Error Propagation in Spatially Explicit Population Models: a Reassessment

Spatially explicit population models (SEPMs) are a means for studying the dynamics and viability of populations on landscapes (e.g., Pulliam et al. 1992; Turner et al. 1995). Because of their capabilities for modeling populations on realistic, heterogeneous landscapes (using geographic information system data, for example), SEPMs have been applied in conservation biology (e.g., Wolff 1994; Schippers et al. 1996; DeAngelis et al. 1998). Concerns have been raised by some authors (Wennergren et al. 1995; Groom & Pascual 1997; Meir & Kareiva 1997; Ruckelshaus et al. 1997), however, that such models are "likely to be highly unreliable given our uncertainties about dispersal behavior and rates" (Wennergren et al. 1995). This pessimism seems to be justified by simulations with an abstract landscape model performed by Ruckelshaus et al. (1997). These authors investigated the effects of input data errors in a model of animals dispersing on a landscape. They deduced from their model that small errors in the mortality associated with dispersal leads to disproportionately large errors in predicting overall survival and conclude "that the dispersal modules of spatially explicit models are staggeringly sensitive to the details of dispersal behavior" (Ruckelshaus et al. 1997). The model of Ruckelshaus et al. (1997) describes a landscape composed of 118 X 118 cells, with patches of suitable habitat randomly distributed over the grid. Model organisms moved randomly over the grid until they reached a patch of suitable habitat. For a given size and shape of the patches and fraction of suitable habitat, 400 individuals were simulated sequentially. This resulted in frequency distributions of steps it takes an individual to reach a patch of suitable habitat. To study the effect of dispersal mortality on overall dispersal success,