Article pubs.acs.org/est
Microbial Community Changes in Hydraulic Fracturing Fluids and Produced Water from Shale Gas Extraction Arvind Murali Mohan,†,‡ Angela Hartsock,† Kyle J. Bibby,§,⊥ Richard W. Hammack,† Radisav D. Vidic,†,§ and Kelvin B. Gregory*,†,‡ †
National Energy Technology Laboratory, Pittsburgh, Pennsylvania 15236, United States Department of Civil and Environmental Engineering, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213, United States § Department of Civil and Environmental Engineering, University of Pittsburgh, Pittsburgh, Pennsylvania 15261, United States ⊥ Department of Computational and Systems Biology, University of Pittsburgh Medical School, Pittsburgh, Pennsylvania 15260, United States ‡
S Supporting Information *
ABSTRACT: Microbial communities associated with produced water from hydraulic fracturing are not well understood, and their deleterious activity can lead to significant increases in production costs and adverse environmental impacts. In this study, we compared the microbial ecology in prefracturing fluids (fracturing source water and fracturing fluid) and produced water at multiple time points from a natural gas well in southwestern Pennsylvania using 16S rRNA gene-based clone libraries, pyrosequencing, and quantitative PCR. The majority of the bacterial community in prefracturing fluids constituted aerobic species affiliated with the class Alphaproteobacteria. However, their relative abundance decreased in produced water with an increase in halotolerant, anaerobic/facultative anaerobic species affiliated with the classes Clostridia, Bacilli, Gammaproteobacteria, Epsilonproteobacteria, Bacteroidia, and Fusobacteria. Produced water collected at the last time point (day 187) consisted almost entirely of sequences similar to Clostridia and showed a decrease in bacterial abundance by 3 orders of magnitude compared to the prefracturing fluids and produced water samplesfrom earlier time points. Geochemical analysis showed that produced water contained higher concentrations of salts and total radioactivity compared to prefracturing fluids. This study provides evidence of long-term subsurface selection of the microbial community introduced through hydraulic fracturing, which may include significant implications for disinfection as well as reuse of produced water in future fracturing operations.
■
INTRODUCTION The Marcellus shale of the Appalachian basin is one of the largest natural gas deposits in the U.S., with estimates ranging from 262 to 500 trillion cubic feet of recoverable gas.1,2 Commercial production of natural gas from deep shale formations, such as the Marcellus, requires stimulation through hydraulic fracturing3,4 where large volumes (∼5000−20 000 m3) of water mixed with chemical additives is injected at high pressure to introduce fissures in the formation and optimize the flow rate of gas to the well bore.3,5 A typical fracturing fluid contains water (90.6%), sand as a proppant (8.95%), and other chemicals (0.45%) including biocides, friction reducers, corrosion inhibitors, and viscosity modifiers.6 After hydraulic fracturing, the pumping pressure is released, and 30−70% of the injected fluid flows back to the surface.4 Water that returns to the surface is referred to as “produced water”. The initial flow rate of produced water is high and diminishes rapidly as the reservoir pressure decreases,5 and it continues to decrease through the gas-production phase. Produced waters are brine © 2013 American Chemical Society
solutions with salts, metals, and organics as well as additives from the hydraulic fracturing solution.5,7−9 Deleterious microbial activity associated with produced waters in the subsurface and in surface storage facilities, such as impoundments and tanks, can lead to adverse environmental impacts and increased production costs.10−12 Microbial activity can cause the souring of natural gas because of sulfide production, a decrease in formation permeability because of plugging, and can corrode equipment and flowlines.10−13 There are several potential sources for microorganisms that emerge in produced water, including the makeup water used to prepare the fracturing fluid and drilling muds used in the horizontal-drilling operations.12,14 Microbial growth and activity in oil and gas environments is commonly Received: Revised: Accepted: Published: 13141
July 2, 2013 September 17, 2013 October 2, 2013 October 2, 2013 dx.doi.org/10.1021/es402928b | Environ. Sci. Technol. 2013, 47, 13141−13150
Environmental Science & Technology
Article
