Executive Summary

Per- and polyfluoroalkyl substances (PFAS) behave differently from many legacy groundwater contaminants because vadose-zone migration can be strongly influenced by air–water interfacial adsorption, solid-phase sorption, transient water content, and episodic recharge. For many sites, groundwater concentrations are not controlled only by source mass and saturated-zone transport. They are also controlled by processes such as PFAS leaching rate through the unsaturated soils, retardation imposed by air–water interfaces, rate limited sorption/desorption, potential preferential flow pathways, potential precursor transformation, and the timing of wetting and drying cycles.

Failure to represent these processes correctly can result in orders-of-magnitude errors in predicted groundwater concentrations.

PFAS Vadose Zone Conceptual Transport

Governing Unsaturated Flow Equations

PFAS transport in the vadose zone is governed by Richards’ equation describing variably saturated flow.

Richards Equation:

$$
\frac{\partial \theta}{\partial t}
=
\frac{\partial}{\partial z}
\left[
K(h)\left(\frac{\partial h}{\partial z} + 1\right)
\right]
$$

θ = volumetric water content
t = time
K(h) = hydraulic conductivity function
h = pressure head
z = vertical coordinate

The hydraulic functions are typically described using the van Genuchten-Mualem model.

Soil Water Retention Curve:

$$
\theta(h)
=
\theta_r
+
\frac{\theta_s – \theta_r}
{\left(1 + |\alpha h|^{n}\right)^{m}}
\qquad (h < 0)
$$

θr = residual water content
θs = saturated water content
α,n,m = empirical parameters
m = 1 − 1/n

This relationship controls moisture distribution and PFAS mobility.

PFAS Transport Equations

Solid-phase sorption term

Linear equilibrium sorption

$$
S_s = K_d\,C
$$

then:

$$
\rho_b \frac{\partial S_s}{\partial t}
=
\rho_b K_d \frac{\partial C}{\partial t}
$$​

Freundlich sorption

$$
S_s = K_f\,C^{N}
$$

then:

$$
\frac{\partial S_s}{\partial t}
=
K_f N C^{N-1}
\frac{\partial C}{\partial t}
$$

This approach tends to be more realistic for certain PFAS concentration ranges, although Kd remains commonly utilized in screening calculations.

Langmuir Sorption

$$
S_s = \frac{S_{\max} K_L C}{1 + K_L C}
$$

Linearized Langmuir form

$$
\frac{S_s}{C}
=
\frac{K_L S_{\max}}{1 + S_{\max} C}
$$

Langmuir model occurs on finite sorption sites, and become important when PFAS concentrations are elevated near source zones, sites approach saturation, and batch tests show nonlinear sorption behavior. At high concentrations sorption capacity becomes limited, retardation decreases, and PFAS migration may accelerate.

Air–water interfacial adsorption

PFAS in the vadose zone can strongly accumulate at the air–water interface, especially long-chain compounds. This is one of the most important vadose-zone-specific mechanisms.

Equilibrium expression

$$
\Gamma = K_{aw}\,C
$$

$$
\frac{\partial}{\partial t}(A_{aw}\Gamma)
=
\frac{\partial}{\partial t}(A_{aw}K_{aw}C)
$$

$$
A_{aw}(\theta)
=
\frac{\rho_w g}{\sigma_0}
\int_{\theta}^{\theta_s}
h(\theta)\, d\theta
$$

Transport equation becomes:

$$
\left(
\theta
+
\rho_b K_d
+
A_{aw} K_{aw}
\right)
\frac{\partial C}{\partial t}
=
\frac{\partial}{\partial z}
\left(
\theta D \frac{\partial C}{\partial z}
\right)

\frac{\partial}{\partial z}(qC)

\lambda_w \theta C
+
\sum R
$$

Retardation factor:

$$
R_f
=
1
+
\frac{\rho_b K_d}{\theta}
+
\frac{A_{aw} K_{aw}}{\theta}
$$​

Thus:

​$$
R_f \frac{\partial C}{\partial t}
=
\frac{1}{\theta}
\frac{\partial}{\partial z}
\left(
\theta D \frac{\partial C}{\partial z}
\right)

\frac{1}{\theta}
\frac{\partial}{\partial z}(qC)

\lambda_w C
+
\frac{1}{\theta}\sum R
$$

More interface area means more binding soil sites where PFAS can accumulate, increasing retention in the vadose zone. PFAS retention often peaks at moderate soil moisture, not at full saturation.

