AMOC response to global warming: dependence on the background climate and response timescale

Abstract

This paper investigates the response of the Atlantic meridional overturning circulation (AMOC) to a sudden doubling of atmospheric CO₂ in the National Center for Atmospheric Research Community Climate System Model version 3, with a focus on differences under different background climates. The findings reveal that the evolution of the AMOC differs significantly between the modern climate and the last glacial maximum (LGM). In the modern climate, the AMOC decreases (by 25 %, 4 Sv) in the first 100 years and then recovers slowly (by 6 %, 1 Sv) by the end of the 1,500-year simulation. At the LGM, the AMOC also weakens (by 8 %, 1 Sv) in the initial 90 years, but then recovers, first rapidly (by 30 %, 4 Sv) over the following 300 years, and then slowly (by 13 %, 1.6 Sv) during the remainder of the integration. These results suggest that the responses of the AMOC under both climates have a similar initial rapid weakening period of ~100 years and a final slow strengthening period over 1,000 years long. However, additional intermediate period of ~300 years does occur for the LGM, with rapid intensification in the AMOC. Analyses suggest that the rapid intensification is triggered and sustained primarily by a coupled sea ice–ocean feedback: the reduction of meltwater flux in the northern North Atlantic—associated with the remarkable sea-ice retreat at the LGM—intensifies the AMOC and northward heat transport, which, in turn, causes further sea-ice retreat and more reduction of meltwater. These processes are insignificant under modern conditions.

Appendices

Appendix 1: A detailed heat and freshwater budget

To quantitatively evaluate the importance of different processes and feedbacks in changing the AMOC in doubling CO₂ experiments, we analyse the heat/freshwater budget for both the northern and southern box in the upper Atlantic Ocean. The time-varying equation for the volume-integrated heat/freshwater budget is

$$\frac{{dM_{\text{S}} }}{dt} = M_{\text{F}} - {\text{div}}M_{\text{T}} + R$$

(3)

where $M_{\text{S}}$ is the heat/freshwater storage, $M_{\text{F}}$ the area-integrated surface flux, ${\text{div}}M_{\text{T}}$ divergence of the transport between the northern and southern boundary of the box, and $R$ is the residual, including diffusion and convection with the lower ocean. The transport across certain latitude $M_{\text{T}}$ can be further separated into the meridional part ($M_{\text{MOC}}$) and the azonal part ($M_{\text{az}}$). Take the freshwater transport as an example,

$$M_{\text{MOC}} = - \frac{1}{{S_{0} }}\int {\overline{v} (z) \cdot \left( {\left\langle {\overline{S} } \right\rangle - S_{0} } \right) \cdot {\text{d}}z}$$

(4)

$$M_{\text{az}} = - \frac{1}{{S_{0} }}\int {\overline{{v(z)^{{\prime }} \cdot S^{{\prime }} }} \cdot {\text{d}}z}$$

(5)

where the reference salinity S ₀ is the averaged salinity of the Atlantic Ocean, 34.7 and 36.5 for the present day and the LGM, respectively. $\overline{v} \left( z \right)$ and $\left\langle {\overline{S} } \right\rangle$ denote the zonally integrated northward velocity and averaged salinity, and $v\left( z \right)^{{\prime }}$ and $S^{{\prime }}$ represent the deviations from their zonal means.

The changes in heat storage, surface flux and convergence of transport in the northern and southern box during each stage are listed in Table 3. For both the MOD-2CO2 and LGM-2CO2, the increase of heat storage (0.04 PW) in the northern box during the initial weakening stage is mainly attributable to the increase of surface heat flux (0.10–0.13 PW). The convergence of heat transport is negative (−0.08 to −0.11 PW), indicating a negative feedback between the strength of the AMOC and heat transport. It works as follows: a weakening of the AMOC reduces the northward heat transport, leads to a cooling, and promotes the deep convection. It is interesting to note that during the intermediate strengthening stage in LGM-2CO2, the convergence of heat transport is positive (0.06 PW) because of the enhanced AMOC, and simultaneously the surface heat flux decreases significantly (−0.05 PW) due to the reduction of the sea-ice insulating effect and the increase of ocean temperature. The former, again, suggests a negative feedback between the AMOC and heat transport, while the latter indicates a positive feedback between the AMOC and the surface heat flux. This positive feedback can act through two different loops: Firstly, intensification in the AMOC transports more heat northward, leads to a warming and enhanced heat loss to the atmosphere. Secondly, a strengthening of the AMOC and heat transport can cause more sea-ice retreat and enhanced heat loss through the sea-ice insulating effect.

