Objective ranges of soil-to-dust transfer coefficients for lead-impacted sites

Abstract

Residential yard soil and indoor dust datasets from eight communities near historical mining, smelting, and refining operations were used to quantify soil track-in, an important factor in evaluating indoor exposures to soil metals and to set residential soil cleanup levels. Regression analyses were used to derive slopes that represent mass soil-to-dust transfer coefficients or MSDs. Lead concentration data were available for all datasets. Arsenic data were available for six of the eight datasets. Cadmium and zinc data were available for one dataset, allowing limited comparison of MSDs for lead with other metals. Covariates that could indicate potential indoor sources of metals, such as house age and indoor heating source, were examined by multivariate regression analysis when available (three datasets). Covariates that could affect soil track-in, such as the amount of bare soil in the yard or having pets, were examined by stratified linear regression analysis when available (two datasets). Most of the R-squared values for lead, cadmium and zinc indicate a good to moderate fit (≥0.25), but for arsenic most indicate a poor fit (<0.25). Significant MSDs for models with a good to moderate fit range from 0.14 to 0.47 for lead, and 0.12 to 0.43 for the other metals (arsenic, cadmium, and zinc). The treatment of outliers was a significant methodological factor affecting the slope of the regressions. Substantial variability is expected among soils at residences due to both physical characteristics of each property and the ways in which residents interact with their home. Survey data providing information on various factors affecting soil track-in help to refine MSD estimates. For three of the datasets, covariate data were available that improved model fit by multivariate or stratified regression analysis for lead. When multivariate or stratified regression analyses were performed, the estimated MSD varied as little as <1% to as great as 200% depending on the dataset, but all estimates were below 0.4. Notably, the MSDs were lowest for the three datasets with the highest soil lead concentrations, i.e., those with average soil lead concentrations greater than 300 mg/kg after outlier removal. For five of the six datasets that had both arsenic and lead sampled, arsenic MSDs were much less than the lead MSDs; however, only two of the sites’ arsenic models had significant MSDs and adequate fit. Cadmium and zinc were only included in one dataset, limiting our ability to draw any conclusions from comparison to those MSDs. The results of our study are consistent with prior studies suggesting that MSDs for metals without internal sources are 0.3–0.4, and application of MSDs in that range will provide more reliable exposure estimates than the 0.7 default value used by the United States Environmental Protection Agency in the Integrated Exposure Uptake Biokinetic (IEUBK) Model.