Rate-limited sorption/desorption

Some PFAS binds to soils quickly, while another fraction becomes trapped in locations where it releases slowly. For PFAS, particularly field-scale legacy source behavior, instantaneous equilibrium might be overly simplistic. A two-site model is often utilized.

$$
S_s = S_1 + S_2
$$

  • S1 = instantaneous sorption domain
  • = kinetic sorption/desorption domain

Even after the original source is removed, PFAS stored in soils may continue to slowly leach into groundwater for many years. This process explains long-term PFAS plume persistence observed at many contaminated sites.

Instantaneous site

$$
S_1 = f K_d\, C
$$

Kinetic site

$$
\frac{\partial S_2}{\partial t}
=
\alpha_s
\left[
(1-f)K_d C – S_2
\right]
$$

  • f = fraction of sorption sites in equilibrium
  • = first-order mass-transfer coefficient [T^

Then the total solid-phase storage term is:

$$
\rho_b \frac{\partial S_s}{\partial t}
=
\rho_b \frac{\partial S_1}{\partial t}
+
\rho_b \frac{\partial S_2}{\partial t}
$$

Transport equation becomes:

$$
\frac{\partial (\theta C)}{\partial t}
+
\rho_b \frac{\partial S_1}{\partial t}
+
\rho_b \frac{\partial S_2}{\partial t}
+
\frac{\partial (A_{aw}\Gamma)}{\partial t}
=
\frac{\partial}{\partial z}
\left(
\theta D \frac{\partial C}{\partial z}
\right)

\frac{\partial}{\partial z}(qC)
+
\sum R
$$

Rate-limited air–water interfacial adsorption

Vadose zone contains extensive air-water interfaces. PFAS retention at these interfaces can significantly slow contaminant migration. Sorption process via AWI may not occur instantaneously; therefore, a kinetic formulation may be utilized as:

$$
\frac{\partial \Gamma}{\partial t}
=
\alpha_{aw}\left(K_{aw}C – \Gamma\right)
$$

$$
\frac{\partial (A_{aw}\Gamma)}{\partial t}
=
A_{aw}\frac{\partial \Gamma}{\partial t}
+
\Gamma \frac{\partial A_{aw}}{\partial t}
$$​

This mechanism is unique to unsaturated soils and may result in large PFAS reservoirs stored above groundwater.

Air–water interfacial area relationship

Aaw changes with saturation and is commonly expressed as a function of water content/saturation levels. A frequently referenced form is

$$
A_{aw}(\theta)
=
\frac{\rho_w g}{\sigma_0}
\int_{\theta}^{\theta_s} h(\theta)\, d\theta
$$

  • = water density
  • = gravitational acceleration
  • σ0 = surface tension of pure water
  • h(θ) = capillary pressure head as a function of water content

This relationship is important because AWI adsorption often becomes strongest at intermediate water saturations, not necessarily at full saturation.

Mechanical dispersion & diffusion

Dispersion coefficient is usually written as:

$$
D = \alpha_L v + D_m \tau
$$

Where:

$$
v = \frac{q}{\theta}
$$

  • = longitudinal dispersivity
  • = molecular diffusion coefficient in water
  • = tortuosity factor

Thus:

$$
D = \alpha_L \frac{q}{\theta} + D_m \tau
$$

Advection predominates during infiltration pulses in many vadose-zone PFAS leaching/transport, while diffusion and mass transfer become more significant during dry periods and redistribution. Greater dispersion can enlarge the vertical footprint of PFAS contamination in the vadose zone and broaden the groundwater plume.

