Creating simple predictive models in ecology, conservation and environmental policy based on Bayesian belief networks

Area of Science:

• Environmental science
• Ecological modeling
• Computational biology

Background:

• Predictive models offer valuable insights for management, conservation, and policy-making but are often complex.
• Existing modeling tools can require significant mathematical and programming expertise, limiting accessibility.

Purpose of the Study:

• Introduce BBNet, a novel R package designed for accessible predictive modeling.
• Enable users with limited programming backgrounds to construct, analyze, and visualize complex models.
• Facilitate ordinal predictions and sensitivity testing for system dynamics.

Main Methods:

• Utilizes modified Bayesian belief networks (BBNs) with a focus on straightforward interaction concepts.
• Models are constructed using basic spreadsheet tools and loaded into the R package.
• Parameterization incorporates data, literature, expert opinion, and surveys, using simple ordinal scales for interactions.

Main Results:

• BBNet provides a simplified approach to building and interpreting predictive models.
• Models can be analyzed, visualized, and sensitivity tested to understand information flow and predict outcomes.
• Ordinal comparison of model outcomes allows for ranking scenarios from 'best' to 'worse'.

Conclusions:

• BBNet democratizes predictive modeling, making it accessible beyond specialized programming expertise.
• The tool supports informed decision-making in environmental and ecological management, with potential applications in diverse fields.
• Modified BBNs offer a flexible framework for ordinal predictions and system analysis.