Climate Dynamics

, Volume 43, Issue 5, pp 1531–1544

The drivers of projected North Atlantic sea level change

Authors

    • NCAS-Climate, Meteorology DepartmentUniversity of Reading
  • J. M. Gregory
    • NCAS-Climate, Meteorology DepartmentUniversity of Reading
    • Met Office Hadley Centre
  • T. Kuhlbrodt
    • NCAS-Climate, Meteorology DepartmentUniversity of Reading
  • R. S. Smith
    • NCAS-Climate, Meteorology DepartmentUniversity of Reading
Article

DOI: 10.1007/s00382-013-1973-8

Cite this article as:
Bouttes, N., Gregory, J.M., Kuhlbrodt, T. et al. Clim Dyn (2014) 43: 1531. doi:10.1007/s00382-013-1973-8

Abstract

Sea level change predicted by the CMIP5 atmosphere–ocean general circulation models (AOGCMs) is not spatially homogeneous. In particular, the sea level change in the North Atlantic is usually characterised by a meridional dipole pattern with higher sea level rise north of 40°N and lower to the south. The spread among models is also high in that region. Here we evaluate the role of surface buoyancy fluxes by carrying out simulations with the FAMOUS low-resolution AOGCM forced by surface freshwater and heat flux changes from CO2-forced climate change experiments with CMIP5 AOGCMs, and by a standard idealised surface freshwater flux applied in the North Atlantic. Both kinds of buoyancy flux change lead to the formation of the sea level dipole pattern, although the effect of the heat flux has a greater magnitude, and is the main cause of the spread of results among the CMIP5 models. By using passive tracers in FAMOUS to distinguish between additional and redistributed buoyancy, we show that the enhanced sea level rise north of 40°N is mainly due to the direct steric effect (the reduction of sea water density) caused by adding heat or freshwater locally. The surface buoyancy forcing also causes a weakening of the Atlantic meridional overturning circulation, and the consequent reduction of the northward ocean heat transport imposes a negative tendency on sea level rise, producing the reduced rise south of 40°N. However, unlike previous authors, we find that this indirect effect of buoyancy forcing is generally less important than the direct one, except in a narrow band along the east coast of the US, where it plays a major role and leads to sea level rise, as found by previous authors.

Keywords

Sea levelClimate modelAMOCOceanCMIP5

1 Introduction

One of the main causes of the global mean sea level rise is thermal expansion, which was responsible for around half of the sea level rise in recent decades (Church et al. 2011), with most of the remainder due to the addition of mass from glaciers and ice sheets. In future centuries, global mean sea level is projected to rise further, as a consequence of anthropogenic climate change (Meehl et al. 2007). Global mean sea level rise due to thermal expansion is the best understood of the projected contributions. The model-mean estimate of thermal expansion by the end of the twenty first century is 0.13 m for scenario RCP2.6, and 0.28 m for RCP8.5 (Yin 2012). These two scenarios bracket the range of CO2 concentration trajectories for which projections have been made by the coupled atmosphere–ocean general circulation models (AOGCMs) of the Coupled Model Intercomparison Project Phase 5 (CMIP5).

Sea level change is not homogeneous regionally and sea level rise is projected to be greater in some areas than others (Fig. 1a, Pardaens et al. 2011; Yin et al. 2010; Yin 2012). Note that Fig. 1a shows the difference of local sea level change, due to ocean density and circulation change, from the global mean sea level rise (due to thermal expansion and change in mass of the ocean); Fig. 1a and all other fields of sea level change shown in this paper have a global mean of zero. Henceforward, all changes are computed as the difference between the last and first decades of the 100 year simulations.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-013-1973-8/MediaObjects/382_2013_1973_Fig1_HTML.gif
Fig. 1

Multi model ensemble mean for the CMIP5 models of a sea level change relative to the global mean (m), b surface heat flux change into the ocean (W/m2), c freshwater flux change into the ocean (mm/day). The change is between the decade of years 90–99 and the first decade of the 1 % CO2 simulations. The CMIP5 models considered are given in Table 1. Positive values (in red) of both the heat and freshwater fluxes will tend to lower the density

In particular, projected regional sea level change displays a meridional dipole pattern in the North Atlantic, with generally higher sea level rise north of the Gulf Stream around 40°N and smaller sea level rise to the south. The spread among models is also great in this area both in CMIP3 (Pardaens et al. 2011) and CMIP5 models (Bouttes et al. 2012; Yin 2012). The dipole pattern in the North Atlantic cannot be explained by the local wind stress change (Bouttes et al. 2012), which induces relatively lower sea level south of 40°N, but not the relatively higher sea level north of 40°N. Other surface fluxes, such as heat or freshwater, could play a role in that region, in setting both the pattern of sea level change and the spread among models.

