Leachate emissions of short-and long-chain per-and poly ﬂ uoralkyl substances (PFASs) from various Norwegian land ﬁ lls †

Restrictions on the use of long-chain per-and poly ﬂ uoralkyl substances (PFASs) has led to substitutions with short-chain PFASs. This study investigated the presence of four short-chain PFASs and twenty-four long-chain PFASs in leachate and sediment from ten Norwegian land ﬁ lls, including one site in Svalbard, to assess whether short-chain PFASs are more dominant in leachate. PFASs were detected in all sites. Short-chain PFASs were major contributors to the total PFAS leachate concentrations in six of ten land ﬁ lls, though not in Svalbard. In sediment, long-chain PFASs such as per ﬂ uorooctanesulfonate (PFOS) and PFOS-precursors were dominant. Short-chain PFAS leachate concentrations ranged from 68 to 6800 ng L (cid:1) 1 (mean: 980 (cid:3) 1800; median: 360 ng L (cid:1) 1 ), whereas long-chain concentrations ranged from 140 to 2900 ng L (cid:1) 1 (mean: 530 (cid:3) 730; median: 290 ng L (cid:1) 1 ). Sediment concentrations, which contained mainly long-chain PFASs, ranged from 8.5 to 120 m g kg (cid:1) 1 (mean: 47 (cid:3) 36; median: 41 m g kg (cid:1) 1 ). National X 28 PFAS release from Norwegian land ﬁ lls to the environment was estimated to be 17 (cid:3) 29 kg per year (median: 6.3 kg per year), which is in the same range as national emissions from the US, China and Germany after normalizing the data to a per capita emission factor (3.2 (cid:3) 5.5 mg per person per year). Results from this study are compared with previous and current studies in other countries, indicating a general trend that short-chain PFASs are dominating over long-chain PFASs in land ﬁ ll leachate emissions.


Introduction
Per-and polyuoroalkyl substances (PFASs) in the environment are of concern because of their general persistence in combination with either potential aquatic mobility, long range transport, bioaccumulation, toxicity, or some combination thereof. 1 PFASs have been used in a variety of industrial processes and in commercial products over the past 60 years, 2,3 due to their unique chemical and physical properties, such as their thermal and chemical stability as well as their hydrophobic/lipophobic behavior. 2,[4][5][6] Regulations of PFAS in recent years 3,7,8 have led to restrictions of peruorooctanesulfonate (PFOS), peruorooctanoate (PFOA) and other so-called "long-chain" PFASs. 9,10 This has resulted in these being replaced on the market with alternatives including so-called "short-chain" PFASs, such as per-uorobutanesulfonate (PFBS), which are considered to be less bioaccumulative. [9][10][11] Short-chain PFASs are dened by Buck et al. (2011) 10 and the Organisation for Economic Cooperation and Development (OECD) as peruoroalkyl carboxylic acids (PFCAs) with a chain length of <C 7 , and <C 6 for peruoroalkane sulfonates (PFSAs). 9,10 Though shortchain PFASs are less bioaccumulative, they are of concern because of their environmental persistence and aquatic mobility, 1 and because currently little is known about their (eco)toxicity. 12 As such, it is important to investigate to what extent these short-chain PFASs are being emitted to and occurring in the environment.
Landll leachate from municipal solid waste are potential emission hotspots for PFASs, [13][14][15][16] though concentrations vary widely. [13][14][15]17,18 There are few available data on PFAS occurrence in Norwegian landll leachates; 19,20 though the rst available data, comprising data from ten landlls from 2003-2007 based on non-target screening, indicated short-chain PFASs were infrequently detected (or analysed for (ref. 21)), and long-chain PFASs clearly dominated. 20 Studies from North American land-lls from 2006 to 2009 reported that specic short-chain PFASs were amongst the most dominant in landlls, 13,15 and this has also been reported in more recent studies from elsewhere such as China, 22 Germany, 17 Sweden, 23 and Spain. 24 It is therefore anticipated based on this shi of usage and data from other countries, that short-chain PFASs are emitted to a greater extent than long-chain PFASs from Norwegian landll leachate.
The aims of this study are three-fold. The rst is to characterize the aqueous (leachate) concentration and composition of PFASs being emitted from a diverse array of Norwegian landlls, alongside PFAS composition in leachate sediment. The second aim is, based on this data, to test the hypothesis that short-chain PFASs dominate landll emissions over longchain PFASs. The third aim is to compare these results with other regions, through the derived national per capita emission factors of short and long-chain PFASs. For this study, leachate and sediment was collected from ten diverse Norwegian landlls, including one in Svalbard, and analysed for PFASs. The collected data is further discussed in terms of the distribution of PFASs between collected sediment and water leachate, the inuence of leachate characteristics, including pH, electrical conductivity (EC), leachate ow rates, dissolved organic carbon (DOC) and meteorological data (24 h and 2 week precipitation).

