An Ecosystem-Scale Model of PFAS Dynamics in Stream-to-Riparian Food Webs

Abstract

Per- and polyfluoroalkyl substances (PFAS) are widespread contaminants in aquatic ecosystems, where they bioaccumulate and trophically transfer through food webs. We derived an ecosystem-scale model assessing Σ₁₂PFAS partitioning, bioaccumulation by larval aquatic insects, bioamplification by adult aquatic insects, and trophic transfer from aquatic insects to riparian spiders using data collected across five streams in a mildly impacted watershed. Partitioning involved the greatest PFAS enrichments (range in posterior median log₁₀ k _d values = 1.5–2.1), with lesser concentration increases from particulates to larval aquatic insects and from larval to adult aquatic insects, and Σ₁₂PFAS concentration decreases from adult aquatic insects to spiders. Among the 12 compounds, PFOS had the largest proportional contribution to Σ₁₂PFAS concentrations for all sample types, with consistently increasing concentrations from surface water to adult aquatic insects. Σ₁₂PFAS concentrations (5.4–51 ppb wet weight), body burdens (0.44–5.2 ng per individual), and bioamplification factors (0.15–4.9) differed considerably among five common aquatic insect taxa. This study presents a robust modeling framework for PFAS trophodynamics and offers insights into the biological controls that contribute to the context dependency of PFAS at the interfaces of aquatic and terrestrial food webs.

1. Introduction

Aquatic ecosystems are susceptible to contaminant inputs via groundwater and point source discharges, terrestrial runoff, and atmospheric deposition. , While research emphasizes contaminant distribution by physical processes, biological processes such as food web interactions are increasingly recognized as important controls over the ecosystem dynamics and potential influence of contaminants. , Contaminants are introduced to aquatic food webs by partitioning to abiotic and biotic compartments that serve as habitat and dietary resources for low trophic level organisms. − Following uptake by primary producers, contaminant distribution among food web compartments is largely determined by the extent to which contaminants are bioaccumulated and exert toxicity among trophic levels. These phenomena have been an area of active research in freshwater systems where merolimnic insects (i.e., aquatic insects that have an aerial/terrestrial life stage following metamorphosis) biotransport nutrients and contaminants acquired from their aquatic habitat to their consumers in riparian habitats, thereby linking aquatic and terrestrial food webs. ,− Aquatic-to-riparian contaminant biotransport can represent substantial proportions of contaminant exposure for spiders, amphibians, bats, and birds that preferentially feed in riparian habitats during peak periods of energy-dense invertebrate emergence. ,, Biotransport by adult aquatic insects has been explored for many chemical types, , including macronutrients and fatty acids (e.g., − ), algal toxins, metals (e.g., − ), pharmaceuticals, pesticides, − and more recently for contaminants of emerging concern such as the per- and polyfluoroalkyl substances (PFAS). −

PFAS are a subgroup of synthetic organofluorine compounds defined as containing at least one perfluorinated methyl group or perfluorinated methylene group. , These highly stable perfluorinated carbons (i.e., C–F bonds) confer thermal and chemical resistance and oil- and water-repellent properties, which collectively govern the widespread distribution and persistence of PFAS in the environment. − Though PFAS contamination of water resources is an increasingly well-known problem, there are numerous outstanding questions concerning the magnitude, extent, management, and health consequences of PFAS contamination that are important to addressing the PFAS problem internationally. − The huge number of compounds (>15,000) and application types (e.g., as components of surfactants, cleaning products, stain and water repellants, pesticides, and firefighting or aqueous film-forming foams (AFFF)) ,, contribute multiple pathways by which PFAS are introduced to freshwater systems. The best known sources include landfills, military, and other installations with historical, ongoing, and/or frequent use of AFFF, industrial sites, and municipal wastewater treatment effluents. , Contamination from these sources occurs via direct inputs or diffuse pathways such as PFAS infiltration to groundwaters at contaminated land surfaces and subsequent groundwater flow path discharge to surface waters. −

Our current understanding of PFAS in linked aquatic and riparian food webs comes from only a few food web assessments performed in ecosystems with elevated levels of PFAS loading from known sources. − From these studies, we understand that primary producers and consumers at the base of aquatic food webs readily bioaccumulate some PFAS compounds, which subsequently move along food chains by dietary trophic transfer. PFAS are detected in adult aquatic insects and their predators in adjacent riparian habitats, indicating the importance of linked aquatic and terrestrial food webs as mechanisms of PFAS transfer from water to land. Both PFAS concentrations and mixture compositions are altered by these trophodynamic processes due to compound-specific properties influencing their food web behavior. − While these studies differ in their estimates of where along food webs we expect to find the highest PFAS concentrations, it is apparent that biota exert substantial control over PFAS fate and transport, determining the levels and mixtures of PFAS compounds that reach riparian and terrestrial consumers.

As with all contaminants, PFAS dynamics and biotransport at the aquatic–riparian interface are functions of ecological, physiological, and toxicological factors that collectively influence the extent to which PFAS are bioaccumulated and retained during metamorphosis and rates of aquatic insect emergence. PFAS biotransport can therefore be concentration-dependent, with potentially distinct patterns observed in ecosystems with high concentrations, where contaminant effects may dominate, versus mildly impaired ecosystems, where ecological and physiological processes are anticipated to be the primary drivers. The goal of the present study was to characterize mixture- and compound-specific concentration patterns in linked stream and riparian food webs of systems with ambient and relatively low PFAS levels, thereby complementing prior studies downstream of PFAS point sources. We developed an ecosystem-scale model of PFAS trophodynamics using PFAS concentration data for surface waters, sediment, detritus, seston, biofilm, larval, and adult aquatic insects and riparian spiders collected from five stream sites in the Farmington River watershed (Connecticut, USA) between 2022 and 2023. We subsequently used the model to (1) compare PFAS partitioning efficiencies of different stream particulate fractions, (2) characterize PFAS concentrations, mixture compositions, and trophic magnification from aquatic to riparian compartments, and (3) assess the extent to which PFAS concentrations and burdens are bioamplified by aquatic insect metamorphosis.

2. Methods

2.1. Study Area and Watershed Survey

This study was based in the Farmington River watershed, a fifth-order tributary (1571 km²) of the Connecticut River spanning 33 towns in northwestern Connecticut and southwestern Massachusetts (Figure S1). Land cover in the Farmington River watershed was primarily forested with ∼20% urban/developed area, 8% forested wetlands, and ∼3% each of agriculture and open water. The watershed encompassed a number of sites that may contribute PFAS to the surrounding environment, including 32 active or closed landfills, wastewater treatment plants, airports, consolidated fire services, toxic release inventory sites, Superfund sites, and U.S. EPA Cleanups in My Community (EPA CIMC) list sites. Amplified concerns about PFAS in the watershed stem from an accidental large-volume release of PFAS-containing AFFF to the Farmington River near the Connecticut River confluence in June 2019. , Groundwater discharge and/or surface water samples were collected from 21 sites (Figure S1 and Table S2) during the first phase of this project to characterize PFAS patterns (concentrations and mixtures; Tables S6 and S7) across the watershed and to guide the selection of sites for intensive food web sampling. Full information about the survey methods and results is provided in the Supporting Information.

