Dimensional approaches to scaling experimental ecosystems: designing mousetraps to catch elephants.

Enclosed experimental ecosystems (mesocosms) are small relative to their natural counterparts, are typically operated for short durations relative to the timescales of a number of important ecological processes, and also often have reduced biological and physical complexity relative to nature. These reductions in time, space, and complexity scales have been cited as sources of unrealistic ecological behavior within mesocosms and raise questions about extrapolating results from mesocosms to nature. Dimensional analysis, a technique widely used by engineers to create scale models, uses compensatory distortion as a means of maintaining dynamic similarity in properties and relationships of interest. Although biological parameters are generally less controllable than physical ones, a variety of dimensional approaches can be taken to maintain such key ecological properties as effective habitat size, environmental variability, vertical and horizontal gradients, interactions among habitats, and control of experimental artifacts. To date, application of dimensional approaches to mesocosm design has been largely intuitive and idiosyncratic. We argue that a more explicit, systematic, and quantitative approach will increase realism and may also provide a critical means of developing, testing, and advancing our understanding of scaling relationships in nature.