Preferential flow formulation

In structured soils caused by fractured media, desiccation cracks, root channels, worm burrows, and unstable wetting fronts  water does not always move uniformly through soil. Instead, infiltration can create fast pathways that bypass much of the soil matrix. Matrix-only flow can underpredict PFAS migration. A dual-domain approach can be used.

$$
\frac{\partial (\theta_m C_m)}{\partial t}
+
\rho_b \frac{\partial S_m}{\partial t}
=
\frac{\partial}{\partial z}
\left(
\theta_m D_m \frac{\partial C_m}{\partial z}
\right)

\frac{\partial}{\partial z}(q_m C_m)
+
\omega (C_f – C_m)
$$

$$
\frac{\partial (\theta_f C_f)}{\partial t}
+
\rho_b \frac{\partial S_f}{\partial t}
=
\frac{\partial}{\partial z}
\left(
\theta_f D_f \frac{\partial C_f}{\partial z}
\right)

\frac{\partial}{\partial z}(q_f C_f)

\omega (C_f – C_m)
$$

  • = concentration in matrix domain
  • Cf = concentration in preferential-flow domain
  • ω = mass exchange coefficient between mobile and less-mobile domains

Ignoring preferential flow may cause models to underestimate how quickly PFAS reaches the underlying groundwater.

Precursor transformation

If PFAS precursors are present, then daughter PFAS formation should be included. Even if measured PFAS concentrations decline, precursor compounds may continue generating new PFAS mass. Transport and transformation of PFAS precursor compounds is given as:

$$
\frac{\partial (\theta C_p)}{\partial t}
=
\frac{\partial}{\partial z}
\left(
\theta D_p \frac{\partial C_p}{\partial z}
\right)

\frac{\partial}{\partial z}(q C_p)

\lambda_p \theta C_p
$$

and daughter PFAS production enters the PFAS transport equation as a source term:

$$
R_{\text{precursor}} = Y \lambda_p \theta C_p
$$

  • λp = precursor transformation rate
  • = stoichiometric yield to the daughter PFAS

Then the daughter PFAS equation becomes:

$$
\frac{\partial (\theta C_p)}{\partial t}
+
\rho_b \frac{\partial S_p}{\partial t}
+
\frac{\partial (A_{aw}\Gamma_p)}{\partial t}
=
\frac{\partial}{\partial z}
\left(
\theta D_p \frac{\partial C_p}{\partial z}
\right)

\frac{\partial}{\partial z}(q C_p)

\lambda_p \theta C_p
$$

$$
R_{p \to i} = Y_i\,\lambda_{p,i}\,\theta\,C_{p,i}
$$

If multiple precursors exist:

$$
\sum_{j} Y_{ij}\,\lambda_{p,j}\,\theta\,C_{p,j}
$$

  • = precursor transformation rate
  • = stoichiometric yield

Ignoring precursor chemistry can significantly underestimate future PFAS groundwater impacts.

PFAS precursor transformation can be represented at different levels of complexity depending on the modeling objective. Precursor transformation equations are useful when presented as lumped reactive-transport approximations adapted to PFAS systems, rather than as universally established PFAS reaction laws. This equation is not a universal or mechanistic depiction of PFAS chemistry because precursor degradation pathways may consist of many intermediate compounds, branching reaction networks, and site-specific environmental constraints. Instead, It is better seen as a simplified modeling framework developed from contaminant transport theory to approximate the net conversion of precursor compounds into terminal PFAS within environmental systems.

Integrated Vadose-Zone PFAS Transport Equation

A practical “advanced” 1-D PFAS vadose-zone equation combining the major mechanisms is:

$$
\frac{\partial (\theta C)}{\partial t}
+
\rho_b \frac{\partial (S_1 + S_2)}{\partial t}
+
\frac{\partial (A_{aw}\Gamma)}{\partial t}
=
\frac{\partial}{\partial z}
\left(
\theta D \frac{\partial C}{\partial z}
\right)

\frac{\partial}{\partial z}(qC)
+
Y \lambda_p \theta C_p

\lambda_w \theta C
$$

Two-site sorption + precursor transformation:

$$
\frac{\partial (\theta C)}{\partial t}
+
\rho_b \frac{\partial (S_1 + S_2)}{\partial t}
+
\frac{\partial (A_{aw}\Gamma)}{\partial t}
=
\frac{\partial}{\partial z}
\left(
\theta D \frac{\partial C}{\partial z}
\right)

