U.S. Geological Survey monitoring programs extensively used two analytical methods, gas chromatography/mass spectrometry and liquid chromatography/mass spectrometry, to measure pesticides in filtered water samples during 1992–2012. In October 2012, the monitoring programs began using direct aqueous-injection liquid chromatography tandem mass spectrometry as a new analytical method for pesticides. The change in analytical methods, however, has the potential to inadvertently introduce bias in analysis of datasets that span the change.

A field study was designed to document performance of the new method in a variety of stream-water matrices and to quantify any potential changes in measurement bias or variability that could be attributed to changes in analytical methods. The goals of the field study were to (1) summarize performance (bias and variability of pesticide recovery) of the new method in a variety of stream-water matrices; (2) compare performance of the new method in laboratory blank water (laboratory reagent spikes) to that in a variety of stream-water matrices; (3) compare performance (analytical recovery) of the new method to that of the old methods in a variety of stream-water matrices; (4) compare pesticide detections and concentrations measured by the new method to those of the old methods in a variety of stream-water matrices; (5) compare contamination measured by field blank water samples in old and new methods; (6) summarize the variability of pesticide detections and concentrations measured by the new method in field duplicate water samples; and (7) identify matrix characteristics of environmental water samples that adversely influence the performance of the new method. Stream-water samples and a variety of field quality-control samples were collected at 48 sites in the U.S. Geological Survey monitoring networks during June–September 2012. Stream sites were located across the United States and included sites in agricultural and urban land-use settings, as well as sites on major rivers.

The results of the field study identified several challenges for the analysis and interpretation of data analyzed by both old and new methods, particularly when data span the change in methods and are combined for analysis of temporal trends in water quality. The main challenges identified are large (greater than 30 percent), statistically significant differences in analytical recovery, detection capability, and (or) measured concentrations for selected pesticides. These challenges are documented and discussed, but specific guidance or statistical methods to resolve these differences in methods are beyond the scope of the report. The results of the field study indicate that the implications of the change in analytical methods must be assessed individually for each pesticide and method.

Understanding the possible causes of the systematic differences in concentrations between methods that remain after recovery adjustment might be necessary to determine how to account for the differences in data analysis. Because recoveries for each method are independently determined from separate reference standards and spiking solutions, the differences might be due to an error in one of the reference standards or solutions or some other basic aspect of standard procedure in the analytical process. Further investigation of the possible causes is needed, which will lead to specific decisions on how to compensate for these differences in concentrations in data analysis. In the event that further investigations do not provide insight into the causes of systematic differences in concentrations between methods, the authors recommend continuing to collect and analyze paired environmental water samples by both old and new methods. This effort should be targeted to seasons, sites, and expected concentrations to supplement those concentrations already assessed and to compare the ongoing analytical recovery of old and new methods to those observed in the summer and fall of 2012.