Design, analysis, and interpretation of field quality-control data for water-sampling projects

Abstract

The process of obtaining and analyzing water samples from the environment includes a number of steps that can affect the reported result. The equipment used to collect and filter samples, the bottles used for specific subsamples, any added preservatives, sample storage in the field, and shipment to the laboratory have the potential to affect how accurately samples represent the environment from which they were collected. During the early 1990s, the U.S. Geological Survey implemented policies to include the routine collection of quality-control samples in order to evaluate these effects and to ensure that water-quality data were adequately representing environmental conditions. Since that time, the U.S. Geological Survey Office of Water Quality has provided training in how to design effective field quality-control sampling programs and how to evaluate the resultant quality-control data. This report documents that training material and provides a reference for methods used to analyze quality-control data. Quality-control data are those generated from the collection and analysis of quality-control samples, and are used to estimate the magnitude of errors in the process of obtaining environmental data. “Bias” and “variability” are the terms used in this report for the two types of errors in environmental data that are quantified by the data from quality-control samples. Bias is the systematic error inherent in a method or measurement system. Variability is the random error that occurs in independent measurements. The types of field quality-control samples discussed in this report include blanks, spikes, and replicates. Blanks are samples prepared with water that is intended to be free of measurable constituents that will be analyzed by the laboratory; blanks are used to estimate bias caused by contamination. Spiked samples are modified by addition of specific analytes; spikes are used to determine the performance of analytical methods and to estimate the potential bias due to matrix interference or analyte degradation. Replicate samples are two or more samples that are considered to be essentially identical in composition. Replicates are used to evaluate variability in analytical results. Various sub-types of these quality-control samples are defined and discussed in this report, and guidance is provided for incorporating the proper samples into the design for a project. The concept of inference space is introduced to help determine where and when quality-control samples should be collected as well as which environmental samples are related to a set of quality-control samples. The recommended basic quality-control design incorporates project-specific considerations, such as the objectives and scale of the study, and hydrologic and chemical conditions within the study area. The report provides extensive information about statistical methods used to analyze quality-control data in order to estimate potential bias and variability in environmental data. These methods include construction of confidence intervals on various statistical measures, such as the mean, percentiles and percentages, and standard deviation. The methods are used to compare quality-control results with the larger set of environmental data in order to determine whether the effects of bias and variability might interfere with interpretation of these data. Examples from published reports are presented to illustrate how the methods are applied, how bias and variability are reported, and how the interpretation of environmental data can be qualified based on the quality-control analysis.