How PFAS moves — and why it stays

“Forever chemicals” is a good headline and a poor explanation. PFAS do eventually move — through the soil you stand on, the unsaturated ground beneath it, the aquifer below that, and out into streams and lakes — but they move slowly, selectively, and differently at every step. This page walks that journey compartment by compartment. Each section explains the controlling mechanism in plain terms, states the one relationship that governs it, and grounds it in numbers from the PFAS models we build and run — labeled honestly as simulated or reported, never invented.

The interactive lesson lives right on this page — press play on 53 years in Section 2, then read the mechanisms behind what you watched.

1. What PFAS are, and why they don’t go away

PFAS — per- and polyfluoroalkyl substances — are built around a chain of carbon atoms whose hydrogens have been swapped for fluorine. The carbon–fluorine bond is the strongest single bond in organic chemistry. Nothing in ordinary soil, water, or biology has the energy to break it at scale: there is no microbe that eats it, no sunlight reaction in an aquifer, no natural weathering that takes it apart. That is the whole of “forever” — not that PFAS is immobile, but that it does not degrade. Whatever is released stays PFAS and simply redistributes.

The second fact that governs everything downstream is shape. A PFAS molecule is a surfactant — a soap. One end is a fluorinated tail that repels water; the other is a charged head (a sulfonate for PFOS, a carboxylate for PFOA) that dissolves happily in it. A molecule that is half water-hating and half water-loving cannot be comfortable anywhere, so it collects at surfaces: the boundary between water and air, water and a soil grain, water and organic matter. Every retention mechanism in the rest of this page is a consequence of that split personality. It is also why PFAS makes foam.

Most of the contamination we model is AFFF — aqueous film-forming foam, the firefighting agent used from the 1960s onward at airports and military bases to smother fuel fires. Decades of fire-training exercises poured concentrated PFAS onto bare ground at the same spots, year after year. Those training areas are the point sources behind most of the groundwater plumes in the United States, including the site this page uses as its worked example: the former Wurtsmith Air Force Base in Oscoda, Michigan.

2. See it in motion — the PFAS Journey

Before the mechanisms, watch the whole story once: press play on 53 years at Wurtsmith Air Force Base — the plume below is the real SWAT+/MODFLOW-6 simulation, year by year, and every particle moves at its compound’s retarded pace on the same clock. Water forgets in years; PFAS remembers for decades.

Each section below explains one leg of this journey — and points you back here to watch it. More hands-on lessons like this one: interactive water lessons.

3. Soil — sorption, chain length, and precursors

In the root zone, the question is how tightly the soil grips PFAS versus how much rides along with infiltrating water. Two grips matter, and both trace directly back to the surfactant shape.

The first is hydrophobic partitioning: the water-repelling tail buries itself in soil organic carbon to get away from water, the same way oil beads out of a glass of water. The longer the fluorinated chain, the stronger this pull — so an eight-carbon PFOS sorbs far more than a four-carbon PFBS. The strength is captured by the organic-carbon partition coefficient, K_oc; multiplying by a soil’s organic-carbon fraction gives its distribution coefficient K_d. The second grip is electrostatic: the negatively charged head is attracted to positively charged mineral surfaces (iron and aluminum oxide edges) and repelled by negative ones. Because that surface charge depends on pH, PFAS sorption is pH-sensitive in a way that neutral organic pollutants are not — lower pH generally means more positive surface charge and stronger anion sorption.

Real soils do not follow a simple straight-line partition. Sorption sites fill up, so the more PFAS is present the smaller the marginal fraction that sticks. The standard description is the Freundlich isotherm:

These are exactly the parameters our coupled model carries per compound. The table below is the sorption set that drives PFAS transport inside our MODFLOW 6 groundwater transport package — derived from the Li et al. (2019) soil-isotherm measurements and log K_oc. Read down the K_f and R columns and the chain-length story is right there: PFOS grips hardest and moves slowest; each carbon you remove loosens the grip.

One complication runs the other way. Much of what was sprayed was not PFOS or PFOA at all but precursors — larger polyfluorinated molecules (fluorotelomer alcohols and sulfonamides) that the soil’s own microbes slowly oxidize, one step at a time, into the stable perfluoroalkyl acids we measure. So a source zone keeps generating PFOA and PFOS for years after the spill, from a reservoir of precursors that routine sampling often misses (Houtz & Sedlak; Adamson et al. 2020). This is one reason source zones stay hot far longer than a single release would suggest.

4. The vadose zone — the mechanism that makes PFAS special

Between the root zone and the water table is the vadose zone — the unsaturated ground, part water and part air. For most contaminants it is just a delay. For PFAS it is the single most distinctive part of the whole journey, because of a mechanism that barely exists for anything else: air–water interfacial adsorption.

