How to Compare Different Conceptual Approaches to Metapopulation Modelling

Models are essential tools in understanding population dynamics and deriving management measuresin the context of population viability analysis. However, very often the question arises which type ofmodel architecture is appropriate for a given situation. Mostly this situation is characterized by a shortage ofdata for model parameterization. In this study, an approach is presented to overcome this lack of real-worlddata by using the output of long-term simulation runs of specific individual-based models. Thus, it is possibleto evaluate the quality of macroscopic model predictions. Furthermore, this setting allows to compare totallydifferent types of metapopulation models. As an exemplary case study, this approach is applied to genericgrasshopper species in highly fragmented habitat landscapes, assessing on the one hand the well-known incidencefunction model and on the other hand a grid-based approach. The results show that predictions of bothmodels have substantial biases. Nonetheless, recommendations can be derived how to obtain more accuratemodel estimates. Finally, the patch-matrix model proves to be more adequate than the grid-based approach.