Third-generation spectral models allow the development of a wave spectrum without any 'a priori' limitations on spectral evolution. One such spectral model is SWAN (Simulation of Waves in the Nearshore). This finite depth model accounts for wind generated waves, whitecapping, bottom friction, refraction, depth induced breaking and shoaling, but does not account for diffraction. The primary goal of this thesis is to determine the suitability of the SWAN model to predict significant wave height, peak period, and wave direction in the southern basin of Lake Winnipeg and Cedar Lake. A quasi nonstationary approach was developed to model storm events for Lake Winnipeg and Cedar Lake. Model predictions were compared to data obtained from an array of waveriders (directional and non-directional) deployed in the south basin of Lake Winnipeg in 1996. The array of waverider buoys allowed an opportunity to examine the temporal and spatial ability of the model to predict wave growth and decay in a relatively shallow lake. Cedar Lake which is the reservoir for the Grand Rapids generating station provided a second opportunity to test the SWAN model on a shallow lake. (Abstract shortened by UMI.)