Determining global background soil PFAS loads and

Science of the Total Environment 651 (2019) 2444–2449

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Determining global background soil PFAS loads and the fluorotelomer-based polymer degradation rates that can account for these loads John W. Washington a,⁎, Keegan Rankin b, E. Laurence Libelo c, David G. Lynch d, Mike Cyterski a,⁎ a

National Exposure Research Laboratory, USEPA, 960 College Station Road, Athens, GA 30605, United States of America Department of Chemistry, University of Toronto, 80 St. George Street, Toronto, Ontario M5S 3H6, Canada c Office of Land and Emergency Management, USEPA, Washington, DC 20460, United States of America d Office of Chemical Safety and Pollution Prevention, USEPA, Washington, DC 20460, United States of America b

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• Here we study perfluoroalkyl substances (PFASs) in soils • We report global background estimates for soil PFAS concentrations and loads • These estimates show that surface soils constitute a major global reservoir for PFASs • Soil & ocean PFAS loads exceed existing emissions load estimates for some PFASs • The global load discrepancy for C10 & C12 can be explained by polymer degradation

a r t i c l e

i n f o

Article history: Received 14 August 2018 Received in revised form 1 October 2018 Accepted 5 October 2018 Available online 10 October 2018 Editor: Jay Gan Keywords: PFAS Background soil concentration Global load Fluorotelomer-based polymer

a b s t r a c t In recent years, fluorotelomer-based polymers (FTPs) have been the dominant product of the fluorotelomer industry. For the last decade, whether FTPs degrade to toxic perfluorocarboxylates (PFCAs) has been vigorously contested, with early studies arguing that FTPs have half-lives N1000 years, and others concluding decadal half-lives. Given this FTP half-life discrepancy of 10- to N100-fold, here we investigate whether environmental loads of long-chain PFCAs might offer an independent approach to assess FTP half-lives. Specifically we: i) use surface soil-PFCA data to estimate terrestrial surface-soil background PFCA concentrations and loads; ii) extrapolate these data to generate global PFCA load estimates; iii) compare these estimates to published ocean-derived and industrial-emissions load estimates, finding agreement for perfluorooctanoate (C8), but an excess in longerchain (C10,C12) PFCAs for ocean- and soil-derived loads relative to emissions; iv) model FTP degradation rates required to reconcile this discrepancy; and iv) compare our modeled estimates to existing experimental results. These findings show agreement for FTP half-lives at the decades-scale supporting existing laboratory studies that report decade-scale half-lives for FTPs. This suggests that global long-chain PFCA loads will increase for decades if legacy FTPs already manufactured are not contained upon disposal. These results suggest that FTPs comprised of novel poly- and perfluorinated alkyl substances (PFASs) now in production might constitute considerable sources to the environment of the new generation of PFASs. Published by Elsevier B.V.

⁎ Corresponding authors. E-mail addresses: [email protected] (J.W. Washington), [email protected] (M. Cyterski).

https://doi.org/10.1016/j.scitotenv.2018.10.071 0048-9697/Published by Elsevier B.V.