Site descriptions
Ten diverse Norwegian landlls, including one site in Svalbard (Table S1, (ESI †)), receiving primarily municipal solid waste (MSW) and in some cases industrial waste and contaminated soil and sewage sludge, were included. Yearly leachate volumes for each landll are given in Table S2. † Due to condentiality reasons, the identities of the landlls are anonymized. Some of the landlls were established in a period where there was few or no requirements for liners, leachate drainage nor control of the inuence of storm water and groundwater dilution of the raw leachate. However, it is our opinion that the selected landlls represent typical Norwegian, and that unpolluted storm water and groundwater have relatively limited inuence on the sampled raw leachate. A description of landlls, the sampling points, their hydrology and whether dilution by storm water/groundwater can be a consideration is presented in the ESI. †

Sampling
As an aim of this study was to characterize PFAS landll emissions and sediment concentrations from all of Norway, available resources were used to sample from many locations rather than obtaining replicates or time series in individual locations. Sampling at each landll was conducted mainly between April to June 2018 by landll operators following sampling protocols and equipment provided by the authors. Sampling dates are given in Table S2. † Sampling of leachate (0.5 L, HDPE bottles, Eurons Environment, Norway) was conducted as close as possible to where leachate leaves the landll (either a retention pond, borehole, pumping station, stream, or underground culvert/leachate pipe accessed by a manhole; see Table S2, ESI †), in order to be representative of landll emissions. Not all samples are considered raw leachate, but raw leachate diluted by storm water and groundwater (see the ESI †).
Sediment samples (500 g, Rilsan bags, Eurons Environment, Norway) were collected from sedimentation ponds, if present, or by sandtraps in the underground culverts. The samples were placed in coolers with cooling elements and bubble wrap and shipped generally overnight and stored cold (ca. 4 C) until analysis. To compliment this data, existing, recent PFAS data was also included when possible, provided by the site owners. These were obtained for four of the landlls from the same sampling points using similar protocols and analysis laboratories as in this study.

Statistical analysis
Statistical analysis was performed with Statistica v. 13.1 (©1984-2016 by Statso, Tulsa, USA). All concentrations were Box-Cox transformed prior to statistical analysis. Physical-chemical properties (pH, DOC, EC and precipitation) were not transformed. Pearson product moment was used for testing for signicant correlations. The signicance level was set at p ¼ 0.05. Principal component analysis was performed exploratively to check for correlations.

