Soil-to-plant transfer of heavy metals and an assessment of human health risks in vegetable-producing areas of São Paulo state

While contaminated food products are known to be a leading source of exposure to potentially toxic elements (PTEs), for the general population, few studies have been carried out to examine PTEs levels in soils and plants in wet tropical regions such as Brazil. While the most commonly used index for estimating PTEs accumulation in vegetables and the subsequent exposure to humans who eat them is the bioconcentration factor (BCF) - the ratio between the concentration of metals in the edible portions of produce and their total concentration in soils - the BCF does not provide an adequate description of soil-to-plant metal transfers. A better understanding of such transfers requires information about the soil attributes that influence the availability of PTEs to plants. The state of São Paulo (SP) is the largest consumer of vegetables in Brazil, as well as the largest and most diversified producer. Studies are therefore needed on PTEs concentrations in soils and vegetables, in order to assess their quality under guidelines established by Brazilian legislation. It is likewise crucial to establish critical limits of these elements in soils, via models that assess risks to human health, based on data that reflect current conditions in the soils of São Paulo. The objectives in this study were: (i) to characterize and to evaluate the relations between the concentrations of Cd, Cu, Ni, Pb and Zn in soils and in vegetables from the \"Green Belt\" of the state of São Paulo, Brazil, taking the limits established by legislation into account; (ii) to develop empiric models to derive appropriate soil screening values and to provide an accurate risk assessment for tropical regions; (iii) to develop proposals for improved human health-based screening values for Cd, Cu, Ni, Pb and Zn in São Paulo soils, using soil - vegetable relations. With the exception of Cd, there was a positive correlation between pseudototais and bioavailable contents of PTEs. Cd and Pb content in plants, moreover, not significantly correlated with any of the variables studied. All models of random forests and trees were good predictors of results generated from a regression model and provided useful information about which covariates were important to forecast only for the zinc concentration in the plant. The soil-plant transfer models proposed in this study had a good performance and are useful for eight of the ten combinations (five metals versus two species). SP data combined with NL data for Cd in lettuce and for Ni and Zn in lettuce and in carrot when pH, organic carbon - OC and clay contents were included in the model. Including such soil properties results in improved relations between PTEs concentrations in soils and in vegetables to derive appropriate screening values for SP State. The model in which pH, OC and clay contents were included gave the most useful results with SP and NL data set combined for Cu, Pb, Zn in lettuce and for Cd and Cu in carrot. Our setup did not work for Ni and for Pb in carrot because the data models gave an inconsistent result and the combination of datasets did not or insufficiently improve the results. (AU)