1. Introduction

J.W. Washington et al. / Science of the Total Environment 651 (2019) 2444–2449

Perfluoroalkylcarboxylic acids (PFCAs) are recalcitrant anthropogenic compounds that are globally distributed in the environment including remote oceanic and terrestrial locations (Wang et al., 2014a, 2014b; Rankin et al., 2015). Some of these compounds are bioaccumulative (Conder et al., 2008), toxic in animal models (Lau et al., 2007; Lau, 2012), prevalently detected in humans (Kato et al., 2015; CDC, 2018), and implicated in a number of adverse effects in humans, in both general and PFCA-elevated populations (Johnson et al., 2014; NTP, 2016). Given these combined traits of persistence, ubiquitous distribution and toxicity, multiple researchers have sought to estimate global environmental loads of these compounds based on industry-emissions or oceanic inventories (Armitage et al., 2006; Prevedouros et al., 2006; Armitage et al., 2009; Cousins et al., 2011; Wang et al., 2014a, 2014b). Whereas there has been reasonable agreement for C8 and C9 estimates, for shorter (C6, C7) and longer (C10– C12) homologues, existing oceanic-derived estimates grossly exceed emission estimates, with no overlap in estimated ranges (Wang et al., 2014a, 2014b). Whether these discrepancies arise from uncertainties in oceanic estimates (e.g., mixing depth in surface waters, retention in surface waters, treatment of non-detects) or industry emissions (e.g., factory-release rates, unaccounted-for degradation of FTPs) has remained unresolved. A recent global study of poly- and perfluoroalkyl substances (PFASs) in soils, using internally consistent sampling, extraction and analytical procedures, reported quantifiable levels in every survey sample, including remote locations on every continent and numerous islands (Rankin et al., 2015). Based on PFCA homologue ratios, Rankin et al. (2015) concluded that the dominant mode of occurrence in remote terrestrial locations is long-range transport of volatile PFCA precursors in the troposphere. Here we use these data to: i) estimate background surface-soil PFCA concentrations and loads, for the first time as far as we know; ii) extrapolate our soil loads and existing oceanic loads to global area to compare with existing emission-load estimates; and iii) explore whether discrepancies among load estimates for C10 and C12 might be resolved by accounting for degradation of commercial fluorotelomer-based polymers (FTPs). 2. Methods To estimate loads from sample surveys, whether and how to include low- and high-concentration data requires careful deliberation. Regarding low concentrations, some value must be placed on non-detects. For their ocean inventory, Wang et al. (2014a, 2014b) used statistical procedures to calculate 95% upper confidence-limit concentrations that varied depending on the fraction of non-detects. For Rankin et al. (2015), detected data fit log-normal distributions (SI2, Figs. S1–S4), so we used extrapolation (in log concentration vs Z-score space) to estimate nondetects following the general approach of Cohen (Cohen, 1959; USEPA, 1992) (SI2, Table S3). High concentration data present a potential risk for load estimates. If these values reflect local or regional sources and the database is insufficiently dense to assure unbiased representation of these inhomogeneities, then load estimates might be skewed high or low by over- or under-representation of these high values. In the case of Rankin et al., high PFCA values of the United States were overly represented relative to the US fraction of global land area. In contrast, low concentrations in Antarctica and other areas with few or no local sources were less densely represented. To avoid bias, high values reflecting regional sources can be culled from datasets, so that resulting load estimates reflect ‘background’ conditions having lesser variability introduced from local sources, as opposed to ‘total’ load. Toward this end, Wang et al. manually selected “samples that were not influenced by coastal water that was potentially contaminated by PFCAs from local sources.” For the Rankin et al. (2015) data, probability plots of untransformed PFAS data yielded inflection points (SI2; Figs. S5–S12), suggesting more

2445

than one source of variation in the data (Sinclair, 1974). Taking high PFAS concentrations as reflecting local or regional sources and assuming long-range tropospheric mixing would yield an approximately normal distribution, we evaluated the normality of each dataset stepwise, discarding the highest concentration at each step until the remaining samples fit a normal distribution for each compound (SI2; Figs. S5– S12). Noting that some volatile PFCA precursors appear to be lower in the southern hemisphere than the northern (Wang et al., 2014a, 2014b), we tested these background data for evidence of hemispheric differences, finding statistically significant differences for C9, C11 and C12, but not for the other compounds. We took the mean of these normally distributed subsets (segregated by hemisphere for C9, C11 and C12) as reflecting the ‘global background.’ The 95% confidence intervals of the background datasets were estimated by bootstrapping (Efron and Tibshirani, 1993) 10,000 simulations of the full datasets; identifying the normally distributed subset for each bootstrap sample (SI2). Global background surface-soil loads were estimated from the background soil concentrations by multiplying by a typical surface-soil bulk density of 1.35 g/cm3 (Foth, 1978), a sampling depth of 10 cm (Rankin et al., 2015), and Earth's land area, 1.5 × 108 km2. For the three PFCAs that tested as having hemispheric differences (C9, C11, C12), background loads were calculated by hemisphere and summed. In turn, seeking to elucidate possible causes for discrepancies in existing global load estimates, we used the surface-soil loads to calculate independent estimates of global background total loads by dividing by 0.3, the fraction of Earth's surface composed of land (SI3). For previously reported ocean-load estimates, to extrapolate to total global area, we divided by 0.7, the fraction of Earth's surface composed of water. There are several obvious assumptions and limitations to both the surface-soil and ocean estimates that will be discussed below.

