Predictions of PFAS regional-scale atmospheric deposition and ambient air exposure

Abstract

Per- and polyfluoroalkyl substances (PFAS) are a large class of human-made compounds that have contaminated the global environment. One environmental entry point for PFAS is via atmospheric emission. Air releases can impact human health through multiple routes, including direct inhalation and contamination of drinking water following air deposition. In this work, we convert the reference dose (RfD) underlying the United States Environmental Protection Agency’s GenX drinking water Health Advisory to an inhalation screening level and compare to predicted PFAS and GenX air concentrations from a fluorochemical manufacturing facility in Eastern North Carolina. We find that the area around the facility experiences ~15 days per year of GenX concentrations above the inhalation screening level we derive. We investigate the sensitivity of model predictions to assumptions regarding model spatial resolution, emissions temporal profiles, and knowledge of air emission chemical composition. Decreasing the chemical specificity of PFAS emissions has the largest impact on deposition predictions with domain-wide total deposition varying by as much as 250% for total PFAS. However, predicted domain-wide mean and median air concentrations varied by less than 18% over all scenarios tested for total PFAS. Other model features like emission temporal variability and model spatial resolution had weaker impacts on predicted PFAS deposition.

Graphical Abstract

1. Introduction

Per- and polyfluoroalkyl substances (PFAS) are a class of human-made compounds that have been used in a variety of industrial and consumer applications for decades¹. Their prevalence and persistence of use, combined with the strength of the carbon-fluorine bond which resists degradation, has led to the detection of PFAS throughout the environment^2–6. Much of the scientific and regulatory focus of PFAS has been in water, specifically drinking water^7–12. In the United States, the Environmental Protection Agency (EPA) has established drinking water Health Advisories since 2009 for PFOA and PFOS¹³ (2009: Provisional Health Advisories, 2016: Lifetime Health Advisories, 2022: Interim Updated Health Advisories), and has recently established Health Advisories for two additional PFAS (HFPO-DA, commonly known as GenX Chemicals, and PFBS)¹². Health Advisories are used to inform the public of potential negative health effects of a contaminant and are non-regulatory and non-enforceable.

Alongside the focus on drinking water, there is growing evidence that air emissions of PFAS are an important environmental contamination route with potential human health impacts. It has been predicted that inhalation is the largest exposure route for occupationally exposed individuals¹⁴. This suggests the need to more broadly understand the inhalation exposure route, with recent work in this area^15–17. Additionally, it has been shown that dermal uptake of gas-phase PFAS can be an exposure route¹⁸, and deposition of atmospheric PFAS can lead to water contamination¹⁹.

Despite the growing evidence of PFAS air emissions on environmental contamination and human health, there is a paucity of work comprehensively examining the emission, transport, and fate of total PFAS in the atmosphere. Existing work is limited to modeling and measuring specific sub-classes of PFAS, for example perfluorocarboxylic and sulfonic acids²⁰, or perfluorocarboxylic acids and their precursors²¹, which are not complete representations of PFAS emissions from most sources. For example, manufacturers may emit a variety of PFAS from combinations of distinct production processes each resulting in a unique mixture of PFAS emissions²², and aqueous film forming foam contains numerous different PFAS²³. Previous work has shown that, while one compound (GenX) may be of particular interest²², excluding the remainder of PFAS emissions would underestimate deposition by a factor of 200²⁴.

In addition to chemical complexity, the role of model spatial resolution and temporal emission rates on modeled PFAS atmospheric fates has not been studied, although previous studies vary in spatial scale. Global- and regional-scale studies have been used to estimate the total atmospheric burden of subsets of PFAS^{25, 26}, and to explain their global spread^{27, 28}. Near-field air dispersion modeling has been performed for major PFAS manufacturing sites²⁹, demonstrating that episodic, elevated concentrations near sources are predicted using highly temporally-resolved emissions and that there is potential for soil and water contamination via atmospheric transport and deposition^30–32.

