Climate models point to Arctic ice loss and ocean shifts at lower warming

The changes identified

The researchers created an automated, reproducible method to identify “strong nonlinear surprises” in CMIP6 projections. They defined these as either abrupt changes over decadal timescales or state transitions that develop over multiple decades and are too large to be explained by the external climate forcing alone without strong internal feedbacks.

Across 54 models and five shared socioeconomic pathways, the method found 73 events in 11 categories. Four categories involved abrupt changes and seven involved state transitions. About 45% of the events concerned sea-ice cover, 19% ocean currents, 29% mixed-layer depth—the depth over which ocean water is stirred—and 7% atmospheric systems such as the Intertropical Convergence Zone.

The analysis also identified linked cascades. In one, changes in sea ice, deep-ocean mixing, temperature and salinity interact. In another, the collapse of the subpolar mixed layer triggers a shutdown of the Atlantic Meridional Overturning Circulation, the large-scale system of currents that includes the flow associated with North Atlantic Deep Water formation. The timing distributions for each category indicated earlier changes and lower global temperature rise than previous reviews had assessed, especially for winter Arctic sea-ice disappearance and the North Atlantic changes.

Why the timing matters

The study suggests that some major ocean and Arctic transitions in climate-model projections appear within lower warming ranges and on earlier timelines than previous reviews indicated. That makes the timing of winter Arctic sea-ice loss, North Atlantic mixing collapse and a transition to a weak Atlantic overturning state important areas for further assessment. The findings describe model-projected possibilities, not confirmed future events.

Evidence and limits

This is a model-based analysis of the CMIP6 archive, covering 54 climate models, five future emissions pathways and regions of at least 1 million square kilometres. An automated algorithm applied stringent criteria to identify abrupt changes over about a decade and transitions developing over several decades. The results are projections rather than observations, and the timing depends on the detection criteria, model simulations and future pathway; the abstract does not establish that every identified event will occur in the real climate system.