We've gotten much better at predicting weather a few days out[0], because not only has our understanding of the Earth's climate improved by leaps and bounds, but we've also had the chance to iterate thousands of times over the last couple decades to see how well our models perform and adjust them based on that evidence.

On the other hand, we've only had the chance to iterate a few times over the last couple decades in predicting longer-term trends several years out, and even though our understanding has increased here as well, we don't have the benefit of very many results to confirm if that understanding is anywhere near enough to predict well over these sorts of time spans. Apparently arctic sea ice has decreased faster than predicted, but average global temperatures are falling out of the "95%" ranges of most of the models of a decade or two ago[1]

[0]http://www.nytimes.com/2012/09/09/magazine/the-weatherman-is... [1]http://www.economist.com/news/science-and-technology/2157446...

Predictions aren't magic. They're based on looking at past data and using that to draw conclusions about future data.

Weather, being extremely variable even within a day, is a lot harder to predict than, say, "It will be colder in January than in August (in the northern hemisphere)." That one's a slam dunk.

Watching the warming trend, it seems more predictable than weekly weather. Even now, with this winder, the north pole is warmer than normal which is part of what's causing the weather systems we're experiencing in North America (as you referenced with the vortex).

Why on earth would the models have something like this vortex detaching built into them? We've never seen it before, or seen it so rarely that nobody built it into the model. That doesn't say we can't predict climate, it just says we didn't predict this particular consequence of climate change.

The idea is that there's a mostly random high frequency medium amplitude component (weather), a somewhat predictable medium frequency high amplitude component (seasons), and a low frequency low amplitude component (climate).

So if you observe for long enough, the higher frequency parts mostly average out to zero and you can get an idea of the low frequency component. Even without being able to predict the other components.

(See also, http://en.wikipedia.org/wiki/Fourier_transform )

That predicting behavior at larger scales is easier than predicting it at small scales is a principle you'll encounter repeatedly in the physical science and engineering disciplines. In the physical sciences, models are often top-down, as opposed to in computing where they are almost always bottom-up. In fluid dynamics, for example, the behavior air in a turbulent flow is chaotic at the small scale. Yet, we can still design airplane wings because we have pretty good approximate models for how such flows behave at larger scales. Or to use a physics example: it's easy to predict where a baseball will land, but very complicated to predict where an electron will be in a molecule.

The analogy I like to think of is a ping-pong ball in the ocean. At any given second, its location will be hard to predict, as at the second-granularity, its location will be dominated by local phenomenon. But, over the course of months and years, we can often accurately predict where it will generally end up in the world due to ocean currents.

I, too, am a bit skeptical, but this is boiling climate science down into too simple an exercise. We do, for instance, know that things are warming up, and we have many different models to predict what effect that will have on, say, weather trends over a longish period of time. It's kind of like the stock market -- I might think that the fundamentals of company ABC are good, and believe that they will be up over the next five years. Some people do make reasonably good predictions of this sort. Predicting the open and close of ABC's stock tomorrow, however, or the high and low over the next three weeks, is another story, and there's reason to believe that we do no better than random chance at this kind of prediction, in aggregate.

See NASA scientist, "the father of climate change," James Hansen's 2009 book entitled Storms of My Grandchildren. He's highly critical of models, though notes that they may be useful in certain specific scenarios. What he focuses more on is the geological record because models can only be as accurate as the information that goes into them.

I want to keep this brief, but I also want to quickly note that the cap and trade for greenhouse gases is business as usual for energy producers. Don't believe a word of it; the short story is that it worked poorly in preventing acid rain and won't work for combating climate change. We need a fee and dividend system that doesn't come back to us in tax breaks. I should note that Hansen comes across as very credible and apolitical; I don't trust anyone else when it comes to climate science.

His most recent work is available here:

http://www.columbia.edu/~jeh1/publications.shtml

It is harder to predict very unusual events.

Also, I am not sure what difference it makes? We want to predict climate to inform environmental policy. But a lack of certainty does negate the damage that pollution does.

There is an enormous difference in localized weather predictions, and whole globe climate predictions. That is, it is possible to predict, say, average temperature across the entire world, while being unable to reasonably predict average temperature in any one place.

It seems to me that people are misunderstanding looking backwards and looking forwards...