The propagation of remotely generated superinertial internal tides constitutes a difficulty for the modelling of regional ocean tidal variability which we illustrate in several ways. First, the M2 tidal solution inside a control region located along the Southern California Bight coastline is monitored while the extent of the numerical domain is increased (up to 512 x 512 km). While the amplitude and phase of sea level averaged over the region is quasi-insensitive to domain size, a steady increase of kinetic energy, predominantly baroclinic, is observed with increasing domain size. The increasing flux of energy into the control region suggests that this trend is explained by the growing contribution from remote generation sites of internal tide which can propagate up to the control region. Increasing viscosities confirms this interpretation by lowering baroclinic energy levels and limiting their rate of increase with domain size. Doubling the grid spacing allows consideration of numerical domains 2 times larger. While the coarse grid has lower energy levels than the finer grid, the rate of energy increase with domain size appears to be slowing for the largest domain of the coarse grid simulations. Forcing the smallest domain with depth-varying tidal boundary conditions from the simulation in the largest domain produces energy levels inside the control region comparable to those in the control region for the largest domain, thereby confirming the feasibility of a nested approach. In contrast, simulations forced with a subinertial tidal constituent (K1) show that when the propagation of internal tide is limited, the control region kinetic energy is mostly barotropic and the magnitudes of variations of the kinetic energy with domain size are reduced. (C) 2012 Elsevier Ltd. All rights reserved.