2.2. PFAS Partitioning and Trophodynamics

Five stream sites (Hop Brook, Russell Brook, Pequabuck River, Ratlum Brook, and Burr Pond Brook) were selected for focused assessment of PFAS partitioning and trophodynamics, including bioaccumulation, metamorphic retention, and trophic transfer from aquatic to riparian food webs (hereafter termed “food web sites”). Sites were selected based on surface water PFAS concentrations and site characteristics (e.g., accessibility, quality of stream, and riparian habitat) to facilitate sampling success and to capture ranges in PFAS concentrations and compounds contributing to PFAS mixtures from diverse potential sources. Selected sites received PFAS from surrounding sources via diffuse, non-point pathways. Anthropogenic activities in the surrounding areas of selected sites with potential influence on PFAS loading included waste treatment and management facilities (compost, wastewater, sewage, landfills), EPA CIMC sites, various industries, fire service facilities, airports and heliports, and ski slopes. PFAS source attribution was outside the scope and motivation of this study.

Hop Brook and Russell Brook were sampled in 2022 (sampling period: June 2–30) and Pequabuck River (April 24–July 24), Ratlum Brook (April 21–July 25), and Burr Pond Brook (April 27–July 26) were sampled in 2023. All sampling activities were in accordance with a Scientific Collections permit from the Connecticut Department of Energy and Environmental Protection. Surface waters, sediment, seston, detritus, biofilm, larval and adult aquatic insects, and riparian spiders were collected from a 30 m stream reach divided into three 10 m segments at each site. Surface waters were grab-sampled using methanol-cleaned 250 mL polypropylene copolymer (PPCO) bottles. Bulk sediment samples were collected from the top 6 cm using a bulb planter or scoopula into Ziploc bags. Seston was collected using a plankton net (80 μm mesh) secured and suspended in the water column in the last third of the sampled segment with rebar; accumulated seston was subsequently transferred to a 250 mL PPCO bottle. Detritus samples were sampled using D-nets (LaMotte, 12 × 6″, 500 μm Nylon mesh), sorted to remove biotic material (e.g., aquatic insects, plant matter), transferred to a falcon tube, and stored frozen until analysis. Biofilm samples (i.e., attached epiphytic algae presumed to include heterotrophic microbes) were gently scraped from rock surfaces using pre-cleaned stainless-steel scalpels into falcon tubes. Larval aquatic insects were collected by using kick nets (BioQuip Products, 9 × 18″ with 500 μm Nytex mesh) and preliminarily sorting by order to falcon tubes. Adult aquatic insects were collected using a combination of floating pyramidal emergence traps (60 × 60 cm frame and ∼0.33 m² capture area, 5 per site, distributed along the 30 m stream segment). Emergence traps were checked every 3–5 days, and accumulated adult aquatic insects were collected into pre-cleaned, empty PPCO sampling bottles. To meet biomass requirements for PFAS and stable isotope analyses, we supplemented emergence trap sampling with evening bankside sampling (within 4 h of dusk) using blacklights to attract insects to white sheets. Insects were collected using forceps or a modified hand vacuum. Tetragnathid spiders were hand-collected from riparian vegetation within 1 m of the stream bank or collected on drop sheets by shaking branches. We used stainless-steel tweezers to transfer spiders to falcon tubes or vials. Samples were collected using methanol-rinsed bottles, scoops, trays, tubes, and tweezers wherever possible. Samples were stored on ice immediately and transferred to 4 °C (waters) or −20 °C (particulates and biota) storage within 12 h of collection upon returning to the laboratory. Water, sediment, seston, detritus, and biofilm were collected at both the beginning and the end of the collection period at each site; aquatic insects and spiders were collected throughout the collection period at each site.

Thawed seston and biofilm samples were subsequently vacuum-filtered onto glass microfiber filters (Cytiva Whatman Binder-Free, grade GF/D, 47 mm) to remove any remaining water. Seston and biofilm samples were subsequently scraped from filters into 2.0 mL centrifuge tubes and frozen prior to analysis. Bulk sediment samples were massed in the laboratory to ∼ 0.5 g wet weight (ww) subsamples and frozen in 2.0 mL centrifuge tubes prior to analysis. Aquatic insects were sorted into composites of ∼ 0.5 g ww by life stage and order to create larval and adult composites of Diptera (midges), Ephemeroptera (mayflies), Megaloptera (e.g., alderflies), Odonata (dragonflies), Plecoptera (stoneflies), and Trichoptera (caddisflies, cases removed). Further, we composited larval aquatic insects by order across instars to meet analytical biomass requirements. Riparian spiders belonging to the family Tetragnathidae (long-jawed orb weavers) were similarly composited for chemical and stable isotope analyses.

2.3. Sample Preparation and PFAS Analysis

All samples were prepared and analyzed according to industry standard, clean field techniques at the University of Connecticut’s Center for Environmental Sciences and Engineering (Storrs, Connecticut). Water samples (250 mL) were analyzed for 14–28 PFAS compounds (Table S1) using a modified version of U.S. EPA Method 537.1. Surrogate solutions (40 μL, EPA-537SS-R1 mix, Wellington) were spiked into 250 mL of agitated water samples followed by extraction using preconditioned UCT SDVB (8 mL, 0.5 g) SPE cartridges and elution with 6 mL of HPLC-grade methanol. Evaporated eluents were reconstituted with methanol and spiked with an internal standard (40 μL, EPA-5371S mix, Wellington) to a final volume of 250 μL. Solid samples were analyzed for 28 PFAS compounds using methods previously validated by our group. , Samples were thawed at room temperature and homogenized. An aliquot of 0.1 ± 0.0001 g ww was weighed in a 2.0 mL polypropylene (PP) centrifuge tube and spiked with a surrogate solution and 500 μL of methanol, followed by sonication and vortexing. 0.05–0.10 g of QuEChERS powder (MgSO₄/NaCl) was then added to each sample, shaken, and vortexed for 10 min at 2500 rpm. Samples were cooled at −20 °C overnight to assist in the sample cleanup process by enhancing the precipitation of unwanted sample matrix interferences and centrifuged the following day at 14,000 rpm for 10 min. A 300 μL aliquot of the supernatant was transferred to a 2.0 mL PP centrifuge tube, and 0.05–0.10 g of QuEChERS cleanup powder (PSA/GCB/MgSO₄) was added. All samples were vortexed for 10 min, centrifuged for 10 min at 14,000 rpm, and an aliquot of the supernatant was transferred to a 300 μL LC vial and spiked with the internal standard.