The net heat flux and fresh water flux into the ocean both increase at high latitudes in the North Atlantic over the twenty first century simulations in the CMIP5 models (Fig. 1b, c for the 1 % CO2 simulations). Since both of these changes tend to make the water less dense (leading to sea level rise), they could play a role in explaining the sea level dipole in that region.

The increase of freshwater and temperature at high latitudes has also an impact on ocean circulation. Because of the warming and freshening, the formation of deep water and the thermohaline circulation are projected to weaken in climate models (Meehl et al. 2007). The CMIP3 models show a reduction of the Atlantic Meridional Overturning Circulation (AMOC) of 50 % or more under the SRESA1B scenario. The AMOC of the CMIP5 models also weakens under the RCP scenarios, with a best estimate of 26 % for RCP4.5 (Weaver et al. 2012).

This change of AMOC could also be responsible for local sea level change in the North Atlantic. Levermann et al. (2005) showed that a reduction of AMOC induced by additional fresh water in the North Atlantic results in higher sea level in that region. Pardaens et al. (2011) highlight a significant correlation between the sea level rise in the North Atlantic and the AMOC decrease among the CMIP3 models for the SRESA1B scenario. Yin et al. (2009) link the twenty first century high sea level rise along the north-east coast of the United States to a weakening of the AMOC.

In this work we use model simulations, forced by surface buoyancy fluxes, including passive tracers of buoyancy anomalies, to analyse the causes of sea level change in the North Atlantic and why the spread among the projections of CMIP5 models is so high in that region. Using these experiments, we distinguish the effects of the change of AMOC and the change of surface buoyancy fluxes on sea level change in the North Atlantic, which cannot be separated using the CMIP5 model results.

2 Models

We carry out our experiments with the FAMOUS AOGCM (Smith et al. 2008), which is a low-resolution version of the HadCM3 AOGCM (Gordon et al. 2000): its atmosphere component runs on a 7.5º longitude by 5º latitude grid with 11 levels and its ocean component on a 3.75º longitude by 2.5º latitude grid with 20 levels. It is structurally almost identical to HadCM3, and produces climate and climate-change simulations which are reasonably similar to HadCM3, but runs about twenty times faster, and is hence particularly useful for investigations involving many integrations. FAMOUS is a “rigid lid” model: the ocean volume is fixed and the sea level change has to be computed. The rigid lid formulation, which removes external gravity waves, has been used in many models such as FAMOUS parent model HadCM3 and other CMIP models. Different methods can be used to compute sea level change with the rigid-lid approximation, which give similar results (Gregory et al. 2001). In FAMOUS the sea level change, which is obtained by calculating the rigid lid pressure, is described in Lowe and Gregory (2006).

We use FAMOUS because it is much faster than other AOGCMs, and has been spun up to a steady state consistent with its surface fluxes. Because FAMOUS is an AOGCM, unlike ocean-only models forced by prescribed fluxes from climatology, it has the advantage of having rapid feedback between sea surface temperature and surface fluxes, which may be important to variability on short timescales, and could influence the mean state. Ocean-only models with restoring boundary conditions could not be used for an investigation of climate change.

As boundary conditions for the experiments described in Sect. 3, we use the first 100 years of the anomalous surface fluxes from experiments with CMIP5 AOGCMs in which the atmospheric CO2 concentration increases at 1 % per year (By “anomalous” we mean the difference from the control climatology). This is an idealised scenario, which is useful because increasing CO2 gives the dominant anthropogenic radiative forcing of the RCP scenarios. We are able to obtain freshwater fluxes from 17 of these models and heat fluxes from 14 (Table 1). The anomalous surface fluxes are computed as the difference between the monthly means of the 1 % CO2 and the pre-industrial control experiments. We consider only those models for which the diagnostic of sea level change was also available. Changes in the AMOC could also be diagnosed for some of these models (Table 1). In Sect. 3.4, we describe experiments with idealised surface freshwater flux perturbations (“hosing” experiments) used to corroborate our interpretation of the experiments with the CMIP5 surface heat and freshwater fluxes.
Table 1

CMIP5 models and experiments used in this study

Models

Simulations with CMIP5 anomalous heat flux

Simulations with CMIP5 anomalous freshwater flux

AMOC change diagnosed from CMIP5 1 % CO2 experiment

ACCESS1-0

x

x

 

CESM1-BGC

x

x

x

CNRM-CM5

x

x

x

CSIRO-Mk3-6-0

 

x

 

CanESM2

x

x

x

FGOALS-g2

x

x

x

HadGEM2-ES

x

x

x

IPSL-CM5A-LR

 

x

 

