An assessment of vertical mixing schemes in comparison with observations in the European shelf.

Using the NEMO-shelf model of the Atlantic Marginal Domain with 7km resolution (AMM7) we examine 6 different turbulent closure structural functions, based on the k-epsilon version of the Generic Length Scale Model. The closures include three different models by the Canuto group (2001, a, b and 2010), two by Kantha and Clayson (1994, 2004) and the Galperin (1988) type closure. Each simulation is conducted for the period of 1996 to 2009. The AMM7 model realistically reproduces both the tides and shelf sea processes in the upper and benthic layers, and the depth of the mixed layer and pycnocline. The results have been compared with observations of temperature and turbulence, velocity shear and Brunt-Vaisala frequency collected during 1998-2009. We evaluate the ability of each turbulence closure to reproduce both the direct turbulence observations and large scale physical oceanographic properties. The latter include: upper and benthic mixed layer depths, potential energy anomaly, depth and thickness of the pycnocline, SST and bottom temperature. All models predict depths of mixed layer and pycnocline relatively well, but fail to predict variability in the pycnocline thickness and depth. All models show high correlations of pycnocline depth and bottom temperature with observations, however the ‘less diffusive’ Kantha Clayson and Galperin models have much smaller biases in bottom temperature, while the more diffusive Canuto models provide a better prediction of the pycnocline depth. All models underestimate the dissipation rate of turbulent kinetic energy in the mixed layer and pycnocline at least by an order of magnitude, but have a good agreement with observations in the bottom boundary layer. We discuss the effects of Stoke’s drift velocity and Langmuir circulation in the upper layer and internal waves in the pycnocline to explain disagreement in model results and turbulence measurements.