\frac{\partial}{\partial z}(qC)
+
Y \lambda_p \theta C_p

\lambda_w \theta C
$$

Lumped sorption + decay of all phases:

$$
\frac{\partial (\theta C)}{\partial t}
+
\rho_b \frac{\partial S_s}{\partial t}
+
\frac{\partial (A_{aw}\Gamma)}{\partial t}
=
\frac{\partial}{\partial z}
\left(
\theta D \frac{\partial C}{\partial z}
\right)

\frac{\partial}{\partial z}(qC)

\lambda_w \theta C

\lambda_s \rho_b S_s

\lambda_{aw} A_{aw} \Gamma
+
\sum R_{\text{sources/sinks}}
$$

Variably saturated porous medium with mobile and immobile / preferential domains, a general 1-D form can be written as:

$$
\frac{\partial (\theta_m C_{i,m})}{\partial t}
+
\rho_b \frac{\partial (S_{i,1} + S_{i,2})}{\partial t}
+
\frac{\partial (A_{aw,m}\Gamma_{i,m})}{\partial t}
=
\frac{\partial}{\partial z}
\left(
\theta_m D_{i,m} \frac{\partial C_{i,m}}{\partial z}
\right)

\frac{\partial}{\partial z}(q_m C_{i,m})
+
\omega_i (C_{i,f} – C_{i,m})
+
Y_i \lambda_{p,i} \theta_m C_{p,i}

\lambda_{w,i} \theta_m C_{i,m}
$$

If PFAS transport through macropores, fingers, fractures, or unstable infiltration pathways is important, then a separate mobile-domain equation is added:

$$
\frac{\partial (\theta_f C_{i,f})}{\partial t}
+
\rho_b \frac{\partial (S_{i,f,1} + S_{i,f,2})}{\partial t}
+
\frac{\partial (A_{aw,f}\Gamma_{i,f})}{\partial t}
=
\frac{\partial}{\partial z}
\left(
\theta_f D_{i,f} \frac{\partial C_{i,f}}{\partial z}
\right)

\frac{\partial}{\partial z}(q_f C_{i,f})

\omega_i (C_{i,f} – C_{i,m})
+
Y_i \lambda_{p,i} \theta_f C_{p,i,f}

\lambda_{w,i} \theta_f C_{i,f}
$$

Where:

  • C = aqueous PFAS concentration
  • θ = volumetric water content
  • t=time
  • ρb = bulk density of porous media
  • Ss= sorbed concentration on solids
  • = maximum sorption capacity of the soil (monolayer saturation)
  • = Langmuir adsorption constant (sorption affinity coefficient)
  • Slope = 1/Smax1
  • Intercept = 1/(KLSmax)
  • Γ = PFAS mass adsorbed at the air–water interface per unit interfacial area
  • D = hydrodynamic dispersion coefficient
  • q = Darcy flux
  • λw,λs,λaw= first-order decay/transformation constants
  • Rsources/sinks = source / sink terms, including precursor transformation, recharge input, etc.
  • Ci,m = Aqueous concentration of PFAS species i in the matrix (soil pore-water) domain
  • Ci,f = PFAS concentration in the preferential-flow domain
  • Cp,i​ = Concentration of precursor compound that can transform into PFAS species i
  • θm = Volumetric water content of the matrix domain
  • Rf = retardation factor

Calibration Framework

Objective Function

Calibration adjusts model parameters so simulated results match observed site data as closely as possible.

$$
\Phi(p) = \sum_{i=1}^{n} w_i \left[ y_i^{\mathrm{obs}} – y_i(p) \right]^2
$$

  • Φ(p) = objective function to be minimized
  • yi^{obs} = observed value
  • = simulated value for parameter set p
  • = observation weight

common weighting

$$
w_i = \frac{1}{\sigma_i^{2}}
$$

where σi2 is the observation-error variance.

PFAS calibration should include more than one target – visually good fit to one dataset can still miss the real retention and release processes controlling long-term groundwater loading.

