The authors sincerely thank A. Polonsky for his thoughtful comments on our study, which we address below individually (Polonsky 2015).

Polonsky correctly points out that the work by A. Sarkisyan (overview given in Sarkisyan 2006) regarding the joint effect of baroclinicity and relief (JEBAR) on the large-scale circulation has made an important contribution to the discussion of Sverdrup balance. The JEBAR term, which arises in the depth-integrated vorticity equation as the Jacobian of the potential energy anomaly and the inverse depth, has been considered in detail by a number of authors including Sarkisyan and Ivanov (1971), Holland and Hirschman (1972), Mellor et al. (1982), Mertz and Wright (1992), and Myers et al. (1996). Cane et al. (1998), however, argue that JEBAR fails to accurately characterize the effect of topography on the circulation because it does not fully account for the dynamical feedbacks between the depth-integrated circulation and the bottom flow. In this view, the vertical velocity at the seafloor gives a more meaningful representation of the impact of bottom topography on the large-scale circulation in general and on Sverdrup balance in particular (Drijfhout et al. 2013). In Gray and Riser (2014), we chose to discuss Sverdrup balance and our results in this framework.

Using two different ways to evaluate the accuracy of the Sverdrup relation [given by (5) in Gray and Riser (2014)], we sought to test the assumption that the vertical velocity w is negligible at some depth h, one of the central tenets of classical Sverdrup balance, rather than assume a priori that it holds. In section 5 (Gray and Riser 2014), the possibility that w does not go to zero anywhere in the water column, in which case w at the seafloor must be taken into account, is thoroughly discussed as one reason why Sverdrup balance may fail in some regions. Estimating the along-isopycnal component of w for specific isopycnals in the manner suggested by Polonsky would in principle provide a complementary way of assessing the assumption that w ≈ 0 at some depth h. However, this calculation requires computing the gradient of isopycnal depth and is accordingly quite noisy. The methods used in Gray and Riser (2014) instead involve an integral quantity and were thus preferred. Furthermore, by stipulating that the vertical velocity must be zero for a depth range of at least 200 db in order for Sverdrup balance to be deemed accurate, the second method used to assess this relation was specifically designed to distinguish between a formal minimization and the case where the wind-driven layer is isolated from the deeper circulation by a relatively quiescent layer. While assessing the vertical velocity at the seafloor from topography and absolute horizontal currents would undoubtedly be a worthwhile effort, it was beyond the scope of the study and not possible using the Argo dataset, which does not include measurements below 2000 db.

Although best efforts were made to provide accurate estimates of the uncertainties in the calculations in Gray and Riser (2014), space limitations did not allow for as comprehensive a discussion as we might have liked. We were remiss in failing to mention the errors associated with the satellite measurements of wind stress. However, these were not included because the uncertainty in the mean wind-derived transport [the right-hand side of (5) in Gray and Riser (2014)] due to these errors was an order of magnitude smaller than the uncertainty in the geostrophic transports calculated from Argo. The QuikSCAT-derived wind stress product obtained from Centre de Recherche et d’Exploitation Satellitaire (CERSAT) was provided on a 0.5° grid, and a simple two-dimensional linear interpolation was used to resample the gridded values to the same 1° grid as the mapped Argo velocity fields.

Citing the large interannual variability in the zonally integrated meridional Sverdrup transport at 35°N calculated by Dzhiganshin and Polonsky (2009), Polonsky writes that a change in the averaging period of a few years should substantially change the Sverdrup transport. We found that using different averaging periods did not produce any change in the overall conclusions, and thus this point was not discussed any further. In the energetic boundary regions, large interannual variability was observed; however, Sverdrup balance did not accurately describe the large-scale circulation in these areas regardless of the averaging period. Our analysis was not conducted using zonally integrated values, unlike that of Dzhiganshin and Polonsky (2009). Instead, a point-by-point assessment of Sverdrup balance was made, and therefore our results for the ocean interior were not influenced by the highly variable boundary currents.

We agree with Polonsky that the validity of Sverdrup balance for large parts of the subtropical and tropical ocean implies that the mean meridional transport induced by mesoscale eddies through the divergence of the eddy potential vorticity flux is not of leading order in these areas. However, given the potential impact of mesoscale eddies on horizontal and vertical mixing as well as heat and freshwater transport, we cannot conclude from this fact alone that these features are not important in the dynamics of the large-scale circulation, even in the regions where Sverdrup balance was found to be accurate.