Fig. 12.4 (a) AMOC strength vs. salinity averaged over the region (45-0° W, 50-60° N) for two particular ocean depths (120 and 1,958 m). Here, AMOC strength is defined as the total meridional volume transport in the Atlantic across 50° N between the surface and 666 m (i.e. near to where maximum transport normally occurs in the model). Dotted curves indicate the two-sided 90% confidence intervals of the regression mean. Solid circles show the points on the curve for the model's normal salinity (higher value), and after it is freshened by 0.5 psu
occurs at depth of 1958m iV/^
black circles in Fig. 12.4). Dependence of this 'AMOC sensitivity' on where in the water column freshening occurs is shown in Fig. 12.5a. Sensitivity increases with depth from the surface down to about 600 m, then decreases to become near-zero at intermediate depths around 1,500 m. Towards abyssal depths the AMOC sensitivity grows again, but there the quadratic fit is hardly useful, as quantified by the R2 curve or as seen in the scatterplot of Fig. 12.4b; by the nature of the perturbations,
Fig. 12.5 (a) AMOC weakening (solid line, lower horizontal axis) in response to a freshening of 0.5 psu, applied at a single spot depth (vertical axis). Dotted curves denote the range based on the uncertainty estimate of the regression at each depth (cf. Fig. 12.4). The dashed curve (upper horizontal axis) shows R2, the fraction of variance that is explained by each quadratic fit at each depth). (b) As in (a), but for 0.1 Sv*year (about 3*1012 m3) of fresh water distributed uniformly between the surface and the indicated depth
Fig. 12.5 (a) AMOC weakening (solid line, lower horizontal axis) in response to a freshening of 0.5 psu, applied at a single spot depth (vertical axis). Dotted curves denote the range based on the uncertainty estimate of the regression at each depth (cf. Fig. 12.4). The dashed curve (upper horizontal axis) shows R2, the fraction of variance that is explained by each quadratic fit at each depth). (b) As in (a), but for 0.1 Sv*year (about 3*1012 m3) of fresh water distributed uniformly between the surface and the indicated depth model states with freshening at greater depths are probably less well-sampled. The results suggest that freshening is more effective in weakening the AMOC if it occurs at shallower depths, and less effective at the depth occupied by the overflow water south of the Ridges (1,000 m and deeper).
One can also ask if a given anomalous freshwater loading is more or less effective in affecting the AMOC if it is spread out over a larger vertical section of the water column. The above analysis was repeated, but now for salinity anomalies averaged between the surface and different depths. The AMOC sensitivity was then determined for a given freshwater anomaly of 0.1 Sv*year, by converting that into a salinity anomaly based on the ocean volume occupied by that part of the water column (effectively diluting it with depth). As shown in Fig. 12.5b, the greatest sensitivity occurs if the fresh anomaly is confined to a shallow layer near the top of the water column. If the anomaly is distributed over depth and the salinity anomaly is smaller, AMOC weakening is reduced. The goodness of the quadratic fit between AMOC and salinity turns out to be particularly strong at depths between 400 and 500 m.
One of the motivations to do sensitivity experiments in the form of 'water hosing' is to quantify the effects on the AMOC of any future increases in freshwater flux that may be missed by models due to model imperfections (Stouffer et al. 2006). These might include, for example, the aforementioned uncertainty in projected precipitation change, or in the melt of the Greenland ice sheet which is not usually simulated directly in GCM climate change experiments (although, recently, several groups have begun to include in their climate simulations some of the processes that affect the Greenland ice sheet mass balance: Ridley et al. 2005; Swingedouw et al. 2006).
It seems appropriate to verify how comparable is the model response to freshwater hosing (typically carried out under pre-industrial greenhouse gas concentrations) to that under anthropogenic climate change, where both surface heat and freshwater fluxes are changing. For this we use several experiments carried out with HadCM3. These include the same freshwater experiments used in the hosing sensitivity study (Fig. 12.3) and in the study of sensitivity to the vertical distribution of freshening (Figs. 12.4 and 12.5). In addition we use data from idealized CO2 and SRES forcing scenario experiments for the 21st century. Decadally averaged data from all these experiments show a close relation between the ocean density of the combined Nordic Seas/Arctic Ocean (averaged over the top 3,000 m), and the AMOC (Fig. 12.6a), similar to what has been found in other studies (Hughes and Weaver 1994; Rahmstorf 1996; Thorpe et al. 2001). The relation is approximately linear for density changes of magnitude less than 0.5 kg m-3. For greater density changes the effect on the AMOC saturates. Crucially, all experiments (hosing, initial perturbations, greenhouse gas) follow the same empirical relation.
If, however, the density changes in this region are decomposed into those stemming from changes in temperature (ApT), and those due to changes in salinity (ApS) the different experiments start to fan out, as described in Fig. 12.6b. For instance, in greenhouse gas experiments (red circles) warm temperature anomalies dominate density changes. In hosing runs (black squares) fresh anomalies dominate density changes, though we also note from this figure that the most-extreme freshening effects on density are those with accompanying warm anomalies. In initial perturbation experiments (black triangles) temperature and salinity changes work in opposite ways, but salinity effects dominate. In Fig. 12.6, we also show data from
Was this article helpful?