Ecotoxicology

, Volume 25, Issue 10, pp 1739–1750

Fracked ecology: Response of aquatic trophic structure and mercury biomagnification dynamics in the Marcellus Shale Formation

  • Christopher James Grant
  • Allison K. Lutz
  • Aaron D. Kulig
  • Mitchell R. Stanton
Article

DOI: 10.1007/s10646-016-1717-8

Cite this article as:
Grant, C.J., Lutz, A.K., Kulig, A.D. et al. Ecotoxicology (2016) 25: 1739. doi:10.1007/s10646-016-1717-8

Abstract

Unconventional natural gas development and hydraulic fracturing practices (fracking) are increasing worldwide due to global energy demands. Research has only recently begun to assess fracking impacts to surrounding environments, and very little research is aimed at determining effects on aquatic biodiversity and contaminant biomagnification. Twenty-seven remotely-located streams in Pennsylvania’s Marcellus Shale basin were sampled during June and July of 2012 and 2013. At each stream, stream physiochemical properties, trophic biodiversity, and structure and mercury levels were assessed. We used δ15N, δ13C, and methyl mercury to determine whether changes in methyl mercury biomagnification were related to the fracking occurring within the streams’ watersheds. While we observed no difference in rates of biomagnificaion related to within-watershed fracking activities, we did observe elevated methyl mercury concentrations that were influenced by decreased stream pH, elevated dissolved stream water Hg values, decreased macroinvertebrate Index for Biotic Integrity scores, and lower Ephemeroptera, Plecoptera, and Trichoptera macroinvertebrate richness at stream sites where fracking had occurred within their watershed. We documented the loss of scrapers from streams with the highest well densities, and no fish or no fish diversity at streams with documented frackwater fluid spills. Our results suggest fracking has the potential to alter aquatic biodiversity and methyl mercury concentrations at the base of food webs.

Keywords

Aquatic ecology Biodiversity Biomagnification Hydraulic fracturing Stable isotopes Marcellus shale 

Introduction

Unconventional natural gas development is on a global increase with the potential to impact terrestrial and aquatic ecology. Advancement of unconventional natural gas development technologies, such as hydraulic fracturing (fracking), are facilitating the exploitation of natural gas reserves in many countries (Garvie & Shaw 2015), including the United States (Entrekin et al. 2011). Despite widespread fracking in the United States, its influence on terrestrial and aquatic ecology is not well understood. Current efforts have shown possibilities of groundwater aquifer contamination (Llewellyn et al. 2015a) and forest fragmentation (Drohan et al. 2012). Surface waters also can be impacted by fracking through changes in physiochemical properties (Entrekin et al. 2011), microbial communities (Trexler et al. 2014a), and mercury (Hg) concentrations in aquatic ecosystems (Grant et al. 2015). This is particularly concerning, because changes in stream physio-chemistry, microbial communities, and soluble metal concentrations can cause cascading changes in aquatic trophic structure and biomagnification rates (Kelly et al. 2003).

Changes in trophic structure and biomagnification rates commonly are assessed by comparing the trophic position of specific organisms, the length of the food chains, and measures of biodiversity (Jardine et al. 2013). Isotopic ratios of 15N: 14N are often used as a surrogate for trophic level (Perkins et al. 2014), and ratios of 13C: 12C isotopes have been used to determine whether a particular aquatic organism’s diet is comprised of terrestrial or aquatic carbon sources (Seifert & Scheu 2012). Knowledge of food sources can aid in understanding differences in the bioaccumulation of contaminants, because terrestrially derived sources often contain lower concentrations of contaminants, such as Hg (Chételat et al. 2011). Relating changes in trophic structure and stream physiochemical characteristics can help identify factors controlling mercury levels within organisms (Chasar et al. 2009). Perturbations changing stream physiochemical conditions may ultimately affect biomagnification of contaminants, such as Hg, and aquatic biodiversity.

The scientific community has recently begun describing how perturbations to aquatic ecosystems may be associated with fracking. In an earlier study, results from a single summer sampling effort documented stream physio-chemistry, aquatic ecology, and total mercury (THg) concentration in biotic compartments in relation to fracking (Grant et al. 2015). However, an understanding of how fracking impacts biomagnification rates of methyl mercury (MeHg), or any contaminant, in streams is not known. Understanding the controlling factors is needed to determine how organisms, communities, and ecosystems respond to fracking and potential contaminant biomagnification through food webs. Here we build upon previous work through a two-year study utilizing δ15N, δ13C and MeHg values to determine whether changes to MeHg biomagnification rates and fish and macroinvertebrate biodiversity were related to fracking. We also explored relationships between MeHg levels, trophic structure, food source, aquatic biodiversity and stream physiochemical characteristics.

Methods

Study sites and sample collection

During June and July 2012 and 2013, we sampled 27 remotely located forested watersheds within the Marcellus shale basin in Pennsylvania (Fig. 1). The streams in these watersheds contained naturally reproducing brook trout populations. We compiled data on drilling and hydraulic fracturing dates from the Pennsylvania Department of Environmental Protection Permits Issued Details Report (2012a, b), and from Pennsylvania Department of Environmental Protection wastewater generation reports (2013) for each unique well permit within the watershed of all targeted study sites. One specific 100-m study site was established at each stream downstream and as close as possible to the permitted well-pad site(s). Streams were categorized in groups according to presence (F, n = 12) or absence (NF, n = 15) of hydraulic fracturing processes within their watersheds prior to our sampling efforts. Watersheds were delineated to show the area upstream of each sampling point. Although some streams were in similar larger drainages, the study sites were located in separate sub-basins (i.e., not downstream or upstream of each other), separated by at least 1200 m of stream-line distance, and not confounded. All sites were sampled under base flow conditions and all stream physio-chemistry, biodiversity, nitrogen, carbon isotope, and mercury measures were averaged across 2 years for each stream. Deer Creek data was not averaged across sampling years because it changed from NF to F group between 2012 and 2013.
Fig. 1

