Introduction

In the context of the Paris Agreement’s global warming targets, it is imperative to evaluate the potential consequences of artificial carbon dioxide (CO2) removal from the atmosphere on both climate and ecosystems1,2,3,4. Many studies have primarily concentrated on examining how global temperatures respond to positive and negative CO2 concentration trends5,6,7,8,9. Recent studies utilizing Earth system models have addressed the climate variability response to net negative CO2 emissions10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29. Europe is well known that the impact of the carbon cycle on climate extremes and agricultural productivity will increase as CO2 emissions rise30,31. Despite significant advancements in understanding these relationships, the effects of climate-carbon cycle interactions on terrestrial productivity in Europe during scenarios involving net CO2 removal remain unclear. To address this issue, we conducted comprehensive simulations using an Earth system model (see Methods) with full complexity. The model was subjected to a 280-year experiment. During this period, the CO2 concentration was increased by 1% per year in the ‘ramp-up phase’ from 2001 to 2140, and then decreased symmetrically by about 1% per year in the ‘ramp-down phase’ from 2141 to 2280. The maximum CO2 concentration quadrupled compared to the initial levels (Fig. 1a), while all other external forcing factors were fixed at their initial values.

Fig. 1: Asymmetry of ecosystems and climate over Europe from CDRMIP.
figure 1

Time series of atmospheric CO2 concentration (a). Differences in accumulated 2 m temperature (b), precipitation (c), total soil moisture within the top 10 cm (d), and GPP (e) over Europe (10°W-40°E, 35°N-70°N) and North Atlantic SST (f) between the ramp-up and ramp-down phases. Negative values indicate a deficit of accumulated climate variables during the ramp-down phase compared to those during the ramp-up phase. Seven-model ensemble mean data from CDRMIP (see Methods) are used for the analysis. Dotted areas represent a 95% significant level.

Full size image

Results

Asymmetry in European GPP from multi-model simulations

To assess an asymmetry in the European climate and ecosystem, we conducted an analysis using multi-model simulation data from the Carbon Dioxide Removal Model Intercomparison Project (CDRMIP)11 (Methods). Figure 1 displays the accumulated differences in climate factors over Europe (10°W-40°E, 35°N-70°N) between the ramp-up and ramp-down phases during boreal summer (June-July-August) using the multi-model ensemble mean data. The results from CDRMIP showed a more pronounced warming and drying climate during the ramp-down phase compared to the ramp-up phase (Fig. 1b, c). The decreased precipitation pattern correlated well with the total soil moisture deficits within the top 10 cm (Fig. 1d). The gross primary productivity (GPP) showed an asymmetry of overall decreases, with the most significant decreases in GPP similar to areas that experienced declines in soil moisture (Fig. 1e). Interestingly, although sea surface temperature (SST) in the North Atlantic showed insignificant change across models, it was cooler, contrary to the expected warming in sub-tropical and polar regions (Fig. 1f). This suggests that SST in the North Atlantic is more closely linked to ocean circulation, specifically the Atlantic Meridional Overturning Circulation (AMOC), known as the ‘warming hole’ during the ramp-up phase32, rather than to the mere release of stored heat during the ramp-down period22,33. To understand the causes of asymmetric climate change in Europe in response to symmetric CO2 variations and to analyze the associated overall decrease in GPP, large-scale ensemble experiments were conducted under the carbon dioxide removal scenario (Methods).

Extremely rapid reduction in GPP during CO2 removal scenario

GPP over Europe (10°W-40°E, 35°N-70°N) during boreal summer exhibited a significant increase during the ramp-up phase (2001-2140) of increasing CO2 concentrations (Fig. 2a, black line), reaching levels as high as 25-30 g C m−2 month−1 (Fig. 2d), which correlated with notably higher 2 m temperatures over Europe (Fig. 2a, red line). Conversely, during the ramp-down phase, GPP experienced a sharp decline and reverted to its original state about 35% quicker than expected, taking approximately 90 years (2230) instead of the anticipated 140 years (2280) (Fig. 2d), despite a symmetrical decrease in temperature compared to the ramp-up phase (Fig. 2a). Our hypothesis suggests that this remarkably rapid reduction in GPP during the ramp-down phase may be attributed to an asymmetrical response in precipitation and associated soil moisture in over Europe. The results reveal that precipitation over Europe showed a consistent decreasing trend during the ramp-up phase, while its recovery during the ramp-down phase was notably slow, particularly between 2141 and 2220 (Fig. 2b). Similarly, soil moisture in 1 m depth exhibited a notable asymmetry between the ramp-up and ramp-down phases. It declined with increasing CO2 emissions, did not recover, and continued to decrease until 2200, despite a 45% reduction in CO2 concentration (Fig. 2c). This lack of recovery of soil moisture was partly carried over from the precipitation and strong soil moisture deficit during the previous spring (Supplementary Fig. 1a, b). Note that the model showed reduced local evaporation during the ramp-up phase and a significant delay in its recovery during the ramp-down phase, implying that local evaporation (Supplementary Fig. 1c, f) was associated with the soil moisture deficit.