In this study, we compare ambient air concentrations predicted by the Community Multiscale Air Quality (CMAQ) model to the inhalation screening level derived from the RfD underlying the U.S. EPA drinking water Health Advisory. We quantify the sensitivity of our previous modeling results to varying model configuration and emissions inputs. Our previous work used highly chemically and temporally specific emissions reported by the manufacturer, and CMAQ was run with fine-scale horizontal resolution (1 km), giving us high confidence in our predicted air concentrations and deposition rates. In the near-term, PFAS modeling studies will likely not have access to a complete accounting of PFAS air emissions magnitudes and timing from key sources, as our previous work did. Thus, we explore simplifying the chemical speciation to investigate the impact of treating PFAS as a class versus modeling individual compounds. We also test the impact of simplifying the temporal allocation of emissions and coarsening the model spatial resolution to understand potential biases for large-scale comprehensive PFAS modeling applications. We quantify the impact all three of these tests have on the predicted PFAS air concentrations and deposition fluxes near the source and downwind. We then discuss how the insights gained from these tests inform priorities for developing PFAS emission inventories and atmospheric models.

2. Methods

2.1. Inhalation Screening Level

In October 2021, the US EPA published a chronic RfD for HFPO-DA, commonly known as GenX chemicals, of 3 ng per kg body weight per day³³. Monnot et al.¹⁵ developed a route-to-route extrapolation method to derive an inhalation screening level from the reference dose (RfD). The method involves multiplying the RfD by body weight (kg) and dividing by the physiological daily inhalation rates (m³ day⁻¹), both from the EPA’s Exposure Factors Handbook³⁴, resulting in an inhalation equivalent concentration as a mass per volume of air. We calculate the inhalation screening level (ISL) individually for males and females aged 2.6 months to 96 years, with average body weights and 95^th percentile physiological daily inhalation rates for each age group. The population-average weight and inhalation rate of all age groups and sexes was then taken to derive the aggregated ISL. See the Supporting Information for more information on the parameters used for this calculation.

2.2. CMAQ-PFAS

The Community Multiscale Air Quality (CMAQ) model is the EPA’s flagship air quality model that solves conservation of mass equations for a given airshed. It is used to support implementation of the National Ambient Air Quality Standards (NAAQS) under the Clean Air Act³⁵, for estimating exposure to hazardous air pollutants as part of the EPA Air Toxics Screening Assessment (AirToxScreen)^{36, 37}, and to estimate criteria air pollutant formation as part of the National Air Quality Forecast Capability³⁸, among many other research and regulatory applications. CMAQ has been publicly available³⁹ for 25 years and is used across the globe by academics, governments, and the private sector⁴⁰. CMAQ v5.3.2 was evaluated with performance consistent with state-of-the-science models for criteria air pollutants⁴¹ and has been expanded to include the transport, multiphase partitioning, and deposition of explicit and lumped PFAS species²⁴. The model is run for the entirety of 2018, to coincide with deposition measurements as discussed in D’Ambro et al.²⁴. The domain is 260 × 244 km at 1 km horizontal resolution, covering much of Eastern NC and Northeastern SC, USA, centered on the Chemours fluorochemical manufacturing facility in Fayetteville, NC. Annual emissions of 53 PFAS were obtained from the facility and totaled 109,393 kg in 2017. Motivated by deposition measurements made by NC Department of Environmental Quality (NC DEQ) near the facility²², we focus on GenX, a family of compounds which we define in our model runs as the sum of hexafluoropropylene oxide dimer acid (HFPO-DA) and hexafluoropropylene oxide dimer acyl fluoride (HFPO-DAF). The Base model configuration (Table 1), discussed in D’Ambro et al.²⁴ included 26 explicit PFAS compounds, accounting for 99.8% of the emissions by mass. Physicochemical properties were obtained from the Open structure–activity/property Relationship App (OPERA) model⁴² which has a demonstrated ability to predict PFAS physicochemical properties⁴³. The remaining compounds representing 0.2% of the emitted mass were lumped into a representative specie (“PFASOTHER”) with mole-weighted properties. The facility shared detailed operations information including tables associating compounds with process units and records documenting which processes ran each day of 2018, allowing us to build a daily-resolved timeseries of speciated emissions. This model configuration discussed in our previous study²⁴ and presented here as the “Base” scenario, has extremely high-quality inputs and thus should give us the most accurate results for air concentrations and deposition. It is unlikely that future studies will have such explicit and reliable emissions information, thus we test the impact on our results of lower quality inputs. These simulations only consider PFAS emitted from the Chemours facility; no background PFAS from other sources are included.

Table 1.

Description of the eight model runs discussed in the main text.

