iii groundwater withdrawals until that time) had been reduced to zero, and to subsequently minimize use of water from aquifer storage. The model indicated that this "debt" to the Rio Grande could be eliminated after about 9 years by using an optimal withdrawal distribution, which is about 9 years earlier than would be achieved using the year-2000 distribution of groundwater withdrawal. The objectives of this model also could be achieved while keeping water-level declines less than about 2.5 feet per year. Results of the optimization models were evaluated for their ability to approximate the water budgets and water-level declines produced by the simulation model, as well as for their applicability to a range of river-aquifer conditions. Comparison between optimization calculations and simulation results indicated that all five optimization models provided a reasonable approximation of the river-aquifer system as represented in the groundwater flow model, despite the nonlinearities inherent in that model. The response functions used in the optimization models, and therefore the results of the optimization models, also were found to be applicable to a reasonable range of river conditions and groundwater withdrawal scenarios. Although the exact solutions to the optimization models would vary with changes in future water demand, the locations of municipal-supply wells, or other factors, the broader implications of model results for the river-aquifer system would remain important to the management of regional water resources. In particular, information learned about the timing and magnitude of effects of groundwater withdrawals in different locations on aquifer storage and on the river system could be applied to multiple management issues in the Middle Rio Grande Basin, and perhaps in other alluvial basins of the Southwestern United States.