Greenland Cooled for Four Decades Before It Began Warming Fast

For most of the second half of the twentieth century, Greenland did something the rest of the planet was not doing. It got colder. While global mean temperature climbed, the annual mean at the surface of the world's second-largest ice sheet fell, significantly, decade after decade, until some time in the mid-1990s the sign flipped and Greenland began to warm fast. That kink has been visible in station records for years and has been a standing puzzle in ice-sheet science, because whatever produced it is presumably still operating.

A team at Shandong Normal University and Shanghai Jiao Tong University has now mapped the kink at a resolution the field has not had before. Working from a station-based reconstruction of surface air temperature on a 0.1-degree grid, roughly 11 kilometres, covering 1950 to 2024, Shujing Wang, Yetang Wang, Yulun Zhang, Zhaosheng Zhai and Shugui Hou set out to answer three linked questions: where in Greenland the change happened, in which seasons, and what was driving the decade-scale swings. Their analysis was accepted Aug. 3 by Environmental Research Letters and posted the same day.

The record they describe divides cleanly in two. Annual mean surface temperature decreased significantly from 1950 to the mid-1990s, after which it shifted to rapid warming, with strong interdecadal variability in every season. The spatial pattern is the part that is new. Warming has been stronger in northern Greenland than in southern Greenland, and the ablation zone, the low-lying margin where the ice sheet loses mass by melting rather than gaining it by snowfall, shows significant year-round warming.

Those two details do more work than they look like they do. Melt is concentrated at the margins, so year-round warming in the ablation zone means a longer melt season at the elevations where melting actually happens, not just warmer air over the high, cold interior where a few degrees changes nothing. And the north-south contrast, meanwhile, is a spatial signature that a coarse average would have hidden entirely.

The shape of the record is not itself a surprise. Greenland's mid-century cooling and its late-1990s reversal have been visible in long station series and in earlier reconstructions, and the correlation between detrended Greenland air temperature and the AMO index has been reported before. What this paper adds is spatial and seasonal detail: which parts of the island, in which months, and how much of the swing each oscillation accounts for.

The driver it settles on is the Atlantic Multidecadal Oscillation. The AMO is a slow see-saw in North Atlantic sea surface temperature that swings between warm and cool phases over roughly 60 to 80 years; the Pacific Decadal Oscillation is its rough Pacific analogue, with a shorter and less regular period. Neither is a forcing in the way that carbon dioxide is. They redistribute heat and reshape atmospheric circulation, and in doing so they can hold a region warm or cold for decades at a stretch.

In this reconstruction the AMO is the primary driver of decadal-scale temperature variability over Greenland, and the effect is strongest in summer. The PDO does not compete with it; it modulates it, most effectively in autumn. That scoping is the authors' and it repays careful reading: what the two oscillations explain is the decade-to-decade variability, the wobble, not the long-term trend. Nothing in the paper offers ocean cycles as an alternative to greenhouse warming; they are what makes the record lurch rather than climb smoothly.

The mechanism the authors describe for the recent decades is a coincidence of phases. Since the late 1990s the AMO has been positive and the PDO negative, and that particular pairing, opposite in sign but acting in the same direction over Greenland, has jointly enhanced anticyclonic conditions over the island. In plainer terms, it has favoured a persistent high-pressure ridge. A blocking high over Greenland clears the sky, lets sunlight onto the ice, draws warm air northward around its flanks, and warms the air that sinks beneath it. Sit one over an ice sheet for a summer and you get a melt year. Sit one there for two decades and you get a trend.

The payoff the authors claim is predictability. Greenland's contribution to sea level runs through its surface mass balance, the annual books of snowfall in against melt and runoff out, and surface air temperature is the dominant term on the losing side. If a large share of the decade-scale variability in that temperature is set by the AMO and tuned by the PDO, then to the extent those indices can be projected even a few years ahead, so can Greenland's near-term temperature. The authors write that the findings significantly improve the predictability of Greenland surface air temperature and support more accurate projections of future surface mass balance.

Two limits sit around that claim. The sea-level step is the standard one, but it is a step: the paper stops at surface mass balance and does not model ice discharge or sea-level contribution. And oscillation indices are themselves only partly predictable, which caps how much foresight the mechanism can buy. What the reconstruction delivers regardless is a map, at roughly 11 kilometres, of where on Greenland the temperature changed and in which season. That is the scale at which melt models operate.