Regularized calibration

When many parameters are poorly identifiable, regularization is often added:

$$
\Phi_{\text{total}} = \Phi_{\text{obs}} + \mu \Phi_{\text{reg}}
$$

Φreg=j=1m(σpjpjpj,prior)2

  • = estimated parameter
  • = prior or preferred value
  • = allowed deviation from prior
  • = regularization weight

Regularization keeps the model from drifting into unrealistic parameter combinations just to fit the data. In PFAS vadose-zone, this is especially useful.

common inverse-model:

$$
(J^{T} W J + \lambda I)\,\Delta p = J^{T} W r
$$

  • = Jacobian or sensitivity matrix
  • = weighting matrix
  • = residual vector
  • = damping factor
  • = parameter adjustment

Selected calibration metrics

$$
\mathrm{PBIAS}
=
100\,
\frac{
\sum_{i=1}^{n}\left(y_i – y_i^{\mathrm{obs}}\right)
}{
\sum_{i=1}^{n} y_i^{\mathrm{obs}}
}
$$

$$
\mathrm{MAE} = \frac{1}{n}\sum_{i=1}^{n}\left|y_i^{\mathrm{obs}} – y_i\right|
$$

$$
\mathrm{NSE}
=
1

\frac{
\sum_{i=1}^{n} \left( y_i^{\mathrm{obs}} – y_i \right)^2
}{
\sum_{i=1}^{n} \left( y_i^{\mathrm{obs}} – \bar{y}^{\mathrm{obs}} \right)^2
}
$$

$$
\mathrm{RMSE} = \sqrt{\frac{1}{n}\sum_{i=1}^{n}\left(y_i^{\mathrm{obs}} – y_i\right)^2}
$$

$$
\mathrm{NRMSE}
=
\frac{\mathrm{RMSE}}{y_{\max}^{\mathrm{obs}} – y_{\min}^{\mathrm{obs}}}
$$

Performance metrics are essential, however, they should not serve as the sole foundation for defensibility. PFAS model may demonstrate a satisfactory statistical fit, yet fail to accurately represent actual long-term PFAS behavior if it excludes factors such as AWI adsorption, preferential flow, slow desorption, or precursor transformation.

Validation

For PFAS vadose-zone modeling, the strongest validation is usually process-based, not just statistical. A robust model should reproduce not only concentrations, but also the characteristic behaviors expected for PFAS.  Model validation is best viewed as a structured test of whether the model can reproduce independent hydraulic and PFAS observations in a way that is consistent with the site conceptual model.

Sensitivity analysis framework

Sensitivity analysis identifies which parameters most strongly control model predictions.

Local coefficient

$$
S_j^{\text{local}} = \frac{\partial Y}{\partial p_j}
$$

Normalized form:

$$
S_j^{*}
=
\frac{p_j}{Y}\,
\frac{\partial Y}{\partial p_j}
$$

Finite-difference approximation:

$$
\frac{\partial Y}{\partial p_j}
\approx
\frac{Y(p_j + \Delta p_j) – Y(p_j – \Delta p_j)}{2\,\Delta p_j}
$$

Morris screening method

$$
EE_i(X)
=
\frac{
Y(X_1,\ldots,X_i+\Delta,\ldots,X_k) – Y(X)
}{\Delta}
$$​

$$
\mu_i^{*}
=
\frac{1}{r}
\sum_{j=1}^{r}
\left| EE_{i,j} \right|
$$

$$
\sigma_i
=
\sqrt{
\frac{1}{r-1}
\sum_{j=1}^{r}
\left( EE_{i,j} – \mu_i \right)^2
}
$$

Sobol global

$$
\operatorname{Var}(Y)
=
\sum_{i} V_i
+
\sum_{i<j} V_{ij}
+
\cdots
$$

$$
S_i
=
\frac{V_i}{\operatorname{Var}(Y)}
$$

$$
S_{T_i}
=
1

\frac{
\operatorname{Var}_{X_{\sim i}}
\left(
\mathbb{E}_{X_i}\left[\,Y \mid X_{\sim i}\,\right]
\right)
}{
\operatorname{Var}(Y)
}
$$

Sensitivity analysis extends beyond theoretical applications. This addresses a key inquiry concerning which field measurements will most effectively reduce prediction uncertainty.

Uncertainty analysis framework

Uncertainty analysis estimates the range of plausible future outcomes after calibration.