Overview map showing general location of 27 streams sampled in northwestern Pennsylvania during summer 2012 and 2013. Insets show watersheds delineated and outlined upstream of each sampling point. Blue outlined watersheds indicate study sites where hydraulic fracturing has not occurred by the time of sampling, and red outlined watersheds indicate study sites where fracking has occurred within the watershed before sampling. Watersheds with a documented frack water spill are denoted by red outline and grey striped fill pattern. For each watershed, permitted unconventional wells that were not fracked by sampling are designated by open circles, while blackened circles indicate unconventional wells that were fracked prior to sampling. Open triangles indicate permitted conventional oil wells within each watershed (see online version for color figures)

We used the watershed tool in Arc- GIS 10.0 to calculate watersheds, using a 1/3 arcsecond (approximately 10 m) resolution National Elevation Dataset (NED), (Gesch 2007) digital elevation model (DEM), and GPS coordinates for each sampling point as pour points. All elevation and slope metrics were calculated using the 1/3 arcsec NED DEM, which is based on LiDAR-derived elevation data for the area studied in Pennsylvania. Once watersheds were delineated for each sampling point, land cover metrics were determined using the PAMAP Land Cover for Pennsylvania (2016) (Table 1). Watersheds were forested, had no evidence of mining legacy, and most had few conventional oil wells. Hydric soils data per watershed were calculated using the SSURGO (NRCS 2016), with soil types either categorized as all hydric or partially hydric. Drainage density was calculated using the area of each watershed (km2) and the stream length (km) within each delineated watershed as measured using the networked streams of Pennsylvania shapefile (PASDA 2016). Stream pH was recorded using a Eutech PCSTestr 35 Multi-parameter test probe that was calibrated weekly using a three point-calibration (pH of 3, 7, and 9). Stream water samples for Hg analysis were collected at each site using 2-L Polyethylene terephthalate copolyester, glycolmodified (PETG) containers following clean hands-dirty hands techniques (Lutz et al. 2008) and trace metal clean techniques. PETG containers were double bagged and stored on ice in darkened coolers until returning to the lab for filtration and analysis of filtered total mercury (FTHg), filtered methyl mercury (FMHg), particulate total mercury (PTHg), and particulate methyl mercury (PMHg).
Table 1

Watershed characteristics of 27 streams sampled in northwestern Pennsylvania during summer 2012 and 2013

 

S

ID

Lat.

Long.

WS area (ha)

WS Elev. (m)

F. wells

Well dens.

nearest well (m)

C. wells

% For.

% Hyd.

S.L. (km)

D. dens. (km/km2)

S. S. (deg.)

Alex Branch

F

1

41.17528

−78.41682

8.82

618.5

5

0.57

72

1

87

46

6.8

0.77

3.27

Little Laurel

F

2

41.18092

−78.47267

8.86

638.4

9

1.02

80

1

79

58

11.8

1.33

2.83

Stone Run

F

3

41.11958

−78.46036

4.97

662.0

13

2.62

261

0

87

41

3.0

0.61

5.05

Indian Run

F

4

41.65257

−78.37559

10.59

619.3

9

0.85

628

1

99

37

12.4

1.17

8.82

Long Run

F

5

41.61109

−78.72559

5.32

599.1

4

0.75

244

15

96

61

5.6

1.06

7.54

Bear Creek

F

6

41.52467

−78.70687

3.22

527.4

2

0.62

372

0

87

81

4.2

1.31

3.80

Cold Stream

F

7

41.12894

−78.41967

10.69

655.4

7

0.65

111

3

90

20

10.9

1.02

7.18

Deer Creeka

F

8

41.15914

−78.30115

6.88

636.4

1

0.15

463

8

95

33

7.1

1.04

4.69

Iron Run

F

9

41.61109

−78.96098

3.91

529.9

1

0.26

669

22

95

62

3.6

0.92

6.46

Laurel Run

F

10

41.34202

−78.78692

12.49

613.6

2

0.16

373

2

89

39

15.6

1.25

7.98

Lick Run

F

11

41.10113

−78.46626

21.24

636.2

10

0.47

193

2

83

51

19.2

1.01

6.73

Little Wolf R.

F

12

41.53778

−78.69437

6.87

552.9

3

0.44

467

2

98

62

6.5

0.94

5.33

Dixon Run

NF

13

41.15534

−78.39382

2.78

648.5

0

0

209

0

86

23

2.0

0.73

4.36

SBNFRC

NF

14

41.26725

−78.93087

27.94

540.5

0

0

288

47

81

67

29.2

1.05

4.16

Straight creek

NF

15

41.5935

−78.51421

14.95

647.9

0

0

81

31

98

57

17.3

1.16

6.11

Trout Run

NF

16

41.14959

−78.40733

33.6

617.4

0

0

191

3

89

40

3.4

1.02

5.09

Bens Creek

NF

17

40.38899

−78.61037

13.19

768.9

0

0

245

0

96

48

15.2

1.15

7.24

Camp Run

NF

18

41.58467

−79.33978

2.71

482.0

0

0

345

18

99

49

2.6

0.97

7.53

Crooked Run

NF

19

41.10461

−78.38187

9.08

568.1

0

0

409

1

94

19

7.8

0.79

6.72

Dead Mans L.