controlled through the use of commercial biocides.6,12,15 However, the elevated concentration of salts, metals, and organics as well as the long residence times of these fluids in the shale can decrease biocide efficacy.16,17 Studies from the Barnett shale have shown that microbial communities recovered in the flowback and produced water are distinct from those in fracturing fluids and that microbial communities are capable of surviving commercial biocides added for microbial control prior to hydraulic fracturing.18,19 These studies provide evidence for the existence of a dynamic and adaptable microbial community in natural gas brines. The lack of Class II deep-well disposal options in Pennsylvania has driven the industry toward the reuse of produced fluids as a component of the makeup water for subsequent hydraulic fracturing.5 Reuse of produced water in hydraulic fracturing reduces water management costs as well as freshwater withdrawals from surface or municipal sources.20 However, microbial populations in flowback and produced water may serve as a seed for the rapid development of microbial communities in the subsequent wells that could potentially accelerate sour-gas production, biofouling, and scale formation.18,19 Despite the importance of microbial communities in the management of waste brines following hydraulic fracturing, few studies are available14,18,19,21 and none consider the unique water-management environment that has emerged during the development of the Marcellus shale region. In this study, we consider a horizontal-drilling and hydraulic fracturing operation of a natural gas well located in the Marcellus shale region of southwestern Pennsylvania. These operations utilized synthetic oil-based drilling mud and a fracturing fluid composed of both freshwater and recycled produced water. These features represent common practices in Marcellus shale and will provide a useful comparison to contrasting practices studied in the Barnett shale where waterbased drilling muds and freshwater are more commonly utilized in drilling and fracturing operations.14,18 We identified microbial communities, estimated diversity and richness, and determined microbial abundance in the hydraulic fracturing fluid and produced water at different time points using 16S rRNA gene-based clone libraries, tag-encoded pyrosequencing, and quantitative PCR. The concentrations of metals and anions and the levels of total radioactivity in these fluids were also determined. The goals of this study were to understand how microbial communities and associated geochemical parameters change in the water utilized and generated from hydraulic fracturing.
(paraffinic solvent), breaker (sodium persulfate), and friction reducer (hydrotreated petroleum distillate) were amended. This mixture is termed ‘fracturing fluid’ and is injected at high pressure to fracture the shale formation. The maximum ingredient concentration in the fracturing fluid by mass was water (88.9%), proppant (10.9%), biocide (0.0043%), breaker (0.0021%), HCl (0.10%), gel (0.0025%), scale inhibitor (0.0022%), and friction reducer (0.02%). Samples of fracturing source water and fracturing fluid were collected from pipelines before the fracturing operation. Produced water was collected on days 1, 7, and 9 from valves located at the well head following which the well was shut off for approximately 5 months before gas production commenced. A produced water sample was collected on day 187 from the gas−water separator. A sample of synthetic oil-based drilling mud was also collected from a closed tank in the drilling rig prior to the drilling operations. All samples were collected in sterile 1 L polypropylene bottles, which were filled without headspace and sealed using screw caps to prevent oxygen intrusion. Samples were stored on ice during brief transportation and frozen at −80 °C within 5 h of sampling. Chemical Analyses. Cations were analyzed by inductively coupled plasma optical emissions spectroscopy (ICP-OES) on a PerkinElmer Optima 3000 Radial View spectrometer (PerkinElmer, Waltham, MA) according to U.S. EPA method 6010C. Anions were analyzed using a Dionex ICS-3000 ion chromatogram with an IonPac AS18 column (Dionex, Sunnyvale, CA) according to EPA method 300.1 with 28 mM KOH as the eluent at a flow rate of 1 mL/min. The total radioactivity in water samples (α and β emissions) was measured in a Packard 2100TR liquid scintillation counter (PerkinElmer). To 4 mL of the sample, 14 mL of Ultima Gold Cocktail (PerkinElmer) was added in a scintillation vial and counts were recorded over a period of 60 min. DNA Extraction, 16S rRNA Gene-Based Clone Library, and Tag-Encoded Pyrosequencing Techniques. Unfiltered water samples were centrifuged at 6000g for 30 