Parameter covariance

$$
C_p = s^2 (J^{T} W J)^{-1}
$$

$$
s^2 = \frac{\Phi(p^*)}{n – m}
$$

  • = parameter covariance matrix
  • = residual variance estimate
  • = number of observations
  • = number of estimated parameters

Parameter standard error:

$$
SE(p_j) = \sqrt{C_{p,jj}}
$$

First-order prediction

$$
Z = g(p)
$$

$$
\operatorname{Var}(Z) \approx g_p^{T} C_p g_p
$$

$$
\sigma_Z = \sqrt{\operatorname{Var}(Z)}
$$

Monte Carlo propagation

$$
p^{(k)} \sim f_p(p), \quad k = 1,\ldots,N
$$

$$
Y^{(k)} = f\!\left(p^{(k)}\right)
$$

$$
\bar{Y} = \frac{1}{N}\sum_{k=1}^{N} Y^{(k)}
$$

$$
\operatorname{Var}(Y)
\approx
\frac{1}{N-1}
\sum_{k=1}^{N}
\left( Y^{(k)} – \bar{Y} \right)^2
$$

$$
PI_{95\%} = \left[ Q_{2.5}(Y),\, Q_{97.5}(Y) \right]
$$

Null-space Monte Carlo

$$
p^{(k)} = p^{*} + V_n\, a^{(k)}
$$

Relevant for complex PFAS models where multiple parameter combinations can fit the same observations.

Bayesian- likelihood weighting

$$
L(p) \propto \exp\!\left(-\tfrac{1}{2}\,\Phi(p)\right)
$$

$$
f(p \mid y^{\mathrm{obs}}) \propto L(p)\,f(p)
$$

Total uncertainty prediction

$$
\operatorname{Var}_{\text{total}}(Z)
=
\operatorname{Var}_{\text{param}}(Z)
+
\operatorname{Var}_{\text{obs}}(Z)
+
\operatorname{Var}_{\text{scenario}}(Z)
+
\operatorname{Var}_{\text{structure}}(Z)
$$

The most significant takeaway is that model uncertainty is more than just parameter ranges. It also incorporates measurement error, future scenario uncertainty, and structural uncertainty due to model simplifications. For PFAS, structural uncertainty can be severe if the model omits AWI adsorption, preferential flow, kinetic desorption, or precursor transformation.

conclusions

PFAS leaching in the vadose zone is influenced by transient flow, adsorption to solids, adsorption at air-water interfaces, and contingent upon site conditions, kinetic retention, preferential flow, and precursor transformation. Models that haphazardly ignore these processes only serve as back-of-the-envelope screening and can substantially misestimate groundwater impacts.

References

Chen, S., & Guo, B. (2025). Semi-analytical solutions for nonequilibrium transport and transformation of PFAS and other solutes in heterogeneous vadose zones with structured porous media. Advances in Water Resources, 206, 105099. https://doi.org/10.1016/j.advwatres.2025.105099

Doherty, J., & Hunt, R. J. (2010). Approaches to highly parameterized inversion: A guide to using PEST for groundwater-model calibration. U.S. Geological Survey Scientific Investigations Report 2010–5169.

Interstate Technology & Regulatory Council. (2023). PFAS guidance document. Washington, DC: Interstate Technology & Regulatory Council. https://pfas-1.itrcweb.org/

Morris, M. D. (1991). Factorial sampling plans for preliminary computational experiments. Technometrics, 33(2), 161–174.

Sobol’, I. M. (2001). Global sensitivity indices for nonlinear mathematical models and their Monte Carlo estimates. Mathematics and Computers in Simulation, 55(1–3), 271–280.

Yan, P.-F., Dong, S., Woodcock, M. J., Manz, K. E., Garza-Rubalcava, U., Abriola, L. M., Pennell, K. D., & Cápiro, N. L. (2024). Biotransformation of 6:2 fluorotelomer sulfonate and microbial community dynamics in water-saturated one-dimensional flow-through columns. Water Research, 252, 121146. https://doi.org/10.1016/j.watres.2024.121146

Zhang, W., Pang, S., Lin, Z., Mishra, S., Bhatt, P., & Chen, S. (2021). Biotransformation of perfluoroalkyl acid precursors from various environmental systems: Advances and perspectives. Environmental Pollution, 272, 115908. https://doi.org/10.1016/j.envpol.2020.115908

PFAS Modeling FAQs

Explore PFAS vadose zone modeling FAQs, including PFAS leaching to groundwater, air-water interfacial adsorption, precursor transformation, calibration, sensitivity, uncertainty analysis.

