To identify oceanic processes responsible for the formation and evolution of large-scale thermal anomalies, a series of initial-value model simulations are carried out using a 10-level primitive equation model in a closed rectangular basin. The model contains time-dependent seasonal forcing by the atmosphere, a parameterization of surface-generated wind stirring and convective overturning, and nonlinear lateral eddy viscosity based on two-dimensional turbulence theory. A “model climatology” is first generated by integrating the model over 240 years of simulated time using climatological atmospheric forcing which varies on the seasonal time scale. This model-generated ocean climatology is shown to compare qualitatively with the large-scale circulation in the North Pacific Ocean. The model is used to simulate oceanic thermal anomalies using both anomalous and climatological atmospheric forcing. The initial conditions for the simulations are obtained by combining a particular temperature anomaly with the climatological temperatures generated by the model for the given time of year. The anomalous part of the initial vertical shear current is geostrophic while that of the vertical mean current is zero. Using climatological atmospheric forcing, the free evolution of an initial sea surface temperature (SST) anomaly is shown to be dependent on its depth of penetration. Using anomalous atmospheric wind forcing derived from the first empirical orthogonal function of monthly mean sea level pressure anomalies (Davis, 1976), realistic SST anomalies develop in a period of 30 days. The mechanism responsible for the anomaly development is horizontal advection of mean temperature by anomalous surface (Ekman) currents. Two simulation experiments are made using observed anomaly data from the atmosphere and the ocean. In one case (fall–winter 1971–72) the temperature anomaly is known only at the surface where in the other case (fall–winter 1976–77) it is mapped from observations down to 400 m by a NORPAX ships-of-opportunity program. Both experiments utilize climatological heating and anomalous wind forcing derived from monthly mean sea level pressure anomalies. In both cases horizontal advection of mean temperature by anomalous wind-driven surface (Ekman) currents is able to explain a large fraction of the observed anomaly development in the upper layers. In the 1976–77 case the model simulation below 100 m is less successful. It is pointed out that improved simulations can be expected from an inclusion of anomalous surface heat fluxes and a more realistic treatment of anomalous wind mixing but these are left for a future study.