Aggregation error in ecological models

Models of complex ecosystems require the representation of a large number of components by a few variables. The exact nature of the limitations imposed by this abstraction process has received surprisingly little treatment. In this study, models with k variables were taken, for the sake of argument, to exactly represent a system. Models with k — n variables were then developed for the same system and differences in output examined. As a result of the study, we suggest some simple rules for minimizing aggregation error, including lumping components with similar turnover times and lumping rare species with common ones.