Use of overlap studies to evaluate method changes in water chemistry protocols

Abstract

Long-term monitoring projects are usually plagued with method changes that occur in the midst of the monitoring record. Such changes can affect the data, resulting in observations of long-term trends that reflect the change in methods rather than the monitored system. This article describes two statistical approaches to evaluate the effect of method changes, illustrated by several examples from the U.S. Environmental Protection Agency's Long-Term Monitoring Project, a study of the effects of acidic deposition on surface water chemistry. Structural regression models or paired t-tests were applied to various overlapping datasets to determine whether statistically significant differences existed between methods. Statistically significant differences between method changes were seen for each of the following: different filter types, a change in anion analysis from colorimetric to ion Chromatographic techniques, and a change in sample collection method from an integrated hose sampler to a Kemmerer sampler. The characteristics under which each statistical approach was applied are discussed, as are considerations regarding calibration of the older portions of the data.