Community dynamics and sensitivity to model structure: towards a probabilistic view of process-based model...

Area of Science:

• Ecology
• Computational Biology
• Environmental Science

Background:

• Ecological and biological predictions often rely on process-based models, but their uncertainty is not fully captured.
• Predictive uncertainty in these models is typically limited to parametric uncertainty, neglecting structural sensitivity.

Purpose of the Study:

• To integrate mechanistic, process-based modeling with statistical inference using Bayesian statistics.
• To develop a practical approach for quantifying predictive uncertainty that includes both parametric and structural sensitivity.

Main Methods:

• A probabilistic framework was developed to incorporate uncertainty from model construction and parametric uncertainty.
• The approach was applied to a predator-prey system using the Rosenzweig-MacArthur model as a proof of concept.

Main Results:

• Structural sensitivity was found to regularly outweigh parametric sensitivity in the predator-prey model.
• The proposed method provides a probabilistic view of predictions, accounting for uncertainty in model choice.

Conclusions:

• The integrated approach enhances the theoretical power of process-based models while addressing predictive uncertainty.
• This methodology offers a pathway for more robust and operational probabilistic predictions in ecological and biological sciences.