Missing brackets when computing the inverse of surface area at u/v points (zs_u_1 and zs_v_1) introduce large errors in the computation of the before scale factors at u/v points (fse3u_b and fse3v_b). The same error occurs during initialisation in domvvl.F90. In both cases:
          zs_u_1(:,:) = 0.5 / e1u(:,:) * e2u(:,:) 
     zs_v_1(:,:) = 0.5 / e1v(:,:) * e2v(:,:) 

should be:
          zs_u_1(:,:) = 0.5 / ( e1u(:,:) * e2u(:,:) )
     zs_v_1(:,:) = 0.5 / ( e1v(:,:) * e2v(:,:) )

This is likely to be a major contributor to the problems reported in #812 which were fixed by avoiding the use on fse3u/v_b in dynspg_ts.F90. That problem should be reinvestigated in the light of this fix.

NOCS 6 - Lateral ocean physics and TOP: Griffies operator for passive tracers

The Griffies's implementation of the Gent and McWilliams eddy transport as a skew flux was implemented in 2010. This has the advantage that its numerical discretization can be written in terms of contributions from quarter cells – ‘triads’. This gives a tighter stencil, disallowing 2-gridpoint numerical noise that is permitted by the advective discretization. Following the the merge of TRA-TRC advection schemes that also occurred in 2010, the Griffies operator can now be tidied and extended to work with passive tracers.

Create a new branch for Neptune effect parameterisation (NOCS).

NOCS.10 — Neptune effect: The Neptune effect (Holloway 1992) is a well established parameterisation of the interaction of eddies with topography. The parameterisation is often seen as essential in Arctic ocean simulations even in high resolution models. The improvments it provide to an ORCA1-LIM simulation is shown in Wang and Holloway (OM, 2009). This implementation concentrates on the bare essentials of the parameterisation and avoids many of the optional extras that appear in later publications.