Fig. 2: Evolution and asymmetry of ecosystems and climate over Europe in CESM.
figure 2

Time series of atmospheric CO2 concentration (a, black), 2 m temperature (a, red), precipitation (b), total soil moisture within the top 1 m (c), GPP (d) averaged over Europe (10°W-40°E, 35°N-70°N), and SST (e, °C) averaged over the North Atlantic Ocean (50°W-20°W, 40°N-60°N, red box in Fig. 2j) during the ramp up (2001-2140) and ramp-down (2141-2280) phases based on the present-day simulation for 200 years. Time series of AMOC (e, maximum values at 26°N, black) during the ramp-up and the ramp-down phases. Eleven-year running mean data are used, and shading represents one standard deviation from the ensemble simulations. A thick line shows ensemble mean data, and dashed line represents zero. The vertical gray line represents the end of ramp-up periods (2140). f–j Differences in accumulated 2 m temperature (f), precipitation (g), soil moisture (h), GPP (i), and SST (j) between the ramp-up and ramp-down phases. Negative values indicate a deficit of accumulated climate variables during the ramp-down phase compared to those during the ramp-up phase. Ensemble mean data are used for the analysis. Dotted areas represent a 95% significant level.

Full size image

The question that arises from these results is what causes this asymmetry in precipitation and soil moisture. It is well known that the North Atlantic Ocean influences Europe’s climate34,35,36,37. We examined the asymmetry in SST in the high latitudes of the North Atlantic Ocean. Figure 2e shows changes in SST over the North Atlantic Ocean between the ramp-up and ramp-down phases, revealing a strong asymmetric response: SST increases continuously during the ramp-up phase, but it rapidly drops and returns to its initial state after 40 years from the maximum CO2 year. Subsequently, it undergoes a further reduction of −3 °C until 2250 and experiences a slight increase towards the end of the ramp-down phase (Fig. 2e, purple line). Interestingly, the temporal evolution of SST over the North Atlantic Ocean resembles that of GPP over Europe (Fig. 2d). Previous studies have suggested that during the ramp-down phase, changes in North Atlantic SST are influenced by the AMOC38,39,40,41. We further investigated the AMOC response to both the ramp-up and ramp-down phases. During the ramp-up phase, the AMOC was significantly weakened by an increase in surface freshwater fluxes and a reduction in horizontal salinity advection39, which is closely related to the positive salinity advection feedback42. However, during the early ramp-down phase, the AMOC did not show signs of strengthening; instead, it continuously weakened due to the salinity advection feedback and freshening by a decrease in evaporation until the year 220039, followed by a gradual recovery (Fig. 2e, black line). The salinity advection feedback contributes to the delayed recovery of AMOC39,41. This extremely delayed recovery in AMOC resulted in a reduction in meridional heat transport to the high latitudes of the North Atlantic39, leading to rapid cooling there during the CO2 removal period.