3. Results and discussion Our estimates of global background surface-soil concentrations are summarized in Table 1. The coefficients of variation for these background data all are less than unity, consistent with an absence of local PFAS sources. Supporting the global distribution of these background subsets, eight samples passing our statistical normality filter (for defining background) for all analytes included sample locations from every continent (SI2; Table S4). Comparing among the analytes, C6, C7, C8 and PFOS all are N2-fold higher than C9 through C12 (Tables 1 and S6). This variation among analyte background concentrations also manifests in global background surface-soil loads, with the lower 95% confidence limits for C6 and C8 exceeding the upper 95% confidence limits for C9 through C12 (Fig. 1; Table S7). So far as we know, these are the first estimates of background surface-soil concentrations (Table 1) and loads (Fig. 1; Table S7) for PFASs in peer-reviewed literature. Compared to existing background ocean load estimates, these background soil load estimates are expected to be robust because: i) the sample sets were drawn entirely from a single study in which remote, uncontaminated soils were the objective (Rankin et al., 2015); ii) single, consistent extraction and analytical methods were used to generate all data; iii) every continent is represented (Table S4); iv) background values were culled from regionally impacted samples statistically (Figs. S5–S12); v) each analyte was evaluated using 19 to 36 samples; vi) there were only 0 to 2 non-detect estimates for each analyte except C12 which was composed of 67% detected values (Figs. S1–S4); and vii) non-detects were estimated with a central tendency approach. In contrast, for ocean load estimates: i) Prevedouros et al. (2006) estimated PFOA load based on limiting values for two reported ranges (Yamashita et al., 2005), comprised of a total of 21 samples; ii) Wang et al. (2014a, 2014b) collected data from a number of sources, likely using a variety of extraction and analytical methods; iii) Wang's dataset commonly had ~50% or more nondetects; and iv) Wang's loads were estimated from these heavily

2446

J.W. Washington et al. / Science of the Total Environment 651 (2019) 2444–2449

Table 1 Background surface-soil concentration summary statistics.a,b Summary statistic

Compound

Normally distributed subset count (n) Global mean (pg/g) Global standard deviation (pg/g) Global coefficient of variation (COV) Northern Hemisphere subset count (n) Northern Hemisphere mean (pg/g) Northern Hemisphere standard deviation (pg/g) Northern Hemisphere COV Southern Hemisphere subset count (n) Southern Hemisphere mean (pg/g) Southern Hemisphere standard deviation (pg/g) Southern Hemisphere COV a b

C6

C7

C8

C9

C10

C11

C12

S8

31 58.3 37.1 0.64

33 54.7 38.7 0.71

19 47.1 28.3 0.60

23

29 11.4 7.7 0.67

26

36

29 55.5 38.1 0.69

19 9.63 4.66 0.48 7 3.57 3.84 1.07

28 8.95 5.18 0.58 8 3.32 1.59 0.48

15 17.9 8.7 0.49 8 8.05 5.44 0.68

These are analytical data, not bootstrapped. Bootstrapping values are similar (Table S6). Whereas PFOS (S8) is extraneous, because the data were available and due to its prominent history among PFASs, we include it here.

Global Bkgd Surface-Soil Load (metric tons)

censored data by an algorithm that estimates the upper limit of the 95% confidence interval, perhaps drawing load estimates upward. Soil- and ocean-based total load estimates (extrapolated to Earth's surface area) are summarized, alongside literature emission-load estimates, in Fig. 2 (and Table S8). Comparing ocean- (from Wang et al. (2014a, 2014b)) and soil-derived (from Rankin et al. (2015) data) estimates, C6 and C7 are reasonably consistent, with the central (median for Wang, mean of laboratory data for Rankin; Table S6) estimates falling within 30% of each other. For C8, the soil-based central tendency estimate is only 42% of the ocean-based estimate, but the estimate ranges overlap substantially, and the ranges of both studies are fully encompassed by the range reported by Prevedouros et al. (2006) (Fig. 2). For the C9–C12 homologues, the soil-estimate ranges all are lower than the ocean-based estimates, falling between 15 and 41% (Fig. 2). Some of these differences might be related to the soil-based loads being derived from central-tendency concentrations of sample sets comprised mostly of detected values (Table S2), whereas the Wang ocean-based loads represent upper 95% confidence-limit concentrations for sample sets commonly containing ~50% non-detects or more. Additionally, there are likely physical causes for the lower soilbased versus ocean-based loads owing to differences in modes of occurrence. Specifically, several PFCAs (e.g., C8, C9, C11) historically had substantial aqueous releases (Wang et al., 2014a, 2014b) that constitute a fraction of the oceanic loads but not necessarily background soil loads which arise primarily from atmospheric transport of PFCA precursors

3000 2500 2000 1500 1000 500 0 C6

C7

C8

C9 C10 C11 C12 PFAS

S8

Fig. 1. PFAS global background surface-soil loads estimated from the surface-soil data of Rankin et al. (2015). Black dots are mean estimated load, blue are bootstrapped lower 95% confidence limits and red are upper 95% confidence limits. See text, SI2 and SI3 for details. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