NF

20

41.5872

−78.44127

6.13

654.2

0

0

445

12

96

54

5.1

0.82

8.79

Dutch Hollow

NF

21

41.21196

−78.00217

6.8

469.6

0

0

451

0

98

29

9.2

1.36

6.09

Findley Run

NF

22

40.4206

−78.98495

15.45

610.1

0

0

440

16

85

35

16.6

1.07

8.30

Moccasin

NF

23

41.25294

−77.94334

9.86

478.6

0

0

536

8

97

12

7.3

0.74

6.54

NBIR

NF

24

41.39724

−78.75048

9.39

593.3

0

0

896

5

93

51

6.2

0.66

9.40

SFWBPC

NF

25

41.6263

−78.44174

3.47

670.4

0

0

249

8

90

68

3.5

1.02

3.51

UNT C. River

NF

26

41.63719

−78.679

3.65

602.9

0

0

223

0

100

54

5.2

1.42

8.23

Vineyard Run

NF

27

41.30624

−78.78811

3.65

594.0

0

0

516

0

91

51

5.5

1.52

6.79

Mean

F

   

9

607

  

316b

12b

90

49

9

1

6

 

NF

   

11

596

  

345b

15b

93

44

9

1

7

P value

    

0.465

0.679

  

0.678

0.604

0.297

0.412

0.944

0.970

0.280

S represents stream status and identifies whether hydraulic fracturing has occurred (F) or has not occurred (NF) within the watershed at the time of sampling. ID represents the number identifier corresponding to a given stream/watershed in Fig. 1. Latitude (Lat.) and Longitude (Long.) are given for the sampling location at each stream, along with total watershed area upstream of sampling point (WS area), and mean watershed elevation (WS elev.). The total number of fracked unconventional wells (F. wells), the nearest permitted well to our sampling point, in meters (nearest well), and fracked unconventional well density (well dens.; calculated by dividing F. wells by WS area) are included for each stream. The number of permitted conventional wells (C. Wells) are also included for each watershed. Additional watershed characteristics, including the percent of watershed that is forested (% For.), the percent of hydric soils in the watershed (% Hyd.), stream length upstream of sampling point (S.L.), stream drainage density (D. dens.), and stream slope (S. S.) are included. Means and P values for the F and NF groups are reported, and P-values reflect Student’s T for all measures except for nearest well and C. wells, where a Mann-Whitney’s U and medians are reported. No significant differences were observed between the F and NF groups

a Deer Creek changed fracking status in 2013

b median reported for nonparametric testing

We determined fish assemblages and collected fish Hg, and isotopic samples at each 100 m reach using unblocked segments and the wadeable electrofishing protocol (Barbour et al. 1998) with a Smith and Root LR 24 backpack electrofisher with pulsed direct currents ranging from 450–700 volts, depending on stream conductivity. Equal effort (i.e., electrofishing time, number participating) was employed for sampling fish at each stream site to allow for accurate cross-site comparison. At the completion of a pass, all fishes were identified to species level and relative abundances were recorded. Field measurements of brook trout included weight (in grams), measured using an OHAUS scale, and total length (in millimeters), measured using a fish board. Three to five brook trout were euthanized, wrapped in aluminum foil and kept on ice for later Hg and isotope analysis. These retained fish were stored on ice in darkened coolers until returning to the lab.

Macroinvertebrate were sampled using a 500 micron screen barrier kick net at four representative riffles in each stream (Barbour et al. 1998). Five crayfish per stream were collected through a combination of electrofishing and hand-capture methods. All collected macroinvertebrates were placed in high density polyethylene bottles with stream water and stored on ice in darkened coolers. Composite periphyton samples (5 spots within each stream) were collected from submerged rocks within the active channel using pre-cleaned Teflon scoops and certified-clean brown glass I-chem vials with Teflon-coated lids.

Laboratory analyses

Leica dissection microscopes were used for all identification of macroinvertebrates to the genus level (Merritt et al. 2008), and Shannon-Wiener diversity index, Index for Biotic Integrity (IBI) scores and Ephemeroptera, Plecoptera, and Trichoptera (EPT) richness were calculated for each stream in 2012 and 2013 according to Chalfant (2012). Macroinvertebrate samples were then categorized and grouped by functional feeding group (shredders, scrapers, collectors, or predators) for each stream, by year, to obtain enough mass for isotopic and mercury analysis. Additionally, subsets of brook trout were examined for stomach contents in 2012 (n = 70) and 2013 (n = 120), when present and discernable, the consumed insects were identified to the order level and categorized as an aquatic or terrestrial food item.

The stream water samples collected for Hg analysis were filtered <24 h after collection at Juniata College and shipped overnight to the United States Geological Survey Wisconsin Mercury Lab, where cold vapor atomic fluorescence spectroscopy was used according to US EPA Method 1631 (Revision E) to determine concentrations of FTHg, FMHg, PTHg, and PMhg in all stream water samples (USEPA Method 1631, 2002).

Brook trout filets, crayfish tail muscle, macroinvertebrate functional feeding groups, and periphyton were homogenized and analyzed for THg concentrations using combustion atomic absorption spectrophotometry according to US EPA method 7473 (USEPA 2007) at the Pennsylvania State University Water Quality Laboratory. All samples were run with multiple solid and liquid standards; the methods yielded an instrument detection limit of 0.002 mg/g Hg wet weight and 0.02 mg/g Hg on a dry-weight basis. Subsets of all samples were analyzed for MeHg and then compared to THg concentrations, to determine ratios to be applied to all total mercury concentrations across trophic groups. The methyl mercury samples were analyzed by Brooks and Rand Laboratories (Seattle, WA, USA) ultra-sensitive cold-vapor atomic fluorescence detectors in conformance with EPA Method 1630. Isotope samples were dried in a Precision Oven at 50–60 °C and masses of approximately 1 mg of tissue were then shipped to the UC Davis Stable Isotope Facility at the University of California, where δ13C and δ15N isotope analyses were conducted with an isotope ratio mass spectrometer.

