After I mentioned a paper by "Weigert" in a posting I have received numerous 
requests for the reference. The correct spelling is Wiegert. My apologies to 
all of you, especially Prof. Wiegert.

I am currently travelling and do not have the reference with me, but I found 
his website and the paper may be

Wiegert, R.G. 1975. Simulation models of ecosystems. Ann. Rev. Ecol. Syst. 
6:311-338.

although I thought it was earlier. The paper describes about half a dozen 
models of a simple salt spring ecosystem, and as I wrote earlier, he found that 
the best performance came with an intermediate level of complexity.

I've been asked in another posting why complex models do not work too well. 
There are many reasons for this, including the need for too many parameters and 
resultant magnification of errors, but in ecological modelling when you get too 
specific and try to model individual species you need to describe the factors 
behind zonation and succession, which is hard to do. In general I find that 
fairly aggregated models work well, especially when aggregated on the basis of 
function rather than taxonomy.

Another problem is that very precise models are more susceptible to problems 
arising from discontinuous processes, such as insect outbreaks and blooms 
(algal, jellyfish, etc.). Unless we know precisely what triggers these events 
and know how to predict the events, the models will not perform well.

I recall a modelling exercise where the components of an estuarine ecosystem 
were modelled by separate groups of scientists. All went well, except that the 
head of the zooplankton group insisted that all four of the Acartia species he 
was studying be modelled individually. They never managed to do this (again, 
problems of zonation and succession) and the project would have been aborted if 
the rest of us hadn't thrown together a simpler but working submodel.

Not everything can be successfully predicted. This is a property of natural 
systems, not just ecological models. I have already mentioned the cod-haddock 
issue on this list. Modelling fish recruitment is tricky because it depends on 
the overlap between larval emergence and plankton blooms, which we generally 
cannot predict. We cannot do very well at predicting earthquakes either. And in 
mathematics, remember Gödel's famous proof that not all true theorems can be 
proven. Failure to do the impossible is not really failure (unless you are a 
Marine!).

Ciência Silvert
www.ciencia.silvert.org

Reply via email to