IPSL-CM5A-MR

x

x

 

MIROC-ESM

x

x

 

MIROC5

x

x

 

MPI-ESM-LR

 

x

 

MPI-ESM-P

x

x

x

MRI-CGCM3

x

x

x

NorESM1-ME

x

x

x

NorESM1-M

x

x

x

Inmcm4

x

x

x

3 Sea level change due to increasing CO2

3.1 Sea level change in FAMOUS forced by buoyancy fluxes

To evaluate the role of heat and freshwater fluxes, we carry out FAMOUS experiments in which the monthly anomalous surface fluxes from the CMIP5 1 % CO2 experiments (Sect. 2) are applied as surface forcings to the FAMOUS ocean, interpolated in time and updated daily, similar to the method of Lowe and Gregory (2006). Apart from these additional surface fluxes, the FAMOUS simulations are run under control boundary conditions, with constant pre-industrial atmospheric CO2. That is, the forcing of sea level change in FAMOUS comes only from the anomalous buoyancy fluxes. Two sets of experiments are conducted: one with FAMOUS forced by the anomalous surface heat flux from each of the CMIP5 models (“heat experiments”), and one with FAMOUS forced by the anomalous surface freshwater flux (“freshwater experiments”). To simulate the freshwater flux, FAMOUS uses a formulation of the virtual salt flux calculated with respect to the local surface salinity (Smith and Gregory 2009). Note that the wind stress change is not considered here as it has already been studied previously (Bouttes et al. 2012) and has a small effect on sea level change in the North Atlantic.

For the simulations with FAMOUS forced by the heat flux, we use a passive tracer to avoid the multiannual-to-decadal feedback from the SST change on the heat flux. The ocean temperature field T, whose initial state is T0, is forced by F + F’, with F the heat flux computed by FAMOUS during the experiment and F’ the monthly anomalous heat flux prescribed from the CMIP5 model considered (Fig. 2a). As opposed to the temperature T, the passive tracer Tc is forced only by F, while it is initialised with the same temperature field T0. Tc is transported by the ocean circulation like the temperature T. Being a passive tracer, Tc is not used for any dynamical computation in the ocean model, but the surface values of Tc supply the sea surface temperature seen by the atmosphere model. In this way, a feedback against the anomalous heat flux F’ via the SST seen by the atmosphere is avoided.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-013-1973-8/MediaObjects/382_2013_1973_Fig2_HTML.gif
Fig. 2

Schematic representation of the experimental design. In both kinds of experiment, CO2 is fixed at its preindustrial value, and sea level change results only from the imposed anomalous surface flux. a The temperature T, whose initial field is T0, is forced by the surface heat flux F + F’, with F the heat flux computed interactively by FAMOUS and F’ the anomalous heat flux from the CMIP5 models. F’ is computed as the difference between the surface heat flux in each CMIP5 1 % CO2 experiment and its corresponding control. The passive tracer Tc is initialized with the initial temperature field T0, is forced by F only and is transported like the temperature. It allows us to evaluate the effect of the change of circulation and is used as sea surface temperature for the atmosphere to avoid the feedback from changing the SST due to the additional heat flux. b Conversely, the passive tracer Ta is initialized with the temperature field T0 and is transported by the circulation like T, but it is forced by F + F’ and hence includes the effect of the anomalous surface heat flux

The passive tracer Tc is similar to the one giving the redistribution term in Xie and Vallis (2011). It also follows a similar approach as in Banks and Gregory (2006) where a passive anomaly temperature (PAT) is used, but the initialization and the surface boundary conditions for the two passive tracers are different. This is because the passive tracers are used in different ways: here Tc is primarily used to avoid the feedback from the ocean to the atmosphere due to the additional heat, while PAT was used to study the penetration of heat into the ocean. PAT could be diagnosed as the difference between T and Tc.

When FAMOUS is forced by the heat flux change, the sea level change also displays a meridional dipole in the North Atlantic that resembles the one in the CMIP5 models in its north–south contrast, but it differs from the CMIP5 ensemble mean in its longitudinal form (Fig. 3a, c). In FAMOUS the positive pole is in the north-eastern Atlantic while it is in the north-western Atlantic in CMIP5. The negative pole has a larger magnitude in FAMOUS. Such a dipole was not simulated by imposing the wind stress change from CMIP5 in FAMOUS (Bouttes et al. 2012).
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-013-1973-8/MediaObjects/382_2013_1973_Fig3_HTML.gif
Fig. 3

Mean sea level change (left column) (m) and corresponding inter-model SD (right column) for a, b the CMIP5 models (all in m) c, d the FAMOUS experiments forced by the heat flux anomalies from the CMIP5 models, e, f the FAMOUS experiments forced by the freshwater flux anomalies from the CMIP5 models