Quantitative analysis

All values of δ15N and log MeHg used to perform linear regressions of the trophic biomagnification rates in the F and NF groups were averaged by feeding group within each stream sampled across both years. We tested for interaction effects to determine if the regression coefficients (slopes) or biomagnification rates were different between F and NF groups. We used ANCOVAs to test for the effects of differences in elevation and slopes on the plotted biomagnification rates between F and NF groups. Steam water Hg values were non-normal and could not be normalized with transformations so Mann-Whitney testing was used to assess differences between F and NF groups; Student’s t tests were used to compare pH, conductivity and total dissolved solids (TDS) between fracked and non-fracked groups (Table 1). Student t-tests were used to test differences in macroinvertebrate IBI scores, EPT richness, Shannon index for macroinvertebrate diversity, Simpson’s fish biodiversity and brook trout abundance between N and NF groups. Additionally, watershed characteristics including watershed size (ha), mean watershed elevation (m), percent of watershed that was forested, percent of watershed that had hydric soils, stream length (km), mean stream slope (degrees), and drainage density (km/km2), for the F and NF groups were compared using Student’s T test (Table 1). We used Mann-Whitney tests to assess differences between the nearest permitted unconventional well to sampling point (m) and the number of conventional wells between F and NF groups.

A cluster of 11 brook trout outliers became evident while generating a three dimensional scatterplot for δ13C, δ15N, and Log MeHg. We used normalized Ward’s minimum variance clustering by Euclidean distance to assess whether δ13C, δ15N, and Log MeHg values for these brook trout outliers differed significantly from their same-stream counterparts. We used ANOVA and post-hoc Tukey Kramer tests between the outgroup of brook trout and the other sampled brook trout to validate the clustering method by examining differences between δ13C, δ15N, and Log MeHg. Students t tests were used to test for differences between averaged physical characteristics (i.e., length, weight, trophic position, and stomach contents) of the clustered outgroup and their same-stream counterpart. Stream and watershed characteristics of the streams with clustered outgroup brook trout were compared to those of the other streams utilizing Student’s t test, while stream mercury concentrations were compared using Mann-Whitney’s U test. We also used a Mann-Whitney U test to determine if values of δ13C differed between predatory macroinvertebrates and brook trout from clustered outgroup streams. The statistical and graphical computing software (R version 3.2.2) was used for all data analysis and to create the figures (R, 2014).
Table 2

Physiochemical characterstics of 27 streams sampled in northwestern Pennsylvania during summer, 2012 and 2013

Site

S

Map ID

pH

Cond.

TDS

Sal.

PMHg

FMHg

PTHg

FTHg

Methy eff.

Alex Branch

F

1

4.73

37.9

26.8

22.2

0.041

0.15

0.878

2.65

5.66

Little Laurel Run

F

2

4.24

45.2

32.1

28.2

0.021

0.22

0.249

1.28

17.19

Stone Run

F

3

5.1

47.7

34.0

27.1

0.01

0.04

0.297

0.54

7.41

Indian Run

F

4

7.04

27.5

19.5

16.5

0.187

0.69

0.749

2.38

28.99

Long Run

F

5

7.42

36.7

25.8

20.5

0.012

0.04

0.371

0.46

8.70

Bear Creek

F

6

5.61

37.9

26.8

21.6

0.157

0.28

1.592

2.36

11.86

Cold Stream Run

F

7

6.43

23.4

18.0

18.0

0.015

0.07

0.455

1.04

6.73

Deer Creeka

F

8

5.79

20.3

15.2

15.5

0.086

0.04

0.079

0.61

6.56

Iron Run

F

9

6.55

59.6

42.3

32.4

0.031

0.07

0.481

1.42

4.93

Laurel Run

F

10

6.7

29.2

20.1

18.6

0.022

0.05

0.604

0.81

6.17

Lick Run

F

11

6.56

25.0

17.8

16.4

0.01

0.05

0.124

0.75

6.67

Little Wolf Run

F

12

6.8

34.1

24.1

19.9

0.015

0.04

0.195

0.59

6.78

Dixon Run

N

13

6.71

18.8

15.1

14.3

0.01

0.04

0.186

0.54

7.41

SBNFRC

N

14

7.34

51.5

36.7

29.2

0.058

0.15

0.687

2.21

6.79

Straight Creek

N

15

7.31

31.8

22.6

19.6

0.009

0.04

0.295

0.74

5.41

Trout Run

N

16

6.8

29.1

20.7

19.5

0.05

0.11

0.862

2.32

4.74

Bens Creek

N

17

7.74

34.7

24.7

21.0

0.362

0.22

5.577

0.74

29.73

Camp Run

N

18

7.58

31.0

23.0

20.8

0.015

0.07

0.455

1.04

6.73

Crooked Run

N

19

5.4

20.3

16.9

15.2

0.017

0.04

0.231

0.41

9.76

Dead Man’s Lick

N

20

7.28

22.4

16.6

16.1

0.014

0.04

0.221

0.51

7.84

Dutch Hollow

N

21

7.13b

32.5

23.0

20.1

0.01

0.04

0.191

0.5

8.00

Findley Run

N

22

7.78

72.5

51.5

40.1

0.01

0.04

0.706

0.31

12.90

Moccasin Run

N

23

8.06b

36.3b

25.8

21.2

0.01

0.04

0.268

0.42

9.52

NBIR

N

24

7.22

50.1

35.5

28.0

0.02

0.12

0.226

0.34

35.29

SFWBPC

N

25

4.78b

23.1b

16.6

17.1

0.195

0.5

2.394

4.42

11.31

UNT C. River

N

26

7.63

44.0

31.3

23.6

0.052

0.04

0.281

0.41

9.76

Vineyard Run

N

27

7.56

85.1

60.4

43.4

0.009

0.04

0.482

0.67

5.97

Mean

N

 

