Motional induction in the ocean by tides has long been observed by both land and satellite measurements of magnetic fields. While these signals are weak (similar to 10 nT) when compared to the main magnetic field, their persistent nature makes them important for consideration during geomagnetic field modelling. Previous studies have reported several discrepancies between observations and numerical predictions of the tidal magnetic signals and those studies were inconclusive of the source of the error. We address this issue by (1) analysing magnetometer data from ocean-bottom stations, where the low-noise and high-signal environment is most suitable for detecting the weak tidal magnetic signals, (2) by numerically predicting the magnetic field with a spatial resolution that is 16 times higher than the previous studies and (3) by using four different models of upper-mantle conductivity. We use vector magnetic data from six ocean-bottom electromagnetic (OBEM) stations located in the Northwestern Pacific Ocean. The OBEM tidal amplitudes were derived using an iteratively re-weighted least-squares (IRLS) method and by limiting the analysis of lunar semidiurnal (M2), lunar elliptic semidinurnal (N2) and diurnal (O1) tidal modes to the night-time. Using a 3-D electromagnetic induction solver and the TPX07.2 tidal model, we predict the tidal magnetic signal. We use earth models with non-uniform oceans and four 1-D mantle sections underneath taken from Kuvshinov and Olsen, Shimizu et al. and Baba et al. to compare the effect of upper-mantle conductivity. We find that in general, the predictions and observations match within 10-70 per cent across all the stations for each of the tidal modes. The median normalized percent difference (NPD) between observed and predicted amplitudes for the tidal modes M2, N2 and O1 were 15 per cent, 47 per cent and 98 per cent, respectively, for all the stations and models. At the majority of stations, and for each of the tidal modes, the higher resolution (0.25A degrees A xA 0.25A degrees) modelling gave amplitudes consistently closer to the observations than the lower resolution (1A degrees A xA 1A degrees) modelling. The difference in lithospheric resistance east and west of the Izu-Bonin trench system seems to be affecting the model response and observations in the O1 tidal mode. This response is not seen in the M2 and N2 modes, thereby indicating that the O1 mode is more sensitive to lithospheric resistance.