Improved data collection techniques as well as increasing computing power are opening up new opportunities for the development of sophisticated models that can accurately reproduce hydrodynamic and biochemical conditions of water bodies. While increasing model complexity is considered a virtue for scientific purposes, it is a definite disadvantage for management (engineering) purposes, as it limits the model applicability to what-if analysis over a few, a priori defined interventions. In the recent past, this has become a significant limitation, particularly considering recent advances in water quality rehabilitation technologies (e. g., mixers or oxygenators) for which many design parameters have to be decided. In this paper, a novel approach toward integrating science-oriented and engineering-oriented models and improving water quality planning is presented. It is based on the use of a few appropriately designed simulations of a complex process-based model to iteratively identify the multidimensional function (response surface) that maps the rehabilitation interventions into the objective function. On the basis of the response surface (RS), a greater number of interventions can be quickly evaluated and the corresponding Pareto front can be approximated. Interesting points on the front are then selected and the corresponding interventions are simulated using the original process-based model, thus obtaining new decision-objective samples to refine the RS approximation. The approach is demonstrated in Googong Reservoir (Australia), which is periodically affected by high concentrations of manganese and cyanobacteria. Results indicate that significant improvements could be observed by simply changing the location of the two mixers installed in 2007. Furthermore, it also suggests the best location for an additional pair of mixers.