Spatial Prediction of Concentrations of Per- and Polyfluoroalkyl Substances (PFAS) in European Soils

Abstract

Currently, little is known about the spatial distribution of per- and polyfluoroalkyl substances (PFAS) in soils. In this study, machine learning was applied to a data set from the Map of Forever Pollution in Europe (MFPE) containing 6697 scattered PFAS soil concentration measurements to comprehensively predict PFAS concentrations in European soils. The model is based on a regression analysis between the PFAS soil concentrations and the distance to presumptive point sources indicated by the MFPE. Generally, decreasing PFAS concentrations were observed with increasing distances to the nearest point sources. Subsequently, on the basis of the regression analysis, a map was generated showing the PFAS concentrations in European soils by interpolating the model predictions at 10 000 random spots where no measurements were available. The map revealed that a significant portion of European soils is potentially contaminated with PFAS at concentrations of >5000 ng/kg. At this concentration, the mobilization of PFAS can lead to seepage water concentrations of 2-5 ng/L surpassing current and proposed drinking water guidelines in Europe. This illustrates the need for lower PFAS soil threshold values. Overall, the produced map provides, for the first time, comprehensive information about European PFAS soil contamination and serves as a basis for assessing its environmental risks.