Post-hoc pattern-oriented testing and tuning of an existing large model: lessons from the field vole

Area of Science:

• Ecological modeling
• Computational ecology
• Agent-based modeling

Background:

• Pattern-oriented modeling (POM) is a strategy for developing and testing ecological models.
• Its application to complex, established models, particularly agent-based models, remains underexplored.
• The ALMaSS framework presents a high-complexity simulation environment for ecological research.

Purpose of the Study:

• To investigate the utility of post-hoc pattern-oriented modeling (POM) for testing, calibrating, and developing a complex agent-based model.
• To assess the feasibility of applying POM to an established, high-complexity simulation model within the ALMaSS framework.
• To evaluate the effectiveness and challenges of using multiple patterns for model validation and refinement.

Main Methods:

• Applied pattern-oriented modeling (POM) to an existing agent-based model of the field vole (Microtus agrestis).
• Utilized multiple patterns observed at different scales and hierarchical levels for model testing and calibration.
• Focused on optimizing model structure and selecting sub-models within the ALMaSS framework.

Main Results:

• The application of POM to the complex agent-based model yielded generally very good results in fitting observed patterns.
• The process of testing and calibrating the model using POM was found to be lengthy and intricate.
• Achieving good correspondence between the model and real-world data often necessitated close replication of the environment within the model.

Conclusions:

• Post-hoc pattern-oriented modeling (POM) is a viable and useful approach for validating and refining highly complex simulation models.
• Caution is advised against over-fitting models to real-world patterns lacking detailed explanatory factors.
• Adopting open-science and open-source methodologies is recommended to address challenges in complex ecological simulation modeling.