Spatial Patterning and Error Propagation in a Stream-Ecosystem Model

摘要

In many cases an aggregated model can adequately represent overall properties of a spatially distributed system. The present study examines whether such a lumped model can provide reliable estimates of the uncertainty associated with predictions. For this purpose we compared standard deviations associated with predictions from a spatially distributed and a spatially lumped model of the same stream ecosystem. We examined the ability of the lumped model to estimate the uncertainties predicted from the distributed model. Three sources of spatial patterning were examined: internal dynamics of the stream, patterns of litter inputs, and patterns in initial conditions. The aggregated model was least reliable when spatial patterns were due to internal dynamics of the distributed system; estimated standard deviations were as much as 2.5 times too large. Ignoring spatial patterns in system inputs and initial conditions caused less serious problems, but results were complicated by nonlinearities and interaction terms in the models. In all cases examined, spatial patterning significantly affected the uncertainty associated with model predictions. (ERA citation 06:029236)