Hybrid modeling approach including regionalization method, entropy technique, and Bayesian multiobjective optimization algorithm is proposed for optimum water monitoring network design. For hydrometric network design, all the components of the hybrid model are used as follows. Robust regionalization method is used to generate streamflow at all possible locations of new stations, and dual entropy-multiobjective optimization methods are used to optimize the number and locations of the new stations. For precipitation (rainfall or snowfall) network design, only the dual entropy-multiobjective optimization modules are used to optimize the number and locations of new stations based on an initial grid points which can be from remote sensing database (e.g. SNODAS) or interpolated ground observations. In addition to joint entropy and total correlation, other constraints such as cost, flow signatures, water vulnerability indicators, can be added to further optimize the number and locations of new stations. The hybrid model can also be applied to classified physiographic units to design optimal minimum network that meets the World Meteorological Organization minimum network standards. Three applications are proposed to assess the effectiveness of the hybrid model. This includes the design of optimum hydrometric networks for the middle St-Lawrence River basin and the St-John River basin; the design of optimum rainfall network for St-John River basin and an optimum snow network for the Columbia River basin. For each case study, the hybrid model appears a robust tool for designing optimum networks. Optimal number of stations and locations are determined for each solution (or optimum network) that is part of the Pareto front. Furthermore the hybrid model appears a robust and flexible method that can be further extended to include groundwater and water quality networks design.