7.09

38.9

28.0

23.7

0.056

0.1

0.871

1.04

11.41

 

F

 

6.08

35.3

25.2

21.4

0.05

0.14

0.506

1.24

9.80

P value

  

0.011

0.559

0.500

0.490

0.119

0.135

0.418

0.067

0.36

S represents stream status and whether the presence (F) or absence (NF) of hydraulic fracturing has occurred within the watershed at the time of sampling. Significant differences in stream pH and FTHg (α = 0.1) were observed between fracked and non-fracked streams. P-values reflect significance of Student’s t test for water quality characteristics and Mann-Whitney’s U for particulate methyl Hg (PMHg in ng/L), particulate total Hg (PTHg in ng/L), filtered total Hg (FTHg in ng/L), filtered methyl Hg (PMHg in ng/L) and methylation efficiency (Meth. Eff.). Methylation efficiency was calculated as FMHg/FTHg*100. Map ID represents the number identifier given to a stream/watershed in Fig. 1. Cond. is conductivity in µS/cm, TDS is total dissolved solids in PPM, and Sal. is salinity in PPM. Bold pvalues are considered significant at α=0.11

a Deer Creek changed fracking status in 2013

b Denotes pH and conductivity sampling only in 2013

Table 3

Individual brook trout stomach contents (% Terr.), morphological characteristics (weight and length), tissue MeHg values (Log MeHg), and δ13C δ15N isotopic values of all fish from streams containing individuals classified in an outgroup with respect to isotopic signatures and mercury levels, based on Ward’s (minimum variance) cluster analysis

Stream name

Group

Year

Status

Fish ID

Weight (g)

Length (mm)

Log MeHg

δ13C

δ15N

% Terr.

Laurel Run

SSC

2012

F

C12TR240142

75.4

202

1.759

−24.5

8.064

83

Straight Creek

SSC

2012

N

C12TR240223

33.4

145

1.783

−25.13

8.033

78

Straight Creek

SSC

2012

N

C12TR240221

75.1

199

2.103

−24.8

9.517

70

Bear Creek

SSC

2013

F

C13TR240201

24.8

131

1.806

−25.05

8.239

25

Bear Creek

SSC

2013

F

C13TR240203

79.4

192

2.082

−24.69

8.794

93

Bens Creek

SSC

2013

N

C13TR110021

23.7

149

1.9

−25.75

2.481

40

Bens Creek

SSC

2013

N

C13TR110022

19.0

132

1.812

−24.9

7.928

100

Deer Creek

SSC

2013

F

C13TR170084

41.1

157

1.81

−24.73

7.75

77

Deer Creek

SSC

2013

F

C13TR170082

45.8

166

1.988

−24.13

7.08

87

Laurel Run

SSC

2013

F

C13TR240144

49.9

170

1.597

−24.1

8.55

18

SBNFRC

SSC

2013

N

C13TR330192

13.4

115

1.406

−26.8

9.09

0

SBNFRC

SSC

2013

N

C13TR330191

14.3

117

1.381

−25.19

9.17

75

Straight Creek

SSC

2013

N

C13TR240225

51.5

160

1.991

−24.85

9.094

0

Laurel Run

COG

2012

F

C12TR240141a

155

235

1.303

−19.54

10.215

N/A

Laurel Run

COG

2012

F

C12TR240145

53.7

176

1.38

−19.58

10.217

50

Straight Creek

COG

2012

N

C12TR240222a

86.9

221

1.296

−19.64

9.818

N/A

Bear Creek

COG

2013

F

C13TR240202

22.0

130

1.38

−24.26

10.215

100

Bens Creek

COG

2013

N

C13TR110025

142.3

240

1.12

−19.16

9.546

33

Deer Creek

COG

2013

F

C13TR170083

31.8

150

1.961

−19.15

10.12

97

Laurel Run

COG

2013

F

C13TR240145

147.4

237

1.015

−19.11

9.59

100

Laurel Run

COG

2013

F

C13TR240142

166.6

259

0.931

−19.31

9.39

80

SBNFRC

COG

2013

N

C13TR330193

185.4

278

1.322

−19.09

8.83

100

Straight Creek

COG

2013

N

C13TR240223

149.6

235

1.122

−19.48

9.866

100

Straight Creek

COG

2013

N

C13TR240224

85.4

201

1.148

−19.45

10.029

100

Mean

SSC

   

42.1

156.5

1.8014

−24.97

7.983

57.4

 

COG

   

111.5

214.7

1.2707

−19.80

9.803

84.4

P value

    

0.0030

0.0021

0.0001

0.0001

0.0033

0.0498

Streams represent a subsample (6) of all 27 sties sampled across northwestern Pennsylvania during the summer 2012 and 2013. Streams are categorized according status indicating the presence (F) or absence (NF) of hydraulic fracturing within the watershed prior to sampling. Group refers to how fish are categorized within a stream, as a clustered outgroup individual (COG) or the same-stream counterparts (SSC). The COG of brook trout demonstrated significantly greater weight (grams), length (mm), δ15N, and δ13C values than the same stream counterparts (SSC). Log MeHg values of the COG were significantly lower than the SSC and the COG had significantly more terrestrial stomach (% Terr.) contents than the SSC. P values indicate significance of Student’s t test

a no stomach contents found upon dissection

Results

Biomagnification rates were obtained from the slope of log MeHg concentration vs. δ15N (inferred trophic level) and were significant for both F and NF groups (Fig. 2; P < 0.001). Log MeHg was predicted by δ15N with a slope of 0.153 (R2 = 0.470) and y-intercept of 0.540 in the NF group, while a slope of 0.142 (R2 = 0.541) and y-intercept of 0.664 was observed for the F group. While no significant differences were observed in biomagnification rates (slopes) between F and NF groups the y-intercept was greater for streams in the F group than the NF group (P = 0.034).

