The need to detect anthropogenic impacts in the natural environment has increased interest in the design of cost‐effective environmental monitoring networks. A variety of statistical models have been proposed for this purpose. This paper examines four statistical models, with varying degrees of complexity, used to represent the underlying characteristics of potential impacts. The models are incorporated into an optimization procedure used to select cost‐effective designs. Aquatic monitoring data from a nuclear power plant are used to test the robustness of the statistical models, analyze their sensitivity to input parameters, and determine the circumstances that require the use of more complex statistical models to design effective monitoring programs.