Article pubs.acs.org/est
Strontium Isotopes Test Long-Term Zonal Isolation of Injected and Marcellus Formation Water after Hydraulic Fracturing Courtney A. Kolesar Kohl,†,‡ Rosemary C. Capo,*,†,‡ Brian W. Stewart,*,†,‡ Andrew J. Wall,†,§ Karl T. Schroeder,§ Richard W. Hammack,§ and George D. Guthrie§ †
Department of Geology & Planetary Science, University of Pittsburgh, Pittsburgh, Pennsylvania 15260, United States National Energy Technology Laboratory−Regional University Alliance, Pittsburgh, Pennsylvania 15236, United States § National Energy Technology Laboratory, 626 Cochrans Mill Rd., Pittsburgh, Pennsylvania 15236, United States ‡
S Supporting Information *
ABSTRACT: One concern regarding unconventional hydrocarbon production from organic-rich shale is that hydraulic fracture stimulation could create pathways that allow injected fluids and deep brines from the target formation or adjacent units to migrate upward into shallow drinking water aquifers. This study presents Sr isotope and geochemical data from a well-constrained site in Greene County, Pennsylvania, in which samples were collected before and after hydraulic fracturing of the Middle Devonian Marcellus Shale. Results spanning a 15-month period indicated no significant migration of Marcellus-derived fluids into Upper Devonian/Lower Mississippian units located 900−1200 m above the lateral Marcellus boreholes or into groundwater sampled at a spring near the site. Monitoring the Sr isotope ratio of water from legacy oil and gas wells or drinking water wells can provide a sensitive early warning of upward brine migration for many years after well stimulation.
■
flowback and produced water.15,17,18 While leakage along well casings and surface spills during disposal are the most likely vectors for produced water release,3,15 concerns have been raised about possible migration of gas and brine in the subsurface resulting from hydraulic fracturing.19−21 Warner et al.22 suggested that Marcellus-derived brines migrated to the near-surface naturally in north-central Pennsylvania on time scales of 104 years, raising the possibility that pathways exist that could be reactivated by hydraulic fracturing. However, the great depths to hydrocarbon-producing Marcellus Shale units (up to ∼3050 m)23 and high capillary tension could act to prevent significant upward movement of injected waters and formation brines.24−26 One limitation has been a general lack of field data documenting the chemistry of nonreservoir formation fluids before and after hydraulic fracturing, in part due to the difficulties in site access and in part due to challenges in observing small changes to high-TDS fluids. In addition, the ability to differentiate among multiple possible contributors to change in water quality can be difficult in dilute groundwaters, particularly if fluid migration takes place over long time periods.27−29 Here we present data from an active Marcellus shale gas production site in Greene County, Pennsylvania.30
INTRODUCTION Exploration companies in the United States, and increasingly throughout the world, use directional drilling combined with hydraulic fracturing technologies to stimulate and produce hydrocarbons.1 Hydraulic fracturing is used when hydrocarbons, usually methane, are held in relatively impermeable rock such as shale, and it involves the injection of millions of liters of water, sand (proppant), and chemical additives into the well to create and maintain open fractures in the rock, allowing gas to flow to the well. Usually more than half of the injected water stays in the stimulated reservoir,2 and the remainder is returned to the surface in the first few days after the well is opened (“flowback water”) and later at lower rates during natural gas production (“produced water”).3 Over the first few hours and days, the concentration of total dissolved solids (TDS) in the flowback water increases rapidly, evolving to Na− Ca−Cl brines with high levels of Ba, Sr, and Br,4−7 as well as relatively high Ra activity.8−11 Produced water tends to be highly saline (up to 345 000 ppm TDS7) which, although produced at relatively low rates, presents challenges for environmentally safe disposal.12−15 The Middle Devonian Marcellus Formation in the northeastern United States contains organic carbon-rich mudrocks that serve as both source and reservoir for significant quantities of natural gas.16 A key requirement for environmentally sound extraction of this gas is to ensure that shallow aquifers and surface waters are not impacted by possible contamination with © 2014 American Chemical Society
Received: Revised: Accepted: Published: 9867
March 4, 2014 July 14, 2014 July 15, 2014 July 15, 2014 dx.doi.org/10.1021/es501099k | Environ. Sci. Technol. 2014, 48, 9867−9873
Environmental Science & Technology
Article
