Fast recovery of North Atlantic sea level in response to atmospheric carbon dioxide removal

Sea level rise, driven by ocean thermal expansion and land ice loss, poses escalating risks to coastal ecosystems, infrastructure, and communities. The IPCC AR6 projects substantial global mean sea level increases by 2100 under a range of scenarios, with large associated socioeconomic impacts. Observations and models show strong regional variability; notably, the Subpolar North Atlantic (SPNA; 45°–65°N, 80°–0°W) tends to rise faster than the global mean, often linked to a slowdown of the Atlantic Meridional Overturning Circulation (AMOC) that reduces northward heat and salt transport. As attention grows on the feasibility and consequences of CO₂ removal, an open question is whether regionally elevated sea levels, particularly in the SPNA, can be reversed on human-relevant timescales and what mechanisms control such reversibility. Given the tight coupling between SPNA sea level and AMOC dynamics, the study tests the hypothesis that SPNA sea level exhibits a nonlinear, hysteretic response to CO₂ removal, potentially enabling a rapid decline relative to the global mean during mitigation.

Prior work establishes that regional sea level change is highly nonuniform, with the North Atlantic exhibiting sensitivity to AMOC variability. Models and observations generally indicate AMOC weakening under warming, associated with amplified SPNA sea level rise and dynamic sea level fingerprints along the U.S. East Coast. Studies on reversibility under CO₂ reduction show that global mean temperature and precipitation can be largely reversible within a century, while subsystems like the AMOC can exhibit hysteresis and overshoot behavior. Earlier analyses (e.g., Sigmond et al.) found continued sea level rise after emissions cessation, with differing regional responses. However, detailed spatial-temporal decomposition of SPNA sea level mechanisms during both ramp-up and ramp-down CO₂ scenarios has been limited. This study extends prior understanding by quantifying the relative roles of thermosteric and halosteric components and mass redistribution, and by linking them explicitly to AMOC hysteresis.

Model: Community Earth System Model version 1.2 (CESM1.2) with CAM5 atmosphere and CLM4 land (~1° × 1°, 30 vertical levels), CICE4 sea ice, and POP2 ocean (60 vertical levels; 1° longitude, ~1/3° latitude at equator gradually increasing to ~1/2° toward poles). POP2 uses the Boussinesq approximation. Experiments: A 900-year present-day control (CO₂ = 367 ppm) was run. From 28 distinct initial conditions sampled from the control, a 500-year CO₂ ramp-up/ramp-down experiment was conducted: CO₂ increased by 1% per year for 140 years to quadruple (1468 ppm), then decreased symmetrically for 140 years back to 367 ppm, followed by an additional 220 years at 367 ppm. This design is akin to the CDRMIP protocol but branched from a present-day state. A two-fold CO₂ ramp test and analyses across CMIP6 CDRMIP models (branched from preindustrial) were also performed for robustness. Diagnostics: Sea surface height (SSH) changes were decomposed under hydrostatic balance into steric (density-driven) and mass redistribution components. Because POP2 is Boussinesq (conserves volume), a globally uniform non-Boussinesq steric correction was applied to approximate global mean steric effects following Greatbatch (1994) and Griffies & Greatbatch (2012). Steric changes were further partitioned into thermosteric (temperature) and halosteric (salinity) contributions using reference-state linearization integrals over the water column. Mass redistribution-induced sea level was inferred as the residual between SSH and local steric in the absence of explicit bottom pressure outputs. The AMOC strength was defined as the average Atlantic meridional overturning streamfunction between 35°N–45°N at 1000 m. Periods: rising period from initial (years 0–10) to global mean sea level maximum (years 221–231); recovery period from that maximum to AMOC maximum (years 308–318). Statistics: Ensemble means over 28 members were used. Spatial significance of differences relative to initial conditions was evaluated via bootstrap resampling (10,000 samples) at each grid; significance at 95% if the 2.5th and 97.5th percentiles share the same sign. Inter-ensemble regressions quantified relationships between AMOC strength and sea level patterns.