Next, we analyzed the horizontal pattern of asymmetry in GPP and its relationship with different climatic factors. Across most of Europe, accumulated GPP was significantly lower during the ramp-down phase compared to the ramp-up phase, and the horizontal pattern of the reduction resembles that seen in CDRMIP (Figs. 1e and 2i). The largest differences in accumulated GPP were observed in regions such as Spain, France, Finland, Serbia and Romania (15-25 g C m−2 month−1). We also examined the differences in accumulated precipitation between the ramp-up and ramp-down phases, and these differences showed negative values for most of Europe except Turkey (Fig. 2g). The spatial pattern of these differences was consistent with that of GPP (e.g. Fig. 2i). The accumulated soil moisture deficit during the ramp-down period was particularly pronounced in southern, central, and western Europe, especially in France, Germany, Spain, Italy, Greece, and England (Fig. 2h), contributing to a significant reduction in GPP in these areas. This decline in southern Europe is possibly attributable to a substantial deficit of precipitation and soil moisture in the boreal spring season (March-April-May) (Supplementary Fig. 1d, e). Such deficits can adversely impact vegetation growth by decreasing photosynthesis. Note that temperature in southern Europe showed weak warming during the ramp-down phase compared to the ramp-up phase (Fig. 1f), suggesting that warming played a minor role in GPP changes due to the overriding effect of severe water limitation there (Fig. 2h and Supplementary Fig. 1b, e). In northern Europe, rapidly cooling temperatures may contribute to a rapid decrease in GPP (Fig. 2f, i), primarily due to the seasonality of vegetation, which exhibits high GPP in summer and low GPP in winter. The temperature effect was particularly pronounced in Northeast Europe, a region where the increase in soil moisture is relatively small compared to Norway (Fig. 2h). Not only GPP but also the overall decrease in precipitation and soil moisture, and even the changes in SST, exhibited patterns similar to those of the multi-model ensemble mean and the inter-ensemble mean (Figs. 1 and 2). This suggests that similar processes are inherent within these patterns.

Furthermore, we examined the changes in SST over the North Atlantic Ocean, revealing intensive cooling in the high latitudes, with a peak located south of Greenland, significantly influencing European climate conditions (Fig. 2j). Overall, the reduction in GPP during the ramp-down phase follows the pattern of precipitation in the central and southern regions of Europe and was strongly impacted by extreme soil water limitations from spring to summer (Fig. 2g–i, and Supplementary Fig. 1d, e). Therefore, we hypothesized that the rapid decline in GPP during ramp down phase is driven by atmospheric teleconnections triggered by North Atlantic SST cooling and the associated slow recovery of AMOC (Fig. 2e, black line). To further explore this hypothesis, we employed both observational data and numerical models in our analysis.

European climate and GPP changes driven by North Atlantic SST

Figure 3 shows the regressed ocean-atmosphere-terrestrial anomalies onto SST over the high latitudes of the North Atlantic Ocean (red box in Fig. 2j) using observational and reanalysis data. To analyze the impact of North Atlantic SST on gross primary productivity, we utilized near-infrared reflectance (NIR) data from 2003 to 201843,44. The NIR data revealed a profound reduction in GPP in central and eastern Europe during periods of cooled North Atlantic SST, with a substantial coefficient of 15-20 g C m−2 month−1 K−1 (Fig. 3d). This reduction accounted for approximately 10-15% of the climatological GPP (Supplementary Fig. 2). It was particularly noticeable in temperate deciduous and coniferous Mediterranean forests, as they are highly sensitive to water availability45. The cooling of North Atlantic SST also led to a significant decrease in precipitation by 70-80% of the average and soil water deficit by 60-70% of climatology in central Europe. These climate conditions are likely to favor decreases in vegetation activity, resulting in a reduced carbon sink46,47,48,49,50,51,52. Note that North Atlantic SST cooling led to positive temperature anomalies over central Europe, but these anomalies did not contribute to the reduction in GPP significantly under strong water limit stress conditions52.

Fig. 3: Relationship between North Atlantic SST and European climate.
figure 3

Regressed JJA observed patterns of surface temperature (a), precipitation (b), total soil moisture within the top 1 m (c), GPP (d), geopotential height at 300 hPa (e, GPH300, with a contour interval of 10 m/K, negative values shown with dashed lines), cloud cover (e, shading) and geopotential height at 850 hPa (f, GPH850, with a contour interval of 2 m/K, negative values shown with dashed lines) and wind (f, vectors) onto North Atlantic SST averaged over 50°W-20°W, 40°N-60°N). The analysis uses data from 2003-2021 and all data are detrended. g–l Difference in the composite of 2 m temperature (g), precipitation (h), total soil moisture within the top 1 m (i), GPP (j), GPH300 (k, with a contour interval of 10 m, negative value shown with dashed lines) and cloud cover (k, shading), GPH850 (l, with a contour interval of 2 m, negative value shown with dashed lines), and wind (l, vectors) between NASST-1°C and NASST + 1 °C model experiments. Twenty ensemble mean data for each experiment are used. Dotted areas represent a 95% significant level. Hatched areas denote a 95% significant level for GPH300 and GPH850.

