Stochastic Approximation Applied to Optimal Irrigation and Drainage Planning

Abstract

Parameter uncertainty in modeling complex hydrologic systems has resulted in development of stochastically based water management models. Optimal solutions to such models, particularly those having parameters with large variance, may be difficult, if at all possible, to obtain, due to intensive computational requirements. This paper discusses application of a stochastic quasigradient (SQG) approximation method to a management model. The model describes the effects of regional irrigation and drainage system planning on shallow saline groundwater behavior and net economic returns to farmers. Due to the complexity of the model, Monte Carlo simulation techniques were used. Despite model complexity and large parameter variance originating from spatial variability in soil hydraulic properties, the SQG method obtained near‐optimal solutions in about 15% of the CPU time required to obtain such solutions using response surface methodology.