Recent studies show that the noise recorded by seismic sensors can be used to characterize the seismic structure in the area where the sensors are located. This is called ambient noise tomography. In this thesis, data from ocean bottom cables at the Oseberg oil and gas field was used to characterize the near-surface in an area used for waste disposal. 32 days of data was made available to us from Equinor (Statoil). By using interferometry, we created virtual sources by taking the cross-correlation between ambient noise recordings made at different stations. Through slowness-frequency analysis, the dispersion curves of surface waves were retrieved from the noise correlations, and inverted to explore the shear wave velocity with depth. The platform is a major source of noise in the area. Below 2 Hz, we have noise coming from all directions. Above 2 Hz, the main source is the platform. Since the array of stations is made mostly of 4 lines in line with the platform, both the noise coming from all directions and the noise coming from the platform was used. The ambient noise with frequencies up to 4 Hz was used for Green's function retrieval, and we see that eight hours of data is sufficient for retrieving the Green's functions. No major change with time was observed, and the Green's functions along the 4 lines are very similar. The Love wave and the Scholte wave fundamental mode and overtone emerged from the cross-correlations of ambient noise. The Scholte wave fundamental mode was used for inversion, and a model of S-wave velocity with depth down to 3000 m was obtained. The velocity increases from 300 m/s at the sea bottom to about 1700 m/s at 3 km depth, similar to the results retrieved from other methods.