planner
Triton
Reckoning
Tessera
Proteus
Steward
Kairos
Pharos
Helm
Touches
The fishermanoperator
This is what I covered today, and this is what it was worth.
2
The directorassigns, checks, signs off
I can hand a line to someone with no training and get back data I am willing to put my name on.
3
The communityholds the record
The thing that describes our sea belongs to us.
3
The municipalitythe sovereign
Evidence we control, on a schedule we set.
2
The scientistpublishes against it
Data honest enough to publish against.
3
The credit buyer or investorbuys the claim
The measurement is not marking its own homework.
1
Island Labmethod, tooling, QC
The protocol is the product, and it travels to the next coast as a config file rather than a rebuild. Does not own the data.
11
operates it reads its output registry status: exists mvp or experimental unregistered, or does less than it says
Twenty-five touches in the matrix. Eleven of them are ours. The six actors the product exists for share fourteen between them, and the credit buyer, whose scepticism the whole arc is built to satisfy, opens exactly one surface. That asymmetry is the finding, not the layout: a product whose density sits in the supplier's column is still a service, and the PRD's claim that the community owns the record is not yet visible in the tools they touch.
The console. Hosts every app, shows their data, and tracks each dataset's path from synthetic stand-in to calibrated local truth.
04 Six stages, and the room that reviews each one
Stage 0The session objectNo demo
Nothing today ties an outing, a vessel, an operator, a planned line, an achieved track and the readings into one thing. Schema and plumbing, and the only stage with no demo, which is exactly why it is easy to skip. Done when a session round-trips to disk and back with its provenance intact.
Lights upPeriplus
Stage 1The demo, on synthetic dataRoom: director + fisherman
The simulated run starts producing a session, and gains the screen the whole story rests on: what I scanned, against what was known. Three panels in the order a person asks for them, with cells nobody visited shown as unsurveyed rather than as agreement. Done when a fisherman watching it says that is roughly my day.
Lights upPeriplusPortolanTessera
Nothing in the tree compares an observation surface to the ISPRA extent cell by cell today. This panel is the stage.
Stage 2Real data, which means automatic labellingRoom: scientist + director
Today the trust pipeline begins with a human typing two columns into a spreadsheet. That is the single hardest blocker, because the premise is an untrained operator whose data flows through without a specialist in the middle. Inference replaces the typing, and every reading carries a confidence instead of a silent verdict. Done when an outing goes device to per-cell map with no hand-editing, and the director can override any label.
Lights upPeriplusTritonReckoning
The labelling this stage needs does not exist. A person types label and in_meadow into a CSV.
Stage 3The sovereign modelRoom: mayor + community
The comparison surface becomes a held record: versioned, immutable once published, archived by the municipality, re-runnable next season against the same lines. Better surveying has to visibly improve the model, or the operator has no reason to survey well. Done when a published session cannot be silently altered and a second survey of the same lines produces a difference.
Lights upProteusTesseraHelm
There is no archive surface in the constellation. What the community owns, and under what licence, is written in no document yet.
Stage 4MonitoringRoom: scientist + credit buyer
Two surveys make a trend. The product stops being a picture and becomes a watch: what changed, where, by how much, with what confidence, measured against ourselves last season rather than against 2017. Done when a cell that changed for a boring reason (worse weather, a faster run) is distinguishable from one that changed because the meadow did.
Monitoring that nobody acts on is a hobby. What the monitoring found gets routed to whoever can act on it: a no-anchor zone proposal, a restoration priority, a change of fishing practice, a budget line. The fleet moves from extractors to stewards because the data they produced is what justifies the action, and they can see that.
Lights upKairosPharosStewardHelm
05 Built, designed, absent
Built and running
Periplus simulates a real working day over real geography: speed, safe speed by depth, fuel burn, sounding quality degrading with speed and heave, coverage accumulating.
Portolan plans and exports fleet routes that avoid land and steer toward coverage gaps.
Triton reports a live fleet of buoys: position, health, latest readings.
Reckoning normalises public feeds and fuses them with the sensor readings.
The trust machinery. Readings are down-weighted rather than discarded, weak singletons are flagged as needing replication, and a bad reading collapses toward zero instead of poisoning the mean.
Designed, not built
Side-scan sonar. Designed and costed. Getting there is a hardware purchase plus a log reader that does not exist.
The session object. Stage 0 is the schema that ties an outing to its track, its readings and its provenance. It is a week of work and nothing after it starts without it.
Per-cell twin. Stage 3 at the earliest. The honest interim is a per-cell observation surface, which is not a simulation.
The sovereign archive. Versioned, immutable once published, held by the comune. No surface implements this.
Absent
Sonar ingest, anywhere. No echosounder, no log reader for any sonar format. The firmware carries three sensors: water temperature, light, GPS. The survey we can honestly run today is temperature and position.
Automatic labelling. A human types label (water, air, bad) and in_meadow into a CSV. Stage 2 cannot start until that is inference.
Cell-level comparison to ISPRA. Nothing currently sets an observation surface against the 2017 extent cell by cell.
Spatial twin. The Posidonia model is a 0-D box model, and the current coupling reduces a whole outing to one number.
Persistence in Periplus. The run evaporates when the tab closes.
What this map does not settle
Three of the eleven surfaces claim survey territory and one of them is not in the registry. The sonar planner lights up at no stage on this page, because until the consolidation decision is made there is no honest answer to which of Portolan, Periplus and the planner a director should open first. That decision is due before stage 1, not after it.
Two more open questions sit under the matrix rather than in it. The director is a named role with sign-off authority in the PRD and a hardcoded provenance string in the code. And sovereignty is claimed in our materials without a document saying whether the community owns the raw logs, the derived maps, or both, and under what licence. Neither is a technical problem, and neither is solved by the tools shown here.
We are not claiming the 2017 map is wrong. We are claiming it is old, and that a community can now check it. Those are different statements and only the second one is ours to make.