Something I've been saying for a while: The ice modeling in the current gold-standard models (like GISS-E) is Just Plain Wrong: it doesn't adequately account for positive feedback in ice-melting, such as the way meltwater changes the ambient environment for ice, or the way that ice melt affects ambient atmospheric properties. A calculation like that is pretty much guaranteed to predict that ice melts only very slowly and adiabatically, instead of quickly and with marked "tipping points".
Conclusion: We're going to have a seasonally navigable North Polar Sea a lot sooner than many people anticipate.
