SAMPLING AND ANALYTICAL TOLERANCE REQUIREMENTS FOR DETECTING TRENDS IN WATER-QUALITY

In designing a water quality sampling program, the detection threshold (equivalent to ''effect size'' in a power analysis) can be calculated as a function of the Type I and II error risks, sampling frequency an d duration, and (if appropriate) autocorrelation. Then the impact of s ampling and analytical errors on the trend that can be detected within that threshold should be considered. A procedure for doing so is desc ribed, enabling a priori prescription to laboratory and field staff of the analytical tolerances expected of them in their work. It assumes that the trend (a linear or step change) can be described by a paramet ric model.