Drinking water can be a major source of poly- and perfluoroalkyl substance (PFAS) exposure for humans. The lack of historic data on PFAS drinking-water concentrations and consumption patterns are a limiting factor for developing estimates of past exposure. Here, in contribution to a community-scale PFAS health effects study near fire training facilities that contaminated a local aquifer with PFASs, we present a novel water-infrastructure, mass-balance mixing model coupled to a non-steady state, single-compartment toxicokinetic model that used Monte Carlo simulations to estimate the start of PFAS exposure in drinking water for individuals within three PFAS-impacted communities in El Paso County, Colorado. Our modeling focused on perfluorohexane sulfonic acid (PFHxS) because median serum PFHxS concentrations in a sample of local residents (n = 213) were twelve times the median observed in the U.S. National Health and Nutrition Examination Survey (2015-2016). Modeling results for study participants were grouped according to their community of residence, revealing a median start of exposure for the town of Fountain of 1998 (25-75% interquartile range [IQR], 1992 to 2010), 2006 (IQR 1995 to 2012) for Security, and 2009 (IQR 1996-2012) for Widefield. Based on the towns' locations relative to an identified hydraulically upgradient PFAS source, the modeled exposure sequencing does not completely align with this conceptual flow model, implying the presence of an additional PFAS source for the groundwater between Widefield and Fountain.