Optimization of Water Quality Monitoring Networks

Abstract

Water quality and biological monitoring provide an indication of the degree to which the natural environment has been affected by anthropogenic‐activities. Currently, this monitoring is both expensive and time‐consuming. A new cost‐effective approach to the design of aquatic monitoring networks is presented. Classical analysis of variance (ANOVA) techniques are reviewed and a modified ANOVA model with control station pairs is suggested. An interactive optimization procedure is presented that incorporates a modified gradient search algorithm to select designs which maximize the statistical power of a network for a specified budget or minimize the cost of a network for a specified statistical power requirement. The sensitivity of the model results are explofed as a function of the cost, the number of sampling stations, replicates, and occasions, the Type I and Type II error, estimates of data variance, and cost components for data describing an aquatic species from a New England power facility. It is shown that for specified power and cost, numerous solutions exist providing the designer with a wide selection of alternatives from which to choose.