Statistical and economic considerations for improving regulatory water quality monitoring networks

Abstract

The assignment of sampling frequencies in regulatory water quality monitoring networks has often been performed with little or no statistical or cost-effectiveness analysis. The research effort described here has attempted to address this problem through the development of appropriate statistical and economic analysis tools which might be applied by regulatory agencies. A technique is presented for predicting confidence interval widths about annual means or annual geometric means for water quality constituents while considering (1) serial correlation and (2) seasonal variation of the quality time series. These two effects are quantified by fitting deterministic seasonal models and time series models of the autoregressive-moving average type to historic water quality records. The statistical tools are illustrated via application to three sets of water quality observations, and the consequences of applying more elementary statistical tools in the determination of confidence interval widths are then examined. A dynamic programming algorithm is developed to assign sampling frequencies throughout a network in order to achieve desirable confidence Interval widths about annual geometric means for selected quality constituents while operating within a fixed budget. The use of the algorithm is illustrated through an application, and a sensitivity analysis is performed to study the effect of variation in values of input variables on the results of the dynamic programming solution.