Mountain rivers in particular are prone to sediment input in the form of pulses rather than a more continuous supply. These pulses often enter in the form of landslides from adjacent hillslopes or debris flows from steeper tributaries. The activities of humans such as timber harvesting, road building, and urban development can increase the frequency of sediment pulses. The question as to how mountain rivers accommodate pulses of sediment thus becomes of practical as well as academic significance. In part 1 <Cui et al., 2003>, the results of three laboratory experiments on sediment pulses are reported. It was found there that the pulses were eliminated from the flume predominantly by dispersion of the topographic high. Significant translation was observed only when the pulse material was substantially finer than the ambient load in the river. Here the laboratory data are used to test a numerical model originally devised for predicting the evolution of sediment pulses in field-scale gravel bed streams. The model successfully reproduces the predominantly dispersive deformation of the experimental pulses. Rates of dispersion are generally underestimated, largely because bed load transport rates are underestimated by the transport equation used in the model. The model reproduces the experimental data best when the pulse is significantly coarser than the ambient sediment. In this case, the model successfully predicts the formation and downstream progradation of a delta that formed in the backwater zone of the pulse in run 3. The performance of the model is less successful when the pulse is composed primarily of sand. This is likely because the bed load equation used in the study is specifically designed for gravel. When the model is adapted to conditions characteristic of large, sand bed rivers with low Froude numbers, it predicts substantial translation of pulses as well as dispersion.