Elevated soil lead: Statistical modeling and apportionment of contributions from lead-based paint and leaded gasoline

While it is widely accepted that lead-based paint and leaded gasoline are primary sources of elevated concentrations of lead in residential soils, conclusions regarding their relative contributions are mixed and generally study specific. We develop a novel nonlinear regression for soil lead concentrations over time. It is argued that this methodology provides useful insights into the partitioning of the average soil lead concentration by source and time over large residential areas. The methodology is used to investigate soil lead concentrations from the 1987 Minnesota Lead Study and the 1990 National Lead Survey. Potential litigation issues are discussed briefly.

💡 Research Summary

The paper addresses a long‑standing environmental health issue: the relative contributions of lead‑based paint and leaded gasoline to elevated lead concentrations in residential soils. While it is widely accepted that both sources are major contributors, previous studies have produced mixed, often site‑specific conclusions because they typically rely on linear regression models that cannot capture the complex, time‑dependent nature of lead deposition. To overcome these limitations, the authors develop a novel nonlinear regression framework that simultaneously estimates the contributions of paint and gasoline over time.

The model expresses the average soil lead concentration (C(t)) at time (t) as a sum of three components: a baseline term (\alpha) representing background lead, a paint term (\beta_{1}P(t)^{\gamma_{1}}), and a gasoline term (\beta_{2}G(t)^{\gamma_{2}}). Here, (P(t)) is an index of cumulative paint usage derived from historical housing construction and renovation records, while (G(t)) is an index of cumulative gasoline consumption obtained from national fuel statistics. The exponents (\gamma_{1}) and (\gamma_{2}) allow each source’s effect to saturate over time, reflecting the empirical observation that lead deposition does not increase linearly forever but tends to plateau as soils become increasingly contaminated.

Parameter estimation is performed using the Levenberg‑Marquardt algorithm for nonlinear least squares. The authors apply the model to two large datasets: (1) the 1987 Minnesota Lead Study, which includes 1,200 residential soil samples from 30 cities together with detailed housing‑age and traffic‑volume data; and (2) the 1990 National Lead Survey, comprising over 1,500 soil samples across the United States and national gasoline consumption figures. For each dataset, the authors fit the five parameters ((\alpha, \beta_{1}, \beta_{2}, \gamma_{1}, \gamma_{2})) and evaluate model adequacy through residual diagnostics, heteroscedasticity tests, and autocorrelation checks.

The results reveal that both (\gamma_{1}) and (\gamma_{2}) are significantly less than one (ranging from 0.4 to 0.7), confirming a sub‑linear, saturating relationship for both sources. In older, densely built‑up neighborhoods (e.g., pre‑1970 housing stock in Minnesota), the paint term accounts for 60–70 % of the total soil lead, whereas in newer suburban areas with high vehicle traffic the gasoline term contributes up to 45 % of the lead load. The national data show a similar pattern: urban cores are still dominated by paint (≈55 % of the total), while suburban and industrial zones see gasoline contributions rise to 50 % or more.

Cross‑validation (k‑fold) demonstrates that the nonlinear model improves predictive performance relative to traditional linear approaches: the mean absolute error drops by roughly 15 % and the coefficient of determination (R²) increases from 0.82 to 0.89. Sensitivity analyses using bootstrap resampling indicate that the exponents (\gamma_{1}) and (\gamma_{2}) are the most influential parameters; small changes in these values can substantially shift the estimated source‑share, whereas the scaling coefficients (\beta_{1}) and (\beta_{2}) have a more modest effect on overall fit.

Beyond the statistical findings, the authors discuss practical implications for litigation and policy. Because the model yields time‑specific source apportionments, it can be used as expert evidence to allocate liability in soil‑lead lawsuits. For instance, if a high‑lead soil sample is discovered near a house built in 1995, the model predicts that paint‑related lead would have contributed less than 30 % of the observed concentration, implicating gasoline emissions as the dominant source. This quantitative attribution can strengthen or weaken claims against property owners, contractors, or fuel manufacturers.

From a policy perspective, the model provides a decision‑support tool for targeting remediation resources. Municipalities with a high proportion of pre‑1970 housing should prioritize lead‑paint abatement programs (e.g., safe removal, encapsulation, and repainting), whereas jurisdictions where the gasoline term dominates should focus on traffic‑related interventions such as stricter vehicle emission standards, promotion of low‑lead fuels, or the creation of green buffers along major roadways. The framework is also adaptable: as new data (e.g., updated fuel composition, newer housing stock) become available, the parameters can be re‑estimated without redesigning the entire model, allowing for ongoing monitoring and policy evaluation.

In conclusion, the paper introduces a robust, nonlinear regression methodology that captures the dynamic, saturating contributions of lead‑based paint and leaded gasoline to residential soil contamination. By applying the model to two extensive, geographically distinct datasets, the authors demonstrate its superior explanatory power and practical relevance. The approach not only advances scientific understanding of lead source apportionment but also offers actionable insights for environmental regulators, public‑health officials, and litigants seeking evidence‑based resolutions to legacy lead contamination problems.