Reconsidering an Agent-Based Model of Food Web Evolution

Samuel Armstrong, Emma Norling

Abstract: This paper reimplements and extends a prior agent-based model of food web evolution. The earlier work attempted to replicate the results achieved by a system dynamics model of food web evolution, but failed to achieve the diversity or realistic dynamics of the system dynamics approach. This work starts by adding spatial diversity to the model, the lack of this being flagged as a potential problem in the original work. This produced some improvement in the results (with more diverse food webs being produced), but there were still patterns commonly found in the resultant food webs that are uncommon in real-world food webs. To further refine the model, a more complex representation of species traits was added, and methods for classifying species based upon the traits. In particular, an unsupervised learning clustering algorithm has been introduced to classify species in the evolving food web. This has resulted in a model which produces abstract food webs that far more closely mimic the patterns found in real-world food webs.