(Rankin et al., 2015) (Table S8). Also, the soil samples were collected in 2006–2007 (Rankin et al., 2015) whilst ocean data included samples collected more recently (Wang et al., 2014a, 2014b), a period during which the highest fluorotelomer production was taking place (Wang et al., 2014a, 2014b). The ocean data do not account for losses to the deep, a loss term that recently has been modeled to have a half-life as short as 2 to 5 years (Zhang et al., 2017). Likewise, the soil-based values do not account for leaching losses to the subsurface. While this undoubtedly had some effect, long-chain PFCAs appear to leach through soils markedly less than shorter chains (Washington et al., 2010), a trend that might exacerbate the discrepancies we report here rather than alleviate them. Despite that the soil- and ocean-based total load estimates both bear the artifact of losses to subjacent media, potentially diminishing their resulting load estimates, both environmental-inventory load estimates independently and mutually support that global environmental loads for several homologues grossly exceed their respective emissions load estimates (Fig. 2). Addressing C6, the lower range of both environmental loads exceeds the highest range of any emission load. For C7, the lower limit of the ocean-based load and the central value for the soilbased load far exceed the upper limit of any emission load. Among all PFCAs, estimates of PFOA (C8) loads are most internally consistent among studies, with the ranges of most studies overlapping. The oddchain-lengths C9 and C11 load estimates are unique in that the oceanbased load exceeds all emission loads, whilst the soil-based load falls in the general range of the emission loads, perhaps reflecting historical aqueous releases of these compounds that would not be reflected in the soil-based load (Wang et al., 2014a, 2014b) (Table S8). In contrast, for the even-chained C10 and C12, the lower limits of both environmental loads exceed the highest estimates of any emissions loads. Considering that the soil- and ocean-based global loads are conservatively low estimates (both excluded high-concentration samples and both are subject to mass losses to deeper zones), these C10 and C12 environmentalinventory excesses compared to emission estimates strongly suggest a source unaccounted for in emission loads or a gross underestimate of historical emissions. Consistent with the environmental-load excesses we describe for C6, C10 and C12, Rankin et al. noted that these compounds were elevated in their study soils relative to other homologues, and argued these excesses were evidence of degradative formation from precursors including FTPs (Rankin et al., 2015). For C10 and C12 compounds, the primary product line was fluorotelomer-based polymers (FTPs), constituting ~80% of the fluorotelomer market of all chain lengths (Prevedouros et al., 2006; Buck et al., 2011), and 88% of the C10 and C12 fluorotelomer market (Wang et al., 2014a, 2014b). In commercial FTPs, C10 and C12 typically constitute ~25% and ~9% of the content, respectively (Wang et al., 2014a, 2014b). Here we modeled the FTP-degradation half-lives

J.W. Washington et al. / Science of the Total Environment 651 (2019) 2444–2449

Fig. 2. Global PFCA loads (abbreviated first author and publication year) segregated by emission- vs environmental-based estimates, black dots are central or singular estimates, red are high-end ranges and blue are low. For the environmental estimates, Prev.'06 and Wang'14 estimates were based on ocean data and Rank.'15 were based on soil data. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

(T1/2) necessary to balance the environmental-minus-emission excess loads (SI3) using a simple first-order model: dM ¼ Φ−kM dt

ð1Þ

where M (metric tonnes) is the global mass of commercial FTP at any time t (years), Φ (tonnes/year) is FTP-product manufacturing rate, and k (year−1) is a first-order degradation constant. Integration of Eq. (1) yields (SI4): M¼

