The objective of this study is to present an optimization approach to determine locations of new groundwater production wells, where groundwater is relatively less susceptible to groundwater contamination (i.e. more likely to obtain clean groundwater), the pumping rate is maximum or the cost of well installation and operation is minimum for a prescribed set of constraints. The approach also finds locations that are in suitable areas for new groundwater exploration with respect to land use. A regional-scale groundwater flow model is coupled with a hybrid optimization model that uses the Differential Evolution (DE) algorithm and the Broyden-Fletcher-Goldfarb-Shanno (BFGS) method as the global and local optimizers, respectively. Several constraints such as the depth to the water table, total well length and the restriction of seawater intrusion are considered in the optimization process. The optimization problem can be formulated either as the maximization of the pumping rate or as the minimization of total costs of well installation and pumping operation from existing and new wells. Pumping rates of existing wells that are prone to seawater intrusion are optimized to prevent groundwater flux from the shoreline towards these wells. The proposed simulation-optimization model is demonstrated on an existing groundwater flow model for the Tahtali watershed in Izmir-Turkey. The model identifies for the demonstration study locations and pumping rates for up to four new wells and one new well in the cost minimization and maximization problem, respectively. All new well locations in the optimized solution coincide with areas of relatively low groundwater vulnerability. Considering all solutions of the demonstration study, groundwater vulnerability indices for new well locations range from 29.64 to 40.48 (on a scale of 0-100, where 100 indicates high vulnerability). All identified wells are located relatively close to each other. This implies that the method pinpoints the best area for new wells both in terms of groundwater quantity and quality. Furthermore, sensitivity analysis results indicate that identification results are insensitive to the selection of DE parameters. (C) 2014 Elsevier B.V. All rights reserved.