Three methods are proposed for the inclusion of inertia when deriving currents from sea surface height (SSH) in the Mozambique Channel: gradient wind, perturbation expansion, and an iterative method. They are tested in a model and applied to satellite altimetry. For an eddy of 25 cm amplitude and 100 km radius, typical of Mozambique Channel rings at 18°S, the error made with geostrophy is 40% for the anticyclones and 20% for the cyclones. Inertia could reach one third of the pressure gradient. Geostrophy underestimates subsurface currents by up to 50 cm s⁻¹, resulting in errors of 30–40%. The iterative method results in errors of <5% for the most part of the structure. The error RMS in velocities based on 8 years of model SSH is in excess of 30 cm s−1 for geostrophy and reduces to about 10 cm s−1 for the gradient wind and iterative methods. The perturbation method is less accurate. Applied to satellite altimetry, the addition of inertia results in a significant increase in velocities for the anticyclones and a decrease for the cyclones. It induces a velocity increase of >50% in Mozambique Channel rings. Geostrophic EKE reaches 1400 cm² s⁻², while it attains 1800 cm² s⁻² when inertia is added. Applied to the Gulf Stream, these methods confirm the hypothesis of Maximenko and Niiler [2006] that centrifugal accelerations should be the main cause for the difference observed between geostrophic and drifter EKE. This methodology should result in a net improvement for operational surface ocean currents.