		Species Lumping			Temporal Emissions Allocation		Horizontal Resolution
Scenario Name	Base	Single Species Particle	Single Species Gas	Single Species Gas/Particle	Uncontrolled Base	Uniform	4 km	12 km
Chemical speciation	Explicit	aerosol-phase specie	gas-phase specie	Gas specie that can partition	Explicit	Explicit	Explicit	Explicit
Temporal Emissions Allocation	Daily, with controls	Daily, with controls	Daily, with controls	Daily, with controls	Daily, no controls	Annually uniform, no controls	Daily, with controls	Daily, with controls
Horizontal resolution	1 km	1 km	1 km	1 km	1 km	1 km	4 km	12 km

2.2.1. Chemical Specificity

In the Base case, 26 of the 53 reported compounds are represented explicitly, with the remaining lumped into a surrogate species. Throughout, we refer to “total PFAS” which we define as the totality of PFAS compounds reported from this manufacturer. We do not mean to imply that this is the totality of PFAS that exist in the atmosphere within our domain, as we are likely missing sources. To quantify the impact of chemical specificity, we perform tests in which total PFAS emissions are lumped into one of three surrogate model compounds: one that is emitted as a gas and allowed to freely partition to the aerosol phase based on mole-weighted average volatility and solubility (Single Species Gas/Particle, SSGP), one that is emitted as a gas and confined to the gas phase (Single Species Gas, SSG), and one that is emitted and confined to the aerosol phase (Single Species Particle, SSP) (Tables 1 and S1).

2.2.2. Emission Rate Temporal Variability

The Base case utilized a process-based inventory of 53 compounds (26 explicit, remainder lumped) allocated daily according to operation records and accounting for incremental emissions controls for specific compounds implemented throughout the year. To test the impact of simplified input data, we ran a simulation where the emissions were distributed in time just like the Base case, but without including compound-specific controls (“Uncontrolled Base”) (Table 1). The controlled compounds account for 5.4 % of the emitted mass and include: HFPO-DA (CAS: 13252-13-6), HFPO-DAF (CAS: 2062-98-8), PAF (CAS: 354-34-7), PMPF (CAS: 2927-83-5), TAF (CAS: 3299-24-9), COF2 (CAS: 353-50-4), PEPF (CAS: 1682-78-6), DA (CAS: 4089-58-1), HFPO-TA (CAS: 2641-34-1), MA (CAS: 4089-57-0), and PFASOTHER (lumped model species). We then ran another simulation where the annual total emissions from the Uncontrolled Base were uniformly distributed across the days of the year (“Uniform”). In all eight cases (Table 1), the emissions are allocated uniformly across each day as the plant operates 24 hours a day. Comparing the “Uniform” and “Uncontrolled Base” simulations isolates the impact of meteorological and emissions co-variability on deposition fluxes. For example, if emissions are focused on a limited number of production events throughout the year, then the relative importance of meteorological events like rain coincident with those emissions increases. The Base simulation provides an estimate of the impact of implementing realistic controls targeting specific compounds and how that compares to variability introduced by production schedules and meteorology.

2.2.3. Spatial Resolution

The Base model configuration is 1 km horizontal resolution, which is well-suited for capturing steep gradients in concentration and deposition near the facility²⁴. To test the impact of spatial resolution on our results, we ran the same model configuration at coarser horizontal resolutions of 4 km, typically used for regional studies, and 12 km, typically used for the contiguous US (Table 1). Land-use and 2018 non-PFAS anthropogenic emissions inputs for each grid were generated independently using corresponding spatial surrogates. Chemical and meteorological boundary condition inputs for the different resolutions of the same North Carolina domain were generated by successive nesting from a 12 km contiguous US domain. The results from the contiguous 12 km domain were subset to the 12 km NC domain, and were also used to generate boundary conditions for a nested 4 km Eastern US domain. The results from this 4 km simulation were then subset to the NC domain and also used to inform boundary conditions for the 1 km domain.

3. Results and Discussion

3.1. Air Concentrations & Inhalation Exposure

Domain-median and -mean predicted air concentrations vary by less than 18% for total PFAS among the eight scenarios. The model assumes PFAS compounds are emitted in the gas phase, and their physicochemical properties drive the bulk of the mass to remain in the gas phase, meaning that even relatively large differences in deposition do not perturb the air concentration substantially (Fig. S1A). GenX concentrations vary by up to 91% across the Base, Uncontrolled Base, Uniform, and horizontal resolution scenarios. For both total PFAS and GenX, decreasing the model’s horizontal resolution decreases the peak concentrations near the facility (Fig. S2A&B). The total PFAS annual average domain-wide peak concentration decreases from 3.7 μg m⁻³, to 0.997 μg m⁻³, to 0.3 μg m⁻³ from 1 km, to 4 km, to 12 km, respectively. For GenX, the peak concentration decreases from 0.025 μg m⁻³, to 0.0065 μg m⁻³, to 0.0022 μg m⁻³, from the 1 km scenario to the 4 km and 12 km, respectively (Table S8, Fig. S2A&B).

