Further work and PFT�� chemical structure additional sensitivity experiments should help clarify this point. Analysis of heat, freshwater and volume transport was done using PAGO

(http://www.whoi.edu/science/PO/pago/). DS thanks Laurent Bopp and Christian Ethé for their help in the setup of the CM5_piCtrl_NoBio simulation. Arnaud Caubel, Sébastien Denvil, Marie-Alice Foujols and the whole team of the “pole de modélisation de l’IPSL” are also acknowledged for their work in carrying CMIP3 and 5 simulations from IPSL. This work has benefited from the support of LEFE-MISSTERRE. The authors are grateful to the reviewers and editor for their valuable comments which helped to improve the manuscript. “

“The publisher regrets that the value for Eq. (3) was not incorporated appropriately for the above paper. The equation should be read as: C10=10-4-0.0160U102+0.967U10+8.058 The publisher would like to apologize for any inconvenience caused. “

“The selleck range of temporal and spatial scales of ocean flows is vast, differing from hours to centuries and metres to thousands of kilometres. The ocean is also full of transient features that can change in both size and/or location; examples include algal blooms, dense water overflows and mesoscale eddies. In an ocean model how much, when and where to place numerical resolution, both spatial and temporal, must be considered and cannot necessarily

be predicted a priori. Adaptive meshes, which coarsen or refine depending on the evolution of the flow,

support efficient use of available computational resources and, in principle, do not require an extensive a priori knowledge of the dynamics (e.g. Behrens, 1998, Jacobs et al., 2013, Munday et al., 2010 and Popinet and Rickard, 2007). Using an adaptive mesh adds another layer of numerical complexity to a model. The performance of such meshes and the implications for the computed flow dynamics therefore require careful consideration. Adaptive mesh techniques have been used relatively widely in computational fluid dynamics (Baker, 1997, Cao, 2005, Frey and Alauzet, 2005, Remacle Urocanase et al., 2005, Speares and Berzins, 1997 and Venditti and Darmofal, 2003), with the use of adaptive meshes in ocean modelling still under development (Piggott et al., 2009). For structured meshes, studies include the application and extension of a quadtree based adaptive structured mesh Navier–Stokes solver (Gerris) to ocean flows (Popinet and Rickard, 2007) and investigation of a general adaptive structured mesh tool (Blayo and Debreu, 1999). For unstructured meshes, the studies have focused predominantly on the shallow-water equations (Behrens, 1998, Bernard et al., 2007 and Remacle et al., 2006) with limited applications in three dimensions (Munday et al., 2010, Piggott et al., 2008 and Power et al., 2006).