Evaluating Detail in Ecological Models

Evaluating the Importance of Detail

The simple fact that SORTIE predicts forest dynamics does not give any insight into which aspects of the model control these predicted behaviors. The experiments with the removal of space and the approximation of species function help identify what detail at the scale of individual trees is relevant to large-scale forest successional development. SORTIE provides a natural framework for a computational investigation into the role of these processes in forest dynamics, because the model is both spatially explicit and empirically based. The detailed (base line) simulations provide a reference set against which the simplified models are compared.

Assessments of the role of spatial interactions are made by removing this spatial coupling. Two processes, local competition and local dispersal, couple the interactions of the trees. The failure of the mean-field models to match the base-line dynamics demonstrates that many major aspects of SORTIE’s community predictions are mediated by spatially localized interactions (Pacala and Deutschman 1995). These insights confirm more abstract modeling and theoretical work, suggesting the strong potential role of spatial interactions (Hastings 1990; Chesson 1991; Durrett and Levin 1994; Levin and Pacala 1997; Bolker et al. in press).

Searching for simplified ways in which to incorporate species interactions in ecological models is difficult. The success of the PCA in identifying a simple two-dimensional (that is, two factor) system from within the larger 10-dimensional system was very encouraging. The PCA description of species function was not only efficient but seemed to capture the essence of the role each species plays in the community. Unfortunately, the simulations based on the two-dimensional species projections failed to match the base-line dynamics. These results send a strong warning about the potential for approximations to lead to biologically plausible but incorrect predictions about the emergence of large-scale patterns from small-scale interactions.

Base line	Mean field	PCA
Base line with disturbance	Mean field with disturbance	PCA with disturbance

Experiments performed on the model to uncover which aspects control large-scale behavior.

These alternative-model formulations represent fundamental steps in identifying the relevant detail in the forest model (Deutschman 1996; Levin et al. 1997). Insofar as this mechanistic model mirrors nature, it provides a reflection of the critical detail controlling the emergence of forest pattern from the interactions of trees. The mechanistic fidelity of SORTIE to the biological nature of tree interactions indicates that even the most basic aspects of natural forests, like total biomass and diversity, are inherently sensitive to local spatial dynamics. This work also demonstrates that approximate descriptions of species are possible and useful in developing a simple description of each species life-history strategy, but incapable of predicting quantitative dynamics. These results, both positive and negative, show convincingly that experimenting with a detailed, mechanistic model, derived from data, is a powerful method for addressing these questions of relevant detail, emergent properties, and scale in ecological systems.