Using the Semantic Web to Integrate Ecoinformatics Resources

Cynthia Sims Parr, Andriy Parafiynyk, Joel Sachs, Rong Pan, Lushan Han, Li Ding, Tim Finin, David WangAAAI 2006

Introduction

Invasive species pose a significant threat to global biodiversity and local economies. When a new species enters an ecosystem, ecologists must quickly determine its potential impact: what will it consume, and what native predators might control its population? Unfortunately, empirical food web studies are resource-intensive and expensive, leaving many ecosystems "dark" to researchers. The SPIRE project addresses this by leveraging Semantic Web technologies to integrate disparate ecological datasets and predict species interactions in previously unstudied environments.

The SPIRE Framework: ELVIS and the Triple Shop

The core of the project is the Ecosystem Location Visualization and Information System (ELVIS). ELVIS is a suite of web services designed to construct species lists for specific geographic locations and synthesize potential food webs.

A critical component of this infrastructure is the SPIRE Triple Shop, a custom SPARQL query engine built on the Joseki service. Unlike traditional, rigid databases, the Triple Shop allows researchers to execute queries across a distributed collection of semantic web documents. It supports various levels of reasoning (from simple subsumption to OWL-Full), enabling the system to infer relationships—such as taxonomic hierarchies—that are not explicitly defined in the raw data.

The SPIRE SPARQL interface (Triple Shop) allows for complex queries across distributed ecological datasets.

Predictive Modeling via the Food Web Constructor

To bridge knowledge gaps, the Food Web Constructor (FWC) uses known empirical links to predict unrecorded interactions. By calculating taxonomic and phylogenetic distances, the FWC weights evidence from existing literature. For instance, if a predator is documented eating a specific genus in one habitat, the FWC can predict its interaction with a closely related species in a new environment.

This integration is made possible through several key ontologies:

  • ETHAN: Represents evolutionary trees and natural history information.
  • SPIREEcoConcepts: Defines specific food web and ecological relationships.
  • OWL Integration: By using the Web Ontology Language (OWL), the system understands that a "Nile Tilapia" is a type of fish; if a predator is known to eat "fish," the system can logically infer a potential interaction with the Tilapia.

Case Study: The Nile Tilapia

The researchers demonstrated the system's utility by modeling the impact of the Nile Tilapia, an invasive species in Florida. The system successfully synthesized data from multiple sources to predict the Tilapia's diet (primarily algae) and identified a lack of local predators, matching observed ecological outcomes in the region.

Visualization of a predicted food web for the Nile Tilapia, showing its interactions with local species like algae and bacteria.

Conclusion

The SPIRE project demonstrates that Semantic Web technologies can effectively break down data silos in ecology. By providing a transparent "Evidence Provider" and an ad-hoc synthesis engine, researchers can move beyond static databases toward dynamic, predictive models. Future work will focus on expanding the volume of ontologically-mapped literature and refining the algorithmic weighting of taxonomic evidence to improve predictive accuracy at a global scale.

Summary updated

BibTeX
@inproceedings{parr2006using,
  author = {Cynthia Sims Parr and Andriy Parafiynyk and Joel Sachs and Rong Pan and Lushan Han and Li Ding and Tim Finin and David Wang},
  title = {{Using the Semantic Web to Integrate Ecoinformatics Resources}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2006},
}
Figure