So Rahmstorf decided to pursue an entirely different type of model. He looked at the annual rate of sea-level rise from the 1880s onwards, and then matched it with air temperatures at those times. He found a simple relationship: the warmer it got, the faster the sea level rose. In 2007, too late to be considered by that year's IPCC assessment, his model predicted up to 1.4 meters of sea-level rise by 2100 — more than twice the IPCC number.
'Semi-empirical' models such as this have advantages: by definition, they accurately model the rise that has already occurred, and they do not require a full understanding of how and why it is happening. But no one knows how long the relationship at the heart of these models will hold, particularly as melting ice sheets become a bigger factor. The models, says Rahmstorf, “could be good for 50 years, or 100 years. We don't know.”
When it comes to making projections, the choice of models has big consequences. Process models generally predict rather less than 1 meter of rise by 2100, whereas semi-empirical models top out at between 1 and 2 meters — enough, at the higher end, to flood the homes of 187 million people. These high-end, semi-empirical estimates are extremely controversial, and the IPCC has low confidence in them. “The only advantage of these models is that they're easy to calculate,” says Philippe Huybrechts, an ice modeler at the Brussels Free University. “I think they're wrong.”
Process-based modelers have made great progress since 2007, thanks to improved understanding of factors such as how much heat is flowing into the oceans — and thus causing the water to expand — and how much groundwater makes its way into the oceans because of people's unquenchable thirst for fresh water pumped up from below. As a result, modelers can now explain all of the observed rise in sea level, particularly that in recent decades.
But that does not guarantee accurate forecasts. Everyone acknowledges that there are still big issues with process-based projections — in particular, modelers have only a tenuous grasp of how the big ice sheets in Greenland and, especially, the Antarctic might behave and whether they will melt and flow catastrophically into the sea. In all, the ice sheets hold enough water to raise sea levels by more than 65 meters in the long term, compared with as much as 0.4 meters from all the world's glaciers and ice caps.
Despite these problems, the IPCC has decided that researchers finally have a good enough handle on ice behavior in Greenland and — to a lesser extent — Antarctica to forecast how ice sheets will respond, at least provisionally, says Don Chambers, a sea-level researcher at the University of Texas at Austin. The latest estimates add between 3 and 21 centimeters to the predicted sea-level rise by 2100, although tens of centimeters more are possible, according to the most recent IPCC report draft.
The end result is set to be a much higher forecast for sea-level rise than in 2007. Direct comparisons are difficult because the latest report uses different time frames and emission scenarios, but the leaked draft puts the range of estimates between 28 and 97 centimeters of rise by 2100. That is still not as high as semi-empirical estimates, but process-based results are edging upwards — and the difference is narrowing. “I consider it something of a vindication,” says Rahmstorf.
How much will it vary?
When Jeff Freymueller, a geophysicist at the University of Alaska Fairbanks, visited Alaska's Graves Harbor more than a decade ago, his marine charts showed three isolated little islands; what he saw, instead, were three grassy peninsulas connected to the mainland. That was because water levels in some parts of Alaska are dropping — by up to 3 centimeters per year.