Water is needed everywhere to satisfy domestic, agricultural and industrial water demands, to maintain navigation systems, and to preserve healthy and sustainable ecosystems. In order to protect us from floods and to reallocate water resources in a man-made environment, the 'hardware', water-related structures, is constructed in many water systems. However, the 'software', suitable control approaches to operationally maintain and regulate these structures, is still a challenge. Model Predictive Control, standing out from various control technologies, has shown its powerful ability to address multiple objectives, large-scale systems, distributed systems and long-term prediction problems in the field of operational water resources management. This thesis aims to answer the following main research question: How can we apply generic Model Predictive Control and its extensions to satisfy operational objectives in water resources management? As the first element of Model Predictive Control, an internal model needs to be formulated to calculate system states at present and in the future. The linearized Saint Venant equations can well describe hydrodynamics of open water systems, which fit a linear internal model. Second, a quadratic objective function needs to be formulated based on actual management goals, which usually require the water level or discharge to be kept around a pre-defined setpoint. Last, physical and operational limitations are considered as constraints in the optimization problem, such as the height of the riverbank or the capacity of the pump. By describing a water resources management problem in terms of one of the elements, Model Predictive Control is able to address problems of flood management, water supply, irrigation and so on. Chapter 5 applies the standard formulation of Model Predictive Control to flood management in the Dutch water system. Instead of continued reinforcing and heightening water-related structures, Model Predictive Control shows the potential of making the most of the existing infrastructure to reduce potential damages and manage the delta in an operational and holistic way. Different from conventional control, the computational complexity of the optimization is an important issue to consider in Model Predictive Control. Chapter 6 proposes two additional schemes, the adaptive prediction accuracy scheme and the large time step scheme, to achieve computational efficiency. These two schemes are practical and advantageous, especially when implementing Model Predictive Control on large-scale problems. In reality, a large number of water resources management problems includes multiple management objectives at the same time, which can be addressed either in a centralized way or a distributed way. Chapter 7 proposes both the centralized and distributed Model Predictive Control for multi-input-multi-output water resources management. Centralized Model Predictive Control includes all the objectives into a single objective function while distributed Model Predictive Control considers the objectives in each subsystem individually and allows adjacent subsystems to communicate. The results of this thesis suggest that Model Predictive Control is a useful technique for different kinds of water resources management problems, especially with constraints and multiple inputs and multiple outputs. The proposed additional schemes can enable Model Predictive Control to fulfill more real-time requirements, such as a large-scale problem.