This research program was aimed at developing an objective methodology for water quality management on a river basin scale. To that end, a mathematical model has been formulated to determine the best configuration of wastewater treatment plants consistent with either fixed-emission standards or prescribed river quality objectives. It will, of course, be appreciated that the latter case is considerably more difficult since this involves not only site selection but also waste load allocation. In the case of waste load allocation it was first necessary to use a process-based river water quality simulation model to predict the impact of different combinations of effluent discharge standards on the river. For reasons of computational efficiency an artificial neural network was employed to replicate the process-based model, which was then used in conjunction with a genetic algorithm to determine both the best sites and individual effluent discharge standards, subject to meeting the required river water quality. The overall model has been applied to the upper Thames basin in southern England, initially for site selection alone and then for site selection with waste load allocation. The results show that the genetic algorithm performs well for both options, thereby providing an efficient means of planning wastewater treatment on a regional basis.