Previously developed suites of models integrate irrigation water supply and demand management for simplified surface reservoir supply systems. A central stochastic dynamic programming model is supported by simulation models, including a soil water-plant growth model. That modeling is extended herein to include decisions about timing and quantity of reservoir water releases into the delivery system, resulting in the integration of supply, demand, and delivery management. Irrigators have rights to percentages of reservoir capacity, reservoir inflows, and downstream tributary flows. Natural rivers are the supply channels in the study area. Large on-farm water storages exist for storing regulated and unregulated river flows available to the irrigators. Farms can be many days flow from the reservoir, requiring orders for reservoir releases to be lodged before the arrival of a previous order. The probability of unregulated flows from tributaries downstream of the reservoir further complicates ordering decisions. Abandonment of irrigated area to rain-fed status occurs at two levels, forced abandonment if insufficient water is available at the farm when the irrigation-trigger soil water deficit is reached, and planned abandonment to save water for possible later use. The need for forced abandonment is determined by simulation models; planned abandonment decisions are derived by stochastic dynamic programming. Results show annual net revenue means and standard deviations as functions of different capacities of the on-farm storages and water supplies from either regulated or unregulated flows, or both. ¿ American Geophysical Union 1993