The impact of streamflow forecasting on reservoir operations is a function of several factors, including the size of the system, the time scale of operation, the relative size of regulation capacity of the system, and the use of the water stored in the system. Flow forecasting techniques have been extensively developed in literature. However, the inclusion of these forecasts in stochastic optimization models have not obtained wide acceptance among engineers responsible for systems operations, in particular, real time flood control in multiple reservoir systems. The premise of this paper is that the reluctance of practitioners to embrace these methods in flood control does not result from theoretical inadequacy of the stochastic optimization models to represent complex multi-reservoir systems. Rather, the problem may be stem from the lack of completeness of the optimization models: inadequate representation of all relevant objectives and constraints in the operation of a reservoir system, and inadequate mechanisms to incorporate risk within the decision-making process. Extensive research has been carried out to determine operating policies of a multi-reservoir system, but the area remains open for further research. One reason research in this field remains active is the number of simplifications which have to be made in order to make a complex system more tractable. An issue that very few papers have dealt with is the combination of both long and short term operation, in which the long term policies are used as boundary conditions for the short term optimization algorithm.