Φ−ðΦ−kM0 Þe−kΔt k

ð2Þ

2447

With the objective of determining the FTP T1/2 required to reconcile the environmental-minus-emission excess loads of C10 and C12, M is evaluated as the mass of fluorotelomers required to explain the environmental-minus-emission excess loads of C10 and C12 PFCAs. Production rate, Φ, has been documented through time (Wang et al., 2014a, 2014b) (SI4), so the only unknown in Eq. (2) is the degradation-rate constant, k. Given that Wang generally estimated the largest emission loads (Wang et al., 2014a, 2014b) (Fig. 2), and the smallest environmental loads are from the soils data, with an objective of defining the minimum mass that remains unexplained by quantifiable sources, we conservatively estimated excess environmental-minus-emission loads of C10 and C12 PFCAs by subtracting Wang's (Wang et al., 2014a, 2014b) median emission range from the soils-based estimate we report here (Table S10). Then we put limits on these environmental-minusemissions excess loads by subtracting the upper emission estimate from the lower soils-based estimate, and vice versa, lower emissions from upper soils-based estimate. Next, we calculated FTP T1/2 as a function of modeled C10 and C12 loads, assuming FTPs degrade first to fluorotelomer alcohols (Washington and Jenkins, 2015a, 2015b; Washington et al., 2015) (FTOHs), and conservatively using some of the highest FTOH → PFCA reaction yields reported in the literature of 10% (Ellis et al., 2004) to 25% (Wang et al., 2009) (Fig. 3; SI4). Using PFCA reaction yields of 25% (Wang et al., 2009) led to FTP T1/2 values of 36–60 years (Fig. 3A) and reaction yields of 10% (Ellis et al., 2004) led to T1/2 values of 9–19 years (Fig. 3B). In Fig. 4, we compare our global-load based fluorotelomer T1/2 estimates to literature FTP T1/2 from laboratory studies. Among reported FTP T1/2 values, the earliest (Russell et al., 2008) (Russ.'08) stands alone as an outlier, with subsequent research calling into question extraction procedures among other concerns for this first effort (Washington et al., 2009). The next two studies (Washington et al., 2009; Rankin et al., 2014), Wash.'09 and Rank.'14, were conducted with noncommercial experimental FTPs. The last two efforts were conducted with commercial FTPs (Washington and Jenkins, 2015a, 2015b; Washington et al., 2015); Wash.'15a was a biodegradation experiment and Wash.'15b was by abiotic hydrolysis. The green shaded area in Fig. 4 highlights the T1/2-estimate range bounded by hydrolysis (upper) and biodegradation (lower) for commercial FTPs (based on the reasoning that hydrolysis defines the lower reaction-rate limit). Our load-estimated fluorotelomer T1/2 estimates for a 10% PFCAreaction yield (T1/2 = 9–19 years) are encompassed within the experimental-FTP range (8–111 years). Likewise, our load-estimated fluorotelomer T1/2 estimates for a 25% PFCA-reaction yield (T1/2 = 35– 60 years) fall in the green field, encompassed within the commercialFTP range (33–89 years). To our knowledge, this documentation of reasonable internal consistence between laboratory kinetics experiments reporting rates of generation of organic compounds with global contaminant loads estimated from surveys of environmental samples is rare and perhaps unprecedented for PFAS compounds. Some perspective on assumptions inherent in this model: We attributed the entire excess C10 and C12 background loads to FTP degradation, whereas 12% of these homologues went to other uses (Wang et al., 2014a, 2014b). Counterbalancing this simplification, we defined the excess load relative to global background. This definition excludes elevated PFAS levels present in extensive, albeit ill-defined, areas of North America, Europe and Asia where these PFASs commonly are present at 10- to 100-fold higher than the threshold limit for “background” concentrations (Figs. S5–S12). Also, there is the issue of timing of degradation and release to the environment. Wang et al. assumed constant degradation rates during the use of FTP-bearing products, as do we in this model. In the past, we have shown that commercial FTPs degrade by hydrolysis when applied to a substrate and wetted (Washington and Jenkins, 2015a, 2015b). Whether FTPs degrade under ‘dry’ conditions is not well understood; however, it is noteworthy that in our past FTP-degradation experiments (Washington et al., 2015), the in-

2448

J.W. Washington et al. / Science of the Total Environment 651 (2019) 2444–2449

Fig. 4. Commercial FTP T1/2 from literature laboratory studies (abbreviated first author and publication year) compared to load-based T1/2 estimated here. Among reported T1/2 values, the earliest (Russ.'08), stands alone as an outlier. The next two studies, Wash.'09 and Rank.'14, were conducted with noncommercial experimental FTPs. The last two were conducted with commercial FTPs - Wash.'15a was a biodegradation experiment and Wash.'15b was by abiotic hydrolysis. The green shaded field covers the range bounded by hydrolysis (upper) and biodegradation (lower) for commercial FTPs. The central T1/2 load estimates for a 10% PFCA reaction yield correspond closely with the lab studies using experimental FTPs, Wash.'09 and Rank.'14 (Fig. S18). The central 25%-yield estimates fall in the T1/2 range reported for commercial FTPs, Wash.'15a and Wash.'15b. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 3. FTP T1/2 calculated as a function of modeled global loads of C10 (green) and C12 (red) for a FTP → PFCA reaction yield of 25% (3A) and 10% (3B). Dashed curves are upper and lower limits. Horizontal lines define laboratory FTP T1/2 range by hydrolysis (Washington and Jenkins, 2015a, 2015b) (blue) and lower biodegradation limit (Washington et al., 2015) (orange). In 3A the load estimates for FTP T1/2 of 36 (C12) and 60 (C10) years agree well with laboratory commercial-FTP studies for biodegradation (T1/2 = 33 to 112 yr) and hydrolytic degradation (T1/2 = 55 to 89 yr). In 3B, the best load estimates for FTP T1/2 of 9 yr (C12) and 19 yr (C10), fall below the laboratory commercial-FTP studies. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

