Chaotic Nature of Cilmate Models

Lorenz next constructed a simulacrum of climate in a simple mathematical model with some feedbacks, and ran it repeatedly through a computer with minor changes in the initial conditions. His initial plan was simply to compile statistics for the various ways his model climate diverged from its normal state. He wanted to check the validity of the procedures some meteorologists were promoting for long-range "statistical forecasting," along the lines of the traditional idea that climate was an average over temporary variations. But he could not find any valid way to statistically combine the different computer results to predict a future state. It was impossible to prove that a "climate" existed at all, in the traditional sense of a stable long-term average. Like the fluid circulation in some of the dishpan experiments, it seemed that climate could shift in a completely arbitrary way.
These ideas spread among climate scientists, especially at a landmark conference on "Causes of Climate Change" held in Boulder, Colorado in August 1965. Lorenz, invited to give the opening address, explained that the slightest change of initial conditions might randomly bring a huge change in the future climate. "Climate may or may not be deterministic," he concluded. "We shall probably never know for sure." Other meteorologists at the conference pored over new evidence that almost trivial astronomical shifts of the Earths orbit might have "triggered" past ice ages.. Summing up a consensus at the end of the conference, leaders of the field agreed that minor and transitory changes in the past "may have sufficed to 'flip' the atmospheric circulation from one state to another."

Until the future actually came, there would be no way to prove that the modelers understood all the essential forces. If an unlucky combination sent the real climate temporarily into one of the unusual states found in some model runs, that could confuse people about what was happening. But it was not likely to change the eventual outcome. What was no longer in doubt was the most important insight produced by the half-century of computer experiments. Under some circumstances a fairly small change in conditions, even something that seemed so slight as an increase of greenhouse gases, could nudge climate into a severely different state. The climate looked less like a simple predictable system than like a confused beast, which a dozen different forces were prodding in different directions. It responded sluggishly, but once it began to move it would be hard to stop.

http://www.aip.org/history/climate/chaos.htm#L_0565