Produced waters from pre-existing wells tapping the overlying Upper Devonian/Lower Mississippian (UD/LM) gas-producing sandstones were collected and analyzed for major elements and Sr isotopes both before and after hydraulic fracturing to assess the possible upward migration of fluids resulting from hydraulic fracturing for more than a year after fluid injection. Radiogenic strontium isotopes (87Sr/86Sr) have been previously shown to be a powerful tracer for fluids at or near the Earth’s surface.31−33 Strontium typically behaves conservatively in aqueous systems, and the 87Sr/86Sr ratio is not significantly affected by mass fractionation (including temperature-dependent and biological effects). Earlier work has shown that produced waters from the Marcellus shale have a relatively restricted range of 87Sr/86Sr ratios compared to other possible water and contaminant sources, including brines from Upper Devonian oil and gas wells, in the Marcellus natural gas producing region.5,34 The narrow 87Sr/86Sr combined with the high concentrations of Sr in Marcellus produced water,4−6,35 makes Sr isotopes a sensitive tracer of interaction between Marcellus and UD/LM produced waters and infiltration into shallow ground waters or surface waters. Because of the large difference between the Sr isotope ratio (87Sr/86Sr) of UD/LM produced water (0.720−0.721) and that of Marcellus Shale produced water (0.711−0.712), and the high (102−103 mg/L) Sr concentration of Marcellus Shale produced water, the 87 Sr/86Sr ratio is significantly more sensitive to brine intermixing than are elemental concentrations commonly measured in groundwater.5 The study site, here referred to as the Greene County site (GCS), is located in the heart of the Marcellus gas play (Figure 1). In this region, the Marcellus Shale is approximately 2470− 2500 m below the surface and is overlain by the Hamilton Formation (including the Tully limestone) and gas-producing sands and shales of the Upper Devonian Venango and Bradford Formations and the Lower Mississippian Pocono Group (Figure 2). Seismic studies show that the Marcellus shale is cut by NE-trending thrust faults that extend into the lower portions of the overlying Upper Devonian units, but not into the gas-producing sands (Figure S1).30 Natural gas has been produced at the site since 2006. The UD/LM producing zone is a siliciclastic interval 700−1400 m below the surface; most of the vertical wells in these units were completed in multiple zones (Figure 2). Two gas wells in the Marcellus shale were completed and hydraulically fractured in 2008 (MW-1 and MW-2), but because these are vertical wells, the fractures have a very limited spatial extent. In 2012, a total of six laterals (horizontal wellbores) were drilled into the Marcellus Shale, three in a northwesterly direction (wells MH-A, B, and C) and three to the southeast (wells MH-D, E, and F; Figure 1). Hydraulic fracturing in laterals D, E, and F, which underlie the UD/LM wells still producing water, began on June 4, 2012, and flowback commenced on June 10, 2012. Strontium isotope data from flowback and produced water from these horizontal wells were reported by Capo et al.35 Here we present the results of geochemical and Sr isotopic analyses of produced waters collected from five UD/LM wells overlying the laterals, a vertical Marcellus gas well, and a nearby spring, spanning a period of ∼4 months before hydraulic fracturing to ∼14 months afterward.
Figure 1. Site map showing the location of horizontal Marcellus gas wells and vertical Marcellus and Upper Devonian/Lower Mississippian gas wells (modified from Hammack et al.30), as well as a sampled groundwater spring. The inset shows the extent of the Marcellus Formation in North America, and the location of Greene County, Pennsylvania, where the study was conducted. Wells UD-1, UD-3, and MW-2 were not producing water at the time samples from this study were collected.
■
MATERIALS AND METHODS Sample Collection. Sample collection at the GCS was arranged through cooperation between DOE-NETL and the site owner/operator. Initial flowback samples from the horizontal Marcellus wells were collected by the site operators and provided to NETL personnel. Although we did not have access to samples of the injection fluid, early flowback Sr isotope and Sr/Ca data reported by Capo et al.35 provide some information about fluid chemistry as it evolved from injected fluid to produced water. After flowback shifted to production, samples were collected by NETL or USGS personnel. Samples were collected from vertical Marcellus well MW-1 and UD/LM wells UD-2, UD-4, UD-5, UD-6, and UD-7; wells MW-2, UD1, and UD-3 were not generating produced water over the duration of this experiment. For all vertical wells, samples were collected from the gas−liquid separator if liquid had accumulated at the time of collection. Otherwise, they were collected from storage tanks at the site of each well. The spring shown in Figure 1 represents the only accessible source of 9868
dx.doi.org/10.1021/es501099k | Environ. Sci. Technol. 2014, 48, 9867−9873
Environmental Science & Technology
Article
higher εSW Sr values up to three years after hydraulic fracturing. The GCS vertical Marcellus well MW-1 provides an opportunity to estimate this Marcellus Shale produced water endmember value, as this well had been in production for ∼4 years at the time the horizontal wells were hydraulically fractured. Figure 3 shows that the trend continues, with εSW Sr
Figure 3. Evolution of the Sr isotope composition over time from beginning of flowback for vertical and horizontal Marcellus wells from the Greene County site. Horizontal well data (MH-D, E, F) are primarily from Capo et al.35 Sample MW-1 is from a vertical Marcellus well that was hydraulically fractured and began production in 2008, four years prior to hydraulic fracturing of the horizontal wells.
