On modeling turbulent exchange in buoyancy-driven fronts

•Comparisons of small-scale processes using two numerical approaches (LES and OGCM).•Two classic test cases are investigated: lock-exchange and mixed layer instability.•The impact of common OGCM modeling choices in the solutions are quantified.•Choice of turbulence closure is important when modeling submesoscale processes.

Our primary objective is to quantify the uncertainty in the solution space associated with mixing and stirring in ocean general circulation models (OGCMs) due to common modeling choices, namely the spatial resolution, tracer advection schemes, Reynolds number and turbulence closures. In many cases the assessment of errors is limited by the observational data set, therefore, large eddy simulations from a spectral element Boussinesq solver are taken as ground truth. First, the lock-exchange problem is used to quantify the temporal evolution of mixing from shear-driven stratified overturns. It is found that mixing in an OGCM is more sensitive to the choice of grid resolution than any other parameters tested here. The results do not monotonically converge towards the ground truth as the resolution is refined. Second, stirring of a passive tracer by submesoscale eddies generated by surface density fronts is considered. We find that using a second-order turbulence closure leads to an accurate representation of the restratification in the mixed layer.