Data Quality Performance for Different Sampling Strategies for Residential Soil Lead (Pb).

Field Sampling, Measurement and Sensor Technology
Oral Presentation

Prepared by

Contact Information: Crumbling.deana@epa.gov; 703-603-0643


ABSTRACT

Thousands of residential and other properties in the U.S. are sampled each year to detect high levels of lead and/or arsenic, due to concern of contamination by past mining or smelting activities or materials. The financial burden of these investigations is an incentive to find procedures that reduce costs while still providing the desired level of confidence that unacceptable exposures and unnecessary cleanups are avoided.

In partnership with Regional Offices, the Technology Integration and Information Branch (TIIB) within EPA’s HQ Superfund program conducted two field studies that examined several sampling design factors relevant to residential sampling for metals. This presentation will show study findings for data variability and decision error rates for composite samples formed from 5, 9, or 30 increments and compared to data from a previous study employing discrete samples.

The following describes the three field studies and their data sets.

CO Pueblo: A screening sampling study performed and published by the Colorado Dept. of Public Health and the Environment (CDPHE) under a Cooperative Agreement with EPA Region 8 evaluated 47 residential properties within the town of Pueblo, CO, portions of which were impacted by Pb and As from past smelter operations. Surficial soil only was sampled June 21-23, 2010, from 81 yard areas (such as a back or front yard, equivalent to a “decision unit” or “DU”). Five grab samples were collected per yard area, and these were treated as equivalent to “replicates” from the standpoint of DU decision-making (since decisions about small soil areas are often based on single sample results). The CDPHE report itself did not include direct measurements of between- or within-sample heterogeneity since the study was for screening purposes only, and commonly accepted sampling techniques were used. For the purpose of the sampling method comparison presented here, data variability measures were calculated from the raw data provided in the report (CDPHE, 2011).

EPA Pueblo: EPA Region 8 & EPA HQ collaborated on a pilot study involving 12 residential properties within the same area of the town of Pueblo, CO. The pilot study was designed to evaluate the performance of 30-point vs. 5-point composites for controlling between-sample soil heterogeneity when sampling yard DUs, and the ability of sample processing to control within-sample heterogeneity. Each yard area was sampled at 4 depth intervals: the surficial top inch (or 2 inches, depending on the presence of vegetation); the 1 (or 2) to 6 inches depth interval; the 6 to 12 inches depth interval; and the 12 to 18 inches depth interval. This produced a total of 155 DUs. All 155 DUs were sampled with independent triplicate 5-point composite samples. Twenty of the 155 DUs were also sampled with triplicate 30-point composites, so that those 20 DUs yielded 6 independent field samples: three 5-point composites and three 30-point composites. All composite samples were processed identically to control within-sample heterogeneity. Sample processing included drying, disaggregation, and sieving to a <250-micron target particle size.

Furnace Creek: EPA Region 7, EPA HQ & the MO Department of Natural Resources collaborated on a pilot study involving 18 residential properties in the Furnace Creek Lead Superfund Site of Caledonia, Washington County, MO (impacted by Pb from mining operations). A total of 23 yard areas (DUs) were sampled for surficial soil. All DUs were sampled with triplicate 5-, 9-, and 30-point composites, so that all DUs produced 9 independent samples each. All samples were processed to control within-sample heterogeneity via drying, disaggregation, and sieving to a <250-micron particle size.

The presentation will summarize and compare data variability and false positive/false negative decision error rates observed with each sampling technique.