We assessed quality assurance and quality control using method blanks, laboratory control samples, and field blanks. All quality control and field-collected samples were analyzed using a Waters Acquity UPLC coupled with an Acquity TQD tandem mass spectrometer (Waters Co., Milford, MA) fitted with a PFAS conversion kit. The Acquity UPLC BEH C18 (1.7 μm, 2.1 × 100 mm) column was heated to 40 °C, and samples were injected at a volume of 5 μL on a 20 μL loop for analyte separation. We employed a mobile phase consisting of 20 mM ammonium acetate in water (solvent A) and methanol (solvent B) for gradient column elution. The total run time was 12 min with a constant flow rate of 0.3 mL min^–1. The detection and quantification of analytes and surrogate compounds was performed in negative electrospray ionization mode (ESI^‑) and multiple reaction mode using the Waters IntelliStart software for analyte signal optimization. Statistical analysis for obtaining calibration and quantification results for all compounds was run using Waters QuanLynx included in the MassLynx software v.4.2. PFAS concentrations in all method blanks were below analyte-specific method reporting limits (Table S1). Instrument performance measures (i.e., calibration, precision, and accuracy) were assessed against those from a methods study recently published by our group (see Tables S2 and S3 in ref ); biosolid calibrations are also reported in Table S3 alongside those for waters. All laboratory control sample recoveries were within the range of 80–114%. Mean internal standard recoveries ranged from 109 to 122% for waters and from 99 to 116% for biosolids. Mean surrogate recoveries ranged from 58 to 97% for waters and from 60 to 70% for biosolids.

2.4. Stable Isotope Analyses

Biofilm and tissues (aquatic insects and spiders) were analyzed for δ¹⁵N and δ¹³C at the U.S. Geological Survey (USGS) Geology, Geophysics, and Geochemistry Stable Isotope Laboratory. Samples were massed (∼1 mg dry mass per sample) into 4 mm × 6 mm tin capsules and crimp-sealed. Isotopic compositions were measured using an elemental analyzer (Thermo Scientific FlashIRMS) interfaced to a mass spectrometer (Thermo Scientific Delta V plus) operated in continuous flow mode. Isotopic data are reported in standard delta (δ) notation as per mil (‰) deviations relative to internationally accepted scales (air and Vienna PeeDee Belemnite [V-PDB], respectively) following normalization to primary isotopic standards (USGS 40, δ¹⁵N = −4.52‰ and δ¹³C = −26.39‰; USGS 41a, δ¹⁵N = 47.55‰ and δ¹³C = 36.55‰). Analytical precision and sample reproducibility were <0.2‰ and ±0.3‰ for both isotopes, respectively; accuracy bias was <0.1‰ for both isotopes.

2.5. Modeling Approach and Data Analysis

The results of intensive sampling at the five food web sites were used to develop an ecological model for estimating concentration enrichment (Table ) and changes in PFAS mixtures along stream-to-riparian food webs. Data are publicly available at ScienceBase. We collected 258 samples across the five food web sites (including water, particulate (sediment, detritus, seston, biofilm), larval and adult aquatic insects, and riparian spiders) and analyzed them for 28 PFAS compounds, resulting in 7224 observations; numbers and types of aquatic insects varied among sites as summarized in Table S4. We removed nine compounds from the resulting data set for which there were no detections (i.e., 0/258 results were above detection) and an additional eight compounds that were detected in less than 5% of samples with the exception of 8:2FTS, which was detected in ∼3% of samples but was retained in the data set because these samples represented four sample types (water, biofilm, larval aquatic insects, and riparian spiders) from a single site and could therefore be modeled across the food web. Water sample concentrations are reported as ng mL^–1, and solid sample concentrations are reported as ng g^–1 ww. Hereafter, we use parts per billion (ppb) units to facilitate concentration comparisons and enrichment calculations across sample types.

1. Enrichment Factors Calculated for This Study.

term	definition
kd, partitioning coefficient	ratio of compound-specific or Σ12PFAS concentration (ppb units) in particulate fraction (numerator) and surface water (denominator); calculated separately with sediment, detritus, seston, and biofilm as distinct particulate fractions
BSAF, biota-sediment accumulation factor	ratio of compound-specific or Σ12PFAS concentration in larval aquatic insects (numerator) and sediment (denominator)
BAF, bioaccumulation factor	ratio of compound-specific or Σ12PFAS concentration in larval aquatic insects (numerator) and particulate fraction (denominator); calculated separately with detritus, seston, and biofilm as distinct particulate fractions
BAMF, bioamplification factor	ratio of compound-specific or Σ12PFAS concentration in adult aquatic insects (numerator) and larval aquatic insects (denominator); calculated at community and taxon-specific scales
TTF, trophic transfer factor	ratio of compound-specific or Σ12PFAS concentration in riparian spiders (numerator) and adult aquatic insects (denominator)
TMS, trophic magnification slope	the slope of the linear regression between the log10 PFAS concentration and the δ15N values of the biotic fraction (biofilm, larval and adult aquatic insects, and riparian spiders)

The final data set had 3096 observations representing 12 PFAS compounds (PFHxA, PFHpA, PFOA, PFNA, PFDA, PFUnA, PFDoA, PFBS, PFHxS, PFOS, 6:2FTS, and 8:2FTS) measured across 17 food web sample types (11 of which are larval and adult aquatic insect taxa) with positive concentrations ranging from 3% of observations for 8:2FTS to 84% for PFOS; this data set included 68 observations with detections below the reporting limit (Table S1) and 2451 observations below the detection limit (non-detects).

Our overall modeling approach used Bayesian hierarchical models with mixture likelihoods to account for non-detections. We used the raw data described above to inform these models and estimate posterior distributions of quantities of interest such as the central tendencies and associated uncertainties of PFAS concentrations. We used Bayesian models because we were interested in quantifying the direct probabilities of our hypotheses given the data we collected (i.e., P­(H|D)), rather than the probabilities of obtaining data as extreme or more extreme than expected under a null hypothesis (i.e., P­(d ≥ D|H₀)). We also used Bayesian analyses due to the ease of calculating derived quantities that cannot be directly measured (and their uncertainty) from the posterior distributions, such as enrichment factors (Table ), body burden ratios between larval and adult insects, site-specific versus site-averaged (i.e., ecosystem-scale) quantities, etc. Detailed descriptions of model structures and assessment are provided in Supporting Information Methods and summarized here.

2.5.1. Modeling PFAS Concentrations

We modeled PFAS concentrations using a generalized linear mixed model with a hurdle Gamma mixture likelihood to account for the remaining zero inflation (i.e., the non-detects) in the response variable (PFAS ppb) following Huang et al. (2019). This model estimates two main quantities simultaneously: (1) a posterior distribution of the mean PFAS concentration when the compound is detected (μ_detect) and (2) a posterior distribution of the probability of detection (p _detect). The third quantity, the overall mean concentration of PFAS (μ_overall), is the primary quantity in which we were interested . It is calculated as the product of the first two quantities: μ_overall = (μ_detect × p _detect). The response variable was PFAS concentration (ppb ww), and the fixed predictor variable was food web compartment (n = 17 levels). We also included varying intercepts and slopes per site and per PFAS compound (n = 12 compounds), allowing us to estimate the three quantities above (μ_overall, μ_detect, p _detect) for each PFAS compound within all food web compartments at each site.