Using the route-to-route extrapolation and population-level parameters described in the supporting information, we derive an inhalation screening level of 7.5 ng m⁻³ from the 2021 chronic RfD of 3 ng kgbw⁻¹ day⁻¹ (Table 2). For comparison, the other three PFAS with drinking water Health Advisories set by the EPA in 2022 are also shown in Table 2. Our method for deriving an inhalation screening level (ISL) is illustrative and should not be interpreted as an authoritative value below which individuals are guaranteed to be safe from health effects. Rather, this is an attempt to categorize the predicted ambient air concentrations relative to potential health impacts. We then compared our GenX model-predicted air concentrations from every hour of the year in the Base scenario at the facility and the two largest urban areas in the domain, Fayetteville (~25 km north) and Wilmington (~125 km SE), to this inhalation screening level. The maximum annual model-predicted GenX air concentrations in the Base scenario are 89 ng m⁻³, 24 ng m⁻³, and 1.6 ng m⁻³, at the facility, Fayetteville, and Wilmington respectively. The modeled air concentrations were discussed in detail previously²⁴ and agree within reason with GenX particle phase measurements⁴⁴. Finally, we compare the ISL to the GenX particulate matter measurement detection limit (“PM MDL”, 1.7×10⁻⁵ ng m⁻³) from Zhou et al.⁴⁴ (Fig. 1A). We are not aware of published gas-phase MDL’s for GenX or other PFAS, so in the absence of gas-phase MDL’s and acknowledging that GenX is predicted to be >50% in the particle phase in some areas of our domain²⁴, the PM MDL is a useful metric to conceptualize our model predictions.

Table 2.

Drinking water Health Advisories, reference doses, and calculated inhalation screening levels for four PFAS.

Compound	Drinking Water Health Advisory (ng L−1)	RfD (ng kg−1)	ISL (ng m−3)	ISL Reference
PFOA	0.004	1.5×10−3	5.3×10−3	Monnot et al., 2022
PFOS	0.02	7.9×10−3	2.8×10−2	Monnot et al., 2022
GenX	10	3	7.5	This work
PFBS	2,000	300	750	This work

Figure 1.

Fraction of the year that the grid cells containing the facility, Fayetteville, and Wilmington, are below the PM PFAS measurement detection limit (gray), above the GenX ISL of 7.5 ng m⁻³ (yellow), and between these two values (green). The GenX PM MDL is 0.017 pg m⁻³, and the total PFAS MDL is 0.005 pg m⁻³. Due to rounding, some percentages do not add to 100.

Figure 1A shows that for a cumulative ~15 days of the year, the area immediately surrounding the facility experiences GenX air concentrations above the ISL, while Fayetteville and Wilmington never experiences air concentrations above the ISL. In comparison, when emissions are uncontrolled, i.e. historical conditions, the area around the facility experiences 25.5 days of the year above the ISL, while Fayetteville and Wilmington are unchanged (Fig. S3). For most of the year at all locations and both the Uncontrolled and Base scenarios, the predicted air concentration is below the measurement detection level and the average yearly exposure is below the ISL. We can further compare our results to formaldehyde, a known hazardous air pollutant⁴⁵ predicted to account for 9.2% of noncancer risk³⁷ with a reference concentration (RfC) of 9,800 ng m^{−3 46}. Dividing the annual average concentration of GenX at the facility by the ISL, and then comparing this to the annual average formaldehyde concentration divided by its RfC, we derive an inhalation screening level weighted exposure for GenX relative to formaldehyde of 0.76:1 at the facility. This result reflects the fact that that while the GenX ISL (7.5 ng m⁻³) is much lower than the formaldehyde RfC (9,800 ng m⁻³), GenX is much less abundant. See the SI for further details and discussion.

