Beschreibung: This paper is devoted to a study on the effects of different parameterisations of vertical turbulence – with a 3D hydrodynamic model COHERENS – on the accuracy of calculated temperature and salinity fields in a hydrodynamically complex test area – the Baltic Sea, Gulf of Finland. Two algebraic parameterisations and k–ε and k-models were used. For k-model four different sets of stability functions were used. Calculated vertical profiles of temperature and salinity were compared against CTD-profiles collected during a measurement campaign in the Gulf of Finland in summer 1996. The dataset has an outstanding spatial and temporal coverage including over 300 measured CTD profiles. The thermocline depth was underestimated throughout summer by all the vertical turbulence schemes. The selection of stability functions had significant effect on the accuracy of the k-model. Generally k-model performed better when the limiting conditions for mixing length were not applied. The k-model with stability functions based on the Munk–Anderson relation without limiting condition for mixing length showed best accuracy in the calculated profiles of temperature and in the thermocline depth. The improvement of the meteorological forcing had an impact on the exactness of the calculated thermocline depth. However, sensitivity tests showed that this impact was relatively small. Generally, calculated salinity was overestimated in the surface layer and underestimated in the bottom layers. Algebraic parameterisations had highest accuracy in the vertical salinity profiles. In the eastern Gulf of Finland the calculated values of salinity were overestimated. The accuracy of initial conditions, river runoff and bathymetry had significant effect on the accuracy of calculated salinity fields. Highlights: We studied the performance of the vertical turbulence parameterisations in the modeling of hydrodynamics in the Baltic Sea. ► We compared the calculated temperature and salinity fields against dataset having outstanding temporal and spatial coverage. ► The vertical turbulence parameterisations showed considerable differences in the accuracy of the calculated temperature. ► The mixed layer depth was underestimated by all vertical turbulence schemes used in this study. ► Selection of stability functions had significant effect on the accuracy of the turbulence closure schemes.