Sampling error associated with collection and analysis of soil samples at TNT‐contaminated sites

This study assessed short-range spatial het- erogeneity of TNT concentrations in surface soils at ex- plosives-contaminated sites. Discrete and composite samples were analyzed by both on-site calorimetric techniques and standard laboratory protocols. Three lo- cations were sampled at each of three installations, and the results were used to estimate the relative contribu- tions of analytical error and sampling error. The major contaminant at seven of the nine sampling locations was TNT, and the on-site calorimetric method provided results that were in excellent agreement with laboratory results from the use of SW846 Method 8330. At four of the seven TNT locations, short-range concentration variations were modest and analyte distribution was sufficiently Gaussian to allow application of normal dis- tribution statistics to fractionate the total error vari- ances. For these four locations, standard deviations due to sampling were greater than the corresponding standard deviations due to analysis by factors ranging from 2.6 to 22.8. This relationship held whether charac- terization was done with the use of on-site analysis or laboratory analysis. For the other three TNT locations, enormous short-range spatial heterogeneity was en- countered and sampling error overwhelmed analytical error. To improve estimates of mean concentrations, sampling error was reduced by the use of composite sampling strategies. Overall, this study indicates that characterization of explosives-contaminated sites with the use of a combination of composite sampling, in- field sample homogenization, and on-site calorimetric analysis is an efftcient method of obtaining accurate and precise mean concentration estimates that are rep- resentative of the area. @ 1997 John Wiley & Sons, Inc. Field Analyt Chem Technol 1:151-163,1997