rising from a maximum value of +32.1 for the horizontal wells 320 days after hydraulic fracturing (Capo et al.35 and Table S3) to a value of +33.9 in MW-1 four years after hydraulic fracturing in the same unit (Table S2; average value prior to hydraulic fracturing of the horizontal wells). The shift in εSW Sr subsequent to hydraulic fracturing (discussed in the next section) suggests that there remains a reservoir of formation water that continues to mix with the initial injection water even after 4−5 years of production. Even so, the increase in εSW Sr observed from one to five years after hydraulic fracturing is relatively minor, and the values obtained from well MW-1 and late-stage produced water from the horizontal wells at this site fall within a narrow range of values (+31 to +36). The total range of Marcellus produced waters reported to date5,22,38 is much larger (+14 to +42) and reflects large-scale geographic and geological variations across the Appalachian Basin. Produced waters from the conventional Upper Devonian/ Lower Mississippian gas wells collected prior to hydraulic fracturing (“prefrac”) of the horizontal Marcellus wells have εSW Sr values from +151.4 to +166.6 (87Sr/86Sr = 0.71990−0.72098), a total spread of about 15 ε units (Table S2). These values are in the range of those reported for other Upper Devonian produced waters in Pennsylvania (+93 to +181).22,34,38,39 Strontium isotope ratios of produced water from each of the UD/LM gas-producing units tends to fall within a narrow range that is isotopically distinct from adjacent wells. In contrast, major cation concentrations are relatively homogeneous across all UD/LM wells (Table S1). Within-well variation in εSW Sr ranges from ±0.4 to ±2.3 (variances given in Table S4). Figure 4 shows the prefrac UD/LM Sr isotope ratios plotted against molar Sr/Ca ratios. While there is some overlap among wells (particularly with produced water from wells UD-2 and UD-6), each defines a relatively distinct field. However, the UD/LM values can be considered rather tightly clustered when compared to produced water from vertical Marcellus well MW-1 (Figure 4 inset), as noted previously by Chapman et al.5 A hypothetical mixing curve between the average for MW-1 and
Figure 2. Schematic geologic cross section from the Greene County site across line A−A′ (Figure 1). Modified from Hammack et al.30
groundwater samples from the site. Samples collected for elemental and Sr isotope analysis were filtered to 0.45 μm with cellulose nitrate filters and preserved by acidification with ultrapure concentrated nitric acid (HNO3) to 2% HNO3 by volume. Elemental and Isotopic Analysis. Aliquots of the acidified sample were taken for elemental and Sr isotope analysis. Concentrations of Ba, Ca, Fe, K, Li, Mg, Na, and Sr were measured on a Spectroflame Modula ICP-AES following a modified version of EPA method 6010C; Sr isotope compositions were determined on a Thermo Neptune Plus multicollector ICP-MS following the procedure of Wall et al.36 Further details of analytical methods can be found in the Supporting Information, Section S.1. We report our data as εSW Sr , which is the deviation in parts per 104 of the 87Sr/86Sr ratio from that of modern seawater:37 ⎡ 87Sr/ 86Sr ⎤ sample εSrSW = 104⎢ 87 86 − 1⎥ ⎢⎣ Sr/ Srseawater ⎥⎦
Our value for 87Sr/86Srseawater is 0.709166 (relative to our normalized value for SRM 987 of 0.710240), based on multiple replicate measurements of seawater and SRM 987. The estimated external reproducibility for repeat runs of the same sample is ±0.2 ε units.
