The objective of this paper is to present a new methodology for determining the optimal fresh water inflows into bays and estuaries to maximize fish harvest while satisfying the freshwater requirements for other components in the system. This new methodology is based upon solving a large-scale nonlinear programming problem formulated as a discrete-time optimal control problem. A hydrodynamic transport model (HYD-SAL) is used to model both the temporal and spatial variability of salinity. The hydrodynamic transport model is used to perform constraint and function evaluations for a nonlinear programming optimizer (GRG2). Chance-constraints are formulated to explicitly incorporate the uncertainty in regression equations for fish harvest. A computer model based on this new methodology has been developed and applied to the Lavaca-Tres Palacio Estuary (Matagorda Bay) in Texas. This model is applicable to any bay or estuary.