With the long-term goal of developing an ensemble forecast system for coastal flooding, we are developing a dynamically-based, numerical model of the global ocean. The model is based on the NEMO framework and has been used to predict global tides and surges in previous studies. This study focuses on the optimization of the joint prediction of both tides and surges, the two main components of total water level that cause coastal flooding. To improve the predictions of the tide we use a modified form of “spectral nudging”. We show this leads to significant improvements in the prediction of the M2 tide in the open ocean, and also in the shallow regions closer to shore where the model is not nudged. The median value of the vector difference of the tidal amplitude based on sea level observations and a data-assimilative model, and the predictions of our ocean model, is reduced from 11.2 cm to 2.66 cm by the nudging. The improvement deteriorates significantly however if additional tidal constituents are included in the model (most notably S2). This is explained in terms of spectral leakage between tidal bands associated with the nudging methodology and a straightforward solution is proposed.