Across all sites, pH ranged from 4.24 to 8.06, with more acidic pH values occurring in watersheds where well densities were higher. Of all tested stream physiochemical and watershed characteristics tested, only pH (P = 0.011), was significantly different between F and NF groups (Table 2). FTHg was also significantly greater for the F group (1.24 ng/L) than for the NF group (1.04 ng/L; PMann-Whitney = 0.067) at α = 0.10. PMHg (PMann-Whitney = 0.390), FMHg (PMann-Whitney = 0.312), and PTHg (PMann-Whitney = 0.961) were not different between F and NF groups. Conductivity (P = 0.56), TDS (P = 0.50) and salinity (P = 0.49) did not differ significantly between F and NF groups. We did not detect significant differences for watershed size (P = 0.465), mean watershed elevation (P = 0.679), percent of watershed that is forested (P = 0.297), percent of watershed that had hydric soils (P = 0.412), stream length (P = 0.944), stream slope (P = 0.280), and drainage density (P = 0.970) between F and NF groups (Table 1). Also, we found no significant differences in the nearest permitted unconventional well to our sampling sites (PMann-Whitney = 0.678), or number of conventional wells in a watershed (PMann-Whitney = 0.604) between the F and NF groups.

Several measures of biological diversity were related to stream status (Fig. 3). Shannon index values of the F group did not differ significantly from the NF group (P = 0.249). However, IBI values for the F group were significantly lower than for the NF group (P = 0.013). EPT index values for macroinvertebrates were lower in the F group at α = 0.1 (P = 0.076). Simpson’s diversity values for fish was similar between stream groups (P = 0.348), and brook trout abundance was not observed to be different between F and NF groups (P = 0.273).
Fig. 2

Biomagnification plot of all 27 streams in Pennsylvania sampled in 2012 and 2013 grouped by the presence (F) or absence (NF) of hydraulic fracturing occurring within their watershed at time of sampling. All δ15N and log MeHg values are means of feeding group per stream and regression of biomagnification rates were done for each group: (NF: adj. R2 = 0.470, P < 0.001, n = 88), (F; adj. R2 = 0.541, P < 0.001, n = 73). The difference in y-intercepts for the F (red dashed line) and NF (blue solid line) groups is significant (P = 0.034). Red squares denote organisms from F group, while blue circles denote organisms from NF group (see online version for color figures)

Cluster analysis of all brook trout, with δ13C, δ15N, and log MeHg as determining variables, revealed a significant outgroup of 11 fish from six unique streams (Fig. 4). The clustered outgroup demonstrated greater δ15N (P < 0.001), greater δ13C (P < 0.001), and significantly less log MeHg (P < 0.001) compared to all other sampled brook trout. The clustered outgroup brook trout were longer (P = 0.002), heavier (P = 0.003), δ15N (P = 0.003), and had higher δ13C (P < 0.001) compared to their same-stream counterparts (Table 3). Mean MeHg values of the clustered outgroup were lower (P < 0.001) than their same-stream counterparts, and percent stomach contents comprised of terrestrial macroinvertebrates was lower in the clustered outgroup than in their same-stream counterparts (P = 0.05). Values of δ13C for predatory macroinvertebrates from clustered outgroup streams ranged from −29.25 to −26.58 while there was little variation in values for brook trout from the clustered outgroup, which ranged from −19.09 to −19.64. Predatory macroinvertebrate δ13C values were significantly lower than brook trout δ13C values (PMann-Whitney < 0.001). No stream or watershed characteristics were significantly different in streams that were inhabited by fish in the clustered outgroup, except stream PMHg at α = 0.1 (PMann-Whitney = 0.078). (Fig. 4)
Fig. 3

Boxplots of macroinvertebrate diversity indices, fish diversity, and brook trout abundance from 27 streams in Pennsylvania sampled in 2012 and 2013, grouped by the presence (F in red) or absence (NF in blue) of hydraulic fracturing occurring within their watershed at time of sampling. P-values reflect Student’s T test significance. a) Ephemeroptera, Plecoptera, and Trichoptera (EPT) macroinvertebrate richness was marginally lower for the F group (P = 0.076). b) Index for Biotic Integrity (IBI) scores for macroinvertebrates were significantly lower for the F group (P = 0.013). c) Shannon-Wiener diversity index for macroinvertebrates was not different between groups (P = 0.249), d) Simpson’s fish diversity was not different between groups (P = 0.348), and e) brook trout abundance was not significantly different between groups (P = 0.273) (see online version for color figures)

Fig. 4

Log MeHg concentration response to δ15N and δ13C in 27 streams (n = 414 organisms) sampled in northwestern Pennsylvania during 2012 and 2013. Six trophic groups are distinguished in the multi-dimensional plot. These include brook trout, predatory macroinvertebrates (Predators), crayfish, collector macroinvertebrates (Collector), scraper macroinvertebrates (Scraper), and shredder macroinvertebrates (Shredder). Ward’s (minimum variance) cluster analysis of the brook trout feeding group and ANOVA using standardized variables identified the clustered outgroup (circled), comprised of 11 trout from six unique streams. Red circles indicated brook trout, light blue squares squares indicate predators, orange triangles indicate crayfish, green diamonds indicate collectors, purple diamonds indicate scrapers, dark blue squares indicate shredders, and orange asterisks indicate periphyton (see online version for color figures)

