“Ecosystems” are not very good “systems”. They have no real boundaries, they don't have steady states, they have all the homeostasis of a Rube Goldberg tooth-brush (except that they don't grind to a halt when you shove something into the works, which real systems tend to do). In short they resemble waterfalls or storms more than they do thermostats, guided missiles or organisms; and it seems we learn more about them by tracing the flow of chlorine than by contemplating the whole.

ref > http://cscs.umich.edu/~crshalizi/notebooks/ecology.html