Distributed acoustic sensing (DAS) is a novel technology, which allows the seismic wavefield to be sampled densely in space and time. This makes it an ideal tool for retrieving surface waves, which are predominantly sensitive to the S-wave velocity structure of the subsurface. In this study, we evaluate the potential of DAS to image the near surface (top 50 m) using active-source surface waves recorded with straight fibers on a field in the province of Groningen, the Netherlands. Importantly, DAS is used here in conjunction with a Bayesian transdimensional inversion approach, making this the first application of such an algorithm to DAS-acquired strain-rate wavefields. First, we extract laterally varying surface wave phase velocities (i.e., “local” dispersion curves [DCs]) from the fundamental mode surface waves. Then, instead of inverting each local DC separately, we use a novel 2D transdimensional algorithm to estimate the subsurface’s S-wave velocity structure. We develop a few modifications to improve the performance of the 2D transdimensional approach. Specifically, we develop a new birth-and-death scheme for perturbing the dimension of the model space to improve the acceptance probability. In addition, we use a Gibbs sampler to infer the noise hyperparameters more rapidly. Finally, we introduce local prior information (e.g., S-wave logs) as a constraint to the inversion, which helps the algorithm to converge faster. We first validate our approach by successfully recovering the S-wave velocity in a synthetic experiment. Then, we apply the algorithm to the field DAS data, resulting in a smooth laterally varying S-wave velocity model. The posterior mean and uncertainty profiles identify a distinct layer interface at approximately 20 m depth with a sharp increase in velocity and uncertainty at that depth, aligning with borehole log data that indicate a similar velocity increase at the same depth.