New modeling frameworks for assessing risk and uncertainty associated with long-term PFAS contamination in groundwater

Abstract

Per- and poly-fluoroalkyl substances (PFAS) are synthetic organic compounds that have emerged as environmental contaminants of concern. The vadose zone has been shown to be a long-term source of PFAS contamination to groundwater because of complex processes that retain PFAS in the vadose zone. The net effect of these processes is a time-dependent loading or leaching of PFAS to groundwater. Rigorous mathematical models describing the effects of these processes on PFAS transport have been developed, but these models are computationally expensive and require detailed prior knowledge of geologic, hydrologic, and transport parameters, limiting their applications to field sites. This makes assessing the impact of future water infrastructure projects on PFAS transport, along with the long-term vulnerability of existing and planned water infrastructure to PFAS contamination, a difficult undertaking, particularly in places where geological features are uncertain. The work presented in this thesis is motivated by French Island, a PFAS-impacted site in west-central Wisconsin, as a test case for developing new and simpler modeling frameworks for forecasting long-term PFAS behavior. Chapter 2 presents a new inverse modeling framework for determining long-term leach-ing rates from the vadose zone. Instead of a highly parameterized process-based forward model, a simple empirical model is fit to PFAS concentrations sampled from wells down-gradient of a source zone. This empirical model treats PFAS leaching from the vadose zone as a source loading function to the saturated zone, where PFAS is transported via the advection-dispersion equation. The method is validated using simulated test cases resembling conditions at French Island. The results show the method is viable but is sensitive to the time well samples were taken relative to PFAS breakthrough. Results are improved by including multiple sample times or a Bayesian prior on the PFAS fingerprint. The model is highly generalizable and can be applied to a variety of other contaminant transport problems. Chapter 3 details the risk assessment of long-term PFAS contamination in a proposed municipal well at French Island. Significant erosion of the Eau Claire aquitard in the vicinity of French Island increases the possibility of pumping-induced contamination in the deep Mt. Simon aquifer, but the geometry of the erosional windows is not well constrained. One hundred realizations of the thickness and geometry of the Eau Claire aquitard were generated using geostatistical kriging and incorporated into numerical flow and particle tracking models for French Island. Particles released from known source zones were not captured by the proposed well. However, particles released from private wells with PFAS levels exceeding EPA MCLs were captured in every simulation. There is a high probability of capturing PFAS within 1000 meters of the well, indicating the vertical hydraulic gradient induced by well pumping can potentially move contaminants deeper into the aquifer. Careful monitoring of the hydraulic head gradients should therefore be undertaken. If strong vertical gradients matching the model are produced, alternative solutions, including decreasing the pumping rate or casing the well at a lower depth, should be explored.