We compare three different models of rotating neutron star spacetimes: the Hartle-Thorne (HT) slow-rotation approximation at second order in rotation, the exact analytic vacuum solution of Manko et al. and a numerical solution of the full Einstein equations. We integrate the HT structure equations for five representative equations of state. Then we match the HT models to numerical solutions of the Einstein equations, imposing that the mass and angular momentum of the models be the same. We estimate the limits of validity of the HT expansion computing relative errors in the spacetime's quadrupole moment Q and in the ISCO radii. We find that ISCO radii computed in the HT approximation are accurate to better than 1%, even for the fastest observed ms pulsar. At the same rotational rates the accuracy on Q is of order 20%. In the second part of the paper we focus on the exterior vacuum spacetimes. We introduce a physically motivated `quasi-Kinnersley' Newman-Penrose frame. In this frame we evaluate the speciality index S, a coordinate-independent quantity measuring the deviation of each model from Petrov Type D. On the equatorial plane this deviation is smaller than 5%, even for the fastest rotating models. Our main conclusion is that the HT approximation is very reliable for most astrophysical applications.

We discuss the generalisation of the so-called traditional approximation, well known in geophysics, to general relativity. We show that the approximation is applicable for rotating relativistic stars provided that one focuses on relatively thin radial shells. This means that the framework can be used to study waves in neutron star oceans. We demonstrate that, once the effects of the relativistic frame-dragging are accounted for, the angular problem reduces to Laplace's tidal equation. We derive the dispersion relation for various classes of waves in a neutron star ocean and show that the combined effects of the frame-dragging and the gravitational redshift typically lower the frequency of a mode by about 20%.