We present a particle tracking model that simulates the dispersal of the oyster Ostrea chilensis Philippi in Tasman Bay, New Zealand. The model was calibrated to reproduce observed vertical distributions of larvae, and was verified against an independent data set of such distributions. The seabed in Tasman Bay contains little hard substrate other than shell hash. By comparing simulated and observed settlement, we infer that larvae must either be able to sense shell hash at a distance and orient onto it, such that settlement success is higher than indicated by the 1-km² resolution relative density (square meter of exposed shell per square meter of seabed), or be settling on material other than shell hash. The modeling indicates that only a very small proportion of those larvae that do settle travel more than 1 km from their natal location. There are plans to endeavor to restore the oyster fishery by provision of new shell hash to the bed. Our simulations suggest that these new beds will be colonized rapidly only if the new beds are placed close to the remnant beds, or if they are artificially seeded with mature oysters. Conversely, if the new beds are artificially seeded, our simulations imply that a large fraction of the spawn from these adults may be retained within the new bed, rather than dispersing into the surrounding, unsuitable soft sediments.