The difference in the location of the dipole can be linked to the convection sites (diagnosed from the mixed layer depth). In FAMOUS, the main convection site is close to Iceland while there is almost no convection in the Labrador Sea, similar to its parent model HadCM3 (Fig. 4a, b). In the CMIP5 models (Fig. 4c–i), the location of the convection sites varies, especially for the Labrador Sea where it can change from a major convection site (e.g. in MPI-ESM-P, NorESM-M) to almost no convection (e.g. CanESM2, MRI-CGCM3). The differences in the localization of convection sites between FAMOUS and the CMIP5 models can explain the difference in the location of the sea level change dipole, especially its position towards the east of the basin in FAMOUS.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-013-1973-8/MediaObjects/382_2013_1973_Fig4_HTML.gif
Fig. 4

Mixed layer thickness (m) in a FAMOUS, b HadCM3 and ci CMIP5 models (average over 100 years of control simulations)

When the freshwater flux is imposed, FAMOUS also displays a dipole in the North Atlantic, although the positive pole has a smaller magnitude (Fig. 3e). However, the spread among models, which is large in the CMIP5 database in the North Atlantic (Fig. 3b) cannot be explained by the different freshwater fluxes (Fig. 3f), whereas the different heat fluxes lead to different sea level change and a large spread in that region (Fig. 3d). The simulations thus indicate that most of the spread is due to the different heat flux change in the CMIP5 models.

3.2 Roles of anomalous flux and circulation change in CMIP AOGCMs

Dynamic sea level change (i.e. determined by ocean dynamics) can be separated into barotropic and baroclinic components (Lowe and Gregory 2006). The barotropic term is due to the change of depth-mean circulation caused by the wind stress change through the Sverdrup balance. It is usually a small term compared to the baroclinic component. The baroclinic term is due to the density gradients, which also partly depend on the winds due to Ekman pumping and suction, and can be well approximated by the steric sea level change, which is obtained from the local depth-integral of the density change and depends on the local temperature and salinity. Steric sea level change is thus a good approximation to the change in dynamic sea level away from the shelf regions, where the change in mass loading is the primary mechanism for dynamic sea level change (Landerer et al. 2007; Yin et al. 2010). It can in turn be partitioned into two effects: the impact of local anomalous surface fluxes on temperature and salinity, and the effect of the anomalous transport of these tracers (redistribution). By considering steric sea level change, we can analyse the influence of buoyancy fluxes in the North Atlantic, and we can separate the thermosteric and halosteric contributions.

The addition of heat (or freshwater) reduces sea water density and thus causes steric sea level rise. We refer to this as the direct effect of the anomalous surface flux. The indirect effect is due to the change of oceanic circulation. If the surface water becomes less dense, deep convection is inhibited and the AMOC weakens, reducing the ocean heat transport to high latitude. This weakening is apparent in the CMIP5 simulations (Fig. 5a) and in the FAMOUS simulations with the additional heat flux (Fig. 5b), and has been studied for the CMIP3 models (Gregory et al. 2005). For both the CMIP5 models and the FAMOUS simulations, there is a large spread of AMOC weakening. After 100 years of simulation, it is less than 1 Sv in some models and as much as 8 Sv in others.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-013-1973-8/MediaObjects/382_2013_1973_Fig5_HTML.gif
Fig. 5

Evolution of the AMOC strength (Sv), defined as the maximum value of the meridional stream function in the North Atlantic a for the CMIP5 models (the models considered are in Table 1) b for the FAMOUS experiments forced by the heat anomalies from the CMIP5 models. The dashed lines indicate the models for which the AMOC evolution was not available in CMIP5

Previous studies have indicated that there could be a link between the AMOC change and the sea level rise in the North Atlantic (Pardaens et al. 2011; Yin et al. 2009, 2010; Levermann et al. 2005). As an index for the dipole of sea level change in the North Atlantic we consider the difference between the maximum sea level change found anyhere within a box between 20°N and 55°N and 70°W and 40°W, and the minimum found south of the maximum and within the same box. Henceforward, the change of sea level, surface fluxes or AMOC is computed as the difference between the first and last decades of the 100 year simulations. For the CMIP5 models, the correlation between the sea level change and the change of the maximum AMOC strength is neither very strong nor significant (r = 0.41, p = 0.15, Fig. 6b). (The correlation between the sea level change and the AMOC change in CMIP5 is slightly lower than for the CMIP3 models; r = −0.61, p = 0.01, Fig. 7; Pardaens et al. 2011). In CMIP5, the correlation between the sea level change and the change in surface heat flux [averaged over (40ºN–70ºN)] is similar to the correlation between the sea level change and the AMOC change (Fig. 6a, r = 0.45, p = 0.05). The correlation of sea level change with surface heat flux change could be caused by either the direct effect of the anomalous surface heat flux or the indirect effect through the circulation change, but in the CMIP5 models it is not possible to distinguish these possibilities.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-013-1973-8/MediaObjects/382_2013_1973_Fig6_HTML.gif
Fig. 6