min in an Avanti J-E centrifuge (Beckman Coulter, Brea, CA) to pellet the cells. DNA was extracted from a 0.25 g cell pellet using MO BIO power soil DNA isolation kit (MO BIO, Carlsbad, CA) according to the manufacturer’s instructions. DNA from synthetic oil-based drilling mud was extracted from 0.25 g of an oil mud mixture using the same procedure. For construction of clone libraries, 16S rRNA gene fragments were amplified using the universal bacterial and archaeal primer sets BAC 338F/907R22,23 and ARCH 344F/915R.24 The composition of the PCR reaction mixture and thermal cycling conditions are detailed in the Supporting Information. PCR products were purified using the QIAquick PCR purification kit (Qiagen, Valencia, CA), and clone libraries were constructed from the purified products using the TOPO TA cloning kit (Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions. Clones were sequenced at Functional Biosciences (Madison, WI) using the M13F primer. For pyrosequencing, 16S rRNA gene fragments were amplified using the universal bacterial and archaeal primer set F515 and R806.25 The primer F515 included a Roche A pyrosequencing adapter (CCATCTCATCCCTGCGTGTCTCCGACTCAG) (Roche Applied Science, Branford, CT) and a sample specific 10 bp barcode (Roche Applied Science). The R806 primer included a Roche 454-B pyrosequencing adapter (CCTATCCCCTGTGTGCCTTGGCAGTCTCAG)
■
MATERIALS AND METHODS Sampling. Samples of synthetic oil-based drilling mud, fracturing source water, fracturing fluid, and produced water at different time points were collected from a horizontal natural gas well in Westmoreland County, Pennsylvania, USA from October 2011−April 2012. The true vertical depth of the well was 8036 feet and a total of 25.4 million liters of water was utilized for hydraulic fracturing. The source water used for fracturing was a mix of fresh reservoir water (∼80%) and produced water (∼20%) from previous fracturing operations. Fresh and produced water were stored in separate pits on-site before mixing. The fracturing source water was pumped through a series of trucks where fracturing additives including proppant (silica sand), scale inhibitor (ammonium chloride), biocide (mixture of tributyl tetradecyl phosphonium chloride, methanol, and proprietary chemicals), hydrochloric acid, gel 13142
dx.doi.org/10.1021/es402928b | Environ. Sci. Technol. 2013, 47, 13141−13150
Environmental Science & Technology
Article
Table 1. Chemical Composition of Source Water, Fracturing Fluid, and Produced Water (PW) Days 1, 7, 9, and 187 concentration (mg/L) element
source water
fracturing fluid
PW day 1
PW day 7
PW day 9
PW day 187
Ba Sr2+ Ca2+ Cl− Br− Mg2+ Na+ K+ total Fe total S NO3− SO42− radioactivity (pCi/L)b
71.9 126 522 3635 35.3 48.3 2953 26.3 BDLa 7.4 4.7 6.2 171
110 209 866 5980 56.6 78.8 4541 47.5 0.7 11.9 7.6 9.1 198
473 473 1885 18 626 161 182 13 899 158 4.2 32.4 13.9 31.7 3062
2118 1859 6179 63 596 485 690 42 146 246 83.3 56.8 11.3 10.7 8634
2077 1910 6071 63 106 492 699 43 094 251 81.6 55.7 13.4 9.3 9031
3169 2687 9994 91 800 876 1255 44 770 294 109 51.7 18.1 9.5 18 300
2+
a
BDL, below detection limit. bTotal α and β emissions.
submitted to GenBank (accession numbers are yet to be assigned). Quantitative PCR (qPCR). Quantification of bacterial 16S rRNA genes was performed on DNA extracts using TaqMan universal PCR master mix (Applied Biosystems, Foster City, CA), primers BAC 1364F/PROK 1492R, and probe TM1389F.32 The composition of the PCR reaction mixture and thermal cycling conditions are detailed in the Supporting Information. Standard curves were generated using genomic DNA from E. coli K-12 (ATCC 33876). The number of 16S rRNA gene copies in standards and samples was determined based on 10-fold serial dilutions of the standards as outlined previously.33
(Roche Applied Science). The composition of the PCR reaction mixture and thermal cycling conditions are detailed in the Supporting Information. PCR amplicons were submitted to the Ohio State University Plant-Microbe Genomics Facility for sequencing on the Roche 454 Genome Sequencer FLX Titanium System (http://pmgf.osu.edu/). Analysis of Sequencing Data. Raw sequences from 454 pyrosequencing were aligned to the SILVA database using the Mothur program26 and were checked for chimeras using the UCHIME program27 implemented within Mothur. Chimeric sequences and sequences with ambiguous bases, shorter than 200 bp in length and >8 bp homopolymer stretch, were discarded. A total of 42 760 sequences were selected for further analysis. Mothur was used to group sequences with ≥97% identity into operational taxonomical units (OTUs) and to calculate Chao and ACE