Fast reduction of Atlantic SST threatens Europe-wide gross primary productivity under positive and negative CO2 emissions

In the context of the Paris Agreement’s global warming targets, it is imperative to evaluate the potential consequences of artificial carbon dioxide removal from the atmosphere on both climate and ecosystems. Prior work has mainly examined temperature responses to positive and negative CO2 trends, and recent Earth system model studies have explored climate variability under net negative CO2 emissions. Europe is expected to see increasing influence of the carbon cycle on climate extremes and agricultural productivity with rising CO2. Yet, the effects of climate–carbon cycle interactions on terrestrial productivity in Europe under net CO2 removal remain unclear. To address this, the authors used a full-complexity Earth system model in a 280-year experiment with 1% yr−1 CO2 increase from 2001–2140 (ramp-up) to 4× initial concentration, followed by symmetric 1% yr−1 decrease from 2141–2280 (ramp-down), with other forcings fixed, to assess European GPP responses and associated climate mechanisms.

The study builds on literature showing asymmetry and hysteresis in climate–carbon responses to positive versus negative CO2 emissions, including global temperature proportionality and irreversibility under negative emissions, and regional hydrological and circulation changes under CO2 ramp-down scenarios. Prior CDRMIP and CMIP experiments documented AMOC weakening under rising CO2 and delayed recovery during CO2 removal, the North Atlantic “warming hole,” and teleconnections affecting European hydroclimate and extremes. Studies have highlighted Europe’s sensitivity of carbon uptake and GPP to droughts and water limitations, with significant impacts during events like the 2003 and 2018 heat–droughts. However, explicit implications of NASST changes and AMOC recovery for European GPP under negative emissions had not been quantified, motivating the present analysis.

The authors combined multi-model analysis, targeted ESM ensembles, AGCM sensitivity experiments, and observations: - Multi-model: Used seven-model ensemble from CMIP6 CDRMIP (ACCESS-ESM1-5, CESM2, CNRM-ESM2-1, GFDL-ESM4, MIROC-ES2L, NorESM2-LM, UKESM1-0-LL) following 1pctCO2-cdr protocol (1% yr−1 increase for 140 years to 4×CO2, then 1% yr−1 removal for 140 years back to preindustrial, then hold). - CESM1.2 large ensemble: Fully coupled ESM (CAM5, POP2, CICE4, CLM4 with C–N cycle). Present-day constant CO2 (367 ppm) spin and control, then varying-CO2 experiment with 28 ensemble members: ramp-up 1% yr−1 for 140 years to 4×CO2 (1468 ppm), immediate symmetric ramp-down 1% yr−1 for 140 years back to 367 ppm, then constant-CO2 restoring for 220 years. Atmospheric and oceanic initial conditions varied (e.g., different PDO/AMO phases). - AGCM sensitivity: CAM5+CLM4 under present-day conditions with imposed uniform SST anomalies over high-latitude North Atlantic (50°W–20°W, 40°N–60°N): NASST +1°C and NASST −1°C; 20-member ensembles each; differences assessed with Kolmogorov–Smirnov tests. - Observations/reanalyses: ERSSTv5 for SST; CRU TS4.04 for near-surface temperature and precipitation; NCEP-DOE R2 for geopotential height, winds, cloud, radiation; GLEAMv3.7 for evaporation and soil moisture; satellite NIRv-based GPP (2003–2018). Analyses focused on Europe (10°W–40°E, 35°N–70°N) and North Atlantic (50°W–20°W, 40°N–60°N). Temporal smoothing with 11-year running means; accumulated differences computed between ramp-up and ramp-down phases; statistical significance assessed at 95% where indicated.

