When it is decided to transform former industrial sites into residential areas with kitchen gardens, site remediation must begin by quantifying the concentration of chemicals transferred from the soil to vegetables, in order to assess human oral exposure and related health risks.

We aimed to construct models for predicting the concentration of 16 polycyclic aromatic hydrocarbons (PAHs) in the edible portion of common garden vegetables grown in aged industrial soils.

Lettuce, carrots and potatoes were grown in 5 soils gradually polluted, for at least 35 years, and each “vegetable-polluted soil” pair was replicated 5 times. The plants were harvested at maturity and prepared as for eating, but not cooked. Two models were then adjusted by linear regression to the observed data, in a natural scale and after logarithmic transformation.

Benzo[a]pyrene concentrations up to 1.3, 13.4 and 29.0 μg/kg (fresh weight) were measured, respectively, in peeled carrots, unpeeled potatoes and lettuce leaves. Significant relationships observed between plant and soil concentrations made it possible to calculate predictive models for 11 PAHs in lettuce leaves and 9 PAHs in peeled carrots. Where the data could not be modeled, in particular for potatoes, both peeled and unpeeled, we propose an alternative to the regression model.

Our models give the mean concentrations of several PAHs on the US EPA list and their 95% prediction intervals in the edible part of lettuce and carrots grown in aged contaminated soils. Thus, they should be useful for conducting quantitative health risk assessments with uncertainty analysis for guiding decision-making, for example, during the process of remediating a polluted site.