Due to the limited pumping capacity in lowland water systems, reduction of system failure requires anticipation of extreme precipitation events. This can be done by Model Predictive Control that optimizes an objective function over a certain time horizon, for which the system behaviour is calculated by a model and a prediction of the inputs to the system. The forecast inputs usually contain large uncertainties. Because the pump constraints make the optimization problem non-certainty equivalent, uncertainties need to be considered to adequately control the water system. In this paper, the way uncertainties influence the control decision is investigated. An information-control horizon and an information-prediction horizon are introduced as time-limits for the sensitivity to future input information and the value of predictions. These horizons need to be considered in the design of a controller. Multiple Model Predictive Control is suggested to deal with the uncertainties in a risk based way.