Real-time monitoring, databases, optimization models and visualization tools have been integrated into a Decision Support System (DSS) for optimal water resources management of two water supply reservoirs, the Daechung Reservoir and the Yongdam Reservoir of the Geum River basin, Daejeon, Korea. The KModSim as a DSS has been designed to provide information on current reservoir conditions to operational staff and to help in making decisions for short- and long-term management. For the physical calibration, the network simulations in seasonal water allocation of both reservoirs are performed for 23 years from January 1 1983 to June 30 2006. Linear and nonlinear operating rules are developed by using the actual reservoir operation data obtained from both reservoirs which are then used in KModSim by the hydrologic state method to estimate optimized target storages of both reservoirs. For validation of hydrologic states in KModSim and scenario testing for the management simulations, the optimal network simulation for the seasonal water allocations from October 1 2002 to June 30 2006 were also performed. The results' simulation by new rules fit the measured actual reservoir storage and represent well the various outflow discharge curves measured at the gauging stations of Geum River. The developed operating rules are proven to be superior in explaining actual reservoir operation as compared to the simulated target storages by existing optimization models.