Complex systems are characterized by surprising switches to new behavi
ours. Evaluating and predicting these changes demands an understanding
of the behaviour of the whole system. The combined ecosystem-climate
system shows chaotic or pseudorandom behaviour, stochastic or truly ra
ndom behaviour, as well as simple bifurcation and semi-stability. Semi
stability involves the sudden change from a destabilized 'attractor' t
o a new stable attractor which may occur after an apparently unpredict
able time delay. We present some recent results for analyzing time ser
ies data and for using simulations of non-linear models to predict the
se changes.