Inferring ecosystem networks as information flows.

Area of Science:

• Ecology and Systems Biology
• Network Inference and Causality
• Ecological Modeling

Background:

• Inferring causal interactions is crucial for understanding complex ecosystem networks in both basic and applied research.
• Convergent Cross Mapping (CCM) has advanced network inference by assessing how well historical data predict other variables' states.

Purpose of the Study:

• To investigate the capability of a novel Optimal Information Flow (OIF) ecosystem model for inferring bidirectional causality.
• To compare the performance of OIF against CCM in ecological network inference.

Main Methods:

• Developed and applied the Optimal Information Flow (OIF) ecosystem model.
• Tested the model using synthetic datasets from a predator-prey system, real-world sardine-anchovy-temperature data, and a multispecies fish ecosystem.
• Compared OIF's performance against Convergent Cross Mapping (CCM) in predicting population and community patterns.

Main Results:

• OIF demonstrated superior performance over CCM in predicting population and community dynamics across all tested datasets.
• OIF achieved a larger gradient of inferred interactions, higher point-value accuracy, and reduced fluctuations in interactions and diversity.
• OIF accurately captured characteristic time delays in ecological interactions and diversity.

Conclusions:

• The Optimal Information Flow (OIF) model offers a more accurate and computationally efficient approach to inferring predictive causal networks in complex ecosystems.
• OIF's ability to consider synchronization, divergence, and diversity enhances its sensitivity and uncertainty assessment.
• Accurate inference of species interactions using OIF has significant implications for predicting biodiversity changes and informing ecosystem management strategies.