Recall the surfactant shape. In unsaturated soil, the water clings to grain surfaces and hangs in menisci, leaving a vast, convoluted internal surface where water meets air. PFAS molecules crowd onto that surface — tail up into the air, head down in the water (the diagram below) — and every molecule parked there is a molecule not moving downward. A conventional dissolved solute ignores the air entirely; PFAS treats each air pocket as one more thing to stick to. The drier the soil and the deeper the water table, the more air–water interface exists, and the more strongly PFAS is held back. Brusseau (2018, 2019) framed this quantitatively and showed it can dominate PFAS retention in unsaturated soils; Guo et al. (2020) built it into a vadose transport model.

The bookkeeping is a three-phase retardation factor — the standard aquifer retardation, plus one extra term for the air–water interface:

Put real numbers in and the travel times separate sharply by chain length. At Wurtsmith, where the water table sits about 3.5 m down, plain water crosses the vadose zone in about 1.4 years — but PFOS, retarded roughly fifteen-fold, needs on the order of 21 years, PFOA about 7.7, and short-chain PFBS about 4.9. (This is exactly what the animation above shows: switch the compound chips and watch the particles change pace while the faint water tracers slide past.)

The water-table depth is the master control. A shallow site drips into the aquifer within a working career; a deep one effectively never does within a human lifetime. Our deepest end-member — the Rogue River basin, with a median depth to water of 31 m — turns the same physics into a wall: retardation-scaled travel times of roughly 1,000 years for PFOS and ~220 years for PFOA. There, diffuse PFAS applied at the surface simply cannot reach groundwater on any management timescale, which is why any PFAS already in that aquifer must be legacy from a direct point source, not modern leaching. Same chemistry, opposite verdict — set entirely by how far the water table sits below the ground.

5. Groundwater — plumes that grow for decades

Once PFAS reaches the saturated zone it becomes a plume, and a plume is governed by three words: advection, dispersion, retardation. Advection carries it with the groundwater; dispersion spreads and dilutes the front; retardation holds it back as it repeatedly sorbs and releases from the aquifer solids.

The retardation factor is the plain-language heart of it — how many times slower than the water itself the contaminant front advances:

Our MODFLOW 6 transport parameterization carries R ≈ 12 for PFOS down to ≈1.6 for short-chain PFHxA (the table in Section 3). A PFOS plume therefore crawls at roughly a twelfth of the groundwater velocity — which is exactly why these plumes take decades to develop and why they keep growing long after the source is gone. The Wurtsmith simulation makes that concrete. Starting from the AFFF era, the modeled plume climbs from 18 contaminated cells in 1971 to 192 by 2024, still expanding thirty years after the base closed in 1993, with the deep aquifer holding on the order of 1,600 ng/L of PFOS long after the shallow source began to fade. Scrub the animation above to 1993 — the pad goes dark — then keep playing and watch the deep layer climb anyway.

This is why pump-and-treat is slow. At any moment only a small fraction of the mass (1/R) is in the water you can pump; the rest is sorbed to the aquifer, feeding slowly back into solution — plus PFAS that diffused into low-permeability silts and clays over the years now back-diffuses out for decades. You can pull water for years and barely dent the sorbed reservoir. The mass accounting at Wurtsmith frames the challenge: the model integrates roughly 13 kg of PFOS released over the operational era, while the single 2021 FT-02 excavation removed about 33 kg of PFAS (all analytes). Removed mass exceeds this single source term because it is a lower bound on total released mass across all source areas and all compounds — the point being that even a major excavation addresses the soil store, not the plume already in the aquifer.

6. Surface water — discharge, foam, and dilution

The plume does not stop at the property line — it surfaces. Where the water table intersects a stream, wetland, or lake, contaminated groundwater discharges as baseflow, and that groundwater-to-stream hand-off is the dominant delivery path for legacy PFAS. It is the reason a river can carry PFAS for decades with no active spill anywhere near it: the aquifer is the source.

At the surface, the surfactant shape reappears in the most visible way. PFAS concentrates at the air–water interface of the water body itself — the same mechanism that held it back in the vadose zone — and wind and waves whip that enriched surface skin into foam that can carry PFAS at hundreds to thousands of times the concentration of the water beneath it. The foam on Van Etten Lake is the vadose air–water interface story told again, in public, at the surface.

The reported monitoring around Wurtsmith shows the gradient from source to receptor:

Dilution helps — but only so far, and the arithmetic is unforgiving because the standard is so low. The EPA’s 2024 drinking-water limit (MCL) for PFOS and PFOA is 4 ng/L: four parts per trillion. A source porewater near 50,000 ng/L would still exceed that limit after being diluted more than 12,000-fold. Against a target that stringent, “it gets diluted downstream” is not the reassurance it sounds like — a marsh at 560 ng/L is still 140 times the limit. And because these compounds bioaccumulate, PFAS concentrations climb through the food web into fish, which is why lakes like Van Etten carry state fish-consumption advisories (Michigan EGLE) even where the open water reads far below the marsh.

