From interesting details to dynamical relevance: toward more effective useof empirical insights in theory construction

A perennial challenge in ecology is to develop dynamical systems models tha t appropriately abstract and characterize the dynamics of natural systems. Deriving an appropriate model of system dynamics can be a long and iterativ e process whose outcome depends critically on the quality of empirical data describing the long-term behavior of a natural system. Most ecological tim e series are insufficient to offer insight into the way organizational hier archies and spatial scales are causally linked to natural system dynamics. Moreover, the classic tradition of hypothesis testing in ecology is not lik ely to lead to those key insights. This because empirical research is geare d almost exclusively toward testing model predictions based on underlying c ausal relationships assumed by theorists. So, empirical research relies hea vily on theory for guidance on what is or is not dynamically relevant. I ar gue here that it is entirely possible to reduce much of this guesswork invo lved with deciding on causal structure by giving empirical research a new r ole in theory development. In this role, natural history and field observat ions are used to develop stochastic, individual-based and spatially explici t computational models or IBMs that can explore the range of contingency an d complexity inherent in real-world systems. IBMs can be used to run simulations allowing deductions to be made about th e causal linkages between organizational hierarchies, spatial scales, and d ynamics. These deductions can be tested under field conditions using experi ments that manipulate the putative causal structure and evaluate the dynami cal consequences. The emerging insights from this stage can then be used to inspire an analytical construct that embodies the dynamically relevant sca les and mechanisms. In essence, computational modeling serves as an interme diate step in theory development in that a wide range of possibly important biological details are considered and then reduced to a subset that is dyn amically relevant.