The project currently encompasses over 14,000 species and 21,000 documented biotic interactions, mapping the complex food webs and ecological dependencies that sustain marine life worldwide.
Methodology: bridging data and discovery
Recognizing that existing ecological data is inherently incomplete due to real-world sampling limitations, the project integrates two core components:
- Empirical Data Integration: We rigorously combine interaction data from authoritative sources, including the Global Biotic Interactions database (GloBI) and FishBase, to capture existing, verified trophic relationships.
- Predictive Modeling with AI: We apply advanced Graph Neural Network (GNN) algorithms—a form of machine learning—to systematically infer and predict missing interactions. This inference is guided by verified relationships, species traits (like size and diet), and taxonomic distances.
This approach generates a robust and complete representation of marine food webs, moving beyond the limitations of raw data to provide a powerful tool for discovery.
Impact and expertise
This work fuses expertise in marine ecology, network science, machine learning, and biogeography. By combining ecological network theory with cutting-edge computational methods, the Metaweb provides a critical foundation for:
- Understanding Community Resilience: Addressing fundamental questions about how marine communities are assembled, how species coexist, and what drives ecosystem stability.
- Ecosystem-Based Management: Offering a platform to predict how environmental changes (like climate change or fishing) propagate through food webs.
- Targeted Conservation: Supporting the identification of ecologically critical species and vulnerable interaction pathways vital for effective marine conservation planning.
Ultimately, this project is creating the missing map required for a data-driven, ecosystem-based approach to managing the health and stability of our global oceans.