A model for optimal allocation of water from a single-purpose reservoir to an irrigation project with pre-determined multiple copping patterns was developed. The model consisted of two modules: (I) the intra-seasonal allocation model (non-linear programming) which is used for allocation of water among different crops for a definite combination of state variables (inflow class, rainfall class, reservoir storage classes at the beginning and at the end of the season) for the non-dormant season to maximize total farm income; and (II) the seasonal allocation model (stochastic-dynamic programming) which is used for the convergent operation policy over seasons for optimal expected farm income over a year. The model was applied to Ardak reservoir dam (I,R. Iran) in an arid region. Low river inflow in the dormant season at the study area could not admit the reservoir class changes for specific combinations of state variables, and therefore resulted in a non-stable result. Imposing a fictitious positive relative net benefit for all possible combinations of reservoir class changes eliminated this problem. It was also shown that rainfall did not play a marked role in the study area, which is an arid region, and stochastic nature can be removed from the model.

Keywords