The use of method detection limits in environmental measurements.

Abstract

Environmental measurements often produce values below the method detection limit (MDL). Because low or zero values may be used in determining compliance with regulatory limits, in determining emission factors (typical concentrations emitted by a given type of source), or in modeling efforts, there is considerable interest in methods for determining detection limits and in procedures for using MDLs in summary statistics. Selection of methods for determining detection limits and treatment of MDL data should depend on the end-use of the data. The MDL should be determined in the environmental matrix that contains the analyte of interest. Treatment of missing values (values below the MDL) should be specified before the initiation of a measurement effort. When calculating a mean value for a source or area, one approach is to set the missing values at the MDL. This approach is appropriate when there are significant risks associated with incorrectly assuming a low level of a toxic compound; this method of summarizing data introduces a positive bias. Missing values may be assumed to equal zero if there is not a high risk of overlooking a hazard; this method produces a negative bias. However, if enough measurement data are available and the distribution of these data can be established then look-up tables can be used to estimate missing values, using statistically based estimates, without introducing high or low biases in a calculated mean.