Table 3 Changes in the heat storage, surface flux and convergence of transport (units: 10⁻¹ PW) in the northern (35°N–65°N) and southern box (0°N–30°N) in the upper North Atlantic Ocean in the doubling CO₂ experiments

The changes in freshwater storage, surface flux and convergence of transport in the northern and southern box during each stage are listed in Table 4. The most important process is the decrease of surface freshwater flux in the northern box (−0.06 Sv for the intermediate stage) caused by the reduction of meltwater flux (Fig. 11b). It could form a positive sea ice–ocean coupled feedback to intensify the AMOC. At the intermediate stage, the meridional part of freshwater transport (−0.12 Sv) acts to increase salinity in the northern box and stabilize the AMOC (a positive feedback) in LGM-2CO2; however, its role is overwhelmed by the azonal transport (0.15 Sv). The decrease of surface freshwater flux (−0.03 to −0.04 PW) in the southern box due to the enhanced evaporation in both climates during the initial weakening stage is another positive feedback, although overpowered by the negative feedback between the AMOC and surface heat flux.

Table 4 Changes in the freshwater storage, surface flux and convergence of transport (10⁻¹ Sv) in the northern (35°N–65°N) and southern box (0°N–30°N) in the upper North Atlantic Ocean in the doubling CO₂ experiments

Appendix 2: Confirmation from the EOF analysis

In order to confirm the crucial role of salinity changes in enhancing the N–S density contrast and, in turn, the intensification of the AMOC in LGM-2CO2, we use the empirical orthogonal function (EOF) analysis to detect the major modes of changes in zonal mean density, temperature and salinity in the doubling CO₂ experiments (Fig. 13). In MOD-2CO2, the EOF1 of zonal mean density (Fig. 13a) resembles the EOF1 of zonal mean temperature (Fig. 13d) very well with an N–S symmetric decrease of density, except for the northern polar region. This quasi-symmetric change of density agrees with insignificant changes in the AMOC at long timescales in MOD-2CO2. In sharp contrast, in LGM-2CO2, the evolution of the zonal mean density in the Atlantic Ocean is dominated by an N–S asymmetric mode (EOF1) with a significant increase of density in the upper 2,000 m and 40°S northward Atlantic Ocean and a comparable decrease in the deep and Southern Ocean (Fig. 13b). EOF1 can explain 92 % of the total variance and the corresponding PC1 (Fig. 13c) suggests that this mode has timescale longer than 1,000 years. This asymmetric mode can produce a stronger NADW and a weaker AABW through the increase of the N–S density contrast and the decrease of vertical stratification in the Atlantic Ocean, which is consistent with the strengthening of AMOC after the initial weakening in LGM-2CO2 (also shown in the initial stage of PC1). In order to find out, the relative contribution of the changes in temperature and salinity, we carry out the same EOF analysis for the zonal mean temperature and salinity in the Atlantic Ocean (Fig. 13e, h). The first EOF of temperature, which can explain 84 % of the total variance, shows the signal of global warming due to the doubling of CO₂. The warming in the upper and mid Atlantic Ocean is larger, which suggests the temperature change makes a negative contribution to the N–S density contrast. The first EOF of salinity, explaining 94 % of the total variance, resembles the asymmetric mode very well, indicating that the change of zonal mean salinity dominates the asymmetric change of zonal mean density in the Atlantic Ocean in LGM-2CO2.

Fig. 13

First EOF of the Atlantic zonal mean potential density in the doubling CO₂ experiments for the modern climate (a) and LGM (b). The corresponding PCs for modern climate (red) and LGM (blue) are shown in c, d, e and f are the same, but for the zonal mean temperature, g, h and i are for the zonal mean salinity. All the EOFs are normalized, such that the magnitude is unit. The variance explained by the first EOF of the potential density, temperature and salinity are 74, 89 and 25 % for the modern climate, and 92, 84 and 94 % for LGM, respectively

The EOF analysis is coherent with the diagnosing of the trend of zonal mean potential density (Sects. 5.1–5.3) and confirms that the contribution from salinity changes plays the dominant role in increasing the N–S density contrast and strengthening the AMOC in LGM-2CO2 at centennial-millennial timescales. Therefore, we can exclude the role of Atlantic surface heat flux. Combining with the diagnose of the freshwater budget for the North Atlantic in Sect. 5.4, we could find, once again, that the reorganization of the Atlantic surface freshwater flux is of essential importance in strengthening the AMOC in LGM-2CO2. EOF analysis (Fig. 13h) also demonstrates that the increase of salinity in the north could be coupled with a decrease of salinity in the south at millennial timescale.