In this study, the logarithmic least-squares collocation technique was applied to model the geoid surface over Egypt, using the satellite only model GOSO03S as a reference field. Firstly, several combinations of the logarithmic covariance fitting parameters were tried and the best parameters were selected to solve for the geoid. Four different types of gravitational data were input into the collocation algorithm. The resulting geoid model was evaluated at GPS/Levelling control and showed a much better spatial precision than the GOCO03S geoid. After the removal of a long wavelength trend, the final geoid model showed a spatial precision of 11 cm. Such precision was found to lie within the mean geoid noise at the check points, which was obtained from the collocation solution. So, the logarithmic covariance model proved to work well with precise satellite only geopotential models, when used in a remove-restore scheme. Furthermore, such technique resulted in consistent spatial precision and estimated noise for the final geoid model.