Discussion

Impacts of fracking on bioaccumulation and biomagnification of Hg

Elevated MeHg concentrations at the base of food web was observed for streams where hydraulic fracturing had occurred within their watershed. While we observed similar slopes of biomagnification rates between F and NF groups in our study, others have shown that any perturbations to aquatic systems may increase biomagnification of mercury and other contaminants (Liu et al. 2011). Since this is the first study to compare Hg biomagnification to hydraulic fracturing, no literature exists for a direct comparison. However, our results are most similar to research by Van der Velden et al. (2013) comparing Hg biomagnification between lacustrine and marine Salvelinus alpinus (Arctic Charr), where they found no difference in biomagnifcation rates, but observed an increase in MeHg at the base of food webs in lacustrine environments. In our study, elevation of MeHg bioaccumulation at the base of the food web in the F group is likely the result of increased Hg assimilation, increased methylation rates (riparian or instream), and increased transport of mercury from riparian environments to the stream.

Stream pH and FTHg appeared to influence assimilation of Hg into the base of food webs. While absolute values of stream water Hg were low, FTHg was increased while pH decreased across streams in the F group. Further, while not statistically significant, average filtered MeHg concentrations were greater for the F group. Lower stream pH can increases the solubility and uptake of Hg in lower trophic-level biota (Ward et al. 2010), and may explain the elevated Hg levels we observed in periphyton for the F group. Additionally, lower pH in stream water increases uptake of Hg across the gills, gut, and exoskeletons in low-trophic level macroinvertebrates (Riva-Murray et al. 2011), and in higher trophic levels through reductions in growth efficiency (Mierzykowski et al. 2008). The form of mercury can also be important in determining its assimilation into food webs, but, we did not observe any significant increase in stream-water methylation efficiency for the F group. Additionally, watershed analyses suggested no change in catchment characteristics (e.g., hydric soils) between groups that are typically associated with increased methyl mercury production and availability for transport to adjacent streams.

Transport of Hg from terrestrial to aquatic environments may also be influencing increased MeHg bioaccumulation at the base of the food webs in the F group. We observed an increase in FTHg that is indicative of elevated transport from riparian to aquatic ecosystems. Other studies have suggested an increase in dissolved organic carbon (DOC) associated with fracking activities (Grant et al. 2015), presumably resulting from forest disturbance associated with fracking development. Further, Hg can form complexes with DOC, allowing for more rapid transport through watersheds and thus potentially higher concentrations available for accumulation in aquatic stream sediments and biota (Dittman et al. 2010).

Impact of fracking on aquatic biodiversity

Differences in macroinvertebrate biodiversity, and differences in fish community were observed between F and NF groups. We observed lower IBI scores and lower EPT richness at streams in the F group. Significantly lower IBI scores and EPT indices at the F group are likely the result of diminished water quality (Baptista et al. 2007, Wallace et al. 1996), because streams in the F group had lower pH and greater levels of stream-water Hg levels. While physiochemical differences between the F and NF group did not appear significantly influence fish biodiversity between groups, a shift in community composition was observed in the F group and high well-density streams. Fish biodiversity was low overall, with brook trout being the only fish captured across all streams (except Little Laurel Run-extirpated). Blacknose dace and longnose dace (Rhinichthys atratulus and Rhinichthys cataractae, respectfully) were abundant only in the NF group and low well-density streams in the F group. This response of fish community composition to acidic waters has been previously documented, as dace are known to be less tolerant of acidic waters than brook trout, presumably due to their inability to regulate ion concentrations at low pH (Jardine et al. 2013).

Biodiversity was most impacted at two streams in the F group with documented “frack water” fluid spills. Frack water fluid spills includes flowback water (a fluid containing clays, chemical additives, and dissolved metal ions, and TDS that flows back to the surface during and after the completion of fracking) and well wash fluid (fluid mixture of water and chemical constituents used to aid in the fracking process). In Little Laurel Run and Alex Branch, pH was lower, fish diversity was nil, brook trout were rare or absent, and no invertebrate scrapers were present. No fish were found at Little Laurel Run (Fig. 1), which is historically classified as a high-quality coldwater fishery containing naturally reproducing populations of brook trout. The PA Department of Environmental Protection documented a well blowout with flowback fluid reaching the Little Laurel Run in 2010 that decreased water quality. This event caused a fish kill (Levis 2011), and no subsequent repopulation of fishes had occurred by the time we sampled in 2012 and 2013. Further, ongoing contamination of Alex Branch (Fig. 1) between 2009 and 2010 resulted in over 7,000 gallons of surfactants, flowback fluid, and well-wash fluid released into the environment, which negatively impacted the stream ecosystem (Levis 2011). The spills into Alex Branch impacted fish communities and reduced brook trout numbers to less than five individuals captured/year. We also did not find any blacknose dace or longnose dace during our sampling. Interestingly, the three most acidic streams in this study had either experienced frack water fluid contamination (Alex Branch and Little Laurel) or had the highest well density of all streams (Stone Run; Fig. 1) included in our study. Brook trout are more tolerant than dace to acidic waters (Jardine et al. 2013), which may explain why brook trout were the only fish we found at these sites. However, zero brook trout at Little Laurel Run, and the low numbers at Alex Branch indicate that stream conditions (acidic or otherwise) are unfit for robust populations to survive. Lastly, macroinvertebrate scrapers were absent from Little Laurel Run and Stone Run, and it is commonly believed that if aquatic communities become impaired as a result of low pH and elevated levels of dissolved heavy metals, more specialized macroinvertebrate feeders, such as scrapers, are among the first to disappear (Barbour et al. 1998).

