A Method to Assess Analytical Uncertainties Over Large Concentration Ranges With Reference to Volatile Organics in Water

Abstract

AbstractThe uncertainty associated with a volatile organic concentration measurement is a function of variability and bias introduced at the various levels of sample handling: collection, storage, and analysis. During the past decade, sampling materials and the development and/or improvement of sampling protocols have been the subject of considerable research activity. As a result, in cases of samples properly handled, the analytical variability can be the dominant source of uncertainty in a given concentration value. Here analytical variability refers to any error that might arise during analysis, including the detector response error and any sample handling errors common to both standards and samples. This can be a particular concern for field analyses by gas chromatography (GC), Well‐established statistical methods are available to estimate analytical uncertainty from linear calibration curves, but these methods are poorly suited for the analysis of volatile organics because organic samples frequently require instrument calibration (usually GC) over several orders of magnitude in concentration. If a single linear calibration curve is used to determine sample concentrations and uncertainties, then unrealistically large uncertainties may be assigned to low concentration samples. However, the methods can be adopted for extended concentration range calibration curves by breaking the overall calibration line down into smaller sub‐calibration lines that span smaller ranges. These can then be examined and used selectively to determine concentrations with more appropriate uncertainties attached. The method of multiple callbration line analysis described here is suitable for programming with any high level computer language. It can be used to calculate meaningful analytical uncertainty values for any substance analyzed over a wide range in concentrations (i.e., an order of magnitude or more).