- The SPNA exhibits pronounced sea level fluctuations tied to AMOC hysteresis under idealized CO₂ ramp-up and ramp-down forcing. - Magnitude and rates: SPNA sea level rises up to ~0.9 m, increasing during the rising period at about 1.5 times the global mean rate. During recovery, SPNA sea level declines at ~0.64 m per 100 years, about 4.5 times faster than the global mean decline (~0.14 m per 100 years). The SPNA sea level drops below the global mean around year ~283, ~73 years after its peak. - AMOC response: The AMOC weakens during CO₂ increase, continues weakening after CO₂ stabilizes/decreases, and later recovers rapidly, exhibiting an overshoot beyond its initial strength. The timing of the AMOC overshoot coincides with the end of the SPNA sea level decrease, indicating tight dynamical coupling. - Spatial patterns: Sea level rise and fall are skewed toward the northwest SPNA. Weakening during ramp-up reduces Labrador Sea deep convection and slows the subpolar gyre, yielding anticyclonic anomalies and elevated sea level; recovery reverses these, lowering sea level. Mass redistribution from steric gradients elevates sea level on continental shelves, amplifying impacts for the North American east coast (particularly north of 35°N). - Mechanisms by component: Steric changes dominate north of 45°N and split into thermosteric and halosteric parts with a latitude-dependent dipole around 50°N. During rising, AMOC weakening suppresses poleward heat and salt transport, increasing thermosteric south of 50°N and halosteric north of 50°N. Direct radiative warming adds basin-wide thermosteric increases, partly offsetting the dipole signal during ramp-up. During recovery, global cooling plus AMOC strengthening reverse these patterns. The rapid SPNA decline in recovery is largely halosteric (salinity-driven). - Robustness and correlations: Inter-ensemble relationships show significant negative correlations between AMOC strength and SPNA sea level (r ≈ −0.90 during rising; r ≈ −0.95 during recovery). Regression maps confirm negative SPNA sea level sensitivity to AMOC strength. A two-fold CO₂ ramp test and CMIP6 CDRMIP models reproduce key patterns, supporting robustness across scenarios and initialization baselines. - Implications: The northeastern coast of North America is especially susceptible to rapid sea level fluctuations associated with SPNA dynamics under CO₂ mitigation, due to combined steric and circulation-driven mass redistribution effects.

The study addresses whether and how regional sea level in the SPNA can reverse under CO₂ removal. The findings demonstrate that SPNA sea level responds nonlinearly and more rapidly than the global mean, governed by AMOC hysteresis. During CO₂ ramp-down, rapid AMOC recovery and overshoot enhance meridional salinity transport, driving a halosteric-dominated decline that outpaces global mean sea level reduction from thermal contraction alone. This mechanistic linkage clarifies why SPNA sea level can recover faster relative to the global average and highlights the central role of ocean circulation changes, not just thermal expansion/contraction, in regional sea level trajectories. The results are relevant for risk assessment along densely populated North American coastlines, where mass redistribution and dynamic adjustments can substantially modulate local sea levels during both warming and mitigation phases. The robustness across ensemble spread, alternative forcing magnitude (two-fold), and CMIP6 CDRMIP models underscores confidence in the qualitative behavior, even though quantitative values are scenario- and model-dependent.

This work shows that under idealized CO₂ ramp-up and ramp-down forcing, the SPNA experiences amplified and rapid sea level changes linked to AMOC dynamics, with rise about 1.5 times the global mean during warming and decline about 4.5 times faster during recovery. The halosteric component chiefly governs the rapid decline, reflecting restored northward salinity transport as the AMOC rebounds and overshoots. The study advances understanding by providing a spatially and temporally resolved decomposition of mechanisms and by connecting ensemble variability in AMOC to regional sea level responses. These insights have practical implications for coastal risk planning in the North Atlantic. Future research should incorporate interactive land ice, particularly the Greenland Ice Sheet, to account for mass input and potential hysteresis, explore model formulation sensitivities (e.g., AMOC overshoot occurrence), refine initialization to capture historical changes, and assess impacts under more realistic mitigation pathways.

- Land ice mass contributions (glaciers and ice sheets) are not included; Greenland Ice Sheet melt could significantly alter SPNA density, AMOC, and regional sea level and may itself exhibit hysteresis. - Idealized quadrupling and subsequent reduction of CO₂ may not reflect real-world mitigation trajectories; though a two-fold CO₂ sensitivity test was performed, realism remains limited. - Simulations were initialized from a present-day state rather than a preindustrial baseline; although CMIP6 CDRMIP analyses support robustness, initialization differences can affect transient responses. - Some climate models do not exhibit AMOC overshoot; model dependence introduces uncertainty in timing and magnitude of regional sea level changes. - POP2’s Boussinesq approximation requires diagnostic corrections for global mean steric effects, and bottom pressure outputs were unavailable, so mass redistribution was inferred residually. - Model resolution and parameterizations (e.g., subpolar gyre dynamics, deep convection) can influence regional patterns and coastal gradients.