Unravelling changing interspecific interactions across environmental gradients using Markov random fields

Area of Science:

• Ecology
• Community Ecology
• Ecological Modeling

Background:

• Inferring species interactions is vital for understanding community assembly.
• Existing ecological models often fail to account for indirect interactions or interactions varying across environmental gradients.
• Markov random fields (MRF) offer a way to estimate interspecific interactions while controlling for indirect effects.

Purpose of the Study:

• To introduce and demonstrate the utility of Conditional Random Fields (CRF) for inferring interspecific interactions.
• To show how CRFs can incorporate environmental covariates to model how interactions change along gradients.
• To provide practical tools for ecologists to apply CRF models.

Main Methods:

• Utilized Markov random fields (MRF) to model interspecific interactions from multispecies occurrence data.
• Extended MRFs to Conditional Random Fields (CRF) by incorporating environmental covariates to analyze interaction variations.
• Applied CRF models to two presence-absence datasets: avian blood parasite co-infections and mosquito-predator co-occurrences along temperature gradients.

Main Results:

• Demonstrated that CRF models can successfully infer interspecific interactions and their variation with environmental factors.
• Showcased how parasite co-infection probabilities correlate with host abundance and how species co-occurrences change with water temperature.
• Highlighted the potential for CRFs in identifying shifting impacts of key species across climate and land-use gradients.

Conclusions:

• Conditional Random Fields (CRF) provide a powerful framework for advancing ecological understanding of species interactions.
• CRFs enable the investigation of how ecological interactions are shaped by and respond to environmental heterogeneity.
• The developed R package facilitates the application and interpretation of CRF models in ecological research.