The effects of grinding methods on metals concentrations in soil.

Multi-increment sampling (MIS) has been most extensively used for munitions constituents at environmental sites where a high degree of contaminant heterogeneity exits. A revised method (USEPA Method 8330B) for explosives that uses MIS was announced in the fall of 2006, but similar guidance has not been reported for testing metals in soils. Questions have been raised as how to prepare representative analytical samples for metals determination from field composites. Three different grinding procedures were used in this study for three soil types to determine if grinding (relative to homogenizing soil without grinding) increases metal concentrations and decreases variability. The performance of these procedures was demonstrated via the analysis of replicates (n=16 for two soil types) using statistical evaluations that included calculations of various descriptive statistics (e.g., medians, means and standard deviations), Kruskal-Wallis (KW) tests for the medians and two tests for the variances (Bartlett's and Levene's test). There was a slight increase in concentrations for several of the metals in the clay loam soil after grinding, although the increases were a small percentage of the concentrations measured. The standard deviations (and variances) for replicate digestions and analyses generally decreased, although anomalies were observed. The grinding methods increased precision overall, however, the data indicated that the roller mill grinding was not as effective as the other grinding methods tested.

Determining "reliable detection limits" (RDLs) for environmental analyses.

It is generally accepted that the method detection limit or MDL (defined in 40 CFR 136, Appendix B) provides protection against false positives 99% of the time. This is correct, but only for the next single measurement performed after the MDL is determined. Subsequent measurements are not protected against false positives with the same degree of confidence, and there is no protection for false negatives. This paper provides a simple cost-effective approach for estimating the "reliable detection limit." Unlike the MDL, the statistic may be used for an indefinite number of future measurements and minimizes false negatives.