Model Predictive Control has been implemented on a large drainage canal system in the Netherlands. This water system can be represented as a reservoir with uncertain inflow due to rainfall runoff and a water level that has to be kept within a certain range by a control flow that is limited in capacity. Tests demonstrate that Model Predictive Control outperforms feedback and feedforward controls. To deal with uncertainty in the expected inflow, Multiple Model Predictive Control (MMPC) is proposed. This controller minimizes an objective function in which the risk of damage is used by applying different scenarios to multiple identical models.