Abstract. The primary goal of this study was to incorporate data-derived harmonic constants into a complex dynamic model using a form of variational data assimilation, with a view to improve the prediction of 7 dominant tidal constituents in the Adriatic Sea. Firstly, harmonic-constant data for 6 Adriatic stations were fed into a steady-state, 3-D, forward/inverse model to furnish optimal boundary conditions (OBCs). Calculated OBCs were then used to derive individual constituent responses, as well as to synthesise seven-constituent boundary conditions for the time stepping, 3-D model. A separate set of 25 stations provided control harmonic constant data. In validating the model output particular attention has been given to the often-ignored tidal currents. To that end 14 current meter data records were processed into tidal current ellipse parameters and used to examine the comparable model output. Comparison to gauge data has shown that the present solution is better than our own previous one, and shows an improvement over recent solutions by other authors. The model accurately reproduces available data with individual station amplitude differences rarely exceeding 1cm, and with the phase error commonly staying well below 10°. For all tidal constituents individual station differences result in RMSE in the 0.33-0.71-cm range for amplitude, and the 5.6°-19.2° range for phase. Semidiurnal currents appear to be modelled better than the diurnal ones (generally over-predicted). High eccentricity of both data and model-derived ellipses often impaired calculating the proper sense of rotation; inclination of the ellipses proved to be the most robust parameter, successfully predicted for most constituents at all depths.