estimates of species richness, nonparametric Shannon-Weaver measure of diversity, and community coverage. Data from pyrosequencing was classified at the class and order level using the Ribosomal Database Project’s Naı̈ve Bayesian Classifier tool.28 Representative OTUs (97% similarity) along with an archaeal outgroup (Genbank accession number AM180088) were used to construct a phylogenetic tree using the Clear-cut program29 implemented within Mothur. The tree, along with an environmental file including OTU abundances, was used for weighted normalized Unifrac analysis.30 Pairwise comparison of individual environments was carried out using the Parsimony test (p test) implemented within Mothur; p values ≤0.001 were considered to represent significantly different microbial communities. Raw sequences from the clone libraries were subjected to similar quality control parameters as the pyrosequencing data. Sequences were trimmed to a similar length, and those shorter than 450 bp were discarded. After quality control, a total of 476 sequences were used for further analysis. Mothur was used to group sequences with ≥97% identity into operational taxonomical units (OTUs) and to calculate Chao and ACE estimates of species richness, nonparametric Shannon-Weaver measure of diversity, and community coverage. The Classifier tool28 was used to classify sequences at the class and order level, and the BLASTn algorithm31 was used to assign specieslevel affiliations to representative OTUs. Representative bacterial sequences from clone libraries were assigned GenBank accession numbers JX296018−JX296111. Representative sequences from pyrosequencing have been
■
RESULTS Aqueous Geochemical Characteristics. Produced waters contained higher concentrations of cations and anions than corresponding prefracturing fluids (Table 1). The concentration of the measured ions increased with time in produced water, a trend described previously for flowback and produced waters from hydraulic fracturing operations in the Marcellus shale.9,34,35 The total dissolved solids (TDS) composition in produced waters was dominated by Na+, Cl−, and Ca2+, and also contained elevated concentrations of Ba2+, Sr2+, Mg2+, and Br − compared to the prefracturing fluids (Table 1). Concentrations were similar to prior reports for flowback and produced waters from the Marcellus and the Barnet shale.5,9,19,34−36 Produced water day 7, 9, and 187 samples had greater salinities (10.4−11.4% on the basis of NaCl concentrations) than prefracturing fluids (0.6−1% NaCl). The concentration of iron, undetected in the fracturing source water, increased to 81−109 mg/L with time in produced water samples. A major fraction of the total sulfur detected in the prefracturing fluids and produced water day 1 was present as sulfate, whereas sulfate formed only a small fraction of the total sulfur in produced water day 7, 9, and 187 samples. The total radioactivity increased with time in produced water samples from 171 to 198 pCi/L in the prefracturing fluids to 18 300 pCi/L in the produced water day 187 sample (Table 1). The radioactivity that is present in the prefracturing fluids is likely from the diluted produced water that was used in the makeup water for hydraulic fracturing. Microbial Community Analysis. The results showed that bacteria constituted the dominant domain in all samples and 13143
dx.doi.org/10.1021/es402928b | Environ. Sci. Technol. 2013, 47, 13141−13150
Environmental Science & Technology
Article
Table 2. Diversity, Richness Estimates, Coverage, And Abundance of Bacterial 16S rRNA Genes in Source Water, Fracturing Fluid, and Produced Water (PW) Day 1, 7, 9, and 187 Samples from Pyrosequencing and qPCRa sample type source water fracturing fluid PW day 1 PW day 7 PW day 9 PW day 187
OTU0.03
coverage percent
Chao richness
abundance-based coverage estimation richness
Shannon diversity index
copies of bacterial 16S rRNA genes/mLb
2941 4014
168 237
98 92
285 435
387 562
3.5 3.5
8.3 × 106 1.6 × 107
7738 4692 2092 21 283
566 161 85 176
94 98 99 99
1373 393 124 489
3165 577 146 902
4.3 3.2 2.9 0.45
6.3 1.0 1.4 4.1
no. of sequences
× × × ×
106 107 107 104
a Diversity and richness estimates are based on ≥97% sequence identity. bqPCR values are an average of triplicate values with an average standard deviation of ±18%.
Figure 1. Relative abundance of bacterial 16S rRNA gene sequences in source water, fracturing fluid, and produced water (PW) day 1, 7, 9, and 187 samples from (a) tag-encoded pyrosequencing (b) clone libraries. The RDP Classifier tool was used to assign sequences to taxonomical classes at a cutoff of 80. For both techniques, classes that constituted