use controls (FTP dried on cotton, capped in air-filled tubes subject to ambient relative humidity; Table S16) had ~300% more 8:2FTOH and 10:2FTOH on day 376 than on day 63, suggesting the possibility of FTP hydrolysis by sorbed water. Wang et al. assumed no release to the environment after proper disposal in landfills (Wang et al., 2014a, 2014b). However, it has become clear landfills emit considerable amounts of volatile PFASs to the atmosphere (Ahrens et al., 2011; Weinberg et al., 2011; Hamid et al., 2018) and dissolved PFASs in leachate (Gallen et al., 2016; Lang et al., 2016; Fuertes et al., 2017; Lang et al., 2017) which commonly are inefficiently removed during conventional wastewater treatment (Pan et al., 2016). Given all this, for this simple model, we effectively assumed constant degradation and release to the environment before and after disposal. Few other studies have sought specifically to elucidate the role of polymer degradation in global loads of PFCAs and related compounds. Among them, Li et al. (2017) estimated the potential contributions of FTP degradation to global loads of PFOA. Using FTP production records and conservative input assumptions (e.g., no

FTP degradation before disposal, FTOH → PFCA product yield of 5%, T1/2 ≥ 75 yr), Li et al. estimated PFOA loads from FTP degradation of 280–350 tonnes by 2040. This C8 load from FTPs is less than the excess C10 load we estimated of ~700 tonnes (Table S10). Fitting to both FTP-production-record and excess-load constraints, and using high but realistic reaction yields, we solved for the longest FTP half-lives that explain the observed excess loads, T1/2 = 9 to 60 years. Even given the considerable differences in inputs between these studies, both the resulting PFCA-load and FTP half-lives fall reasonably close to each other. Seeking to assess the potential impact of abiotic FTP hydrolysis on global PFCA loads, Washington and Jenkins (2015a, 2015b) estimated future PFCA loads up to 8-fold higher than present. Considered pragmatically, these studies, as well as a study addressing related sulfonamide/sulfonamido sidechain polymers (Wang et al., 2017), all suggest FTPs constitute a considerable potential source to global loads of perfluorinated compounds far into the future. Fluorotelomer manufacturers in the U.S., Europe and Japan voluntarily agreed with the EPA to stop manufacture of long-chain fluorotelomer products by 2015 so emissions of C10 and C12 are expected to slow in the coming years (Li et al., 2017). However, the FTP-degradation modeling results reported herein, which we show to be internally consistent with laboratory FTP-degradation experiments using all conservative assumptions (SI4), suggest that legacy FTPs will continue to degrade to form long-chain PFCAs and related compounds for decades to centuries into the future, consistent with the conclusions of other studies (Washington and Jenkins, 2015a, 2015b; Li et al., 2017; Wang et al., 2017). The degree to which FTPs disposed in landfills will release PFCAs and related compounds to the environment, either as volatile FTOHs and other intermediates in gas emissions, or soluble PFCAs and related charged compounds in treated leachate, still remains only loosely resolved (Li et al., 2017). The potential magnitude of these emissions is several-fold current global loads (Washington and Jenkins, 2015a, 2015b), so this issue is worthy of continued study and vigilant scrutiny. Acknowledgements This research was funded by the U.S. Environmental Protection Agency. The views expressed in this article are those of the authors and do not necessarily represent the views or policies of the U.S. Environmental Protection Agency.

