Interesting response.

I've less climate science knowledge, but I can't help but throw out my knee jerk reaction: If all the proposed answers are so messy, then might that be a sign that we're asking the wrong questions?

The stock market I think provides a useful analogy. As a phenomena, it's easily quantifiable, and the numbers can guide you, but there's no winning formula to reliably predict future stock prices. Such efforts at "technical analysis" will only take you so far -&even then it's considered more of an art than a science - but its limitations make complete sense: People own companies that are valuable; companies are valuable b/c they make products that people like; and there's no way to predict how much people will like x product over y amount of time, and even if there was, there'd be no way to discern the effect of the company's changing profit on its future stock price.

(Or consider other historical scientific dead ends, such as phrenology or astrology. Areas such as those failed not b/c the models weren't complex enough, but b/c they were barking up the wrong tree.)

Just b/c you have some #'s - be it temperatures or the dow - doesn't
automatically mean that you should be able to come up w/a complex model
to predict them.

Ultimately the issue is how can we deal with complex systems, in cases
where we can't simply calculate what we want to know from fundamental
principles. So the question is, are there simple, quantitative general
principles of complex systems at higher organization which we could
use? Not necessarily - there may not be any simple, elegant way to
grapple with these systems, and then we're stuck with the messy tools
we currently use.

You might try repeating those statements but instead of referring to climate, refer to any quantifiable aspect out of daily life, such as the number of times a human being sneezes in a year, the number of ants living in the world, or the number of leaves that will fall come November. In all of these instances you have well-established simple laws that can be synthesized to come up with a rough guesstimate, and if you're lucky a quantifiable amount of error in either direction. If there were a need to answer the latter questions with a greater amount of exactness, then certainly scientists could *try* to lead the effort. They might start by comming up with similiarly complex models.