Modeling PFAS Transport in Groundwater - Exploring current approaches and evaluating parameter importance

Abstract: PFAS contamination in drinking water is a current problem, and new regulations for drinking water limits were recently implemented in each member country of the EU in January this year. Understanding the spread of PFAS in groundwater is therefore important to prevent it from reaching drinking water sources. Groundwater modeling is a valuable tool for this purpose. However, due to the specific characteristics of PFAS, such as sorption, there is a knowledge gap regarding the optimal implementation of this tool. Therefore, this thesis aims to investigate the currently used modeling strategies and examine the parameters that seem to have the most significant impact on PFAS transport. This research also seeks to understand the importance of identifying the specific PFAS compounds present in the contaminant. A literature review was conducted, revealing that the most commonly used method involved transport simulations using MODFLOW. As a result, a similar approach using FEFLOW was adopted with the aim of conducting a sensitivity analysis to identify which parameters are most crucial in determining the obtained PFAS concentration in each node. A total of 500 simulations were performed for four different scenarios, considering two different aquifers and the transport of two different PFAS compounds. The results of these simulations were used to train a random forest regression model, which exhibited a high level of accuracy in predicting the resulting concentration at specific nodes. A sensitivity analysis was conducted on the model, revealing that hydraulic conductivity was the most important parameter in steady-state modeling, followed by recharge for sandy aquifers and longitudinal dispersivity for sand and gravel aquifers. Sorption did not have a significant impact on the results in this context. However, in transient modeling, sorption was found to be of great importance, suggesting that the specific PFAS compound may play a significant role in such scenarios. Lastly, the results differed significantly between steady-state and transient-state modeling, indicating that the choice of modeling approach is also of great importance.