Growers, farmers and harbours

How the signals work

The five signals, what each reads, the difference between levels and activity words, how confidence is set, and why a model value may come from a cell several km from your site.

Last reviewed 31 Oct 2026

The five signals

Each site page shows five signals, worked out every day from open ocean-model, satellite and weather data, plus your own logger where you have one. They are computed by fixed rules and published methods. No language model is involved in computing them.

Signal	What it reads	What it reports
Operability	Forecast wave height (with period and direction where your activity is swell-limited), wind and the predicted tide, against the limits you set for each activity, in 6-hour steps over the next 7 days (days 6 and 7 always low confidence)	Favourable, marginal or unfavourable for each named activity, and the condition that decided it
Thermal stress	Sea temperature near the surface: your own logger where you have one, else the regional ocean model, against the thresholds of each species you stock	A level (low, watch, stress, critical) per species, and dated marine heatwaves
Bloom risk	Satellite and model chlorophyll-a, the sea-temperature anomaly, stratification, calm days and catchment rain	A regional 0 to 100 index with a level (low, watch, stress, critical), always marked “bay not verified”
Runoff shock	The 48-hour drop in model surface salinity and the 24-hour rain over the bay's catchment, checked against Sentinel-2 turbidity when a cloud-free scene exists	A level (low, watch, stress, critical) and the salinity drop in psu
Growth	Sea temperature against the species' optimum, daylight, day length and, for seaweed, nitrate	A 0 to 1 growing-conditions index and the season stage

Levels and activity words

Two kinds of words appear on the cards, and they mean different things.

Favourable, marginal, unfavourable are only ever said for an activity you named, such as longline work or trestle access. They come from the operability signal and your own working limits.
Low, watch, stress, critical are levels of a condition: sea temperature against a species threshold, a salinity drop, or the bloom index. A level is not a verdict on any activity.

The coloured dot on a card uses the same three colours for both, so the page reads at a glance: low shows green, watch amber, stress and critical red. The words on the card always say which scale it is.

Thresholds start from published literature where we found a source, and otherwise from starting values that are awaiting specialist review. You can set your own for a site in the limits editor; every change is recorded with its old and new value.

The bloom index

Conditions that favour phytoplankton growth, regional model, not a measurement of toxic species. Bulk chlorophyll shows conditions that favour phytoplankton growth; it cannot see low-biomass toxic species. So the bloom card says its own level and never favourable, marginal or unfavourable, and it is never advice for harvesting.

For shellfish harvesting and food safety, follow the official closures and the Marine Institute HAB bulletin. Nothing in this app replaces them.

Confidence

Every signal value carries a confidence of high, medium or low. It starts from how far ahead the value is:

up to 48 hours ahead: high; up to 96 hours: medium; further ahead: low;
sea temperature from your own logger is high.

It is then lowered, never raised, by what we know about the model at your site:

If nothing verifies the model at the site, confidence is at most medium. Your own logger always verifies it. A buoy or station verifies it only within 15 km of the site; farther away the card says unverified at this site and no bias correction is applied.
If the model has been far from the buoy or logger over the last 30 days, it drops one step.
For operability and growth, a missing input drops it one step.
Runoff is at most medium unless a Sentinel-2 scene confirmed the event.
If the model cell used for the site is displaced (see below), it drops one step.

The tooltip of the confidence badge on a card names what verified the value.

Model cell distance

The regional ocean model works on a grid of cells about 3 km across. Near the shore, or inside a narrow inlet, the cell covering your site may be land in the model, so the app reads the nearest sea cell, or a sea point configured for the site. That cell can sit several km seaward of your site. The card then shows a line such as Model cell <distance> km seaward, <depth> m deep, unverified at this site, filled in with the distance and depth of your site's cell.

When the cell is more than 5 km from the site, comes from a configured sea point, or lies more than 2 cells from the point searched, confidence drops one step. Only your own logger and records describe the site itself.

Scale: regional, satellite, site

Every value says where it comes from:

regional: ocean, wave and weather models, describing a model cell, not your exact longline;
satellite: observed from space, with gaps on cloudy days;
site: your own logger and records.

More on the assistant, citations and the AI notice: How it works. Where each input comes from: Data sources and licences.

Everyone

Data sources and licences

Where every value comes from, at which scale, and the attribution each licence requires when you reuse the data.

Last reviewed 31 Oct 2026 Agencies, research bodies and insurers

Agency data access

How agencies, research bodies and insurers request scoped, read-only access, the MCP and REST endpoints, the tools, and the limits and audit log.