Search

Abstract: We investigate mixing processes under stratified conditions on the Northwest European Continental shelf using a numerical model (POLCOMS). Our results indicate that convection induced by vertical shearing of horizontal density gradients (‘shear-induced convection’) is a regularly occurring feature in the bottom and surface boundary layers in this open shelf-sea situation. Two types of turbulence models are investigated to study their capability for reproducing the observed location of tidal mixing fronts, and the physical processes occurring in seasonally stratified waters. The first model is a one-equation variant of the Mellor–Yamada model, whereas the second model combines a more recent second-momentum closure with a two-equation model. It is found that generally mean frontal positions (as estimated from ICES data) are predicted more accurately by the two-equation model. The one-equation model reproduces the mean frontal locations to 18.1km (<3 grid spacings) and the two-equation model to 17.1km; although in the Celtic Sea the accuracy is ∼33 and ∼12km, respectively. Comparison with historical tide gauges, current metres, CTD stations, and thermistor chain data from the North Sea Project all show an improvement in accuracy when the two-equation model is used. This is particularly apparent in the model''s ability to reproduce the spring–neap variability during stratification. We find that in the presence of shear-induced convection the routinely applied clipping of the turbulent length-scale, previously thought to be a minor ingredient in a turbulence model, has a dramatic effect on the results: if the length-scale clipping is not applied, substantial over-mixing is observed to occur. The causes and possible remedies of this effect are investigated. Overall our results demonstrate a sensitivity to the details of the turbulence model that is significantly greater than previously thought. [Copyright &y& Elsevier]