7. Case study: Wurtsmith Air Force Base — the plume, in 3D

Every mechanism on this page plays out at one real site. From 1971 to 1993, fire-training exercises at Wurtsmith poured AFFF onto the same sandy pad, and the site’s geology set the terms: sandy soil over a water table only 3.5 m down meant fast delivery to the aquifer (Section 4), while sorption built the massive source-zone reservoir that is still there (Section 3). Five decades on, the simulated plume has crawled — retarded, dispersing — to where the aquifer discharges at Van Etten Lake and Clark’s Marsh, whose reported monitoring numbers you saw in Section 6. The model behind it is nested: the HUC12 watershed model (250 m) resolves regional flow and hands boundary conditions to a 30 m site model that resolves the plume — the full methodology →

Interactive 3D: the coupled SWAT+/MODFLOW-6 simulation of the Wurtsmith PFOS plume — drag to orbit, scrub the decades, hover wells and plume cells for their values. Simulated model output (uncalibrated research demonstration); groundwater observations from public agency reports. Best on a desktop browser.

8. How we model all of this — coupled SWAT+ / MODFLOW 6

Everything above lives in one coupled model. SWAT+ runs the land phase — rainfall, snow, evapotranspiration, runoff, the source loading at the training area, and the deep percolation that leaves the root zone. That percolation, and the PFAS it carries, becomes recharge to MODFLOW 6, which solves the three-dimensional groundwater flow and — through its Groundwater Transport (GWT) and Mobile Storage and Transfer (MST) packages — the PFAS transport in the aquifer, with Freundlich sorption using exactly the K_f/n set in Section 2. The two models exchange fluxes so that surface water and groundwater are solved as one system, not two disconnected halves.

To our knowledge this is the first coupling of SWAT+ with MODFLOW 6 that carries PFAS fate and transport natively in the groundwater engine — no bolt-on WASP, RT3D, or third-party water-quality code. The aquifer itself is not a cartoon: it is built by kriging hydraulic properties from state well-log databases, then validated against measured static water levels. Every model on the platform starts from that same automated pipeline; the PFAS sites are then finished by hand where it counts — the source term, the hydrostratigraphy, the site-specific parameters. Platform-generated, expert-finished.

References

1. Rafiei, V. & Nejadhashemi, A.P. (2023). Watershed-scale PFAS fate and transport model for source identification and management implications. Water Research 240, 120073. doi.org/10.1016/j.watres.2023.120073
2. Rafiei, V. (2026, under review). Watershed-scale PFAS fate and transport across surface water and groundwater: a two-way coupled SWAT+/MODFLOW 6 model quantifies the legacy groundwater contribution to in-stream loads. Water Research.
3. Rafiei, V. (2026, in preparation). When does the vadose zone control watershed PFAS delivery to groundwater? A vadose end-member analysis (Wurtsmith, Gabreski, Rogue).
4. Li, F. et al. (2019). Adsorption of perfluorinated acids onto soils: Kinetics, isotherms, and influencing factors. Science of the Total Environment 649, 504–514. doi.org/10.1016/j.scitotenv.2018.08.209
5. Higgins, C.P. & Luthy, R.G. (2006). Sorption of perfluorinated surfactants on sediments. Environmental Science & Technology 40(23), 7251–7256. doi.org/10.1021/es061000n
6. Brusseau, M.L. (2018). Assessing the potential contributions of additional retention processes to PFAS retardation in the subsurface. Science of the Total Environment 613–614, 176–185. doi.org/10.1016/j.scitotenv.2017.09.065
7. Guo, B., Zeng, J. & Brusseau, M.L. (2020). A mathematical model for the release, transport, and retention of PFAS in the vadose zone. Water Resources Research 56, e2019WR026667. doi.org/10.1029/2019WR026667
8. Adamson, D.T. et al. (2020). Mass-based, field-scale demonstration of PFAS retention within AFFF-associated source areas. Environmental Science & Technology 54(24), 15768–15777. doi.org/10.1021/acs.est.0c04472
9. U.S. EPA (2024). PFAS National Primary Drinking Water Regulation — maximum contaminant level (MCL) 4 ng/L for PFOA and PFOS. 89 FR 32532. www.epa.gov/sdwa/and-polyfluoroalkyl-substances-pfas
10. Michigan EGLE / MPART. Wurtsmith (Oscoda) PFAS response: surface-water and foam monitoring, Van Etten Lake and Clark’s Marsh. Public agency reports. www.michigan.gov/pfasresponse

All model parameters and simulated site numbers on this page come from the SWATGenX coupled SWAT+/MODFLOW 6 PFAS models and their manuscripts; agency and literature values are labeled reported and cited above. First distributed watershed-scale PFAS model: Rafiei & Nejadhashemi (2023), Water Research.