To estimate Σ₁₂PFAS concentrations (the sum concentration of all 12 PFAS compounds), we first summed PFAS for each food web compartment and site in the raw data. We then fit a generalized linear mixed model with Σ₁₂PFAS ppb as the response variable, food web compartment as the fixed predictor, and a varying intercept and slope per site. We added 0.0001 to each concentration prior to fitting to account for the small number of samples (21/258) with Σ₁₂PFAS = 0. To estimate Σ₁₂PFAS concentrations per insect taxon, we fit a similar model but added varying intercepts and slopes per taxon.

2.5.2. Modeling PFAS Burdens in Aquatic Insects

To convert estimates from a concentration (ng g^–1 ww) to a body burden (ng individual^–1), we fit a generalized linear model with individual wet mass as the response variable, insect taxon as a fixed predictor, and varying intercepts and slopes per site and life stage (larval vs adult). Raw individual wet masses were derived by dividing the composite masses by the number of individual organisms per composite. We then multiplied 4000 posterior draws of insect mass by corresponding draws of posterior PFAS concentrations to generate distributions of body burdens in units of ng individual^–1. This method also ensured that body burden accounted for uncertainty in both insect mass and PFAS concentration.

2.5.3. Modeling Trophic Magnification and Trophic Overlap

We defined trophic magnification (Table ) using the slope of the relationship between log₁₀[PFAS] and δ¹⁵N values from biofilm, larval and adult aquatic insects, and riparian spiders (i.e., the change in log₁₀[PFAS] per unit change in δ¹⁵N); using δ¹⁵N in our regressions allowed us to include biofilm as the primary producer in our assessment of concentration change in the food web rather than using biofilm as the primary producer basis for trophic position calculations as is the case when calculating trophic magnification factors. To estimate trophic magnification, we first fit a general linear model estimating taxon-specific δ¹⁵N with mean-centered δ¹⁵N as the response variable, taxon as the predictor variable, and varying intercepts and slopes per site. The likelihood was Gaussian because δ¹⁵N is a continuous quantity that can be negative and positive. From the posteriors of this model, we estimated the median δ¹⁵N values per taxon and site and used those values as predictors in a second model with standardized (z-score) log₁₀[PFAS] as the response variable, median δ¹⁵N as the predictor variable, and varying intercepts and slopes per site and per PFAS compound. This model also had a Gaussian likelihood. To estimate the trophic magnification of Σ₁₂PFAS, we fit a similar model with standardized (z-score) log₁₀[Σ₁₂PFAS] as the response variable, median δ¹⁵N as the predictor variable, and varying intercept and slopes per site. To visualize the trophic overlap of spiders and adult aquatic insects, we fit a multivariate model with δ¹⁵N and δ¹³C as the response variables and the taxon as the predictor variable with varying intercept and slopes per site. The likelihood was Gaussian.

2.5.4. Model Specification

All models were specified in R (v. 4.4.2) using brms and fit using Hamiltonian Monte Carlo in rstan. Models were fit using 4 chains each with 2000 iterations and the first 1000 discarded as warm-up. We ensured chain convergence using diagnostic plots and by ensuring that all R-hats were <1.01. Model fit was assessed using posterior predictive checks and Bayesian p-values (SI Methods, Figure S2 and Table S5). The hierarchical models above estimated the average PFAS concentrations or insect mass for all sites, taxa, and life stages. All subsequent quantities, such as enrichment factors (Table ), PFAS burdens and burden ratios between larval and adult aquatic insects, and ecosystem-scale averages were estimated as derived quantities from the posterior distributions of the fitted models. In other words, ecosystem-scale results were derived by averaging posterior distributions of individual site results, thereby incorporating uncertainty across sites.

3. Results and Discussion

PFAS from diverse sources aggregate in aquatic ecosystems as a result of point source loading and physical processes including atmospheric deposition, land surface runoff, and groundwater flow. ,, The preliminary survey of PFAS contamination in groundwater discharges and surface waters across the Farmington River watershed (SI Methods and Results) indicated that PFAS are commonly detected at concentrations characteristic of areas without point sources (Tables S6 and S7). We used the results of the watershed survey to subjectively select five stream sites (Table S2) with varying surface water concentrations and mixture profiles for the intensive sampling of PFAS trophodynamics in streams and from stream to riparian food webs. The results of this multi-site field campaign were integrated using a Bayesian hierarchical model for estimating patterns of PFAS partitioning and dietary enrichment (Table ) in streams and biotransport to riparian food webs at the ecosystem scale. Specifically, we used the model to (1) compare PFAS partitioning efficiencies among stream particulate fractions, (2) characterize PFAS concentrations, mixture compositions, and trophic magnification from aquatic to riparian compartments, and (3) assess the extent to which PFAS concentrations and burdens are bioamplified by aquatic insect metamorphosis.

3.1. PFAS Partitioning Involves Greatest Concentration Enrichments

Organic contaminants including PFAS are primarily introduced to aquatic food webs via partitioning (i.e., adsorption, bioconcentration, and/or bioaccumulation processes) to abiotic and biotic compartments that serve as habitat and dietary resources for low trophic level organisms. , Numerous factors relating to contaminant and biological properties and the physicochemical environment influence contaminant partitioning and bioavailability. The degree of enrichment can therefore vary substantially in place, time, and among particulate types. − As a result, contaminant partitioning can be highly variable within and across systems. Here, we compared PFAS partitioning from water to sediment, detritus, seston, and biofilm as four particulate compartments important to contaminant dynamics at the base of stream food webs (Figure ). Posterior median particulate Σ₁₂PFAS concentrations ranged from 1.8 ppb (95% credible interval (CrI) = 1.1 to 3.4 ppb) in sediment to 8 ppb (3.7 to 22 ppb) in seston (Figure a, Table S9). Accordingly, the posterior median log₁₀ Σ₁₂PFAS partitioning coefficients (k _d) were highest for seston (2.1, 1.7–2.6) = biofilm (2.1, 1.7–2.4) > detritus (1.8, 1.3–2.3) > sediment (1.5, 1.1–1.8) (Figure , Table S10). The subsequent enrichment processes of aquatic insect bioaccumulation, metamorphosis, and trophic transfer were of smaller magnitudes relative to partitioning, such that partitioning processes involved the greatest increases in Σ₁₂PFAS concentrations (Figure , Table S10).

1.

Σ₁₂PFAS concentration enrichment or dilution with partitioning, bioaccumulation, metamorphosis, and trophic transfer. Bars represent 95% credible intervals with posterior medians marked by larger black points. Smaller points represent raw data of individual sites. k _d: partitioning coefficient, BSAF: biota-sediment accumulation factor; BAF: bioaccumulation factor; BAMF: bioamplification factor; TTF: trophic transfer factor. Enrichment factors are defined in Table .

