Emulating complex causal socio-ecological models in digital twins of ocean

Informacje na temat projektu

ECOTWIN

Identyfikator umowy o grant: 101157369

DOI

10.3030/101157369

Data podpisania przez KE 21 Maja 2024

Data rozpoczęcia 1 Września 2024

Data zakończenia 31 Sierpnia 2027

Finansowanie w ramach

• Climate, Energy and Mobility

Koszt całkowity

€ 2 979 801,25

Wkład UE

€ 2 979 801,25

Koordynowany przez

THE PROVOST, FELLOWS, FOUNDATION SCHOLARS & THE OTHER MEMBERS OF BOARD, OF THE COLLEGE OF THE HOLY & UNDIVIDED TRINITY OF QUEEN ELIZABETH NEAR DUBLIN
Ireland

Periodic Reporting for period 1 - ECOTWIN (Emulating complex causal socio-ecological models in digital twins of ocean)

Okres sprawozdawczy: 2024-09-01 do 2026-02-28

Podsumowanie kontekstu i ogólnych celów projektu

To fully realise the potential of the European Digital Twin Ocean (DTO) to empower citizens, entrepreneurs, scientists and policymakers with accessible knowledge about ocean and coastal domains, new models, tools and methods are needed to understand and assess the complex interactions of social, economic and ecological factors. ECOTWIN delivers innovative graph theoretic solutions, developed using a multi-actor approach with policy makers, local communities, marine and coastal industries and communities, to address the challenges of modelling marine socio-ecological systems and integrating these models into the emerging DTO framework. Innovations include four different classes of novel socio-ecological models based on concepts from mathematics (graph theory), control engineering (stability), computer science (generative AI), statistics (Bayesian theory) and information science (entropy), combined with dynamic participatory feedback from stakeholders, and methodological protocols enabling seamless integration with DTOs by accounting for the interoperability, accessibility, reliability and sustainability of transdisciplinary data. The models comprise: 1) quantitative causal graph theoretic models, 2) qualitative causal graph theoretic models, 3) network models
with participatory feedback and 4) parallel generative AI models. Through validation in four use cases in the North Sea (Belgium, Germany), Celtic Sea (Ireland), Thracian Sea (Greece) and Waterford Harbour (Ireland), tools will be developed for assessing scenarios including the relationships between factors including ecosystem services and health,
geopolitical factors, fishing job losses, tourism and marine renewables development. Results will enable what-if scenario analysis and impact assessments, exploring the interconnectedimpacts of environmental changes, human pressures and policy implementation to empower better decision-making to support ocean health social prosperity and a thriving blue economy.

Prace wykonane od początku projektu do końca okresu sprawozdawczego oraz najważniejsze dotychczasowe rezultaty

The following work have been performed as part of the technical WPs 2-5:

1. Developed data and AI driven models (RECLASS, TEAM, MST GEM) to learn spatio temporal ecological features and construct causal dependency graphs from ECOSMO simulation data.
2. Developed modelling framework and applied ECOSMO E2E ecosystem simulations to assess offshore wind and fisheries impacts, supporting integrated socio ecological analysis.
3.Integrated oil spill simulations, ecosystem modelling (Ecopath/Ecosim), and machine learning workflows to analyse fisheries, climate impacts, and marine ecosystem dynamics.
4.Adapted and calibrated EwE ecosystem model to represent Belgian fleet dynamics and fishing impacts in the southern North Sea.
5. Developed a calibrated Ecopath–Ecosim–Ecospace socio ecological modelling system for the Thracian Sea, integrating environmental, fisheries, and spatial data for scenario analysis.
6. Established a unified multi class socio ecological modelling framework integrating ecosystem, socio economic, and AI based components across use cases.

The main achievements are:
1. Developed data and AI driven causal graph models for marine ecological analysis.
2. Generated spatio temporal feature representations enabling ecological dependency analysis.
3. Developed participatory modelling framework for stakeholder co-design processes.
4. Simulated offshore wind impacts using integrated ecosystem modelling.
5. Integrated oil spill and ecosystem modelling for Thracian Sea analysis.
6. Developed Python workflows enabling reproducible ecosystem modelling and scenario analysis.
7. Adapted EwE model to represent Belgian fleet dynamics.
8. Improved Ecosim calibration fitting historical biomass and catch data.
9.Developed an integrated socio ecological modelling framework with ECOSIM and Ecospace.
10. Enabled cumulative impact analysis of fisheries, energy and environment.
11. Established unified multi class socio ecological modelling and integration framework.
12. Developed AI enabled tools for interpreting complex socio ecological model outputs.

Innowacyjność oraz oczekiwany potencjalny wpływ (w tym dotychczasowe znaczenie społeczno-gospodarcze i szersze implikacje społeczne projektu)

The following results have been achieved beyond the state of the art:
1. Developed a data and AI-driven analytical pipeline (WP2) that complements mechanistic frameworks (ECOSMO, ECOSIM, ECOPATH) by transforming simulation outputs into reusable learned representations rather than rerunning costly process-based models. Key beyond-state-of-the-art achievements include reduced computational burden versus repeated mechanistic runs, robust generalisation with minimal hyperparameter tuning across scenarios, and — most significantly — learned representations that directly enable causal discovery (DAG construction) and socio-ecological coupling, shifting the paradigm from simulation-driven to representation-aware, causally interpretable marine analysis.
2. Conducted preliminary scenario analyses for Use Case 2 (North Sea) using the ECOSMO-E2E framework, representing offshore wind farms (OWFs) as external atmospheric forcing alongside fisheries displacement. The beyond-state-of-the-art contribution lies in new graph-theoretical representations of marine and socio-ecological systems that enable systematic coupling between physical–biological dynamics and socio-economic risk assessment (linking with WP3), bridging quantitative simulation with qualitative stakeholder knowledge.
3. Integrated oil spill simulations with Ecopath-based ecosystem modelling into a unified workflow for the Thracian Sea, resolving biomass, trophic interactions, and fisheries exploitation across 40 functional groups. The principal methodological advance is a Python-based interface for EwE models that replaces manual graphical workflows with reproducible, script-driven execution, enabling systematic scenario analysis, sensitivity testing, and direct coupling with external data and oil-spill hazard assessments.
4. Constructed a multi-class socio-ecological modelling framework that simultaneously evaluates offshore energy development, fisheries activity, and environmental disturbances within a single analytical structure — capturing both direct and indirect trophic and spatial interactions. The calibrated Ecosim model combined with a fully implemented Ecospace environment provides novel capability to assess long-term ecosystem change and spatial redistribution of fishing effort under complex management and infrastructure scenarios.
5. Delivered the foundations of EcoTwin under WP5, advancing the state of the art in four ways: (1) a unified multi-class socio-ecological modelling framework integrating quantitative, qualitative, participatory, and AI-driven methods in one workflow rather than treating them separately; (2) a causal graph-based abstraction layer bridging detailed ecosystem simulations and socio-economic analysis to support cross-domain reasoning and scenario evaluation; (3) a framework for AI-enabled interaction with complex socio-ecological models — including natural-language querying, automated narrative generation, and scenario exploration for non-expert users (implementation planned for phase two); and (4) explicit structuring of operationalisation and usability as core design objectives, positioning AI as an interpretability and interaction layer for policymakers within the European Digital Twin of the Ocean.