Sand waves are large bed waves on the seabed, being a few metres high and lying hundreds of metres apart. In some cases, these sand waves occur in navigation channels. If these sand waves reduce the water depth to an unacceptable level and hinder navigation, they need to be dredged. It has been observed in the Bisanseto Channel in Japan that the sand waves tend to regain their shape after dredging. In this paper, we address modelling of this regeneration of sand waves, aiming to predict this process. For this purpose, we combine a very simple, yet effective, amplitude-evolution model based on the Landau equation, with measurements in the Bisanseto Channel. The model parameters are tuned to the measured data using a genetic algorithm, a stochastic optimization routine. The results are good. The tuned model accurately reproduces the measured growth of the sand waves. The differences between the measured weave heights and the model results are smaller than the measurement noise. Furthermore, the resulting parameters are surprisingly consistent, given the large variations in the sediment characteristics, the water depth and the flow field. This approach was tested on its predictive capacity using a synthetic test case. The model was tuned based on constructed predredging data and the amplitude evolution as measured for over 2 years. After tuning, the predictions were accurate for about 10 years. Thus, it is shown that the approach could be a useful tool in the optimization of dredging strategies in case of dredging of sand waves.