A number of parameterisations for the simulation of mixing processes in the thermocline are compared and tested against the microstructure data of the PROVESS campaigns, conducted in the northern part of the North Sea during the autumn of 1998. The transport term in the turbulent kinetic energy equation is parameterised via the introduction of a third stability function Sk for turbulent energy diffusion. The formulations are compared with a simpler scheme based upon limiting conditions for turbulence variables. Improved results are obtained with a new form of Sk. The best agreement is, however, found with the simpler limiting scheme. This is explained in terms of a turbulence length scale theory for stably stratified turbulence. In agreement with previous laboratory and ocean data it is found that the ratios of the Thorpe and Kolmogorov scales to the Ozmidov length scale approach critical limiting values in the thermocline. The first of these conditions is satisfied when limiting conditions are implemented into the scheme, providing the necessary minimum value for the dissipation rate, whereas the schemes without limiting conditions fail to produce this critical ratio. The basic reason for this failure is that the Thorpe scale is overestimated, which is shown to be connected to an even larger overprediction of the dissipation rate of temperature variance. To investigate the impact of non-resolved advective processes and salinity stratification on the turbulence predictions, additional numerical experiments were conducted using a simple scheme for data assimilation. The best agreement is found again with the limiting scheme, which is able to make reasonable predictions for the dissipation rate without knowing the detailed shape of the mean stratification profile. It is shown that advective transport due to tidally and wind-driven motions has a non-negligible impact on vertical mixing. This is seen in the data and the models by periodic enhancements of turbulent mixing inside the thermocline.