2.

PFAS concentration and composition differences among water, biofilm, larval and adult aquatic insects, and riparian spiders. (a) Ecosystem-scale model estimates of Σ₁₂PFAS concentrations by sample type. Black points represent posterior medians ± 95% credible intervals, and colored shapes represent individual aquatic insect taxa. (b) Compound-specific concentrations by sample type. Colored points represent posterior medians ± 95% credible intervals, and gray points represent individual sample results. (c) Compound-specific proportional contributions to sum PFAS concentrations. W: water, B: biofilm, L: larval aquatic insects, A: adult aquatic insects, S: spiders.

We most commonly detected PFHxA PFOA, PFNA, and PFOS in both water and particulates in a manner that led to relatively consistent and robustly informed estimates of partitioning enrichment for these compounds (Figure S3). For all other compounds, infrequent detections in paired sample types (e.g., water and sediment or water and biofilm) supported less robust estimates of concentration enrichment primarily informed by priors and hierarchical model structures (and are noted in Figure S3 using partial transparency). As with the Σ₁₂PFAS concentration enrichment patterns described above, we estimated efficient partitioning of PFHxA, PFOA, PFNA, and PFOS from water to particulates relative to the degree of concentration enrichment or dilution with bioaccumulation, metamorphosis, and trophic transfer from adult aquatic insects to spiders (Figure S3 and Table S10). PFHxA, PFOA, PFNA, and PFOS more efficiently partitioned to seston (range in posterior median log₁₀ k _d values = 2.4–2.7) and biofilm (2.3–2.7) than to sediment (0.98–1.9) or detritus (1.7–2.4), suggesting greater enrichment to particulate types with greater amounts of biotic material. For example, the extracellular polymeric substances produced by biofilm contain proteins, lipids, polysaccharides, and nucleic acids with functional groups that are thought to serve as PFAS binding sites. , The efficient partitioning to biotic particulates observed in this study is consistent with other studies, , and prior research has also observed greater PFAS partitioning to biofilm compared to other basal resources, such as detritus, plants, phytoplankton and algae, and submerged macrophytes. , These results collectively suggest substantial variation in PFAS enrichment at the base of aquatic food webs, with especially efficient partitioning to biofilm, and that the degree of enrichment may be concentration-dependent as has recently been shown for PFOS experimentally. Hereafter, we emphasize biofilm as the particulate fraction to characterize the overall food web dynamics, as biofilm is a primarily biotic particulate fraction (i.e., representative of uptake at the base of stream food webs), and biofilm Σ₁₂PFAS concentrations (posterior median = 7.1 ppb, 95% credible interval = 4.7–11 ppb) were intermediate to those in sediment (1.8 ppb, 1.1–3.4 ppb), detritus (3.4 ppb, 1.5–11 ppb), and seston (8 ppb, 3.7–22 ppb) (Table S9).

3.2. PFAS Concentrations and Compositions Shift from Streams to Riparian Consumers

The structural diversity of PFAS (i.e., variation in compound-specific carbon chain lengths, functional groups, and membrane partitioning coefficients) , forecasts that PFAS concentrations and mixtures will be altered by the toxicokinetic processes underlying dietary transfer and metamorphosis. Our second objective was therefore to characterize PFAS concentrations, mixture compositions, and trophic magnification from the base of stream food webs to riparian spider predators of aquatic insects. Posterior median Σ₁₂PFAS (ppb) concentrations ranged 4 orders of magnitude among sample types, from 0.0059 ppb (95% CrI: 0.0034–0.13 ppb) in surface water to 46 ppb (27–93 ppb) in adult aquatic insects (Figure a, Table S9). Σ₁₂PFAS concentrations were successively enriched on average through the processes of partitioning to a biofilm, bioaccumulation by larval aquatic insects, and metamorphosis from larval to adult aquatic insects (Figure a, Table S9). We estimated the greatest Σ₁₂PFAS concentration increases with partitioning from surface water to biofilm (posterior median log₁₀ k _d = 2.1, 95% CrI = 1.7–2.4), with lesser Σ₁₂PFAS concentration increases from biofilm to larval aquatic insects (BAF = 4.2, 2.5–7.3) and from larval to adult aquatic insects (BAMF = 1.5, 0.77–3.3). These estimates correspond to >0.99 probability of increasing Σ₁₂PFAS concentrations in biofilm relative to water and in larvae relative to biofilm, and 0.89 probability of increasing Σ₁₂PFAS concentrations in adult relative to larval insects. In contrast, we estimated that Σ₁₂PFAS concentrations were slightly lower (corresponding to a probability of 0.74) in spiders than in adult aquatic insects (TTF = 0.75, 0.31–1.8) (Figure a, Table S10).

Our estimate of decreased spider concentrations (TTF = 0.75, 0.31–1.8) accounts for considerable variation among sites (Figure S4a) and likely captures site-specific patterns of spider reliance on aquatic vs terrestrial prey with assumed, but not measured, influence over PFAS exposure (i.e., terrestrial prey with lower concentrations could contribute to diluted PFAS concentrations in spiders). Further, our sampling indicated differences in aquatic insect community composition among sites (Table S4), with Diptera and Trichoptera being the most consistently sampled taxa. The Σ₁₂PFAS model using only these taxa estimated lesser Σ₁₂PFAS concentrations in larval and adult taxa and a minor increase in Σ₁₂PFAS concentration with trophic transfer from adult aquatic insects to spiders (results not shown). This result is similarly uncertain under both model structures, with the full community model estimating a 75% probability of a decrease in the Σ₁₂PFAS concentration with trophic transfer and the restricted model (Diptera and Trichoptera only) estimating a 50% probability of the Σ₁₂PFAS concentration decrease. We therefore conclude that there is no strong support for substantial concentration enrichment at the aquatic–terrestrial interface in this study system. This is in contrast to prior work in systems with higher PFAS concentrations. For example, Kotalik et al. reported increasing ΣPFAS concentrations with trophic transfer from adult caddisflies (Trichoptera) to riparian spiders (Tetragnathidae) at sites both upstream and downstream of a PFAS point source, with 6:2 FTS contributing for 77–89% of spider ΣPFAS concentrations in both cases. 6:2 FTS contributions to food web ΣPFAS concentrations in the present study were minimal (<2.5%, Table S11), with PFOS, PFHpA, and PFOS contributing most to riparian spider PFAS profiles. It is likely that differences in trophic transfer patterns at aquatic–riparian interfaces among studies reflect differences in PFAS mixture compositions (and the associated PFAS sources) among sites.

