Although seismic methods using S waves can offer high-resolution images of the shallow soil layers, the use of body S-wave tomography for near-surface water monitoring remains underexplored, and the quantitative interpretation of any observed changes in S-wave velocity (VS) in the field conditions is challenging. We conduct a time-lapse S-wave tomography experiment on a field-scale test dike with controlled water levels, allowing for detailed examination of how VS responds to water infiltration. Our results demonstrate that VS decreases progressively, starting from the high-water-side slope and extending across the dike, as the water level rises, with the most significant changes occurring in the sand body and not in the clay cover. The maximum reduction in VS is approximately 40-60 m/s, corresponding to approximately 25%-30% reduction from the initial condition. We use the squared velocity ratio to evaluate the relative contributions of bulk density and shear modulus to VS changes. In the initially unsaturated zone, both factors contribute significantly to the observed VS changes as the zone becomes fully saturated. In fully saturated zones, we assess the changes in the effective stress using the squared VS ratio. Although the low-water side of the dike shows stress changes that are consistent with numerical modeling, the high-water side shows larger stress changes than expected, possibly due to excess pore pressure during the dynamic flow conditions. These findings highlight the potential of body S-wave tomography for high-resolution, near-surface hydrologic monitoring, and provide insights into the complex interactions between physical properties that influence VS changes under varying water levels in field environments.