Distributed acoustic sensing is an ideal tool for ambient noise tomography owing to the dense spatial measurements and the ability to continuously record in harsh environments, such as underwater. We demonstrate the ability to image a complex underwater basin using ambient noise recorded on a fiber deployed offshore Greece. A two-dimensional shear-wave velocity model was derived by analyzing Scholte-wave dispersion. In addition, extremely detailed frequency-dependent resonance and wave propagation characteristics were revealed by computing power spectral densities (PSD) and auto-correlations (AC), respectively. These observations provide crucial information on lateral and vertical wave propagation, and were used to further constrain the velocity model. The analysis reveals significant lateral variations across the short 2.5 km long fiber segment, including basin edge effects and scattered waves. Waveform simulations further support the obtained model. Our results demonstrate the advantages of incorporating PSD and AC observations into ambient noise-based imaging.