Nowadays, uncertainties related to the determination of ocean tides remain a major issue for the exploitation of altimetry data in coastal areas. Using Sea Surface Height (SSH) observations from a new GNSS-based system mounted on an Unmanned Surface Vehicle (USV), we develop a crossover methodology to assess tide models under altimetry tracks. To this purpose, we address the Pertuis Charentais area, a semi enclosed sea located in the centre of the Bay of Biscay (France), as a field and modelling case study. We have developed a barotropic model configuration, based on SCHISM platform, using tidal elevations of an up-to-date regional atlas as boundary conditions. To test the impact of boundary conditions, we propose a second configuration where we applied uniform empirical biases in phases and amplitudes on M3 and MN4 constituents. In addition, the survey was designed to highlight the contribution of third and fourth-diurnal waves that are strongly amplified on the shelf and is used to assess model performances under the pass 216 of Sentinel-3A. Our results show that the second configuration reduces the Root-Mean-Square Error (RMSE) of the survey crossover residuals by more than 60%, leading to half of the residuals below 2.5 cm. This improved solution also reduces by 20% the RMSE computed with data from tide gauges located in the inner part of the Pertuis Charentais. Therefore, our study reinforces the importance for coastal tide modelling of an accurate tidal forcing, especially for shallow water waves. We finally discuss the impact of the remaining M4 error on crossover residual heights. By introducing an empirical correction term based on M4 observations at tide gauges, we further reduce the RMSE of crossover residuals by 15-25%. With this innovative study, we demonstrate the interest of combining crossover validation methods and USV systems to spatially extend our understanding of coastal areas dynamics. This will be crucial in the scope of the future SWOT mission, for which the tide correction accuracy must be assessed over the large-extent areas covered by swaths observations.