The partitioning, bioaccumulation, bioamplification, and trophic transfer patterns observed for Σ₁₂PFAS (Figure a) were consistent with compound-specific patterns except in the case of PFHpA and PFDoA for which we observed higher concentrations in riparian spiders than in adult aquatic insects (Figure b). In contrast, the relative contribution of individual PFAS compounds to Σ₁₂PFAS concentrations varied substantially among sample types (Figure c, Table S11). PFOS had the largest proportional contribution (≥22%, Table S11) to Σ₁₂PFAS concentrations for all sample types, with consistently increasing concentrations from surface water (posterior median = 0.0078 ppb, 95% credible interval = 0.0063–0.0096 ppb) to adult aquatic insects (19 ppb, 15–24 ppb) (Figure b,c, Tables S9 and S11). Compound-specific contributions to surface water Σ₁₂PFAS concentrations were relatively distributed, ranging from 1.1% for 8:2FTS (posterior median, 95% CrI = 0.16%–3.5%) to 22% for PFOS (95% CrI = 17%–26%). In comparison, PFOA, PFNA, PFUnA, and PFOS dominated PFAS profiles in stream food web compartments, collectively contributing 87%, 77%, and 79% of Σ₁₂ PFAS concentrations in biofilm and larval and adult aquatic insects, respectively (Table S11). PFBS, PFHxS, and 6:2FTS, which contributed 10–14% of surface water mixtures, were minor components of Σ₁₂PFAS concentrations in the biota. At the aquatic–terrestrial interface, PFOA (posterior median = 15%, 95% CrI = 10%–21%) and PFOS (33%, 25%–41%) were similarly important to Σ₁₂PFAS concentrations in spiders as they were in aquatic insects, whereas the contributions of PFHxA (8.9%, 5.1–14%) and PFHpA (17%, 8.9%–28%) increased in spiders relative to stream biota (Table S11).

Consistent with other assessments of PFAS in aquatic food webs, ,, Σ₁₂PFAS and compound-specific trophic magnification slopes (TMS) indicated that PFAS modeled in this study trophically magnify in linked aquatic–riparian food webs (i.e., all TMS >0). The probability of a positive slope was >0.99 for all TMS regressions, with a Σ₁₂PFAS posterior median TMS of 0.43 (95% CrI = 0.26–0.59, posterior median R ² = 0.84, 95% CrI = 0.83–0.85; Table S10). We highlight TMS patterns for PFOA, PFNA, PFUnA, and PFOS (Figure a) as the four compounds contribute at least 10% of the Σ₁₂PFAS concentrations in biofilm, larval aquatic insects, adult aquatic insects, and riparian spiders (i.e., compartments represented by δ¹⁵N values on the x-axis), respectively. Posterior median TMS for these compounds ranged from 0.52 to 0.69, with posterior median R ² values ranging from 0.58 to 0.77 (Figure b). Differences in base concentrations among these compounds (leftmost y-axis values; Figure ), but consistent slopes, reinforce similarities in their bioaccumulation (range in BAF_Biofilm = 2.4–3.0), bioamplification (range in BAMFs = 1.2–2.1), and trophic transfer (range in TTFs = 0.28–0.80) in streams and to riparian consumers (Table S10); these patterns included decreases in concentrations from adult aquatic insects to spiders, which reduced slope heights (Figure b, Table S10).

3.

Trophic magnification regressions showing relationships between centered δ¹⁵N values and log₁₀-transformed PFAS concentrations (ppb ww). (a) Σ₁₂PFAS trophic magnification regression. (b) Compound-specific trophic magnification regressions for PFOA, PFNA, PFUnA, and PFOS. The inset text boxes report posterior median (95% credible interval) trophic magnification slopes (TMS) and (posterior median 95% credible interval) R ² values of linear relationships. Symbols represent biota types, including biofilm, riparian spiders, and larval and adult aquatic insects.

Riparian spiders including Tetragnathids position themselves at the interfaces of aquatic and terrestrial habitats and preferentially consume aquatic insect prey during periods of adult aquatic insect emergence, with aquatic insects representing medium to high proportions of their diets. ,, They are increasingly sampled to inform the potential for contaminant transfer from aquatic to riparian food webs and are established as sentinels aquatically derived polychlorinated biphenyls, selenium, and mercury. The present study is among the first to assess PFAS transfer to riparian spiders. , Across sites and at the ecosystem level, stable isotope analyses support the assumptions that Tetragnathids feed on adult aquatic insects (i.e., overlapping δ¹³C and δ¹⁵N values; Figure S5) and that adult aquatic insects were a dietary source of PFAS to spiders. However, spider Σ₁₂PFAS concentrations were lower than those in adult aquatic insects, as were concentrations of all compounds except PFHpA and PFDoA (Figure b, Table S9). As a result, we report distinct patterns of compound-specific contributions to Σ₁₂PFAS concentrations between adult aquatic insects and spiders (Figure c, Table S10). These observations introduce questions for further study, including the degree to which aquatic vs terrestrial prey influence riparian spider PFAS profiles and whether compound or concentration-specific toxicokinetic processes mediate PFAS retention in spider tissues.

Despite the emerging consensus that PFAS bioaccumulate and trophically transfer through aquatic food webs, studies differ considerably in their predictions of which biota will have the highest PFAS concentrations and whether concentrations will be enriched in riparian/terrestrial consumers of aquatic insects. Our integrative model estimated the highest Σ₁₂PFAS concentrations in adult aquatic insects (Figure b). This finding was in line with the observations from three of the five sites informing the model, whereas higher concentrations were observed in riparian spiders at the other two study sites (Figure S4a); this pattern did not appear to be driven by differences in Σ₁₂PFAS concentrations or PFAS mixture composition among sites as compound-specific contributions to Σ₁₂PFAS concentrations were relatively consistent among sites with the exception of Hop Brook where PFOS dominated among biotic samples (Figure S4b). Studies with similar designs have previously observed highest Σ₁₂PFAS concentrations in larval aquatic insects, , or riparian spiders and birds known to consume aquatic insects. This context dependency of PFAS transfer through aquatic–riparian food webs captures differences in PFAS compounds present and their concentrations, organisms sampled (e.g., types of aquatic insects and riparian consumers), and analytical methods and advancements (e.g., detection limits) among studies that illustrate the complexities of developing a generalized understanding of ecosystem-scale PFAS dynamics.

