Lag-1 correlation for inflow between consecutive periods is considered into reservoir model in many publications. This consideration greatly improves the reservoir models to incorporate the uncertainty of inflow. However, no publication was found to incorporate a correlation relationship for releases between consecutive periods to improve the operating policy for reservoir operations. This paper presents a methodology of autoregressive decision rule for an aggregated reservoir operation as a surrogate of a multireservoir system of the Upper Colorado River Basin. The method incorporates a lag-1 correlation for the releases between consecutive periods with the optimal operating policy solved by stochastic dynamic program. The decision rules with and without incorporation of the autoregressive correlation for the releases were then used in simulated operation of the reservoir with historical inflow records to evaluate their relative effectiveness. The results showed that the autoregressive decision rule yields more stable and higher reliability of annual water supply for the aggregated reservoir operations. ©ASCE.