Explore every drop.

We're building a living early warning system for the ocean, reading the sea's smallest signals in real time, so no coastline is ever caught blind.

Read the water See it in action
scroll
Hundredsof users already contributing
Millionsof plankton images collected
Worldwidefrom coastlines around the globe

Why it matters

The ocean runs on plankton,
and we're watching it blind.

~40%

of the world's oxygen is produced by plankton. They also drive the carbon pump.

70%

of the planet is ocean, and most of it goes unwatched in real time.

Weeks

is how long bloom confirmation still takes, while a bloom can turn toxic in hours.

The vision

A living early warning system for the ocean.

Imagine every coastline monitored in real time: plankton imaged at the water's edge, classified and validated by experts, and checked against known thresholds the moment something shifts. Not sampled once a month and mailed to a lab. Read continuously, everywhere, together.

That's what Pelagica is building: the tools to turn the ocean's earliest, smallest warnings into action.

Early warning

What early warning looks like.

The moment a monitored species climbs past its threshold, an alert goes out: to a dashboard, an inbox, a phone.

Watch0s ago
Pennate diatom
Harbor Mouth · North Coast
Climbing toward the alert threshold
41k
cells L⁻¹ observed
50k
alert threshold
Delivered to EmailDashboardAPI
Advisory3s ago
Heterosigma akashiwo
Bay Inlet · Central Coast
Above the alert threshold
1.4M
cells L⁻¹ observed
1.0M
alert threshold
Delivered to SMSEmailDashboard
Watch9s ago
Tintinnid
Offshore Shelf · South
Climbing toward the alert threshold
38k
cells L⁻¹ observed
50k
alert threshold
Delivered to EmailDashboardAPI

Illustrative preview. These are demonstration alerts built from synthetic sample data to show how the system will work. Not live measurements or official advisories.

Explore

Latest from the coast.

Every observation is an imaged plankton cell, geolocated by the PlanktoScope that captured it and confirmed by experts.

Monitored HAB taxa · live status
Pseudo-nitzschia
Pseudo-nitzschia
detected · 72
Akashiwo sanguinea
Akashiwo sanguinea
detected · 24
Dinophysis
Dinophysis
detected · 116
All taxa 29 Pseudo-nitzschia 10 Akashiwo sanguinea 9 Dinophysis 10
All sourcesBenchmark nodesCommunity
✓ validated
Pseudo-nitzschia
Pier 9 SF · Benchmark node
86%
✓ validated
Dinophysis
Monterey Bay · Community
92%
✓ validated
Akashiwo sanguinea
SF Bay · Community
78%
✓ validated
Pseudo-nitzschia
Bodega Head · Benchmark node
88%
✓ validated
Dinophysis
Pillar Point · Community
84%
✓ validated
Akashiwo sanguinea
Half Moon Bay · Community
80%
✓ validated
Pseudo-nitzschia
Trinidad Head · Benchmark node
91%
✓ validated
Dinophysis
SF Bay · Benchmark node
83%
Click any cell to see where it was captured ›

Illustrative preview. The imaged cells are real; counts, locations, and validation states are sample data.

Ground truth

Satellites see the bloom.
We see what's in it.

Every PlanktoScope sample sits beside live satellite chlorophyll (NASA GIBS) and CeNCOOS stations, so a real cell count can confirm, or correct, what the satellite sees from orbit. The map loads external data only when you scroll to it.

6 monitoring stations across the network. Live satellite chlorophyll © NASA EOSDIS GIBS; basemap © OpenStreetMap, CARTO. Sample points are illustrative.

Compare

How does your water compare?

Your readings beside a benchmark node for the same waters, over the last two weeks.

Pennate diatom Dinoflagellate Heterosigma akashiwo Tintinnid
Benchmark node Your readings advisory level
80k 40k 0 cells L⁻¹ advisory 50k 14 days 7 days today

Illustrative comparison. Synthetic sample data.

How it works

Every confirmation sharpens the shared eye.

A shared model proposes what each imaged cell is; experts confirm or correct it; those corrections train the next model for every instrument on the coast.

27k
plankton cells classified
40%
expert validated by scientists
85%
mean model confidence
v2.3
current shared model
1

Observe

Contributors and benchmark nodes image real plankton, drop by drop, across the coast.

9 contributors · 6 nodes
2

Classify

A shared model proposes a taxon and a confidence for every imaged cell, the same model for everyone.

291 awaiting review
3

Validate

Experts confirm the right calls and correct the wrong ones. Each decision becomes a labelled example.

192 labels · 3 experts
4

Improve

Those labels train the next model. Sharper detection returns to every instrument on the network.

32 corrections fed back

The loop closes: a better shared model raises every future reading's accuracy. Today it classifies and validates. Over time, the same growing record becomes forecasts that anticipate blooms before they start.

Counts are illustrative. The validation states, confidence, and model version are the real fields the platform will track.

The instrument

A whole ocean lab in a shoebox.

PlanktoScope is an open source plankton imager with onboard AI. Field ready, lab grade, and affordable enough to put on every dock and every hull, it's the sensor that turns a vision into a network.

A PlanktoScope imaging a live water sample
A PlanktoScope imaging a live sample.

Who it's for

For everyone who needs to know what's in the water.

The people and institutions that act on what the ocean is doing, long before it reaches the shore.

Aquaculture Fisheries Researchers & agencies Coastal communities Climate science Water quality

The field

Advancing ocean observation, alongside the world's leading institutions.

MBARI EMBL ETH Zürich WHOI NOAA National Oceanography Centre Max Planck University of Oslo Columbia University NIOZ Sorbonne Université

Get involved

We're building this in the open.

Follow the work, or reach out if you want to help make monitoring at ocean scale real.

Get in touch Back to top