Levels of emergence in individual based models: Coping with scarcity of data and pattern redundancy

Abstract

Abstract Pattern Oriented Modelling (POM) is a well-known framework for designing, parameterizing, and analyzing individual based models (IBM). It assesses individual or agent based complex systems by applying inverse modelling on patterns observed in the real system of interest. However, scarcity of field data is an underlying problem in many environmental modelling projects and the relationship between the amount of available data and the domain of applicability of IBMs has hardly been considered. We argue that choosing the appropriate decision process, and the traits that represent it, when modelling individuals is fundamental for the model to be able to generalize when few data are available, and therefore not representative of the intended domain of applicability. A pattern being the expression of a given comportment, we define the notion of level of emergence, and propose to order comportments according to these levels. This approach may aid in selecting the appropriate trait for generalization when designing an IBM. We also show how levels of emergence can be used to assess the redundancy in patterns. A pattern is redundant with respect to other patterns when it does not bring new information about the process which generates it. Since POM makes use of patterns to calibrate the model, one should ensure that the considered patterns are independent, because redundancy might lead to non-optimal results. Our recommendations thus provide a way of avoiding some of the potential pitfalls of the POM framework and contribute to the on-going development of a standard approach for individual-based modelling.