3.3. PFAS Concentrations, Body Burdens, and Metamorphic Enrichment Vary among Taxa

Aquatic insect taxa are expected to differ in their contaminant profiles owing to taxonomic variation in habitat use and functional feeding strategies that may influence PFAS bioaccumulation. ,, We assessed differences in PFAS concentrations among larval and adult aquatic insects belonging to five common taxonomic orders: Diptera, Ephemeroptera, Odonata, Plecoptera, and Trichoptera (as we did not capture adult Megaloptera in our sampling, we excluded this order from taxonomic comparisons). Posterior median Σ₁₂PFAS concentrations in larval aquatic insects ranged 9.4-fold, from ∼40–50 ppb in larval Plecoptera (51 ppb, 95% CrI = 27–120 ppb), Ephemeroptera (41 ppb, 21–94 ppb), and Odonata (39 ppb, 21–83 ppb) to ∼5–20 ppb in Trichoptera (22 ppb, 14–40 ppb) and Diptera (5.4 ppb, 2.8–12 ppb) (Table ). We interpreted PFOS as driving this trend, as taxonomic patterns for PFOS mostly aligned with Σ₁₂PFAS concentration patterns (Table S12) and PFOS was the dominant contributor to community-level larval aquatic insect Σ₁₂PFAS concentrations (Figure c). Dietary sources and feeding behaviors vary within, as well as among, taxonomic orders, yet the elevated Σ₁₂PFAS concentrations in larval Plecoptera, Odonata, and Ephemeroptera corresponded with the elevated δ¹⁵N values for these taxa, indicating their relatively high food web positions among the sampled aquatic insects (Table S13). Σ₁₂PFAS variation was similar among adult aquatic insects (10.5-fold, approximately 1 order of magnitude), with the highest concentrations in adult Plecoptera (62 ppb, 33–130 ppb) and the lowest concentrations in Odonata (5.9 ppb, 2.1–14 ppb) (Table ). Adult aquatic insect patterns likely reflected taxonomic differences in larval life stage feeding behaviors as well as differences in PFAS retention through metamorphosis, as no feeding occurs during metamorphosis and adult aquatic insects captured with emergence nets and via bankside sampling were assumed to be captured prior to any feeding in the riparian habitat.

2. Posterior Median (95% Credible Interval) Σ₁₂PFAS Concentrations (ppb ww), Aquatic Insect Masses (g ww) and Σ₁₂PFAS Burdens (ng PFAS per Individual Insect), Bioamplification Factors (BAMFs), and Adult/Larval Ratio of Σ₁₂PFAS Body Burdens by Aquatic Insect Taxon .

taxa	larval			adult			BAMF	estimated change in Σ12PFAS burden (adult/larval)
	Σ12PFAS concentration (ng g–1 ww)	estimated mass per insect (g ww)	Σ12PFAS burden (ng per individual)	Σ12PFAS concentration (ng g–1 ww)	estimated mass per insect (g ww)	Σ12PFAS burden (ng per individual)
Diptera	5.4 (2.8–12)	0.17 (0.058–0.32)	1.3 (0.52–3.5)	26 (14–57)	0.0010 (4 × 10–4 to 0.0023)	0.032 (0.014–0.092)	4.9 (1.8–13)	0.026 (0.0078–0.088)
Ephemeroptera	41 (21–94)	0.0096 (0.0038–0.026)	0.44 (0.19–1.2)	18 (8.9–42)	0.01 (0.002–0.029)	0.23 (0.096–0.68)	0.43 (0.16–1.2)	0.53 (0.15–1.8)
Odonata	39 (21–83)	0.1 (0.031–0.22)	5.2 (2.3–13)	5.9 (2.1–14)	0.084 (0.044–0.19)	0.5 (0.15–1.6)	0.15 (0.044–0.4)	0.095 (0.021–0.38)
Plecoptera	51 (27–120)	0.014 (0.0021–0.085)	0.59 (0.24–1.8)	62 (33–130)	0.017 (0.0024–0.093)	0.33 (0.15–0.8)	1.2 (0.47–3.2)	0.56 (0.13–2)
Trichoptera	22 (14–40)	0.067 (0.016–0.15)	1.7 (0.8– 4.1)	43 (26–84)	0.009 (0.0022–0.022)	0.49 (0.23–1.2)	2 (0.97–4.1)	0.29 (0.12–0.68)

^aCompound-specific concentrations are reported in Table S12.

Aquatic insect metamorphosis involves coupled changes in contaminant concentrations and aquatic insect biomass that collectively determine whether contaminant burdens (contaminant mass per individual) increase, decrease, or remain relatively consistent. Metamorphic changes in contaminant concentrations are associated with tissue catabolism and reorganization, including redistribution of lipids and proteins to which PFAS bind, ,,, whereas metamorphic changes in insect biomass reflect that aquatic insects can lose as much as 80% of their body mass during the transition from larval to adult life stages. We first assessed changes in PFAS concentrations attributed to metamorphosis by comparing Σ₁₂PFAS and compound-specific concentrations in larval and adult aquatic insects across taxa (i.e., at the community level). Σ₁₂PFAS concentrations were higher in adult aquatic insects (posterior median = 46 ppb, 95% credible interval = 27–93 ppb) than in larval aquatic insects (30 ppb, 21–46 ppb) with a Σ₁₂PFAS metamorphic bioamplification factor (BAMF) of 1.5 (95% CI = 0.77–3.3) (Figure a, Tables S9 and S10). These Σ₁₂PFAS patterns agreed with compound-specific BAMFs, ranging from 1.1 to 2.6 (Table S10). We estimated the greatest metamorphic enrichment for PFHxA (posterior median BAMF = 2.6, 95% CrI = 0.34–8.7), PFNA (2.1, 1.0–4.0), PFHxS (2.1, 0.21–7.2), and PFUnA (2.0, 0.77–4.1) (Figure S3, Table S10). As with Σ₁₂PFAS concentrations, Σ₁₂PFAS metamorphic enrichment patterns varied among aquatic insect taxa. Σ₁₂PFAS BAMFs ranged from 0.15 to 4.9 such that posterior median Σ₁₂PFAS concentrations decreased from larval to adult life stages of Odonata and Ephemeroptera but increased with metamorphosis in Plecoptera, Trichoptera, and Diptera (Table ). Σ₁₂PFAS BAMF patterns countered Σ₁₂PFAS concentration patterns in that BAMFs were highest in the aquatic insect taxa (Diptera and Trichoptera) with the lowest larval Σ₁₂PFAS concentrations. Differences in PFAS concentration changes among taxa likely reflect variation in physiological processes influencing PFAS toxicokinetics (i.e., absorption, distribution, metabolism, and excretion). , For example, results from this study preliminarily indicate that there are differences in metamorphic PFOS enrichment among taxa, with decreases in the proportional contribution of PFOS to Σ₁₂PFAS from larval to adult Diptera and Plecoptera, which contrast with increasing PFOS contributions for Ephemeroptera, Odonata, and Trichoptera (Figure a).

4.

Differences in PFAS composition and burdens by taxa and life stage. (a) Compound-specific proportional contributions to Σ₁₂PFAS concentrations by aquatic insect taxa, with larval aquatic insects shown on left and adults shown on right. (b) Compound- and taxa-specific adult/larval aquatic insect body burden ratios. Vertical line shows ratio = 1, with distributions falling to the right of the vertical line indicating higher PFAS body burdens in the adult life stage and distributions falling to the left indicating higher body burdens in larval life stages. Density areas are identical, with density width reflecting more (taller) or less (flatter) certain estimates. D: Diptera, E: Ephemeroptera, O: Odonata, P: Plecoptera, T: Trichoptera.

