The utility of a new technique for assimilating long-term hydrographic information into eddy-permitting ocean models is demonstrated using the Parallel Ocean Program (POP) applied to the North Atlantic. Robust diagnostic and standard prognostic simulations of the North Atlantic yield results similar to those obtained in earlier studies and the differences between them emphasize the need for data assimilation. The basic idea of the new technique is to add correction terms to the model equations that directly influence the model solution only in prescribed frequency and wave number bands, leaving the variations outside of these bands free to evolve prognostically. For this reason the technique is referred to as spectral nudging. We consider two approaches for constraining eddy-permitting models based on observed long-term hydrographic conditions. In the first approach, the model's temperature and salinity climatologies are spectrally nudged toward observed values using restoring terms in the tracer equations; in the second, correction terms are added to the momentum equations. Both approaches result in significant improvements in the model's climatology, as expected. Both approaches also result in more realistic meso-scale eddy fields. However, for the simulations considered here, spectral nudging in the tracer equations generally provides better results than nudging in the momentum equations. An examination of the relationship between the two approaches reveals that this result might be expected to also occur in other model simulations but this is achieved at the cost of increased constraints placed on model dynamics within the nudged frequency and wave number bands.