Leachate concentrations
PFASs were detected in all landll leachate samples. The total sum of 28 PFASs X 28 PFASs ! per landll ranged from 320 to 11 000 ng L À1 (mean AE standard deviation: 1700 AE 2900; median: 630 ng L À1 ), see Table 1 (data for individual PFAS is presented in Tables S4 and S5 †). For the four short-chain PFASs these were from 68 to 6800 ng L À1 (mean: 980 AE 1800; median: 360 ng L À1 ), whereas the 15 long-chain PFASs ranged from 140 to 2900 (mean: 530 AE 730; median: 290 ng L À1 ). Hence, a substantial variation in leachate concentrations, more than two orders of magnitude, can be found at these diverse Norwegian landlls.
For four of the landlls (I, II, III and VI), there was data available from previous, recent sampling campaigns for comparison. The concentrations for landlls I, II and III were quite similar from the current and previous campaign (generally within a factor 1.5 of each other, or not statistically different), though landll VI exhibited substantially higher concentrations in the current April 2018 data compared to the previous September 2017 data. The only landll with a time series available was Landll I, having 6 time points from 2017 to 2018. This time series found more variation in long-chain PFAS concentrations (relative standard deviation, rsd, of 81%) compared to short-chain PFASs (rsd of 14%). Time trends in PFAS leachate concentrations were studied for a Canadian landll by Benskin et al., 13 this study observed a peak in leachate emissions around mid-March to April for long-chain PFASs, but not for short-chain PFASs, which tended to be consistent throughout the year. Though the mechanisms of this are complex, and related to meteorological conditions (e.g. snow melt and precipitation), sorption and water properties (e.g. pH and ionic strength), less variability for short-chain PFASs than long-chain PFASs in landll I is consistent with the Canadian landll studied by Benskin et al. 13 However, what is inconsistent is the apparent difference in both short-chain and long-chain PFASs observed in landll VI between September 2017 to April 2018, where the September 2017 concentrations seems more of a diluted version compared to the April 2018 sample; and may simply be due to a dilution event e.g. storm water. Later in this manuscript a correlation analysis is presented to other leachate parameters (pH, DOC, EC and precipitation).
Regarding the second-aim of this study to see if short-chain PFAS dominate in leachate emissions, in six out of ten landlls, there were higher concentrations of short-chain PFASs than long-chain PFASs, with short-chain to long-chain ratios ranging from 0.28 to 3.1 (Table 1). Overall, the short-chain PFBS contributed most to the X 28 PFAS-concentration (30% based on mean concentrations). In leachate from Svalbard (landll X), the major contributor was the long-chain PFOS (30%). At other landlls, PFOS contributed 1 to 20% of the X 28 PFASs. Relatively lower overall PFOS-abundance and higher abundance of shortchain PFASs (especially PFBS) could indicate that short-chain PFASs are now the major contributors in landll leachate, supporting the study's hypothesis. The role of time trends could also be considered here, as mentioned above, Benskin et al. 13 noticed a drop in long-chain PFASs aer April, but not shortchain PFASs. Since sampling was mainly done in April, shortchain PFAS may have dominated further if sampling occurred later in the year. Sorption may also play a role. PFOS and other long-chain peruoroalkyl acids sorb stronger to organic solids than some of their short-chain analogues; 25-27 for instance, average observed organic carbon partition coefficient, K oc of PFOS and PFBS is 3.0 and 2.2, respectively. 28,29 Thus one would expect long-chain PFASs to leach slower than short-chain PFASs. The relatively high abundance of short chain PFAS, particularly PFBS, in leachate demonstrates that considerable amounts of PFBS (and PFBS precursor) containing waste have been deposed of in Norwegian landlls. The presence of PFOS in the leachates shows that the phase-out in new commercial products has not reduced their concentrations in leachate entirely; due to both the presence in waste being currently deposited at the landlls, and the lag-time of leaching from waste previously landlled.
There were signicant correlations between  (Table S6 †), indicating that these substances have correlating concentration and emission pathways across landlls.
In Table 2, leachate concentrations from the literature are compiled, in which it was possible for us to calculate the shortchain and long-chain concentrations and ratios in a manner similar to this study, because similar PFASs were analysed and the raw data was available. General caveats however with such comparisons are that the analysis methods can differ, PFASs analysed can differ, as can the type of leachate sampled (raw, diluted, treated), as indicated in Table 2, thus preventing exact comparisons, Eggen et al. 19 using non-target analysis ( Table 2) reported relatively high for raw leachate in Norwegian landlls sampled in 2006 (which in comparison is higher than all landlls in this study except landll VI), and a short-to long-chain ratio of 0.16; thus, indicating that short-chain PFASs have become more dominant in Table 1 PFAS concentrations (ng L À1

View Article Online
Norway recent years (though this data is inuenced by a nontarget analysis method being used).
Benskin et al. 13 reported higher  Table 2). The short-to longchain ratio we calculate from their data was 1.4 (Table 2), which agrees with the general dominance of short-chain PFASs in landlls from this study. Four landlls from northern Spain were also sampled in 2015, 24 and the concentration of s in raw leachate was in the same range as PFASs in this study ( Table   2). The short-to long-chain ratio from the data in Fuertes et al. 24 was 1.1 (  PFASs this study. PFOA and PFBS were the most abundant. 22 It appears from this review that short-chain PFBS and PFBA have been a dominant component of leachate in some areas since 2006, though in the more recent studies this seems to be more typically the case, as in Norway.

