The Arctic is warming at a rate more than three times faster than the global average, a phenomenon known as Arctic amplification. This research analyses the response of the Arctic climate system to the cessation of carbon dioxide (CO2) emissions using six-member ensembles of the NASA Goddard Institute for Space Studies (GISS) modelE2-1-G-CC2. The experiments involve an idealised one-per-cent-per-year increase in CO2 concentrations from pre-industrial levels until cumulative emissions reach totals between 800 and 2000 PgC, at which point emissions are instantaneously set to zero.

Arctic ZEC and associated climate responses

The research findings indicate that while global mean temperatures stabilise following the cessation of emissions, the Arctic exhibits distinctly nonlinear and bifurcating behaviour. In scenarios with lower cumulative emissions (800–1500 PgC), global surface air temperatures (SAT) show an initial cooling followed by a recovery toward cessation levels after approximately 60 years. In contrast, higher-accumulation scenarios (1700–2000 PgC) demonstrate persistent cooling of 0.4–0.6 K in the century following cessation.

The Arctic displays even more pronounced sensitivity than the global average. After emissions cease, the 800–1000 PgC scenarios continue to warm slightly by approximately 0.5 K. However, scenarios exceeding 1700 PgC undergo large-scale cooling of 2–3 K over the subsequent century. This temperature change is mirrored by changes in Arctic sea-ice area (SIA). While the 800 PgC scenario loses an additional 0.5 Mkm2 of ice, the 2000 PgC scenario gains roughly 2.5 Mkm2 within 100 years post-cessation. Notably, despite higher global temperatures, the 2000 PgC experiment reaches nearly the same Arctic SAT as the 800 PgC case after a century, yet with substantially greater ice cover. These results suggest an emergent critical tipping threshold between 1500 and 1700 PgC, beyond which Arctic climate trajectories bifurcate.

Causes of Arctic tipping

The divergent Arctic climate trajectories are primarily associated with the response of the Atlantic Meridional Overturning Circulation (AMOC). In lower-accumulation scenarios, the ensemble-mean AMOC strengthens after emissions cease. Conversely, in higher-accumulation scenarios, the AMOC declines steadily or collapses. The tipping point occurs when atmospheric CO2 concentrations decrease past approximately 650 ppm. In terms of cumulative emissions, this threshold is identified between 1500 and 1700 PgC.

An energy budget analysis reveals that the collapse of the AMOC reduces poleward ocean heat transport (OHT) into the Arctic. In the 2000 PgC scenario, where the AMOC nearly collapses, the OHT response is strongly negative, reaching -3 Wm-2. This reduction in heat transport contributes significantly to the observed Arctic cooling and sea-ice expansion. This cooling is further amplified by secondary feedbacks, specifically the lapse-rate and surface-albedo feedbacks. In the 2000 PgC case, the magnitudes of these feedbacks far exceed those in lower-accumulation scenarios, serving as the leading amplifiers of the temperature response.

Discussion

The existence of this tipping threshold carries significant implications for international mitigation policy. The threshold of approximately 1700 PgC corresponds to a global warming level of 2.7°C at the time of cessation, which exceeds the 2°C target established by the Paris Agreement. The results demonstrate that if CO2 mitigation is delayed beyond the point where the Paris Agreement target is breached, the Arctic is increasingly likely to transition into an alternative climate state characterised by cooling and sea-ice expansion.

The study also notes that individual realisations of the simulations show stochastic divergence due to internal variability. For example, one member of the 1500 PgC ensemble underwent a near-complete AMOC collapse, while one member of the 1700 PgC scenario showed AMOC recovery. This indicates that passing the tipping threshold is modulated by stochastic factors, making regional climate projections and socioeconomic planning across the polar domain more complex. Future research should use additional models and larger ensembles to better quantify the likelihood of passing these thresholds.

Methods

The study utilised an emissions-driven configuration of the NASA-GISS ModelE2.1-G-CC2, which includes fully interactive carbon cycle components coupled with the atmosphere, ocean, and land. The atmospheric component has a horizontal resolution of 2° latitude by 2.5° longitude with 40 vertical layers. The ocean component, GISS Ocean V1, has a resolution of 1° latitude by 1.25° longitude with 40 vertical layers. Radiative feedbacks were quantified using the radiative kernels method, employing CAM5 kernels implemented via the ClimKern package.