J.W. Washington et al. / Science of the Total Environment 651 (2019) 2444–2449

Appendix A. Supplementary information Supplementary data to this article can be found online at https://doi. org/10.1016/j.scitotenv.2018.10.071. References Ahrens, L., Shoeib, M., Harner, T., Lee, S.C., Guo, R., Reiner, E.J., 2011. Wastewater treatment plant and landfills as sources of polyfluoroalkyl compounds to the atmosphere. Environ. Sci. Technol. 45 (19), 8098–8105. Armitage, J., Cousins, I.T., Buck, R.C., Prevedouros, K., Russell, M.H., MacLeod, M., Korzeniowski, S.H., 2006. Modeling global-scale fate and transport of perfluorooctanoate emitted from direct sources. Environ. Sci. Technol. 40 (22), 6969–6975. Armitage, J.M., MacLeod, M., Cousins, I.T., 2009. Comparative assessment of the global fate and transport pathways of long-chain perfluorocarboxylic acids (PFCAs) and perfluorocarboxylates (PFCs) emitted from direct sources. Environ. Sci. Technol. 43 (15), 5830–5836. Buck, R.C., Franklin, J., Berger, U., Conder, J.M., Cousins, I.T., de Voogt, P., Jensen, A.A., Kannan, K., Mabury, S.A., van Leeuwen, S.P.J., 2011. Perfluoroalkyl and polyfluoroalkyl substances in the environment: terminology, classification, and origins. Integr. Environ. Assess. Manag. 7 (4), 513–541. CDC, 2018. Fourth National Report on Human Exposure to Environmental Chemicals; Updated Tables, March 2018. C. f. D. Control. Atlanta, GA. One: 841. Cohen, A.C., 1959. Simplified estimators for the normal distribution when samples are singly censored or truncated. Technometrics 1 (3), 217–237. Conder, J.M., Hoke, R.A., De Wolf, W., Russell, M.H., Buck, R.C., 2008. Are PFCAs bioaccumulative? A critical review and comparison with regulatory lipophilic compounds. Environ. Sci. Technol. 42 (4), 995–1003. Cousins, I.T., Kong, D., Vestergren, R., 2011. Reconciling measurement and modelling studies of the sources and fate of perfluorinated carboxylates. Environ. Chem. 8 (4), 339–354. Efron, B., Tibshirani, R., 1993. An Introduction to the Bootstrap. Chapman & Hall/CRC, Boca Raton, FL. Ellis, D.A., Martin, J.W., De Silva, A.O., Mabury, S.A., Hurley, M.D., Andersen, M.P.S., Wallington, T.J., 2004. Degradation of fluorotelomer alcohols: a likely atmospheric source of perfluorinated carboxylic acids. Environ. Sci. Technol. 38 (12), 3316–3321. Foth, H.D., 1978. Fundamentals of Soil Science. NY, John Wiley & Sons, New York. Fuertes, I., Gomez-Lavin, S., Elizalde, M.P., Urtiaga, A., 2017. Perfluorinated alkyl substances (PFASs) in northern Spain municipal solid waste landfill leachates. Chemosphere 168, 399–407. Gallen, C., Drage, D., Kaserzon, S., Baduel, C., Gallen, M., Banks, A., Broomhall, S., Mueller, J.F., 2016. Occurrence and distribution of brominated flame retardants and perfluoroalkyl substances in Australian landfill leachate and biosolids. J. Hazard. Mater. 312, 55–64. Hamid, H., Li, L.Y., Grace, J.R., 2018. Review of the fate and transformation of per- and polyfluoroalkyl substances (PFASs) in landfills. Environ. Pollut. 235, 74–84. Johnson, P.I., Sutton, P., Atchley, D.S., Koustas, E., Lam, J.C.W., Sen, S., Robinson, K.A., Axelrad, D.A., Woodruff, T.J., 2014. The Navigation Guide - evidence-based medicine meets environmental health: systematic review of human evidence for PFOA effects on fetal growth. Environ. Health Perspect. 122, 1028–1039. Kato, K., Ye, X., Calafat, A.M., deWitt, J.C., 2015. PFASs in the General Population. Toxicological Effects of Perfluoroalkyl and Polyfluoroalkyl Substances. Humana Press, Switzerland, pp. 51–76. Lang, J.R., Allred, B.M., Peaslee, G.F., Field, J.A., Barlaz, M.A., 2016. Release of per- and polyfluoroalkyl substances (PFASs) from carpet and clothing in model anaerobic landfill reactors. Environ. Sci. Technol. 50 (10), 5024–5032. Lang, J.R., Allred, B.A., Field, J.A., Levis, J.W., Barlaz, M.A., 2017. National estimate of perand polyfluoroalkyl substance (PFAS) release to U.S. municipal landfill leachate. Environ. Sci. Technol. 51 (4), 2197–2205. Lau, C., 2012. In: Luch, A. (Ed.), Perfluorinated Compounds. Molecular, Clinical and Environmental Toxicology. Springer, Basel, Switzerland. Lau, C., Anitole, K., Hodes, C., Lai, D., Pfahles-Hutchens, A., Seed, J., 2007. Perfluoroalkyl acids: a review of monitoring and toxicological findings. Toxicol. Sci. 99 (2), 366–394.

