The Atlantic Meridional Overturning Circulation carries warm water north and cold water south. If it collapses, northern Europe loses its heating system. The question everyone asks is: how much warming can it take?
Van Westen, Börner, and Dijkstra show it's the wrong question. The AMOC can tolerate +5.5°C of warming — if you get there slowly, at +0.5 ppm CO2 per year. The ocean adjusts. Salt and heat redistribute. Stabilizing feedbacks have time to operate. Under realistic emission scenarios, the same circulation collapses at +2.2°C to +2.8°C. Same ocean, same physics, same destination. The difference is how fast you approach it.
The mechanism is a stabilization process that operates on long timescales — longer than the present rate of forcing increase. When warming is slow, the ocean can continuously readjust its salinity structure to maintain the density gradient that drives overturning. When warming is fast, the salinity adjustment lags, the density gradient weakens before it can be replenished, and the circulation falls into the collapsed state. The stabilizing feedback exists but can't keep up.
This is a rate-dependent tipping point. The threshold isn't a temperature; it's a temperature reached at a specific speed. The same mountain road has a safe speed and a speed that sends you off the cliff, even though the destination is the same town.
The implication cuts through the usual climate policy framing. Most discussions center on limiting peak warming — the Paris Agreement's 1.5°C and 2.0°C targets. This paper says the speed of getting there matters as much. A world that warms 3°C over 500 years might keep its Atlantic circulation. A world that warms 2.5°C over 100 years might not. The safe temperature depends on the pace.
Rate-dependent thresholds are harder to communicate than fixed thresholds. “Don't warm past 2°C” is a bright line. “Don't warm past 2°C too quickly” requires specifying a timescale that varies with how much of the stabilization capacity has already been consumed. The complexity is real. But the alternative — treating the threshold as fixed and discovering it moved because we approached too fast — is worse.