Graphical Methods for Assessing Changes in Water Quality

Abstract

The purpose of this paper is to demonstrate seasonal subseries and multipercentile graphs for analyzing trends in environmental data time series (Tukey 1977; Cleveland 1985). These simple graphical methods do not require the user to make assumptions regarding properties of the data, such as normality, constant variance, and serial independence. A modest number of missing values (up to 30, assuming they occur at random intervals) should not affect the plots enough to alter the interpretation. The plots allow one to see more clearly what is happening at the extremes, and seasonal effects are revealed. The methods are illustrated by using five day BOD data, measured at monthly intervals from 1977 to 1992 at the DePere station on the Fox River in Wisconsin.