The melting in the Arctic has set a new record. On Sunday, according to the National Snow and Ice Center, sea ice covered just 24 percent of the surface of the Arctic Ocean, or 1.32 million square miles. That shattered the previous low set in 2007, when sea ice covered just 29 percent of the ocean.

Via NASA: “Satellite data reveal how the new record low Arctic sea ice extent, from Sept. 16, 2012, compares to the average minimum extent over the past 30 years (in yellow).”

The ice appears to have reached its minimum this year and will now begin re-forming and expanding again as winter approaches. (The NSIC waited a few days to report its results in order to verify that the ocean was actually refreezing.) But as humans continue to warm the planet, and as the melting in the Arctic continues apace, many scientists have been speculating on how soon we could see ice-free summers in the Arctic Ocean.

Predictions can vary wildly—some scientists say the summer ice could be gone within four years. Last week, ocean physicist Peter Wadhams told the Guardian that global warming was driving a "collapse" in the Arctic sea ice. "This collapse, I predicted would occur in 2015-16 at which time the summer Arctic (August to September) would become ice-free,"  he said. "The final collapse towards that state is now happening and will probably be complete by those dates."

Other scientists are more circumspect. Lars-Otto Reiersen, head of the Arctic Monitoring and Assessment Program, told Reuters that most climate models don't expect ice-free summers in the Arctic for another 30 or 40 years. "But," he cautioned, "there are models that indicate 2015 as an extreme."

Why is there such a wide variation? One difficulty is that the computer simulations used to predict the Arctic melt have trouble capturing the full complexity of what happens as the ice melts and thins out. That uncertainty isn't always good news. A 2007 study from the National Center for Atmospheric Research found that sea-ice models at the time had failed to capture how quickly the Arctic sea ice was receding in September.

Old models, circa 2007—far too conservative:

Source: Nevens Arctic Sea Ice Blog. Model runs are from Stroeve et al (2007). Brown line added in shows satellite observations of Arctic sea ice extent for September.

Those earlier climate models had often underestimated certain feedback loops. As the Arctic sea ice withers, more and more more of the ocean is exposed to sunlight. Since the darker ocean surface absorbs more sunlight than the bright ice, this warms the region even further. That, in turn, can accelerate the pace of ice loss. What’s more, a 2011 study in the Journal of Geophysical Research found that earlier models had low-balled the rate at which the ice was thinning out and drifting off into warmer waters, speeding up the melt.

More recent climate models, however, appear to have improved. An assessment published in Geophysical Research Letters this year found that the newest simulations seem to track satellite observations much more closely. But the ice is still melting a bit faster than the average model predicts. And, the study warned, the results "do not appear to have appreciably reduced uncertainty as to when a seasonally ice-free Arctic Ocean will be realized." Some models pointed to ice-free conditions by 2020, others toward the end of the century.

Newer models, circa 2012—better, but still a tad conservative:

The updated models. Source: Stroeve et al 2012. Colored lines show model runs. The red line shows satellite observations of September sea ice extent. Notice that the red is slightly outpacing the mean model prediction, shown by that thick black line in the middle.

Arctic sea ice also has the potential to rebound at times, due to year-to-year variability in weather conditions. After the ice reached a dramatic low in 2007, Mark Serreze, the director of the National Snow and Ice Data Center, predicted that the Arctic was headed for a "death spiral." Surprisingly, though, the ice didn't set a record low for another five years. Later studies found that unusual winds had driven the 2007 melt, pushing ice into the Canadian Basin and opening up an unusually large patch of open ocean. As temperatures rise in the Arctic, the ice becomes more vulnerable to those sorts of freak weather events. (A major storm may have helped drive the 2012 melt.) But it's hard to predict when freak weather events will occur.

These predictions aren't just of interest to scientists. A wide variety of nations and companies are now swooping in to exploit the Arctic's oil, gas and minerals. But even here, the unpredictability of the Arctic ice can cause havoc. Royal Dutch Shell had sought permits to drill for oil in the Chukchi and Beaufort Seas, north of Alaska, this summer. But even though ice was disappearing in much of the Arctic, the sheets remained unusually thick off the Alaskan coast this spring, forcing Shell to postpone drilling for three weeks. (In the end, Shell didn't succeed in drilling any wells this year.)

So that's where we are. Scientists agree the Arctic sea ice is melting at an unprecedented rate—indeed, reconstructions going back 1,450 years show just how dramatic the current melt is. They agree that human activity is driving the melt. And they tend to concur that the disappearance of the sea ice could have major impacts on everything from local wildlife habitats to, potentially, weather in North America and Europe. But there remain plenty of mysteries about how the ice will melt—and when, exactly, we'll see ice-free summers.

Related: Arctic sea ice just hit a record low. Here's why it matters.

Update: Andrew Revkin of the New York Times also has a terrific discussion on "pondering the path to an open polar sea."