Bill Cotton , a Professor at the Department of Atmospheric Science at Colorado State University (and a colleague of mine) has given a lecture
Cotton, W.R., 2010: Is Climate Really Predictable on 10-50 Year Time Scales? International Symposium on Prediction, San Diego, CA, July 2010.
The entire set of slides is worth viewing. Selected conclusions that he reports are
i) It is often claimed that climate is predictable because it is a boundary value problem(that is, only changes in external forcing is needed).
ii) But, we noted that deep ocean variability occurs on time scales of 100’s of years
iii) Thus initialization of deep ocean circulations is needed for forecasts on decadal time scales.
iv) This means that decadal climate prediction is both an initial value problem and boundary value problem
Considering the stochastic external forcing parameters(eg. volcanoes), uncertainties of solar variability forcing, and the tendency for strong model biases on time scales of 2-5 years let alone 10 to 50 years, I see no evidence that climate is predictable on these time-scales nor will it be for decades to come (a forecast!).