Emissions
The yearly amount of leachate generated at each landll were provided by the landll operators. This was used for estimation of the yearly PFAS release from each site, assuming the concentrations in Table 1 were consistent all year round (Table  S7, † illustrated in Fig. 1); though as indicated above, sampling was done at time when concentrations of long-chain PFAS, in particular, are highest. The annual X 28 PFAS release at each landll was estimated to range from 9.2 to 510 g per year (mean: 160 AE 160 g per year; median: 100 g per year). For short-chain PFASs these were from 2.2 to 310 g per year (mean: 87 AE 98;     Table S2 †); and therefore leaching from waste is also slowed. It is important to bear in mind that there are several limitations to the emission estimates provided in the present study. These are (1) most sites had one sampling day, so seasonal, climate and hydrological factors could not be evaluated; (2) sampling occurred around April, which as discussed above is potentially when peak emissions for long-chain PFAS occur; 13 (3) discharge volume is provided by the operators and their accuracy may vary; and (4) the leachate samples could to some degree be inuenced by storm water and groundwater.
Comparisons with emission estimates in the literature need to take into account different PFASs being considered, as well as annual leachate volumes, which can vary tremendously. Therefore comparisons with emissions in the literature presented below must be considered with these limitations in mind. The study on Norwegian landlls in 2006 (ref. 19) reported emissions of 2.1 kg per year, which is higher than any of the landlls in the present study, due to relatively higher PFAS concentrations (Table 2), as well as a relatively large annual loading of 345 000 m 3 leachate per year. 19 The study of the Canadian landll in 2010 13 reported annual X 24 PFAS emissions from 8.5 to 25 kg per year (mean: 16 kg per year) which is considerably higher than the landlls in this study. This is mainly due to the high annual volume of leachate produced at the Canadian site, of 2.2 Â 10 6 m 3 per year, 13 compared to volumes from 2.2 Â 10 4 to 4.6 Â 10 5 m 3 per year in this study (Table S2 †).
To extrapolate these results to the national level, which introduces new uncertainties, a previous report on leachate emissions in Norway concluded that a total volume of up to 1.0 Â 10 7 m 3 is emitted from Norwegian landlls, with 55% of emissions being sent to wastewater treatment plants (WTP) and 45% to the environment. 30 Based on the mean concentrations provided in Table 1  National estimates have also been presented for other countries, such as China, 22 Germany, 17 and the U.S. 16 These were derived by multiplying average emission rates for landlls included in their study by the number of landlls in the country, rather than based on leachate volumes as here.   (Table S2 †) and the concentrations in Table 1) of long-chain and short-chain perfluorinated sulfonates (PFSAs) and perfluoroalkyl carboxylic acids (PFCAs) for the Norwegian landfills in this study (Table S7 † Compared to total per capita emission factors from all sources of PFAS, not just landlls, based on river and water treatment plant data in the literature, the contributions of landlls are relatively low. For instance, considering just PFOS, total per capita emissions from all sources for the EU in 2009 were estimated at 9.9 mg per year per person. 31 In our 2018 study, Norwegian landll emissions contribute roughly 0.35 mg per day per person emissions of PFOS, which would be 3.5% of the European per capita emissions in 2009. Total ux of sewagederived PFOS from Japan was in 2008 estimated to be 3.6 tonnes per year, corresponding to 28 mg per year per person, 32 which is almost two orders of magnitude higher than the per capita emission factors derived here. Part of this may be attributable to PFOS emissions having declined in recent years, since these other per capita emissions were derived.
Long-chain PFASs are more likely to adsorb onto surfaces and partition into soils and sediments. 25,26 Eggen et al. 19 reported that long-chain PFCAs, FOSA and EtFOSA were detected at relatively high concentrations in sediment samples from a Norwegian landll sampled in 2006, with total PFAS concentrations (dry weight) ranging from 1.3 to 382 mg kg À1 (mean: 61 mg kg À1 ), which is in the same range as this study (Table 3), indicating little change. According to Table 3 Average PFAS concentrations (mg per kg dry weight) sediments from landfill I, IV-VII and IX-X (sampled in 2018). Values < LOQ are excluded from the calculations. n.a. PFASs ranged from approximately 0.0013 to 0.024 mg per kg d.w., 20 which is lower than the present study. However, it is noted that PFAS substances such as EtFOSAA and MeFOSAA were not measured in that study, whereas these were measured at relatively high concentrations in this study (Table 3).