Relationship in CMIP5 and FAMOUS experiments between the dynamic sea level change in the North Atlantic and either the AMOC change (maximum value of the meridional stream function in the North Atlantic) or the heat flux change [average over (40°N–70°N)]: a, c sea level change as a function of heat flux change, b, d sea level change as a function of the AMOC change. The change is computed as the difference between the last and first decades of the 100 year simulation

https://static-content.springer.com/image/art%3A10.1007%2Fs00382-013-1973-8/MediaObjects/382_2013_1973_Fig7_HTML.gif
Fig. 7

Relationship in CMIP3 experiments between the dynamic sea level change in the North Atlantic and the AMOC change (maximum value of the meridional stream function in the North Atlantic). The change is computed as the difference between the last and first decades of the 100 year simulation. The CMIP3 models considered are cccma_cgcm3_1_t47, csiro_mk3_0, gfdl_cm2_0, gfdl_cm2_1, giss_aom, giss_model_e_r, iap_fgoals1_0_g, ipsl_cm4, miroc3_2_hires, miroc3_2_medres, miub_echo_g, mpi_echam5, mri_cgcm2_3_2a and ukmo_hadcm3. The outlier with the highest sea level change is miroc3_2_medres

3.3 Roles of anomalous flux and circulation change in FAMOUS experiments

To quantify the role of the two processes, we use the FAMOUS experiments described in Sect. 3.1. In the FAMOUS simulations forced by the heat flux, there is a strong correlation both between the sea level change and the AMOC change (Fig. 6d, r = −0.67, p < 0.01), and between the sea level change and the heat flux change (Fig. 6c, r = 0.52, p = 0.03). The correlation is strong in both cases because the AMOC change and the heat flux change are highly correlated: when the same model is forced with different heat fluxes it reacts in the same way but with different magnitudes. When different models are forced with the same heat flux change, their change of circulation can be different, even if they share some common features. This can be seen in simulations with different models forced by the same freshwater flux (Stouffer et al. 2006). It has also been studied for 1 % CO2 runs by Gregory and Tailleux (2011). Thus in CMIP5 the spread of AMOC response is large despite the common forcing. This could explain why for the CMIP5 models the correlation between the AMOC change and the heat flux change is smaller than in FAMOUS.

To distinguish the direct and indirect influence of the anomalous surface heat flux on sea level change we use the passive tracer Tc that has been added to FAMOUS, whose evolution is due to the redistribution of the existing heat reservoir by the perturbed circulation. Computing the steric sea level change using Tc will thus show the indirect effect. We have also run additional experiments with another passive tracer (Ta) to diagnose the direct effect of the anomalous flux of heat on sea level, excluding its effect via circulation change. In these simulations the temperature T is forced by the heat flux F computed during the experiment by FAMOUS, while the passive tracer Ta is forced by F + F’, where F’ is the anomalous heat flux prescribed from the CMIP5 model considered (Fig. 2b). In this way, computing the steric sea level change using Ta instead of T will show the direct effect of the additional heat flux, which warms the water locally. The anomalous heat flux is passive in these simulations, so the AMOC is not affected by it; the AMOC strength shows unforced variability as usual, but does not have a trend.

The correlation of the steric sea level change in FAMOUS is high with both the heat flux change (Fig. 8a, r = 0.58, p = 0.01) and the AMOC change (Fig. 8b, r = −0.72, p < 0.01). These correlations are higher than with dynamic sea level change in Fig. 6c and d. However, when the change of circulation alone is taken into account, using the passive tracer Tc, the correlation becomes small (Fig. 8c, d, r = −0.28 and p = 0.16 with the heat flux, r = 0.00 and p = 0.50 with the AMOC change). That is, although the AMOC does weaken in response to the heat flux forcing in these experiments (Fig. 5b), the effect on sea level of the redistribution of heat does not correlate with the AMOC change or the applied anomalous heat flux. On the other hand, for the sea level change due to the direct effect of the anomalous flux, using Ta, the correlation is large with the heat flux (Fig. 8e, r = 0.81 and p < 0.01). Hence most of the spread in the sea level change is directly due to the different amount of heat added into the North Atlantic and the corresponding local thermal expansion.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-013-1973-8/MediaObjects/382_2013_1973_Fig8_HTML.gif
Fig. 8

