Studies were carried out to determine the circulation of water in the northern basin of Howe Sound, a small fjord on the mainland coast of British Columbia, and to determine the extent of the influence of the winds, the tide and river runoff on the circulation. In one experiment, surface-layer drogues were tracked by radar for four periods each of approximately three days duration. Data sere recorded photographically, then digitized for computer processing. Cubic spline interpolation was used to produce positions, velocities and accelerations at one minute intervals along every drogue track. The interpolated data were averaged in a suitable manner to produce pseudo-Eulerian estimates of velocity. Near the head of the fjord, both wind and tide appeared to cause temporal fluctuations in the surface current of magnitude similar to the expected mean flow due to the river. The river was the cause of spatial inhomogeneity in the flow, but did not appear to be a significant source of temporal variations. Farther down the inlet, wind forcing was the dominant cause of temporal variations in the surface-layer flow of about five times the magnitude of the expected mean river-driven flow. At no distance along the fjord was the velocity observed to be laterally uniform. Lateral gradients of long-channel velocity were strong at the inlet head and decreased away from it, indicating that the fresh water from the river was slowly mixing across the inlet. Another experiment using drogues at three depths in the upper 6 m of the water indicated that the velocity structure was not uniform, either laterally or with depth. Analysis was done on data from six current meters moored in the northern basin of Howe Sound. The mean currents from these meters showed a surface-layer outflow and a return inflow in the waters just below. A mean down-inlet current was seen at 150 m, 80 m below sill depth. Spectra of the currents showed dominant peaks at diurnal and semi-diurnal periods. The wind was coherent with the currents at 3m for periods longer than 10 hours. Below this depth, no consistent relationship was seen. In the diurnal band, the currents were strongest at the surface, indicative of forcing from the surface by the wind. In the semi-diurnal band, the currents were strongest at 10m depth. Both bands also showed a phase variation with depth indicative of a baroclinic structure. These results were compared with some models for surface-layer behaviour. The first model assumed that the wind momentum input was distributed uniformly throughout the surface layer and that the layer was not frictionally coupled to the deeper waters. Drag coefficients calculated from the wind stress and drogue acceleration gave values of 1 to 2x10-3, similar to values measured in other ways. This model was only valid for the first few hours after the onset of the wind. Another model, developed by Farmer (1972), analysed the behaviour of the surface layer of a semi-infinite canal under the influence of a steady wind stress. It predicted correctly the length of time of wind dominance of the flow, the magnitude of the velocity change and the magnitude of the acceleration of the water. A baroclinic tidal model in a two-layer fjord, adapted from Rattray (1960), correctly predicted the phase of the surface-layer currents near the head of a fjord with respect to the height of the tide. / Science, Faculty of / Earth, Ocean and Atmospheric Sciences, Department of / Graduate