- European GPP asymmetry: Boreal-summer GPP increased during ramp-up, reaching about 25–30 g C m−2 month−1 above baseline, but during ramp-down it declined sharply, returning to near-initial levels ~35% faster than a symmetric expectation (≈90 years by ~2230 vs 140 years by 2280), despite a roughly symmetric temperature decrease. - Hydroclimate drivers: Precipitation decreased during ramp-up and recovered very slowly during ramp-down (notably 2141–2220). Soil moisture (top 1 m) showed strong hysteresis: it declined with increasing CO2 and continued decreasing until ~2200 despite ~45% lower CO2, consistent with spring deficits and delayed recovery of local evaporation. - North Atlantic SST (NASST) cooling: During ramp-down NASST rapidly cooled, returning to initial values within ~40 years after CO2 peak, then cooling further by about −3°C until ~2250 before slight recovery. The temporal evolution of NASST paralleled European GPP decline. - AMOC mechanism: AMOC weakened during ramp-up due to freshwater flux and reduced horizontal salinity advection; during early ramp-down it continued weakening until ~2200 due to salinity-advection feedback and reduced evaporation, then gradually recovered. Reduced meridional heat transport led to high-latitude NASST cooling. Surface net heat fluxes contributed little to NASST change in ramp-down. - Spatial patterns: Accumulated GPP during ramp-down was significantly lower across most of Europe, with largest reductions over Spain, France, Finland, Serbia, and Romania (≈15–25 g C m−2 month−1). Accumulated precipitation and soil moisture also showed widespread deficits (except parts of Turkey), aligning with GPP reductions. Southern Europe’s GPP decline was dominated by severe water limitation (spring–summer precip and soil moisture deficits); northern Europe’s rapid temperature cooling additionally reduced GPP. - Observational and AGCM evidence: Regressions onto NASST (observations 2003–2021) showed reduced GPP in central/eastern Europe with coefficients ~15–20 g C m−2 month−1 K−1, decreased precipitation by ~70–80% of climatology and soil water by ~60–70% during cool NASST states. AGCM NASST−1°C vs +1°C experiments reproduced European high-pressure anomalies, reduced moisture transport, decreased precipitation and soil moisture, and reduced GPP. - Robust linkages: In CESM ensembles during ramp-down, AMOC–NASST correlation r≈0.95 (P<0.01) and NASST–European GPP correlation r≈0.62 (P<0.01). Ensemble analyses showed significant positive relationships between GPP and precipitation and between GPP and soil moisture. - Sensitivity to CO2 peak: A reduced maximum CO2 experiment (2×CO2 peak) largely diminished the GPP asymmetry compared with the 4×CO2 case, indicating strong dependence on the magnitude of NASST cooling.

The study shows that under symmetric CO2 ramp-up and ramp-down, European terrestrial productivity exhibits pronounced hysteresis: GPP rebounds much more slowly and even declines rapidly during CO2 removal due to hydroclimatic deficits linked to North Atlantic dynamics. The delayed AMOC recovery reduces poleward heat transport, causing high-latitude NASST cooling. This triggers a southeast-tilted pressure dipole (low over the subpolar North Atlantic, high over Europe), easterly anomalies and moisture transport reduction, suppressed precipitation, and soil moisture deficits, leading to rapid GPP reduction. The findings address the central question by identifying AMOC-mediated NASST cooling as the key driver of asymmetric European GPP response. The results are consistent across multi-model (CDRMIP), large CESM ensembles, targeted AGCM sensitivity tests, and observational regressions. Socio-environmental implications are substantial: warmer and drier European conditions with flash-drought potential can diminish agricultural productivity and carbon sink capacity during negative emissions. During the subsequent constant-CO2 restoring period, an AMOC overshoot can elevate NASST and partially reverse hydroclimate stress, increasing European precipitation, soil moisture, and GPP, underscoring the dynamical ocean’s control. The magnitude of asymmetry depends on the peak CO2 level, with weaker asymmetry for 2×CO2, implying risk is highest for large overshoots before removal.

This work demonstrates that during CO2 removal, delayed AMOC recovery induces rapid North Atlantic SST cooling, leading to European precipitation and soil moisture deficits and a 35% faster-than-expected decline in GPP under a symmetric CO2 pathway. The AMOC–NASST–GPP linkage is supported by multi-model CDRMIP results, CESM ensembles (r≈0.95 AMOC–NASST; r≈0.62 NASST–GPP), observational regressions, and AGCM sensitivity experiments. The study provides actionable insight that large-scale negative emissions could unintentionally exacerbate terrestrial productivity losses in Europe, affecting agriculture and carbon sinks. Future work should: (i) conduct emission-driven experiments coupling interactive land and ocean biogeochemistry to capture carbon–climate feedbacks; (ii) investigate the mechanisms and impacts of AMOC overshoot during the restoring period; and (iii) assess scenario dependence on the magnitude and rate of CO2 overshoot and removal, including regional adaptation strategies.

- Forcing design: CO2 concentrations were prescribed rather than emission-driven; interactions with land and ocean biogeochemical cycles were not included, potentially affecting atmospheric CO2 trajectories and feedbacks. - Phase comparability: Variables during the constant-CO2 restoring period are not directly comparable to ramp-up/down due to differing forcings and AMOC asymmetries, complicating interpretation. - Model dependence and biases: Although results are robust across models and ensembles, specific spatial details (e.g., extent of GPP reductions) show deviations in the AGCM relative to observations. Representation of soil moisture–evaporation feedbacks and cloud–radiation processes may contribute to uncertainties. - Mechanistic simplifications: Surface net heat fluxes were found negligible for NASST changes in ramp-down in this setup; alternative model physics or internal variability states could alter relative contributions. - Sensitivity to CO2 peak: Asymmetry strength depends on maximum CO2 level; generalization to other pathways and rates warrants further testing.