Relationships in FAMOUS experiments between the steric sea level change in the North Atlantic and either the AMOC change (maximum value of the meridional stream function in the North Atlantic) or the heat flux change [averaged over (40ºN–70ºN)]: a steric sea level change as a function of the heat flux change, b steric sea level change as a function of the AMOC change, c steric sea level change (diagnosed from Tc) due to the change of circulation as a function of the heat flux change, d steric sea level change (diagnosed from Tc) due to the change of circulation as a function of the AMOC change, e steric sea level change (diagnosed from Ta) due to the anomalous surface flux as a function of the heat flux change. The change is computed as the difference between the last and first decades of the 100 year simulation

Considering the geographical distribution of sea level change in the North Atlantic (Fig. 9a, b), we see that the sea level rise north of 40°N is mainly due to the anomalous flux (Fig. 9c). The ocean circulation change has the opposite effect on sea level in most places (Fig. 9d): sea level falls nearly everywhere in the North Atlantic, and this explains the net decline in sea level south of 40°N. Thus, anomalous heating and redistribution are together responsible for the dipole pattern of sea level change in the North Atlantic. However, the circulation change leads to a rise in sea level in a narrow band along the east coast of the US (Fig. 9d). Indeed, our simulations demonstrate that the sea level change in this band is due mainly to the effect of the change of circulation, as suggested by Yin et al. (2009).
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-013-1973-8/MediaObjects/382_2013_1973_Fig9_HTML.gif
Fig. 9

Sea level change (m) in the Atlantic in the heat (first row) and hosing (second row) experiments: a, e dynamic sea level change, b, f steric sea level change, c, g steric sea level change due to the anomalous surface flux (diagnosed from Ta) and d, h steric sea level change due to the change of circulation (diagnosed from Tc)

The roles of the anomalous flux and of the change of circulation shown by FAMOUS are in agreement with the results from the HadCM3 model forced by its own heat flux change (Lowe and Gregory 2006). With HadCM3, the anomalous heat flux also leads to higher sea level in most of the North Atlantic while the redistribution effect is generally of the opposite sign, especially in the eastern part of the North Atlantic.

The sea level rise due to the anomalous heat flux (Fig. 9c) is purely thermosteric since the additional heat flux has no effect on the salinity field, so there is no halosteric contribution. The change of circulation has a more complex effect because it modifies both the temperature and salinity fields. Figure 10b, c show the thermosteric and halosteric contributions to the sea level change due to redistribution shown in Figs. 8d and 9a. As studied before, the thermosteric and halosteric terms are often of opposite signs in many places in the North Atlantic and tend to compensate each other (Pardaens et al. 2011; Lowe and Gregory 2006). Due to the compensation between the thermosteric and halosteric sea level change, the resulting sea level change is of smaller amplitude, for example along the coast of the US where the magnitude of the higher sea level rise is smaller in the steric sea level change compared to the thermosteric one.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-013-1973-8/MediaObjects/382_2013_1973_Fig10_HTML.gif
Fig. 10

Steric, thermosteric and halosteric sea level change (m) due to the change of circulation only (with Tc) for the heat (first row) and hosing (second row) experiments: a, d steric sea level change, b, e thermosteric sea level change and c, f halosteric sea level change

3.4 Hosing experiments

To obtain further information about the effect of circulation change on sea level, we have also run simulations where an idealized freshwater flux is added to the North Atlantic. Such “hosing” experiments are often used to evaluate the effect of an AMOC change in models (Stouffer et al. 2006; Levermann et al. 2005; Smith and Gregory 2009; Hawkins et al. 2011). In our hosing experiments, a freshwater flux of 0.2 Sv is uniformly distributed over the North Atlantic between 50°N and 70°N (cf Stouffer et al. 2006). An advantage of imposing this idealized freshwater flux, instead of the freshwater flux from the CMIP5 models, is that it is confined to the North Atlantic where the deep convection takes place, so the source of the perturbation is unambiguous, whereas in the CMIP5 simulations the freshwater change is large in many other places as well (Fig. 1c). For the heat experiments, the interpretation is clearer anyway because the North Atlantic is the region where the magnitude of the heat flux change is greatest (Fig. 1b).

