We perform the ambient noise Scholte and Love waves phase-velocity tomography to image the shallow subsurface (a few hundreds of metres) at the Valhall oil field. Seismic noise was recorded by multicomponent (north, east and vertical) ocean bottom cable from the Valhall life of field seismic network. We cross-correlate six and a half hours of continuous recording of noise between all possible pairs of receivers. The vertical–vertical and the transverse–transverse components cross-correlations are used to extract the Scholte and Love waves, respectively. We combine more than 10 millions of interstation correlations to compute the average phase-velocity dispersion curves for fundamental mode and first overtone. Then, a Monte Carlo inversion method is used to compute average 1-D profiles of VSV and VSH down to 600 m depth. In the next step, we construct 2-D Scholte and Love waves phase-velocity maps for fundamental mode using the eikonal tomography method. These maps are then jointly inverted to get the 3-D distribution of VSV and VSH from which the radial anisotropy and the isotropic velocity (VS) are estimated. The final model includes two layers of anisotropy: one in the shallow part (above 220 m) with a significant negative radial anisotropy (VSH < VSV) due to vertical cracks because of subsidence and another in the deeper part (between 220 and 600 m) with a positive radial anisotropy (VSH > VSV) due to the stratification at that depth.