Abstract: To establish a high-precision and regionally adaptive depth datum for the Bohai and Yellow Seas, this study integrates satellite altimetry data with multiple global tidal models in order to develop a hybrid modelling and accuracy evaluation framework. Based on long-term observations from 23 tide gauge stations, the applicability of four global tidal models (DTU16, EOT20, FES2014, and FES2022) was systematically assessed. Results indicate that, among short-period tidal models, the FES2014 model exhibited the best performance, with the lowest root mean square error (RSS) value of 4.57 cm for the eight major tidal constituents. To address the limited accuracy of existing tidal models in representing long-period constituents such as the annual (Sa) and semi-annual (Ssa) tides, along-track harmonic constants of Sa and Ssa were extracted from TOPEX/Poseidon and Jason satellite altimetry data (1993-2021) without inverse barometric correction. Validation against tide gauge records shows that the root mean square errors of Sa and Ssa amplitudes are 3.02 cm and 2.89 cm, respectively, demonstrating the reliability of this method for modelling long-period tides. Combining 11 short-period constituents from FES2014 with long-period constituents derived from satellite altimetry created a comprehensive hybrid tidal model to generate a depth datum surface covering the Bohai and Yellow Seas. The resulting datum values range from 51 cm to 530 cm, exhibiting significant spatial variation. Compared with the model based solely on FES2014, the hybrid model shows a notable improvement in accuracy, reducing the root mean square error (RMSE) from 19.44 cm to 11.66 cm at tide gauge locations (− 40% improvement). This enhancement is primarily attributed to the improved treatment of long-period constituents and the strong consistency between uncorrected satellite altimetry and in situ observations. Comparison with existing depth datum values at tide gauge stations reveals persistent regional deviations, reflecting spatiotemporal inconsistencies arising from differences in constituent selection and computational methods used in historical practices. These findings highlight the need to implement localized data assimilation strategies in key areas to improve compatibility with existing datums. The proposed hybrid tidal modelling approach offers a systematic solution for constructing high-precision depth datums in the Bohai and Yellow Seas. It demonstrated the effectiveness of integrating multiple data sources to improve model accuracy and regional adaptability, providing a theoretical and technical foundation for hydrographic surveying, vertical datum unification, and marine engineering applications.