We have run two hosing experiments, with treatment of surface fluxes and passive tracers analogous to the heat experiments, but for salinity instead of temperature. In one experiment, the passive tracer Sc (Fig. 11a), which is analogous to Tc, is used to infer the impact of the change of ocean circulation (redistribution effect). Sc is initialized with the salinity field S0 and is transported like the salinity S. While S is forced by F + F’, which is the sum of precipitation-evaporation (P-E) computed by the model (F) and the prescribed additional freshwater in the North Atlantic (F’), Sc only receives the freshwater flux computed by the model (F). Note that this is simpler than for the corresponding heat experiment because there is no SST feedback on the atmosphere to be avoided. In the other experiment, S is forced only by the P-E computed by the model (F), so that the circulation is the same as in the control experiment, and the additional passive tracer Sa is treated in a similar way to Ta (Fig. 11b). Sa is initialized with S0, but Sa is forced by F + F’, and can be used to evaluate the effect of the anomalous freshwater flux without the change of circulation.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-013-1973-8/MediaObjects/382_2013_1973_Fig11_HTML.gif
Fig. 11

Schematic representation of the experimental design for the hosing experiments. In both experiments, CO2 is fixed at its preindustrial value, and sea level change results only from the imposed anomalous surface flux. a The salinity S, whose initial field is S0, is forced by the surface freshwater flux F + F’, with F the freshwater flux computed interactively by FAMOUS (precipitation-evaporation) and F’ the anomalous freshwater flux added uniformly in the region [50ºN–70ºN] (hosing). The passive tracer Sc is initialized with the initial salinity field S0, is forced by F only and is transported like the salinity. It allows us to evaluate the effect of the change of circulation. b Conversely, the passive tracer Sa is initialized with the salinity field S0 and is transported by the circulation like S, but it is forced by F + F’ and hence includes the effect of the anomalous surface freshwater flux

In the hosing experiments, the sea level change pattern in the North Atlantic also displays a dipole with higher sea level north of 40°N and lower sea level south (Fig. 9e, f). This dipole is similar to the one observed in the CMIP5 mean sea level change. The simulations with the additional passive tracers for salinity confirm the results of the heat experiments: most of the sea level rise in the North Atlantic is due to the anomalous surface flux (in this case the additional freshwater flux, Fig. 9g) while the circulation change leads to a sea level fall (Fig. 9h), except along the coast of the US where both effects result in higher sea level rise. The direct effect of adding fresh water locally is to increase sea level, similar to the effect of adding heat. The effect of the circulation change is also similar to the heat experiments.

The magnitude of the simulated sea level change is bigger than in the FAMOUS experiments with the CMIP5 freshwater flux. This is due to the intensity of the water flux, which is only 0.03 Sv in the CMIP5 ensemble mean (integrated over the same North Atlantic region, but with a non-uniform distribution). The anomalous idealised freshwater flux has a purely halosteric effect on sea level, while the separated thermosteric and halosteric contributions due to the induced circulation change (Fig. 10e, f) have similar patterns to those induced by the CMIP5 anomalous heat fluxes.

3.5 Changes of temperature and salinity

Thermosteric and halosteric sea level change are due to the change of ocean temperature and salinity respectively. In both heat and hosing experiments, the same qualitative processes explain the changes of temperature and salinity.

The direct effect of adding heat or freshwater locally is to increase temperature or decrease salinity in the surface (Fig. 12). This increase of buoyancy is then transported downwards and is responsible for the sea level increase. The pattern of temperature change due to the anomalous heat flux in the heat experiments is more complicated than the salinity change due to the anomalous salinity flux for the hosing experiment because there are widespread and non-uniform changes in surface heat flux in the 1 % CO2 CMIP5 experiments, whereas the simple hosing scenario has a homogeneous freshwater flux added in the North Atlantic only.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-013-1973-8/MediaObjects/382_2013_1973_Fig12_HTML.gif
Fig. 12

Effect of the anomalous surface flux obtained using the passive tracers (Ta and Sa): a temperature (°K) change due to the anomalous heat flux for the FAMOUS heat experiment with the heat flux anomaly from HadGEM2-ES and b salinity (ppt) change due to the anomalous freshwater flux from the hosing experiment (0.2 Sv). The change is computed as the difference between the first and last decades of the 100 year simulation

The indirect effect from the change of ocean circulation includes all processes of tracer transport, such as advection, convection or diffusion. The local increase of buoyancy in the sites of deep convection reduces the formation of deep water. At high latitudes, in the convection sites, this results in less mixing, cooling and freshening in the surface and warming and saltier water in the subsurface (Fig. 13). The AMOC strength is reduced and shows a pronounced shoaling after 100 years of simulation (Fig. 14a, c). NADW occupies a smaller volume while AABW extends upwards. The gyres are also modified with a slower circulation linked to weaker winds (Fig. 14b, d). The gyre circulation transports warm salty water northwards in the western boundary current, and cold fresh water southwards on the eastern side of the basin. Because the circulation is weakened, heat and salt tends to accumulate at low latitudes and on the western boundary; the thermosteric contribution dominates, and increases sea level locally.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-013-1973-8/MediaObjects/382_2013_1973_Fig13_HTML.gif
Fig. 13