We next assessed the relationships between differences in larval versus adult aquatic insect masses and the PFAS concentration changes described above to estimate PFAS burdens by life stage and their variation among taxa. Estimates of contaminant burden complement those of contaminant concentrations when metamorphosis or other physiological processes involve substantial changes in body size because they reconcile how coupled changes in concentrations and biomass affect the total contaminant mass per individual and inform the contaminant dose (contaminant mass consumed per prey item) for their consumers. , In the case of aquatic insects, changes in contaminant body burdens between larval and adult life stages further inform contaminant distribution among aquatic and neighboring riparian/terrestrial ecosystems.

Σ₁₂PFAS burdens for all taxa were lower on average in adults than in larvae but were also highly uncertain for Ephemeroptera and Plecoptera (i.e., corresponding to probabilities of 82% and 84%, respectively, that concentrations were lower in adults) (Table ), suggesting variation in the consistency of body burden changes among taxa. Similarly, we estimated variable taxon-specific body-to-body burden ratios for PFOA, PFNA, PFUnA, and PFOS (Figure b). We predict from these results that Odonata and Trichoptera are the most efficient PFAS biotransporters (highest Σ₁₂PFAS burdens, ng individual^–1) from aquatic to riparian compartments. In the case of Odonata, high adult life stage mass offset concentration decreases (Σ₁₂PFAS BAMF = 0.15), whereas Σ₁₂PFAS concentrations increased in adult Trichoptera (posterior median BAMF = 2.0, 95% CrI = 0.97 to 4.1) and were counteracted by mass losses via metamorphosis (Table ). This finding highlights the need for assessments of diverse riparian consumers to account for consumer-specific prey preferences. We note that aquatic insects were composited by order and across larval development stages for this study, and we did not achieve family- or species-specific estimates of metamorphic enrichment or changes in PFAS burdens between life stages. Additionally, factors important to the interactions between aquatic insect metamorphosis and contaminant exposure, such as whether larval aquatic insects experience PFAS toxicity in a manner that would influence their metamorphosis and emergence (i.e., emergence magnitude or timing, emergent aquatic insect community composition), , were not included in our assessment. While our results offer initial insights, future studies emphasizing lower levels of biological organization would be needed to advance our understanding of how physiological differences among aquatic insect types influence the potential for metamorphic enrichment and PFAS transfer to riparian food webs.

3.3.1. Insights for Studying PFAS in Food Webs

Environmental PFAS monitoring and regulations emphasize water sampling as the basis for determining whether contaminants represent elevated risks for biota and whether waterbodies are impaired. However, it is increasingly recognized that surface water sampling does not provide sufficient information for predicting the concentrations or bioaccumulation potential of PFAS in aquatic food webs. The high frequency of PFAS non-detections in water sampling contributes to this challenge as there is no concentration basis for predicting concentrations in biota. The use of passive samplers (e.g., polar organic chemical integrative samplers) for surface water sampling may address the issue of non-detections, especially in areas with relatively low concentrations, as they have been shown to accumulate PFAS in a manner that represents water column mixtures and at higher concentrations due to integrating exposures over the deployment period. On the other hand, passive sampler-based analyses offer insufficient predictions of PFAS mixture compositions in biota owing to the importance of dietary pathways for some compounds.

The results of this study, together with those of previous studies, − demonstrate that aquatic food webs alter PFAS concentrations and mixtures and thereby mediate exposures for consumers in both aquatic and terrestrial habitats. We found that PFAS partitioning at the base of the food web is the most substantial, yet variable, concentration enrichment process. This suggests that particulate sampling could provide a more robust basis for predicting PFAS in stream food webs. We explored this further by comparing both the probability of PFAS detection and PFAS concentration relationships among compartments. Compound-specific posterior median probabilities of detection in surface water ranged from 0.11 for PFUnA to 0.94 for PFBS (Table S14, Figure a). Probabilities of detection were higher in surface water than in biofilm, larval aquatic insects, and adult aquatic insects for all compounds except PFUnA and PFOS, and probabilities of detection in spiders were elevated relative to stream biota for all compounds except PFBS. Only PFOS had both high and similar probabilities of detection in surface water, biofilm, larval and adult aquatic insects, and spiders (range in posterior median probabilities of detection = 0.89–0.98, Figure a, Table S14). We also observed poor relationships between PFAS concentrations in water and biofilm for all compounds except PFBS, such that PFBS was unlikely to be detected in biofilm but had concentrations predicted by those in surface water when it was detected (Figure a,b). In summary, surface waters overpredicted detection probabilities at the base of aquatic food webs (i.e., probability of compound detection in water > probability of compound detection in biofilm; Figure a) while underpredicting PFAS concentrations (Figure b), making it a poor predictor of PFAS food web dynamics.

5.

Correlations of PFAS detection probabilities (“P­(Detect”)) and concentrations between water and biofilm (top row) and biofilm and larval aquatic insects (bottom row). Dots represent 1000 draws for each PFAS compound from the posterior distribution of the detection parameter or estimated concentration from the hurdle Gamma model. The dashed line shows the 1:1 relationship.

In contrast, we observed strong predictive relationships between the probabilities of detecting PFAS compounds and compound-specific concentrations in biofilm and larval aquatic insects (Figure c,d). Larval aquatic insect patterns subsequently predicted compound-specific probabilities of detection (Figure S6c) and concentrations (Figure S7c) in adult aquatic insects. Finally, we report that probabilities of PFAS detection in adult aquatic insects underpredicted the probabilities of PFAS detection in riparian spiders (Figure S6d), yet PFAS concentrations in adult aquatic insects overpredicted PFAS concentrations in riparian spiders (Figure S7d). This is consistent with the finding that the Σ₁₂PFAS concentrations decreased from adult aquatic insects to spiders (TTF <1, Figure a, Table S10) and that PFAS mixture compositions varied between these two sample types. These results align with prior research concluding that biofilm sampling provides an integrative measure of PFAS bioaccumulation and exposure potential at the base of aquatic food webs. Despite our conclusion that biofilm is an efficient basis for PFAS monitoring, site-specific variation in the communities of organisms comprising field-collected biofilms may contribute to the context dependency of PFAS dynamics among sites and studies. Additional research would also be needed to disentangle the concentration- or compound-specific influences over PFAS transfer and retention at the interface of stream and riparian food webs.

Data are archived on the US Geological Survey ScienceBase repository at 10.5066/P139C2SN. Associated code is available at https://github.com/jswesner/farmington_pfas.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.est.5c16942.

• Information on PFAS compounds, study sites and map, additional analytical and modeling methods and associated QA/QC, watershed surface water and groundwater methods and results, food web study result tables, and supporting figures (PDF)

¶.

State of Washington Department of Ecology, Lacey, WA 98503, USA

This work was supported by the U.S. Geological Survey 104­(b) Program (Project #CT_2021_Brandt) and U.S. Geological Survey Grant G22AP00016-01 (Project # 2021CT003PFAS).

The authors declare no competing financial interest.