In the absence of toxicological information for each PFAS compound we modeled, we compare ambient concentrations of total PFAS to the ISL for GenX, and the median MDL for all PFAS with analytical detection limits from Table S2 of Zhou et al.⁴⁴ (median of 34 compounds = 0.005 pg m⁻³) (Fig. 1B). The Facility experiences ~3 months above the GenX ISL for total PFAS emitted, Fayetteville experiences ~1 month, and Wilmington experiences 11 days (Fig. 1B). This comparison is for illustrative purposes only but is not necessarily an upper bound estimate as future studies will improve our knowledge of the mechanisms and extent of PFAS toxicity. Additionally, recent ambient measurements in North Carolina have recorded PFOA air concentrations (14 pg m^{−3 44}) above the ISL (5.3 pg m^{−3 15}), despite PFOA being phased out in the US ~20 years ago. And finally, our model only contains one source of PFAS within the domain, likely resulting in an underestimate of ambient PFAS air concentrations. For example, a recent study has predicted numerous PFAS drinking water contamination sources within our domain⁴⁷, some of which (industrial sites²⁴, waste water treatment plants⁴⁸) have corresponding air emissions.

3.2. Deposition pathway constraints

PFAS deposition is of particular concern because it has been shown to lead to drinking water contamination¹⁹. Predicted deposition rates (Fig. S1B) of total PFAS are dramatically altered by simplifying the chemical speciation of emissions. PFAS emitted from this facility are, in general, volatile and insoluble. Eighty percent of the emissions are from just three compounds (TFE, HFPO, and HFP) that are highly insoluble (Henry’s law constants of 8.6×10⁻³ – 6.5 M atm⁻¹) and volatile (c* ~10¹⁰ μg m⁻³)²⁴. This results in a mole-weighted Henry’s law constant that is largely insoluble (3.3×10⁻² M atm⁻¹, Table S7) and lower than two of the three compounds, resulting in decreased deposition from these two compounds (TFE and HFP). This, combined with decreased deposition from the remaining compounds with high solubility yet lower emission rate, results in decreased dry deposition for all three lumped scenarios (SSG, SSP, and SSGP). Wet deposition also decreases strongly and plays little role in the removal of the majority of total PFAS from the atmosphere in the SSG and SSGP cases. In the SSP case, wet deposition processes via in-cloud particle activation and below-cloud scavenging are stronger than the dry deposition route, resulting in a greater than doubling of the domain-wide deposition rates compared to the Base case: 5,200 kg yr⁻¹ Base versus 13,000 kg yr⁻¹ SSP.

Next, we simplified the temporal allocation of emissions to investigate the impact of imperfect emissions inventory data. Figure 2A shows the monthly-averaged domain-wide daily deposition of the sum of total PFAS by month. The Uncontrolled Base results are nearly identical to those of the Base for total PFAS because only a minority of the total PFAS mass was targeted for control (Fig. S4A). A sharp decrease in deposition in October reflects a temporary plant closure. In the Uniform case, the trends are smoothed and the decrease in deposition is not as strong in October.

Figure 2.

The impact of temporal allocations on domain-wide cumulative deposition. A) Cumulative domain-wide deposition of total PFAS by month for the different emissions allocation scenarios. B) Cumulative domain-wide deposition of GenX (HFPO-DA + HFPO-DAF) by month.

While there is not a significant impact on domain-wide deposition of total PFAS between the three emission scenarios (Base, Uncontrolled Base, and Uniform), a larger range is seen for GenX (Fig. 2B). GenX includes HFPO-DA, a soluble carboxylic acid, and HFPO-DAF, a sparingly soluble acyl fluoride. Both were targeted by emission controls throughout 2018 (Fig. S4B). The total GenX emissions reductions in the Base scenario are not a uniform percent reduction from the Uncontrolled Base for each month (Fig. S4B), due to the varying contributions and control efficiencies of HFPO-DA versus HFPO-DAF (Fig. S5A&B). HFPO-DA is more efficiently controlled: 61–100% reduction, compared to 13–42% reduction for HFPO-DAF, from the Uncontrolled Base to the Base scenario (Fig. S5C). The GenX emission controls that start in June result in a decrease in deposition rates from June onward in the Base scenario (Fig. 2B). Nine other compounds were targeted for control (PAF, PMPF, TAF, COF2, PEPF, DA HFPO-TA, MA, PFASOTHER, see D’Ambro et al.²⁴, Table S12), and lower deposition rates were also observed for the sum of these nine compounds plus GenX in the Base scenario relative to the Uncontrolled Base (Fig. S6). However, the deposition rate decrease for the sum of these controlled compounds is not as substantial (up to 44% decrease) relative to that of just GenX (up to 94% decrease, Fig. 2B), due to the lower solubility of the other controlled compounds within the total.

