A critical analysis of published data to discern the role of soil and sediment properties in determining sorption of per and polyfluoroalkyl substances (PFASs)

Widespread usage of per- and polyfluoroalkyl substances (PFASs) has caused major environmental contamination globally. The hydrophilic and hydrophobic properties of PFASs affect the sorption behaviour and suggest organic carbon may not be the only factor affecting sorption. We reviewed the quality of all data published in peer-reviewed literature on sorption of PFASs to critically evaluate the role organic carbon (OC) and other properties have in sorption of PFASs in soils or sediments. The largest data sets available were for perfluorooctanoic acid (PFOA, n = 147) and perfluorooctane sulfonic acid (PFOS, n = 178), and these analyses showed very weak correlations between sorption coefficient (Kd) and OC alone (R2 = 0.05-0.07). When only laboratory-derived Kd values of PFASs and OC were analysed, the R2 values increased for PFOA (R2 = 0.24, n = 42), PFOS (R2 = 0.38, n = 69), perfluorononanoic acid (PFNA, R2 = 0.77 n = 12), and perfluorodecanoic acid (PFDA, R2 = 0.78, n = 13). However, the relationships were heavily skewed by one or two high OC values. Similarly there was no significant relationship between Kd values and pH for PFOS (R2 = 0.06) and PFOA (R2 = 0.07), across a range of environmental pH values. Our analyses showed sorption behaviour of a range of PFASs could not be explained by a single soil or sediment property. Multiple regression models better explained the sorption behaviour of a number of PFASs. Regressions of OC and pH together explained a significant proportion of the variation in Kd values for 9 out of 14 PFASs and 8 of these regressions had ≥10 data points. This review highlighted that at least OC, pH and clay content are properties having significant effect on sorption. There is a clear need for more data and studies with thorough characterisation of soils or sediments to better understand their role in PFASs sorption. Current assessments based on OC alone are likely to be erroneous.

Yasong Li, Danielle P. Oliver, Rai S. Kookana