Full size image

The cooling of SST leads to the development of a surface low-pressure system in the North Atlantic53. In the southeastern counterpart of this surface low, there is a surface high-pressure system present over Europe (Fig. 3f). It is observed that the easterlies along the southern of this high-pressure system act as a drought source, inducing dry air and warming (Fig. 3a), reducing precipitation (Fig. 3b) and soil moisture there (Fig. 3c). Note that local evaporation in Europe was linked to reduced precipitation (Supplementary Fig. 3). A large-scale high-pressure system developed in Europe, while a low-pressure system was observed at the high latitudes of the North Atlantic Ocean (Fig. 3e, contour). This southeast-tilted pressure system further reduced rainfall over Europe. The increased cloud cover in the North Atlantic Ocean (Fig. 3e, shading) leads to less shortwave radiation (Supplementary Fig. 4) sustaining the cooling of SST, which may contribute to European high-pressure system53 resulting in the reduced terrestrial productivity over Europe52.

To further investigate causality, we employed an Earth System Model (ESM) in our study. To quantify the impact of North Atlantic Sea Surface Temperature (NASST) on terrestrial productivity across Europe, we conducted an idealized simulation with the ESM. In this simulation, we imposed nearly uniform +1°C anomalies on the climatological SST on the high latitudes of the North Atlantic region (‘NASST + 1 °C’) and, conversely, reduced SST by -1°C (‘NASST-1°C’) (red box in Fig. 2j, see Methods). Figure 3g–l illustrates the differences in climate and ecosystems between NASST-1°C and NASST + 1°C. The model results successfully replicate the reduction in precipitation (Fig. 3h), corresponding soil moisture (Fig. 3i), and decreased GPP (Fig. 3j) over Europe due to NASST cooling. Note that the model exhibits a slight deviation in simulating the extent of reduced GPP, extending slightly farther southward when compared to observations. Furthermore, the model successfully reproduces the dipole pattern of the pressure system, with anomalous high pressure over Europe and anomalous low pressure in the North Atlantic Ocean (Fig. 3k, l). Additionally, the model demonstrates its capability to simulate the reduced moisture transport from the North Atlantic Ocean to Europe (Supplementary Fig. 5a), resulting in decreased rainfall and soil moisture in Europe. Notably, changes in local evaporation mainly occur in southern Europe, suggesting that it did not significantly contribute to the observed soil moisture deficit (Supplementary Fig. 5b).

To further confirm the robustness of the link between climate conditions and European terrestrial productivity, we analyzed the large ensemble simulation data (see Methods). The results from the ensemble simulations show a significant positive relationship between GPP and precipitation response (Supplementary Fig. 6a), consistent with the findings observed in the difference between the ‘NASST-1°C’ and ‘NASST + 1°C’ experiments. Moreover, the results indicate a strong correlation between soil moisture and GPP (Supplementary Fig. 6b). In summary, the cooling of SST at high latitudes leads to decreased land productivity and carbon sink through its influence on European climate conditions. Therefore, the rapid reduction in GPP during the CO2 removal period (Fig. 4a), which we advocate in this study, can be explained by the rapid cooling of NASST.

Fig. 4: Hysteresis of GPP and AMOC corresponding to CO2 concentration.
figure 4

Hysteresis of GPP change over Europe (10°W-40°E, 35°N-70°N) based on the present-day simulation for 200 years (a) and AMOC (b) corresponding to CO2 concentration, respectively. Black line is the ensemble means. Dots indicate full ensemble spread during the ramp-up (pink), ramp-down (blue), and restoring phases (green), respectively. All lines and dots denote annual means smoothed by the eleven-year running mean.

Full size image

Atlantic meridional ocean circulation-induced sea surface temperature cooling

It has been known that NASST is affected by ocean circulation such as the AMOC, net surface fluxes from the atmosphere, and interaction with Arctic sea ice in high latitudes39,40,54,55. The AMOC was weakened with increasing CO2 concentration and tended to recover very slowly during CO2 removal periods (Fig. 4b). However, the AMOC-induced NASST cooling in the high latitudes of the North Atlantic Ocean was revealed by weakened meridional oceanic advection from the mid-latitudes of the North Atlantic Ocean39,41. The oceanic meridional heat advection was weakened during the ramp-up phase and further weakened during the ramp-down phase, suggesting that this mainly contributes to SST cooling. Note that surface net heat fluxes were negligible for changes in SST during the ramp-down phase39.

