Coupling compositional PCA with t-SNE radar signatures for characterizing a PFAS contaminated fractured chalk aquifer exploited for water supply.

This study aims to improve the understanding of the groundwater contamination by per- and polyfluoroalkyl substances (PFAS) at a strategic abstraction site located in a fractured chalk aquifer in Wallonia (South Belgium), to better secure and protect future exploitation and management. A hydrogeochemical investigation was conducted at the groundwater abstraction site, based on a sampling campaign and supported by an integrated multivariate statistical framework. This framework combines compositional data analysis, including compositional Principal Component Analysis (PCA), based on centered log-ratio (CLR) transformation, and advanced visualization techniques such as ternary diagrams, t-distributed Stochastic Neighbor Embedding (t-SNE), and radar plots to identify and characterize PFAS contamination patterns independently of absolute concentration effects. The results reveal the presence of multiple contamination sources and highlight three main PFAS pollution facies: Ensemble 1, driven by PFOS pollution but at low concentrations; Ensemble 2, grouping sampling points dominated by sulfonic compounds, mainly PFBS, and having the highest concentrations; and Ensemble 3 corresponding to surface water samples and samples collected as part of the monitoring of a contamination caused by a nearby motorway accident in 2019. On the whole, this integrated approach provides new insights into the complexity, spatial distribution, and potential origins of PFAS contamination in groundwater systems, and demonstrates the added-value of combining compositional PCA with t-SNE and radar plots visualization.