I don't have time to really think too hard about this story, but it is making the rounds, so I'll at least acknowledge it. Some NCAR simulations now predict essentially no late-summer ice in the Arctic by 2040. See the story at BBC [LINK] or at Nature [LINK]. The actual paper is in Geophysical Research Letters [doi:10.1029/2006GL028024].
What does this mean? Well, I haven't had a chance to look closely at the paper, but I have some first impressions. I did get a sneak-peek of these results this summer, and at the same time was introduced to some of the details of the sea-ice model used in CCSM (NCAR's climate model), so maybe I'll be able to say something halfway meaningful. The paper itself does not predict an ice-free Arctic in 2040, so let's just get that out of the way. This paper is really about the possibility of abrupt decreases in sea-ice in a changing climate, and the current generation of climate models suggest a real possibility of large reductions in perennial ice coverage in the first half of the 21st century. The main focus is a set of CCSM simulations using one of the emissions scenarios from IPCC. They also take a look at some of the results from other IPCC models. The CCSM always has what the authors call "abrupt reductions" in Arctic ice, and several of the other models also show large reductions.
I am willing to accept these results, but I think some skepticism has to be exercised still. First off, this is a GRL paper, which is a journal of short, usually preliminary, work focusing on "sexy" results. The peer-review process for GRL is sometimes thought to be a little lax, and sometimes the quality of the work is questionable. That does not seem to be an issue for this paper; the CCSM is a respected climate model, the authors are top-notch climate scientists, and this work is presented well. That said, this is not the last word on this project; I'm sure that the authors are doing more detailed work and are planning a longer, more careful analysis for another journal (e.g., Journal of Climate, Climate Dynamics). The best thing that could do would be to better quantify what "abrupt changes" really are, and the physical processes that trigger them, which is a big open question in this paper. They say the abrupt changes are driven by thermodynamics, but don't really present evidence of this; I assume they mean that wind patterns/ocean currents are changing to just move ice out of the Arctic, but it is not explained. The other thing to keep in mind is that even in the current generation climate models, the sea-ice models are fairly crude. I don't mean that in a bad way, the people working on these models are doing the best they can. Ice processes are quite complicated, and to properly model sea-ice, much like "properly" modeling clouds, the simulations need to be run in much higher resolutions. That kind of resolution is too expensive right now, and even if the resources were there, it would be a tough sell to dedicate it to the sea-ice component rather than better atmospheric and oceanic components. This particular climate model is known to be fairly sensitive, and when it gets knocked out of equilibrium, the sea-ice is one of the things known to respond fairly erradically. So while I think the CCSM, and several other high-end climate models, can get a lot of important changes correct, we still can't trust the details of these fully coupled simulations. My interpretation is then something like this: in the near future (50 years), it is likely that rapid reductions in perennial Arctic sea-ice will be observed, associated with (but not well-correlated with) increasing atmospheric greenhouse gases.