More solid support for the AMOC control over SST comes from our ESM-based ensemble simulations in which different initial conditions for the AMOC were imposed over the North Atlantic region. When the AMOC was strongly weakened, the SST was more cooled39,41. Figure 5a shows the positive relationship between AMOC and NASST during the ramp-down phase is very significant with a correlation coefficient of 0.95, suggesting that AMOC controls the NASST cooling. We further explored NASST and GPP over Europe and the result showed a strong positive link between NASST and GPP with a correlation coefficient of 0.62 (Fig. 5b). These additional findings provide further support for our hypothesized link between terrestrial GPP and North Atlantic SST, promoted by favorable climate conditions for terrestrial productivity. Cumulatively, these results suggest a strong linkage between weakened AMOC and fast reduction in European GPP in CO2 removal experiments (Fig. 4).

Fig. 5: Contribution of AMOC to NASST and European GPP.
figure 5

a Scatter plot between AMOC strength and North Atlantic SST (50°W-20°W, 40°N-60°N) during ramp-down (lowest period of both SST and AMOC) from the model ensemble simulation. b Same as a but for GPP over Europe (10°W-40°E, 35°N-70°N). The mean values for 2201-2230 (shading in Fig. 2e) from each ensemble member were used. Twenty-eight ensemble members are used.

Full size image

Discussion

In conclusion, understanding the link between NASST cooling and its impact on terrestrial ecosystems is crucial for effective climate adaptation and sustainable resource management. The cooling of NASST, driven by the delayed recovery of the AMOC during the ramp-down phase, leads to significant precipitation and soil moisture deficits. These adverse climate conditions result in a 35% faster reduction in European GPP than anticipated under a symmetric CO2 path (Fig. 4a), making it one of the largest GPP deficit areas in the world (Supplementary Fig. 7). The rapid decline in GPP is robustly supported by both inter-ensemble and multi-model analyses. This information has vital implications for social and economic policymakers, as the warmer and drier climate not only triggers extreme weather events like flash droughts but also diminishes agricultural productivity.

We further examined the impact of asymmetry in AMOC on the changes in GPP corresponding to the intensity of the AMOC between 2001-2180 and 2181-2360 (Supplementary Fig. 8). All variables in the recovery period, which is a constant CO2 forcing period for 2281-2500, are difficult to compare with the ramp-up and ramp-down phases. In particular, the recovery of the AMOC is not symmetric, which is consistent with previous studies. Notably, there is not only a delayed recovery under transient forcing but also a more pronounced AMOC overshoot compared to the initial state. This phenomenon is consistent across different experiments and is attributed to the increased salinity gradient between the subtropical and subpolar regions during the ramp-up and ramp-down periods. This initiates a salinity advection feedback that transports higher salinity from the subtropical to the subpolar regions, thereby strengthening the feedback mechanism and leading to a faster recovery of the AMOC, ultimately resulting in a stronger AMOC, known as an overshoot. During the period of constant CO2 recovery (after 2281), the AMOC overshoot increases the heat transport to the North Atlantic, which raises the North Atlantic SST. This, in turn, induces increased precipitation over Europe, leads to an increase in soil moisture, and ultimately boosts GPP. The impact of the AMOC overshoot on GPP will be investigated in further studies.

The asymmetry in the European GPP in response to CO2 forcings may be sensitive to the maximum CO2 level, due to the delayed recovery of AMOC and the associated nonlinearly response of North Atlantic SST to the maximum CO2 level. To examine the effect of maximum CO2 on the asymmetry in GPP, we conducted an additional experiment with a maximum doubling of CO2 levels (about 740 ppm). The results show that the asymmetry in European GPP is almost decayed compared to those of the experiment with a maximum of 4xCO2 levels (Supplementary Fig. 9), suggesting that the rapid reduction in GPP during the CO2 removal scenario could occur under significant SST cooling over the North Atlantic region.

In this study, we prescribed CO2 concentrations and did not consider their interactions with the ocean and land ecosystems. To address this limitation and achieve a more realistic representation of the global carbon cycle and its impact on climate and terrestrial productivity, we need to conduct additional model experiments using CO2 emission-driven scenarios. These scenarios account for changes in atmospheric CO2 concentrations resulting from land and ocean biogeochemical processes. This issue will be explored in further studies.

