A stochastic groundwater management model is formulated to account for prediction uncertainty when maximizing regionally distributed groundwater production yet obeying regulations to maintain the hydroecological balance in wetland areas. The water table elevation in sensitive wetland areas is lowered by the withdrawal of groundwater at supply wells. Substantial uncertainty exists because drawdowns depend on both the unknown spatial distribution of hydraulic conductivity and regional boundary conditions. Planning in the face of uncertain predictions of water table changes means that optimal production must be prudently reduced. A stochastic simulation-optimization formulation is developed that provides a robust water production plan. Prediction uncertainty is dealt with through stochastic simulation-optimization using a multiple-realization approach. On the basis of analyses involving solution of over 8 million aquifer models and 36,000 stochastic-optimization solutions, the nature and reliability of the optimal groundwater production scheme are inspected to determine the effects of uncertainty in spatially variable hydraulic conductivity, conditioning on local measurements, and the type of boundary conditions imposed in the nonlinear aquifer model. We propose a new measure to predict the expected reliability of meeting water level constraints in wetland areas. Monte Carlo simulations based on numerous optimal groundwater production schemes confirm that the expected reliability is a quantifiable function of the number of hydraulic conductivity realizations included in the stochastic-optimization formulation and the variance of log hydraulic conductivity.