Interactive response surface approaches using computationally intensive models for multiobjective planning of lake water quality remediation

Recent work has demonstrated the utility of interactive response surfaces in integrating decision making and complex process-based modeling of environmental systems. Specifically, we focus on problems that require computationally expensive dynamic models where a limited number of simulations are possible and so traditional meta-modeling might be not an option for the optimization process. Within this constraint, the interactive global response surface method aims to develop a complete picture of how the system responds to a planning decision on the basis of a restricted number of simulations. The interactive local response surface method utilizes the current best alternative to determine improvements on the basis of relatively small changes in the current decision. We outline the use of both methods to a real example of remediation in a water supply reservoir, demonstrating the key advantages and disadvantages of each method and describe a new, interactive cascade approach that provides an improved solution to the problem.