A series of vertical mixing schemes implemented in a circumpolar coupled ice-ocean model of the BRIOS family is validated against observations of hydrography and sea ice coverage in the Weddell Sea.Assessed parameterizations include the Richardson number-dependent Pacanowski-Philander scheme, the Mellor-Yamada turbulent closure scheme, the K-Profile Parameterization, a bulk mixed layer model and the Ocean Penetrative Plume Scheme.Combinations of the Pacanowski-Philander parameterization or the Ocean Penetrative Plume Scheme with a simple diagnostic model depending on the Monin-Obukhov length yield particularly good results.In contrast, experiments using a constant diffusivity and the traditional {\it convective adjustment} cannot reproduce the observations.An underestimation of wind-driven mixing in summer leads to an accumulation of salt in the winter water layer, inducing deep convection in the central Weddell Sea and a homogenization of the water column.Large upward heat fluxes in these simulations lead to the formation of unrealistic, large polynyas in the central Weddell Sea after only a few years of integration.We conclude that an adequate parameterization of both wind-induced mixing and bouyancy-driven convection is crucial for realistic simulations of processes in seasonally ice-covered seas.