Monitoring design and data analysis for trend detection

Abstract

This study was undertaken to assess the effectiveness of alternative water quality monitoring programs in detecting water quality trends over time, identifying differences in water quality between water bodies, and examining spatial variability in water quality within a single water body. Monthly historical data on phosphorus, nitrogen, and specific conductance from the North Carolina Division of Environmental Management and the U.S. Army Corps of Engineers were analyzed. The historical data on nutrients and specific conductance were first examined for deterministic patterns. In most cases, the concentration varied seasonally according to a cyclical pattern that repeated each year. In addition, concentration at riverine sampling sites was often inversely related to streamflow. Following removal of the deterministic patterns from the data, autocorrelation was found in a few cases. Once it was concluded that only white noise remained in residuals for a particular data series, the required sample size for a given power and significance was calculated. In many instances, a large number of samples were required to detect changes of 25 percent or less with low error rates, because of the high level of background variability.