Food source impacts on mercury accumulation and trophic position

The eleven brook trout from six unique streams in the clustered outgroup had the highest δ15N and trophic level of all fish in the study. These individuals were from streams with and without hydraulic fracturing occurring within their watershed (NF = 3 streams, F = 3streams). The significant increase in δ13C values and terrestrial stomach contents for this clustered outgroup suggests a shift in food source. The low variation in δ13C values for brook trout suggest that they obtained their prey from the same (or very similar) source. The significant difference between δ13C values for predatory macroinvertebrates and brook trout suggests that the brook trout are not obtaining their prey from an aquatic environment. Research has shown that values above −30 δ13C often indicate a terrestrial food source, while values below −28 δ13C are likely to have derived carbon from aquatic environments (Seifert & Scheu 2012). Additionally, PMHg was slightly higher in streams in which clustered outgroup individuals were observed, perhaps indicating more overland flow and greater riparian connectivity, which could increase transport of terrestrial invertebrates to the stream. Since aquatic food sources generally contain higher Hg levels than terrestrial food sources, more consumption of terrestrial food at higher trophic levels could explain the lower concentration of MeHg for the brook trout in the clustered outgroup (Chételat et al. 2011).

Conclusion

Differences in MeHg assimilation at the base of the food webs and shifts in aquatic biodiversity observed in this study are likely the result of fracking development occurring within the watershed. More specifically, fracking activities to extract natural gas have already produced over six billion liters of flowback water in Pennsylvania alone (Ward et al. 2010), and the mismanagement of flowback has been suggested as a significant threat to surface water resources (Entrekin et al. 2011). Fracking and flowback fluids can contain various highly acidic agents (Kharaka et al. 2013), organic and inorganic compounds, and even Hg (NYSDEP 2011, Ferrar et al. 2013). The flowback fluids can reach nearby streams through leaking wastewater hoses, impoundments, and lateral seepage and blowouts, as well as by backflow into the wellhead (Llewellyn et al. 2015a, Rahm 2011, Peduzzia & Harding 2013). Flowback water reaching streams can directly impact stream physio-chemistry, as well as decrease aquatic biodiversity by limiting suitability for more sensitive taxa. Further, lowered stream pH increases Hg solubility, leading to increased bioaccumulation in food webs (Kelly et al. 2003, Jardine et al. 2012), and likely resulting in increased MeHg concentrations at the base of food webs.

Our findings are particularly important due to the rapid international expansion of fracking associated with unconventional natural gas development and the limited amount of research addressing ecological impacts to stream ecosystems. This work builds upon previous findings by documenting increased MeHg at the base of food webs resulting from increased acidity, elevated stream water FTHg, and increased Hg transport to steam occurring at streams with hydraulic fracturing occurring in their watershed. Additionally, decreases in macroinvertebrate IBI and EPT at streams with hydraulic fracturing in their watershed, and low or zero fish diversity, and low brook trout abundance at sites with past frackwater contamination (or highest well-density number) are indicators of stream response to highest impact areas. Pre/post-fracking ecological data at streams could allow investigators to decipher the mechanisms of impact. Future efforts toward partitioning effects of land disturbance from infrastructure (well pads, pipelines, roads), fracking activities (fracking and flowback production), and fracking spills is necessary to establish pathways of contamination toward development of best management practices to assure for ecologically sound extraction of natural gas.

Acknowledgments

The authors would like to thank the Colcom Foundation for providing primary funding for this project.

The authors would also like to thank Regina Lamendella, Nicole Marks, Elliot Perow, Alexander Weimer, Ryan Trexler, Alyssa Grube, Jacob Oster, Morgan Decker, and Krista Leibensperger who helped with field sampling and lab analysis. The authors would like to thank Kristen Brubaker for her help with GIS analysis, and Roy Nagle and Kimi Cunningham Grant for their thorough review of this manuscript which greatly improved its quality for submission. They would also like to thank the anonymous reviewers who significantly contributed to the focus and tact of the discussion. This work would not have been possible without the approval for sample collection by the PA Fish and Boat Commission and the Institute for Animal Care and Use Committee (IACUC) at Juniata College. They would like to thank the Keystone Elk County Alliance in helping to provide housing for fieldwork. Finally, the authors would like to thank the many people they encountered in small towns, at hunting camps, on the roads, and in the woods, for their directions to sampling sites and their interest in and support of our research.

Funding

This study was funded by the Colcom Foundation (grant #1201603).

Compliance with ethical standards

Conflict of interest

The authors declare that they have no competing interests.

Ethical approval for vertebrates

All applicable international, national, and/or institutional guidelines for the care and use of animals were followed. We have an approved Institute for Animal Care and Use Committee (IACUC # 20120503) through Juniata College, and collected all specimens under Christopher Grant’s scientific collector permit (#604) from the PA fish and boat commission.

Ethical approval for humans

This article does not contain any studies with human participants performed by any of the authors.

Copyright information

© Springer Science+Business Media New York 2016

Authors and Affiliations

  • Christopher James Grant
    • 1
  • Allison K. Lutz
    • 2
  • Aaron D. Kulig
    • 1
  • Mitchell R. Stanton
    • 3
  1. 1.Juniata Collegevon Liebig Center for ScienceHuntingdonUSA
  2. 2.Biology DepartmentGeorgia Southern UniversityStatesboroUSA
  3. 3.Utah Division of Wildlife ResourcesVernalUSA

Personalised recommendations