Methods

Diagnosis of the observed data

To obtain the monthly mean SST, we used the National Oceanic and Atmospheric Administration Extended Reconstructed SST version 5 (ERSST)56. Near-surface temperature and precipitation are obtained from Climatic Research Unit CRU TS4.0457 (https://crudata.uea.ac.uk/cru/data/hrg/cru_ts_4.04/). Total cloud fraction, wind, radiation, and geopotential height are from the NCEP2-DOE reanalysis data58 (https://psl.noaa.gov/data/gridded/data.ncep.reanalysis2.html). For evaporation and soil moisture, we used the GLEAMv3.7 data59,60 (https://www.gleam.eu). For the gross primary productivity (GPP) over land, we used near-infrared reflectance (NIR) (g C m−2 month−1), which is based on vegetation absorbing blue- and red-light energy to fuel photosynthesis (2003-2018)43,44.

CO2 removal experiment using CESM

This study utilized the Community Earth System Model version 1.261 (CESM1.2), which consists of the following components: the atmosphere (Community Atmospheric Model version 5, CAM5), ocean (Parallel Ocean Program version 2, POP2), sea ice (Community Ice Code version 4, CICE4), and land models (Community Land Model version 4, CLM4). The atmospheric model has a horizontal resolution of approximately 1° and comprises 30 vertical levels. The ocean model features 60 vertical levels with a longitudinal resolution of 1° and a latitudinal resolution of 0.3° near the equator, gradually increasing to 0.5° near the poles. The land model incorporates the carbon-nitrogen cycle.

Two idealized CO2 pathways were employed (Fig. 1a): the constant and varying scenarios. These scenarios closely followed the protocol of the carbon cycle reversibility experiment (CDR-reversibility)11, except for the initial CO2 level. The constant CO2 experiment spanned 900 years, with a constant CO2 concentration of 367 ppm, representing the Present-Day period (PD). Subsequently, the varying CO2 scenario, comprising 28 ensemble members, was implemented. In this scenario, CO2 concentration increased by 1% per year for 140 years, reaching quadruple the initial level (i.e., 1468 ppm; ramp-up period). This was immediately followed by a symmetric decrease (approximately 1% per year) for 140 years until the CO2 concentration returned to the initial level (367 ppm; ramp-down period). Finally, the model simulated a constant CO2 scenario (367 ppm), representing net-zero emissions, for 220 years (restoring period). The 28 ensemble members were identical except for their atmospheric and oceanic initial conditions, which included different phases of the Pacific Decadal Oscillation62 and the Atlantic Multidecadal Oscillation during the PD period.

Sensitivity of the SST over North Atlantic on climate and terrestrial productivity

To investigate the response of European GPP to changes in SST, we conducted experiments using the Atmospheric General Circulation Model (AGCM), which consists of CAM5 and CLM4, under present-day conditions. SST data were obtained from the monthly climatology of HadlSST for the years 1982-2000, with prescribed SST anomalies of +1°C (NASST + 1°C) and -1°C (NASST-1°C) uniformly over the North Atlantic region (as shown in the red box in Fig. 2j, 50°W-20°W, 40°N-60°N). Twenty ensembles with twenty different initial conditions were used for each experiment. The significance level of the difference between NASST + 1°C and NASST-1°C conditions was determined using a Kolmogorov-Smirnov two-sample test.

CDRMIP model simulation data

The climate and carbon cycle reversibility experiment from the CMIP6 Carbon Dioxide Removal Model Intercomparison Project (CDRMIP)11 was analyzed to investigate the GPP response to large-scale CO2 removal. This experiment, branched from the 1pctCO2 experiment, involves increasing the CO2 level at a rate of 1% yr−1 from pre-industrial levels for 140 years, until it quadruples. Then, a 1% yr−1 removal of CO2 from the atmosphere is prescribed for 140 years until the pre-industrial CO2 level is reached and then held for as long as possible (1pctCO2-cdr simulation). We used seven ESMs (ACCESS-ESM1-5, CESM2, CNRM-ESM2-1, GFDL-ESM4, MIROC-ES2L, NorESM2-LM, and UKESM1-0-LL), which were coupled with the full carbon cycle and performed the CDR-reversibility experiment.