Optimal design of a groundwater pump and treat system is a difficult task, especially given the computationally intensive nature of field-scale remediation design. Genetic algorithms (GAs) have been used extensively for remediation design because of their flexibility and global search capabilities, but computational intensity is a particularly difficult issue with GAs. This paper discusses a new competent GA, the hierarchical Bayesian Optimization Algorithm (hBOA), which is designed to reduce the computational effort. GAs operate by assembling highly fit segments of chromosomes (potential solutions), called building blocks. The hBOA enhances the efficiency of this process by using a Bayesian network to create models of the building blocks. The building blocks are nodes on the network, and the algorithm uses the network to generate new solutions, retaining the best building blocks of the parents. This work compares the performance of hBOA to a simple genetic algorithm (SGA) in solving a case study to determine if any benefit can be gained through the use of this approach. This work demonstrates that hBOA more reliably identifies the optimal solution to this groundwater remediation design problem. Copyright ASCE 2005.