Polar amplification

Why does the Arctic warm two or three times faster than the planet as a whole?

The stock answer arrives before the question finishes: bright ice melts, dark ocean absorbs more sunlight, the Arctic warms faster. It is a real feedback, genuinely at work. But it makes a prediction we can check, and the check fails. If sunlight absorbed by newly dark water were the engine, amplification should be strongest in summer, when there is sunlight to absorb. Observed amplification is strongest in autumn and winter, and close to absent in midsummer. What is doing the warming during the polar night?

Before reaching for ice at all, notice that a cold place amplifies warming for a reason that has nothing to do with ice. A surface sheds heat to space roughly as the fourth power of its temperature, so the extra radiation emitted per extra degree is the derivative of that — four times the Stefan-Boltzmann constant times temperature cubed. Put in a global emitting temperature near 255 kelvin and you get about 3.8 watts per square metre per degree; put in a warm surface near 288 kelvin and you get about 5.4. The cold region is worse at shedding extra energy, so the same imbalance forces roughly 1.4 times more warming there. That factor comes out of two arithmetic evaluations, and it would exist on a planet with no ice whatsoever. This is the Planck feedback, routinely omitted from the popular story.

The second mechanism, which recent feedback decompositions identify as the largest single contributor — larger than albedo — concerns where in the air column the warming lands. In the tropics the atmosphere is convectively stirred and follows a moist adiabat, so surface warming is carried aloft and the upper troposphere warms more than the surface. That upper warmth radiates efficiently to space, damping the surface response: a negative lapse-rate feedback. The Arctic atmosphere is the opposite — strongly stratified, often with a surface temperature inversion, convection weak. Extra energy stays low down, so the same added energy produces more surface warming, a positive lapse-rate feedback arising purely from the shape of the temperature profile.

Only now does ice earn its place, and its role is stranger than advertised. Summer sunlight absorbed by open water does not immediately raise air temperature much, because most of it goes into melting ice — a latent heat sink at fixed temperature — and into warming a mixed layer with enormous heat capacity. The Arctic Ocean stores that energy. Then autumn comes, the ocean must lose it to refreeze, and releases it upward as longwave radiation and turbulent heat flux into an atmosphere with no sunlight of its own. That is the season the amplification appears in. The albedo feedback is real, but delayed and displaced: absorbed in summer, expressed in winter. The seasonal mismatch that falsified the naive story is exactly what the corrected one predicts.

Two further contributors round it out: more moisture and heat transported poleward by the atmosphere, and Arctic clouds, which over a dark surface warm more than they cool. Cloud effects remain among the least certain terms.

Now the number in the question. "Two or three times" is the older figure, and it depends heavily on how the Arctic is defined and over what period. A 2022 analysis using the region north of 66.5 degrees over 1979-2021 found warming close to four times the global rate. Both can be right about different domains and intervals — a caution about the whole genre of amplification ratios rather than a correction of either.

Where it is genuinely contested: the ranking of lapse-rate against albedo varies between models and between decomposition methods, and the decomposition is somewhat arbitrary because the feedbacks are not independent — sea-ice loss is part of why the Arctic boundary layer restructures. Separately, whether Arctic amplification is destabilising mid-latitude weather by weakening the jet stream is a live dispute, with Arctic researchers advancing the linkage and atmospheric dynamicists finding the causal evidence weak in both observations and targeted model runs.

Where this sits next to its neighbours: the collection's piece on ice-age triggers leans on the ice-albedo feedback as the amplifier that converts a weak orbital nudge into kilometres of ice. That is the same feedback at a different fixed point — there it operates over millennia, where its summer timing and the slow growth of ice sheets are exactly right; here, on a decadal signal read season by season, it turns out to be second in line and to deliver its effect out of season.

Two refinements matter. "Amplification" is a statement about the ratio of local to global warming, not about a local energy source: the Arctic is not receiving extra forcing, it is converting a shared forcing into more degrees. And the feedbacks are not additive independent boxes despite being reported that way — the decomposition is a convenience, and its numbers should not be read as a ranking with sharp edges.

The misconception to name plainly: "the Arctic warms fast because melting ice exposes dark ocean" is not false, but as a complete account it is wrong, and the evidence against its sufficiency is in plain sight. Amplification appears in the dark half of the year, and models with no sea ice at all still amplify.

• How feedback decompositions are computed, and why their rankings disagree between methods.
• Whether Arctic amplification measurably alters mid-latitude circulation, and why that debate has stayed unresolved.

Every term the collection defines is gathered in the glossary.