PFAS vadose zone fate and transport modeling evaluates how PFAS moves through unsaturated soil above the water table and whether it will remain stored in soil or leach downward to groundwater over time.

PFAS transport is different because many PFAS compounds strongly adsorb at the air-water interface in unsaturated soils, which can significantly retard transport and create long-term contaminant storage in the vadose zone.

A defensible PFAS vadose zone model may include variably saturated flow, advection, dispersion, solid-phase adsorption, air-water interfacial adsorption, rate-limited desorption, precursor transformation, and preferential flow where site conditions justify those processes.

Preferential flow should be considered when a site has fractured media, structured soils, perched conditions, rapid infiltration behavior, or field evidence of early breakthrough that matrix flow alone cannot explain.

Rate-limited desorption is important because it helps explain why PFAS can continue to leach long after the main source is gone and why contamination may decline more slowly than predicted by simple equilibrium models.

PFAS precursor transformation is the conversion of precursor compounds into terminal PFAS in the subsurface. It can sustain or increase long-term PFAS loading to groundwater even when measured terminal PFAS concentrations appear modest at the start of an investigation. Precursor transformation modeling is most appropriate when site history, chemistry, or analytical evidence indicates that precursors are present and may materially affect future PFAS loading.

PFAS precursor transformation is usually most defensibly represented as a lumped reactive-transport approximation of net precursor loss and daughter-PFAS generation rather than a universal mechanistic reaction. This captures the net effect of precursor degradation and daughter-PFAS formation without resolving every intermediate compound or pathway.

A weighted least-squares objective function is a mathematical way of quantifying mismatch between observed and simulated values while giving more influence to more reliable or relevant data. While, regularization is a calibration approach that helps keep parameter estimates within realistic ranges when the model contains poorly constrained parameters such as air-water interfacial terms or kinetic mass-transfer coefficients.

Sensitivity analysis evaluates which model parameters most influence predictions such as time to groundwater arrival, peak concentration, retained vadose zone mass, or long-term mass discharge. It helps identify which assumptions matter most and which field measurements would most improve model confidence. Uncertainty analysis estimates the plausible range of model outcomes after considering uncertainty in parameters, observations, assumptions, future scenarios, and model structure. It is important because a single best-fit model can hide major uncertainty in long-term PFAS leaching, plume persistence, and groundwater impact.

Red flags include, but are not limited to, neglecting air-water interfacial adsorption, using generic PFAS parameters without justification, relying on one-dimensional assumptions for clearly heterogeneous sites, disregarding sensitivity and uncertainty analysis, and overstating precursor transformation certainty.

Common limitations include uncertain interfacial area estimates, limited precursor data, poorly constrained kinetic parameters, transferability of lab parameters to field scale, and incomplete representation of heterogeneity or preferential flow.

Vadose zone modeling can estimate whether continued leaching from unsaturated soils is likely to sustain groundwater concentrations even if the saturated plume appears stable or partially controlled. PFAS vadose zone modeling can help estimate long-term source loading, evaluate source-zone remedies, compare remedial scenarios, assess whether monitored retention is plausible, and improve long-term groundwater protection strategies.

Screening-level models use simplified assumptions and fewer parameters, while advanced models may include transient unsaturated flow, interfacial adsorption, kinetic desorption, precursor transformation, and preferential flow. Advanced modeling is often warranted for large or complex sites, litigation support, remedial design, high-stakes risk decisions, heterogeneous geology, precursor-rich sources, or situations where long-term groundwater loading is a key concern.

AA GeoEnvironmental provides independent technical review of PFAS conceptual site models, governing equation selection, parameter defensibility, calibration strategy, sensitivity analysis, uncertainty evaluation, precursor transformation interpretation, and long-term leaching to groundwater.

Request PFAS Modeling Review