Observations from 1995 and 1996 described the seasonal evolution of the threedimensional density field in the western Irish Sea. A cold, dense pool flanked by strong nearbed density gradients was present from May until October. Temperature had the dominant effect on density from June onwards. The trajectories of 55 satellite-tracked drifters defmed the full spatial extent of the cyclonic circulation that is the western Irish Sea gyre. Several distinct recirculation paths were observed and drifter speeds were in good agreement with geostrophic calculations based on the observed density field. The existence of such organised, baroclinic flows in shelf seas demands that coastal ocean models should reproduce their dynamics correctly, if the models are to be useful as environmental management tools. One such model, ECOMsi, was applied to the study area and results from seasonal simulations were compared with the observations. A new technique was developed to perform quantitative comparisons between modelled and observed flow fields. The model successfully reproduced the three-dimensional temperature structure throughout the seasonal simulations, and also predicted the cyclonic, near-surface residual circulation of the gyre. The model demonstrated conclusively that the gyre is density-driven and revealed the same recirculation paths that were visible in the drifter tracks. The vertical structure of the modelled density-driven flow confirmed the geostrophic nature of the currents and emphasised the important dynamical role of sharp density gradients near the bed (bottom fronts). A quantitative comparison of different model runs identified the critical parameterisations and forcing quantities for this application. An accurate specification of air temperature over the sea region was required for the model to achieve the correct timing of the stratification breakdown. During this phase, convective cooling at the surface was seen to be as important as the mixing by autumnal winds in eroding the density structure. The possibility of a seasonal reversal in density-driven flow along the east coast of Ireland was also identified. A new interaction between the wind and the density field, which could defme where the strongest currents in the gyre are to be found, is described. The model is now considered to be sufficiently well tested to use in a predictive capacity and for biological transport studies. This work highlights the benefits that can be obtained using high quality spatial and temporal field observations in the critical testing of numerical models, and furthermore suggests that shelf seas are the perfect location for such tests to be performed.