Reducing the model horizontal resolution by increasing the grid cell sizes from 1 km (Base) to 4 km and 12 km decreased the peak deposition rates for both scenarios (4 and 12 km), and decreased the total PFAS mean and median deposition rates in the 4 km scenario. However, the mean and median deposition rates for the 12 km scenario were higher than the Base (1 km) scenario (Table S9). The annual deposition for total PFAS is overestimated in the 12 km scenario relative to the 1 km Base scenario for most of the domain except due north and south of the facility (Fig. S2D). The deposition over prediction throughout much of the domain is likely a result of the decreased deposition within 12 km of the facility, due to decreased air concentration, resulting in higher air concentrations and therefore deposition further afield, over a larger area (Fig. S7), reflected in the higher domain-wide median and mean (Table S9). Our results represent a conservative estimate of the impact of going from relatively near-source to regional scale: with increasingly complex terrain, multiple PFAS sources, etc, the impact of coarser horizontal resolution will be stronger.

4. Implications & Future Directions

In this work, we extrapolate the RfD’s underlying EPA’s drinking water Health Advisories for select PFAS to inhalation screening levels, and then compare these to our CMAQ-predicted air concentrations. We find that for 54–89% of the year, the GenX air concentrations at the facility and the two largest urban areas, Fayetteville and Wilmington, are below the measurement detection limits (0.017 pg m^{−3 44}). Fayetteville and Wilmington never experience GenX air concentrations above the inhalation screening level, but the area around the facility experiences ~15 days above this level. Furthermore, we find that the inhalation screening level weighted relative exposure of GenX to formaldehyde, a known hazardous air pollutant, is 0.76:1 at the facility. GenX represents 1% of the total PFAS emissions from this facility²⁴, with no toxicological information on many of the other compounds. Future research on the inhalation risk of total PFAS in ambient air is needed.

The air concentration of total PFAS within 150 km of the manufacturing facility is generally not very sensitive to varying the chemical composition, temporal emissions allocation, or horizontal resolution (Table S8). This is unsurprising: the majority of the emitted mass is volatile and insoluble²⁴, preventing compounds from partitioning to aerosol or depositing significantly. The exception is that the lower horizontal resolution scenarios under predict the near-field peak concentrations, although their domain-wide means and medians are similar to the high spatial resolution results (Table S8). This implies that 12 km resolution (and higher) models may be helpful for estimating regional-scale impacts of PFAS air transport, but 4 km resolution or higher is appropriate for estimation of near-field elevated episodes.

Deposition varies substantially among the model scenarios we tested (Fig. 3). For future studies where emissions are exclusively volatile, insoluble PFAS compounds, or the emissions are lumped into a single species confined to the gas phase (SSG), a deposition rate of 1.3% of total emissions within 150 km of the facility is suggested by the lumped scenarios (Fig. 3). However, if soluble or semi-soluble compounds are present, more care must be taken in constraining model inputs. Gas/particle partitioning, dependent on physicochemical properties, is critical for predicting deposition, thus models require accurate compound-specific representations of vapor pressure and Henry’s law coefficients with dependence on aerosol- and cloud-water pH. Simplifying the temporal allocation of emissions and decreasing model horizontal resolution to 4 km results in small differences in total deposition. Decreasing the model horizontal resolution further, to 12 km, increases the percent of emissions deposited by 40%. In all scenarios, ~2.4 – 3x as much mass is deposited within the first 48 km from the facility as is deposited in the next 48 km radius from the facility (Fig. S7). It’s worth noting that none of these model scenarios included gas- or particle-phase oxidation of PFAS compounds, which may impact our results by altering compounds’ atmospheric lifetimes and physicochemical properties.

Figure 3.

The percent of annual PFAS emissions deposited via wet (blue) and dry (gray scale) deposition within the domain, i.e within ~150 km of the facility, for each of the model scenarios tested herein.

These results indicate that the chemical speciation is the most important input that we tested for representing the air transport and fate of PFAS. For future modeling work, it would be beneficial if emission reporting included speciated PFAS with a focus on capturing differences in solubility and volatility. Alternatively, individual compounds of particular concern/interest can be inventoried, with the remainder grouped by volatility and solubility. Otherwise, if total PFAS emitted from a point source is reported, model runs should include bounding studies to understand the variations in deposition versus transport. Ultimately, it would be better to have total PFAS emissions from a point source rather than one or a select few compounds of concern, which would likely underestimate total ecosystem burdens and human health impacts.