It is true that it would be an interesting experiment to take a section of some species population from an environment, dump it in a different, but habitable environment and predict the effects of evolution. Thing is, the species could evolve in an unexpected manner. How accurate would you have to be to pass the test? And the experiment would take a long time.
Has this been tried at all?
For example, I know the pathway for naphthalene degradation in a species of Pseudomonas. I can predict what will happen if I feed it a non-natural halogenated naphthalene. After doing this type of thing for awhile, my predictions get real good. But this only an incremental step in the modification of a species (small step).
If, however, I take that same bacterium and feed it benzpyrene, my predictions are likely to be wrong unless they are very general (big step).
"It is true that it would be an interesting experiment to take a section of some species population from an environment, dump it in a different, but habitable environment and predict the effects of evolution."
Thing is: that would be a very long experiment, requiring several human generations at least. It's a good one, though. Perhaps you can set up the initial group and see that it is funded for the next few thousand years.