2449

Li, L., Liu, J., Hu, J.B., Wania, F., 2017. Degradation of fluorotelomer-based polymers contributes to the global occurrence of fluorotelomer alcohol and perfluoroalkyl carboxylates: a combined dynamic substance flow and environmental fate modeling analysis. Environ. Sci. Technol. 51, 4461–4470. NTP, 2016. National Toxicity Program Monograph. Systematic Review of Immunotoxicity Associated With Exposure to Perfluorooctanoic Acid (PFOA) or Perfluorooctane Sulfonate (PFOS). 140. US Department of Health and Human Services, Research Triangle Park, NC. Pan, C.G., Liu, Y.S., Ying, G.G., 2016. Perfluoroalkyl substances (PFASs) in wastewater treatment plants and drinking water treatment plants: removal efficiency and exposure risk. Water Res. 106, 562–570. Prevedouros, K., Cousins, I.T., Buck, R.C., Korzeniowski, S.H., 2006. Sources, fate and transport of perfluorocarboxylates. Environ. Sci. Technol. 40 (1), 32–44. Rankin, K., Lee, H., Tseng, P.J., Mabury, S.A., 2014. Investigating the biodegradability of a fluorotelomer-based acrylate polymer in a soil-plant microcosm by indirect and direct analysis. Environ. Sci. Technol. 48, 12783–12790. Rankin, K., Mabury, S.A., Jenkins, T.M., Washington, J.W., 2015. A North American and global survey of perfluoroalkyl substances in surface soils: distribution patterns and mode of occurrence. Chemosphere 161, 333–341. Russell, M.H., Berti, W.R., Szostek, B., Buck, R.C., 2008. Investigation of the biodegradation potential of a fluoroacrylate polymer product in aerobic soils. Environ. Sci. Technol. 42 (3), 800–807. Sinclair, A.J., 1974. Selection of threshold values in geochemical data using probability graphs. J. Geochem. Explor. 3, 129–149. USEPA, 1992. Statistical Analysis of Ground-water Monitoring Data at RCRA Facilities: Addendum to Interim Final Guidance. USEPA, Office of Solid Waste, Washington, D.C. Wang, N., Szostek, B., Buck, R.C., Folsom, P.W., Sulecki, L.M., Gannon, J.T., 2009. 8-2 Fluorotelomer alcohol aerobic soil biodegradation: pathways, metabolites, and metabolite yields. Chemosphere 75 (8), 1089–1096. Wang, Z., Cousins, I.T., Scheringer, M., Buck, R.C., Hungerbuehler, K., 2014a. Global emission inventories for C-4–C-14 perfluoroalkyl carboxylic acid (PFCA) homologues from 1951 to 2030, part I: production and emissions from quantifiable sources. Environ. Int. 70, 62–75. Wang, Z., Cousins, I.T., Scheringer, M., Buck, R.C., Hungerbuehler, K., 2014b. Global emission inventories for C-4–C-14 perfluoroalkyl carboxylic acid (PFCA) homologues from 1951 to 2030, part II: the remaining pieces of the puzzle. Environ. Int. 69, 166–176. Wang, Z., Boucher, J.M., Scheringer, M., Cousins, I.T., Hungerbuehler, K., 2017. Toward a comprehensive global emission inventory of C4–C10 perfluoroalkanesulfonic acids (PFSAs) and related precursors: focus on the life cycle of C8-based products and ongoing industrial transition. Environ. Sci. Technol. 51, 4482–4493. Washington, J.W., Jenkins, T.M., 2015a. Abiotic hydrolysis of fluorotelomer-based polymers as a source of perfluorocarboxylates at the global scale. Environ. Sci. Technol. 49 (24), 14129–14135. Washington, J.W., Jenkins, T.M., 2015b. Abiotic hydrolysis of fluorotelomer polymers as a source of perfluorocarboxylates at the global scale. Environ. Sci. Technol. 49, 14129–14135. Washington, J.W., Ellington, J.J., Jenkins, T.M., Evans, J.J., Yoo, H., Hafner, S.C., 2009. Degradability of an acrylate-linked, fluorotelomer polymer in soil. Environ. Sci. Technol. 43 (17), 6617–6623. Washington, J.W., Yoo, H., Ellington, J.J., Jenkins, T.M., Libelo, E.L., 2010. Concentrations, distribution, and persistence of perfluoroalkylates in sludge-applied soils near Decatur, Alabama, USA. Environ. Sci. Technol. 44 (22), 8390–8396. Washington, J.W., Jenkins, T.M., Rankin, K., Naile, J.E., 2015. Decades-scale degradation of commercial, side-chain, fluorotelomer-based polymers in soils & water. Environ. Sci. Technol. 49 (2), 915–923. Weinberg, I., Dreyer, A., Ebinghaus, R., 2011. Landfills as sources of polyfluorinated compounds, polybrominated diphenyl ethers and musk fragrances to ambient air. Atmos. Environ. 45, 935–941. Yamashita, N., Kannan, K., Taniyasu, S., Horii, Y., Petrick, G., Gamo, T., 2005. A global survey of perfluorinated acids in oceans. Mar. Pollut. Bull. 51 (8–12), 658–668. Zhang, X., Zhang, Y., Dassuncao, C., Lohmann, R., Sunderland, E.M., 2017. North Atlantic deep water formation inhibits high Arctic contamination by continental perfluorooctane sulfonate discharges. Glob. Biogeochem. Cycles 31, 1332–1343.