Performance of a second-order moments advection scheme in an Ocean General Circulation Model

The reliability of Ocean General Circulation Models (OGCMs) strongly depends on the quality of their tracer advection schemes. For the sake of simplicity and computing time, tracer advection schemes most commonly used in large-scale OGCMs tend to be low-order schemes, which suffer from spurious numerical diffusion and dispersion that result in distorted solutions. The application of high-order schemes would reduce numerical errors, but at a considerable cost in terms of computing time. An alternative to the use of high-order methods is the implementation of algorithms that take into account the sub-grid distribution of tracers. One such method is the Second-Order Moments (SOM) scheme of Prather (1986), which is more accurate than a fourth-order scheme, but at the time consumption of a second-order algorithm. This article presents results from coarse-resolution, global-ocean simulations with very low explicit diapycnal mixing, in which active and passive tracers were advected with the SOM method. We compare the performance of the method with that of more traditional schemes, namely, the FCT (flux corrected transport) and QUICKer (quadratic upstream interpolation for convective kinematics) schemes. In general, the use of the SOM method significantly improves tracer distributions and transports compared to FCT and QUICKer, thus leading to a better representation of ocean currents, notably boundary currents and frontal systems. While model simulations employing the FCT and QUICKer schemes recreate a global overturning circulation with strong upwelling occurring in low latitudes, the SOM simulations admit a circulation pattern closer to that known as the "reconfigured conveyor belt'' (Toggweiler and Samuels, 1993), in which the bulk of the global ocean upwelling occurs in the Southern Ocean.