a Temperature (°K) and b salinity (ppt) change due to the change of circulation in the hosing experiment (0.2 Sv) obtained with the passive tracers (Tc and Sc)

https://static-content.springer.com/image/art%3A10.1007%2Fs00382-013-1973-8/MediaObjects/382_2013_1973_Fig14_HTML.gif
Fig. 14

a Atlantic Meridional Oceanic Circulation (AMOC) and b barotropic streamfunction at the beginning of the simulation (0–9 years) and c AMOC and d barotropic streamfunction at the end of the simulation (90–99 years), in Sv

4 Conclusions

Most CMIP3 and CMIP5 models simulate a meridional dipole pattern of sea level change in the North Atlantic when CO2 is increased in the atmosphere. The dipole is characterized by relatively greater sea level rise north of the Gulf Stream (around 40°N) and relatively smaller sea level rise south. We have analysed the causes of this feature of sea level change in CMIP5, using experiments with FAMOUS, a low-resolution AOGCM. We find that the sea level change pattern is mainly due to the change of surface heat flux in the North Atlantic i.e. a reduction of heat loss, giving a net heat uptake, in a warmer climate. The heat uptake in the North Atlantic has itself been linked to the AMOC change (Rugenstein et al. 2013): a larger AMOC decline results in relatively more heat uptake in the North Atlantic. The freshwater flux change (increase in precipitation minus evaporation in the North Atlantic) has a similar effect on sea level but with smaller amplitude. Note that the melt water flux from the ice sheets is not taken into account in the CMIP5 models.

The spread in the surface heat flux change among the CMIP5 models accounts for most of the spread in the sea level change. In earlier work, we have shown that wind stress change is not the cause of the North Atlantic dipole in sea level change. While the surface buoyancy fluxes can explain the occurrence of the sea level change dipole and its magnitude, the location of the dipole depends on the sites of deep convection because this will govern where the heat penetrates. Hence on large scales the main features will be the same, but for local details models that accurately simulate deep convection, especially in the Labrador Sea, can be expected to be more reliable.

The change of surface buoyancy flux (whether heat or freshwater) has two effects on the sea level change: the addition of buoyancy causes local steric sea level rise, and it also modifies the buoyancy-driven ocean circulation, especially the AMOC, leading to redistribution of density and consequent sea level change. The direct effect of the local addition of anomalous heat or freshwater is the dominant cause of the meridional dipole in sea level change, and is generally opposed by redistribution. The direct effect accounts for most of the higher sea level rise north of 40ºN, which is mitigated by the change of circulation; the latter, however, explains the lower sea level rise south of 40ºN. Unlike earlier authors, we find that the weakening of the AMOC change is not the cause of the sea level change in most of the North Atlantic. However, along the east coast of the US, there is a narrow band in which the effect of the change of circulation is of the same sign as the effect from the addition of heat or freshwater, resulting in higher sea level rise; future sea level rise on the US coast is thus related partly to AMOC change, as concluded by previous authors.

Acknowledgments

For their roles in producing, coordinating, and making available the CMIP5 model output, we acknowledge the climate modelling groups (listed in Table 1 of this paper), the World Climate Research Programme’s (WCRP) Working Group on Coupled Modelling (WGCM), and the Global Organization for Earth System Science Portals (GO-ESSP). The research leading to these results has received funding from the European Research Council under the European Community’s Seventh Framework Programme (FP7/2007-2013), ERC grant agreement number 247220, project “Seachange”. We thank the reviewers for their comments which helped improve the manuscript.

Supplementary material

382_2013_1973_MOESM1_ESM.jpg (5.6 mb)
Supplementary material 1 (JPEG 5765 kb)
382_2013_1973_MOESM2_ESM.jpg (5.7 mb)
Supplementary material 2 (JPEG 5853 kb)
382_2013_1973_MOESM3_ESM.jpg (4.7 mb)
Supplementary material 3 (JPEG 4798 kb)
382_2013_1973_MOESM4_ESM.jpg (5.5 mb)
Supplementary material 4 (JPEG 5586 kb)
382_2013_1973_MOESM5_ESM.jpg (5.4 mb)
Supplementary material 5 (JPEG 5579 kb)
382_2013_1973_MOESM6_ESM.jpg (5.4 mb)
Supplementary material 6 (JPEG 5564 kb)
382_2013_1973_MOESM7_ESM.docx (12 kb)
Supplementary material 7 (DOCX 11 kb)

Copyright information

© Springer-Verlag Berlin Heidelberg 2013