Digital Twin Ocean (DTO)

Imagine you are a marine protected area manager. You receive an alert: water temperatures near your artificial reef have risen three degrees above normal. What happens next? Do you wait for monthly reports? Do you send a dive team next week?

With a Digital Twin Ocean, you open your laptop. You see live temperature data streaming from sensors on the reef. An AI model shows you a forecast for the next 72 hours. You run a “what if” scenario: If the heatwave continues, how will oxygen levels change? Within minutes, you decide to adjust fishing restrictions and alert local fish farmers.

That is the promise of a Digital Twin Ocean (DTO). It is a virtual representation of a real-world marine ecosystem that stays synchronised with its physical twin through continuous data streams. Unlike a static map or a one-off report, a DTO lives and breathes with the ocean. It combines four essential layers:

The EU Mission Ocean has two phases: first, development and piloting (by 2025); second, deployment and upscaling (2026–2030). PHAROS sits exactly in between. The DTO built in Gran Canaria and Iceland is not a one-off experiment. It is designed to be replicated – a blueprint that other Atlantic and Arctic regions can adopt.

Too many projects collect excellent data that no one can find or reuse. PHAROS committed to the FAIR principles from day one: Findable, Accessible, Interoperable, Reusable. Every dataset (from nutrient sensors to eDNA samples) follows the same metadata standards, uses the same vocabularies, and is published through EMODnet and the European Digital Twin Ocean (EDITO) platform.

The DTO is not a final report you file away. It is a continuous feedback loop: Sensors feed data into models, models generate predictions, managers take action, action changes the real ocean. New data confirms whether the action worked, and the loop begins again. This is adaptive management at scale.

All harmonised data lives in a hybrid cloud storage system managed through the EDITO platform. The architecture has three tiers:

• Raw tier: Original, unprocessed data – Auditing, reprocessing
• Processed tier: Cleaned, validated, flagged – Daily analysis, visualisation
• DTO ready tier: Aggregated, model compatible – AI training, scenario testing

Backups occur weekly. Versioning is enabled for every file. Role-based access control ensures that sensitive data (such as real-time locations of protected species) is only visible to authorised users.

This is where the DTO becomes intelligent. PHAROS uses three complementary modelling approaches:

1. Numerical models (MOHID for hydrodynamics and biogeochemistry, WW3 for waves, RAINDROP for underwater acoustics). These simulate physical processes based on the laws of physics.

2. AI-driven detection models (Hydrotwin for acoustic classification, SmartFISHER for visual species identification). These are pre-trained on large datasets and continuously fine-tuned with project-specific data.

3. Surrogate models (neural networks trained on sensor time series). These are fast, lightweight approximations of the full numerical models. They run in seconds instead of hours – perfect for real-time forecasting.

Data assimilation techniques (specifically, Ensemble Optimal Interpolation) combine sensor measurements with model outputs to correct biases and improve accuracy.