Simple models of complex systems: The paradox of modeling biodiversity

Background/Question/Methods

Models are simplified purposeful representations. A model’s purpose determines what should be represented in detail, what can be aggregated, and what can be ignored altogether. Depending on the question addressed, we decide whether a certain process or structural element is relevant for answering this question. Model analysis then reveals whether or not our simplifying assumptions are sufficient to make the model reproduce observed phenomena. As a result we ideally obtain a model that is parsimonious and only includes key processes and structures. However, we know from studies of ecosystem resilience that regime shifts can occur due to changes in environmental conditions or disturbances. A regime shift implies that a new set of mechanisms takes over control, but these mechanisms were likely to be ignored in our model of the system functioning in its original regime. We thus have the paradox that the fundamental philosophy of modeling, to focus on key processes, may prevent us from capturing the role of biodiversity because potentially important mechanisms are not included in the model. To solve this paradox, simply representing every potentially important mechanism is not an option, but what can we do instead?

Results/Conclusions

To solve this paradox of modeling biodiversity, several approaches need to be combined: (1) Pattern-oriented modeling: go for structural realism by making a model simultaneously reproduce several patterns observed at different scales and levels of organization. (2) First principles: base the representation of the system’s building blocks, the individual organisms, on general principles of physiology, metabolism, and adaptive behavior. If these representations are parameterized and tested for a wide range of environmental conditions, the system’s model may be able to predict the response to new environmental conditions. (3) Emergence: do not impose system-level properties, but let them emerge from first principles and the interaction and behavior of individuals. (4) Gradients and assembly: apply the model to gradients of environmental conditions, and let communities and ecosystems emerge. (5) Traits and functional types: since it never will be feasible to model all species of a system, identify the spectrum of relevant traits and functional types and implement them in the model. Try different trait classifications and combinations. (6) Design for regime shifts and ecotones: Making a model capturing the functioning of a certain system under “normal” conditions is not sufficient. Focus also on explaining known crashes of the system (regime shifts) and boundaries to other systems (ecotones).