Deciphering Probabilistic Species Interaction Networks

Area of Science:

• Ecology and Evolutionary Biology
• Network Science
• Conservation Biology

Background:

• Species interactions are often represented deterministically, masking inherent uncertainties.
• Probabilistic interaction data capture various uncertainty sources, including model predictions, expert opinion, and empirical variability.
• Lack of guidelines for estimating and documenting probabilistic interaction data hinders robust ecological understanding.

Purpose of the Study:

• To review definitions of probabilistic species interactions across different spatial scales.
• To investigate the distinct statistical properties of local networks and metawebs.
• To develop methods for inferring binary interactions from probabilistic data and identify potential biases.

Main Methods:

• Review of probabilistic interaction definitions at local (spatiotemporal realization) and metaweb (feasibility) scales.
• Comparative analysis of statistical properties of local networks and metawebs using European host-parasite data.
• Development and application of two inference approaches for binary interactions from probabilistic data.

Main Results:

• Local networks and metawebs exhibit different spatial and temporal scaling properties.
• Direct inference of local networks from metawebs introduces systematic biases.
• Probabilistic interaction data require rigorous descriptions specifying conditional variables and uncertainty sources.

Conclusions:

• Distinguishing between local networks and metawebs is crucial for accurate ecological network analysis.
• Standardized methods for defining, estimating, and documenting probabilistic species interactions are needed.
• Improved characterization of probabilistic interactions enhances our understanding of ecological dynamics and uncertainty.