Sediment-leachate distribution
The ratio of C sediment to C leachate , which will be referred to as Q sed/leachate (units L kg À1 ), measured at the specic landlls is presented in Table S10. † It is noted that the Q sed/leachate value cannot be considered an equilibrium distribution coefficient, In principle, factors such as pH, EC and DOC would inuence the sorption of ionic PFASs. Thus, PCA biplots were made for the Q sed/leachate À values of PFOS and PFOA, where Q sed/ leachate is considered a proxy for sorption. As their Q sed/leachate À values were strongly correlated (r ¼ 0.9, p ¼ 0.001), only the PCA-biplot with Q sed/leachate À values for PFOS is shown in Fig. S2. † Q sed/leachate and yearly leachate production volume were positively correlated, implying higher sorption to the sediment phase with increasing leachate volume. This could be accounted for by considering a Freundlich like sorption behaviour (i.e. sorption increases with increasing water to sediment ratios, leaving behind the more strongly sorbed residues). There were no signicant correlations between Q sed/ leachate and pH or DOC. However, there was a negative correlation between PFOS Q sed/leachate and EC, which indicates relatively less partitioning to the sediment phase with increasing EC. This could be accounted for by increasing competition for anionic sorption sites. It is not uncommon for leachate sediments to be rich in metal oxides, 35 which can be positivelycharged and therefore contain anion exchange sites, depending on the pH and salt composition. Studies by Wang et al. 36,37 indicated that the sorption of PFOS and PFOA on the aluminium oxides boehmite and alumina decreased with increasing ionic strength. In contrast, negatively-charged clays like the phyllosilicate bentonite show negligible sorption of PFAS, 38 and therefore would be less inuenced by EC than positively-charged metal oxides. Thus, the stronger sorption with higher ow volumes and lower EC seems to match well with mechanistic expectations for a Freundlich like behaviour: dilution would lower EC and PFAS concentrations, and therefore increase sorption via less competition for anion-exchange sites to e.g. metal oxide surfaces.

Environmental implications
This study estimated that the release of X 28 PFASs from Norwegian landlls was in the range 3.2 to 110 kg per year (mean: 17 AE 29; median: 6.3 kg per year); though, it should be kept in mind there were several assumptions used to make this data, ranging from the (limited) sampling campaign to the assumption that the obtained data were representative of yearly emissions. Future sampling campaigns should address this. As hypothesized, due to the shi towards short-chain PFAS chemistry, emissions of short-chain PFASs appear to dominate over emissions of long-chain PFASs. Short-chain PFASs are generally more mobile, as made evident by the difficulty in this study in obtaining sediment concentrations of PFBS, despite it being the most dominating PFAS in leachate. Future studies could conrm this trend, and relate emission levels with current production levels, uses and disposal of specic PFASs. Landlls are one of the many sources of PFAS emissions. Here we presented a very rough estimation based on a literature comparison that landlls contribute approximately 3.5% of all European per capita emissions of PFOS. A similar estimation for other PFASs could not be made. Future studies could seek to quantify these values better through comparison of different emission sources, e.g. households, re training facilities, airports, and industries, to their respective contributions to wastewater treatment plants, groundwater, rivers, and other recipients. Even though the composition of PFASs in commerce and in landlls change over time, landlls can continue to release contaminants like PFASs for years to come. If the emissions in Norway were consistent for 100 years at 17 kg per year, then there would be 1.7 tonnes emitted from one, relatively small country. To put this into context, a recent global emission inventory of PFHxS and PFDS has estimated that between 1958 and 2015, 120-1022 and 30-378 tons, respectively, have been emitted. 39 To prevent landlls from contributing to future PFAS emissions, proper management strategies are key, such as developing low cost leachate treatment facilities, including PFAS in leachate monitoring, better understanding of leaching mechanisms from waste, and evaluating how concentrations of PFAS in leachate changes with PFAS levels in deposited waste.

Conflicts of interest
There are no conicts to declare.