Fig. 12.6 (a) AMOC strength at 45° N (vertical) against density, averaged over the top 3,000 m in the Nordic Seas/Arctic Ocean. Circles indicate greenhouse gas experiments, squares hosing runs and triangles 'initial perturbation' experiments. (b) Density anomalies (relative to long-term mean of control run) caused by temperature (Ap T , vertical) and salinity (ApS, horizontal). Legend as in (a), but greenhouse gas runs are in red, control run is in green. The dotted line denotes values aPt = aPS

30 decades of the control run (green circles) to allow comparison with anomalies associated with internal variability (e.g. stemming from centennial oscillations of the AMOC (Vellinga and Wu 2004) ). Hosing, greenhouse gas and initial perturbation experiments all cluster in their own regions of the ApS - ApT plane. The three types of experiments sample distinct model states. This result suggests that each class of experiments might involve fundamentally different feedbacks. To what extent this is the case requires further analysis. At this stage we can only suggest that care be taken in transferring conclusions about feedbacks in one class of experiments to those of another.

12.4 Cutting-Edge Questions and Implications for Future Work

As regards future model improvements, there exists a large choice of plausible numerical schemes, parameterizations, parameter values, etc. that could be used in climate models. This kind of uncertainty is inherent to modelling, and can only be quantified using observational constraints (e.g. Knutti et al. 2002; Bony et al. 2006). Our suggestions for 'future work' are therefore by no means exclusive, but we base them on the two results just described (Figs. 12.5b and 12.6) since they are novel, summarise the results of a wide range of model experiments, and seem to pose clear questions for the observer- and modelling-communities that are of more-than-local significance.

0 0

Post a comment