[1] We compare predictions of a coupled, wave-averaged, cross-shore wavescurrents-bathymetric evolution model to observations of onshore and offshore nearshore sandbar migration. The observations span a 10-and 44-day period with onshore/ offshore bar migration at Duck, North Carolina, and at Hasaki, Kashima Coast, Japan, respectively, a 3.5-month period of onshore bar migration at Duck, and a 22-day period of offshore bar migration at Egmond, Netherlands. With best fit parameter values the modeled temporal evolution of the cross-shore bed profiles agrees well with the observations. Model skill, defined as 1 minus the ratio of prediction to no-change error variances, ranges from 0.50 at Egmond to 0.88 for the prolonged onshore bar migration at Duck. Localized (in time and space) reductions in model skill coincide with alongshore variations in the observed morphology. Consistent with earlier observations, simulated offshore bar migration takes place during storms when large waves break on the bar and is due to the feedback between waves, undertow, suspended sediment transport, and the sandbar. Simulated onshore bar migration is predicted for energetic, weakly to nonbreaking conditions and is due to the feedback between near-bed wave skewness, bedload transport, and the sandbar, with negligible to small effects of bound infragravity waves and near-bed streaming. Under small waves and conditions, when breaking and nonbreaking conditions alternate with the tide, the sandbar is predicted to remain stationary. The intersite differences in the optimum parameter values are, at least partly, induced by insensitivity to parameter variations, parameter interdependence, and errors in the offshore wave forcing.