■
RESULTS AND DISCUSSION Endmember Characterization. For all produced and spring waters collected as part of this study, elemental concentrations are reported in Table S1, and Sr isotope ratios are reported in Table S2. Capo et al.35 found that produced waters from horizontal Marcellus shale gas wells, including some from this site, continued to evolve, albeit slowly, toward 9869
dx.doi.org/10.1021/es501099k | Environ. Sci. Technol. 2014, 48, 9867−9873
Environmental Science & Technology
Article
Figure 5. Sr isotope composition of Upper Devonian/Lower Mississippian produced waters before and after hydraulic fracturing of the horizontal Marcellus wells (MH-D, E, F), with the time of hydraulic fracturing indicated by the vertical gray bar. Circles represent samples from separators, triangles samples from storage tanks. The shaded region represents the 2σ variation (96% confidence interval) of prefrac and postfrac samples.
Figure 4. Plot of molar Sr/Ca ratios against Sr isotope composition for prefrac Upper Devonian/Lower Mississippian and vertical well MW-1 Marcellus produced waters. The inset shows a hypothetical mixing curve between Marcellus brines and UD/LM produced waters, with the numbers on the curve indicating the amount of Marcellus brine in the mixture. The main plot focuses in on the variation in the UD/LM samples (bottom right corner of inset), with the same mixing curve indicated as a solid line. The dashed line shows the linear correlation of all UD/LM prefrac samples; this correlation is thought to represent within-well mixing of produced waters from multiple units.
present study. Analysis using a Mann−Whitney U test yields pvalues >0.05 for all UD/LM wells except UD-4 (0.03; Table S4), indicating that the null hypothesis, no change after hydraulic fracturing, is likely. The low p-value for UD-4 largely results from a single prefrac datum that is slightly higher than, but within error of, most postfrac values (Figure 5). For the isotopic shifts to be considered significant enough to suggest an incursion of Marcellus-derived fluid, the mean εSr values would need to decrease by 1−3 ε units after hydraulic fracturing, depending on the constancy of the prehydraulic fracturing measurements. For the Upper Devonian reservoirs at this site, this corresponds to an addition of only 0.1−0.3% Marcellus produced water. For those major cations that show the greatest difference between the UD/LM and Marcellus produced waters (Ba, Li, Sr; Table S1), there is no systematic shift in UD/LM produced waters after hydraulic fracturing (Figure S2). The isotopic composition of the spring waters (samples SPW-#), representing local groundwater, lie between values for the Marcellus brines and the UD/LM brines, with an average 87 86 SW εSW Sr value of +59.3 ( Sr/ Sr = 0.71337). Spring water εSr values shift systematically slightly above and below the mean on a semiannual basis (Figure 6). Because the produced water has much higher Sr concentrations than spring water (102−103 vs 5 ε units, and a 0.05% addition would shift it by 20 ε units, compared to a total observed annual variation of ±0.8 ε units. Similarly, a 0.01% addition of UD/LM brine would shift the groundwater value upward by nearly 10 ε units, and an addition of 0.05% would shift it by over 30 ε units. The spring water does not shift unidirectionally either upward toward UD/LM values or downward toward Marcellus values, but the natural variation in spring water isotope values (±0.8 ε units) provides a basis for evaluating possible contamination in other situations. In the
the average prehydraulic fracturing UD/LM produced water demonstrates the sensitivity of Sr isotopes to incursions of Marcellus fluid into shallow conventional gas reservoirs (Figure 4). An addition of less than 2% of Marcellus fluid to the UD/ LM reservoirs could produce the total range in εSW Sr values. However, most of the UD/LM samples fall off the Marcellus mixing curve, and in fact the prefrac values define a strong linear correlation (R2 = 0.80; dashed line in Figure 4). This suggests that most of the prefrac geochemical variation reflects within-well mixing of produced water among the different UD/ LM reservoirs tapped by each well. Effects of Hydraulic Fracturing on Overlying Units. Elemental concentrations and Sr isotope compositions of produced waters collected after hydraulic fracturing of the underlying Marcellus (all collection dates 6/28/12 and later) are reported in Tables S1 and S2. Because the tank samples represent averaging over a longer period, samples collected from storage tanks generally showed less variation than those from the gas−liquid separator. Nonetheless, there is generally good agreement between tank and separator εSW Sr values. Both pre- and postfrac samples from UD/LM wells are plotted in Figure 5. The shaded regions represent the 95% confidence interval (2σ) for variations within for the pre- and postfrac sample groups. For three wells (UD-4, UD-6, and UD-7), only two prefrac data were available; the confidence interval was calculated the same way as the others, with the recognition that additional samples, had they been available, would likely have changed the size of the uncertainty envelope. We emphasize that the small number of prefrac data limit our ability to carry out rigorous statistical analysis, but there is no compelling evidence that the UD/LM reservoirs were measurably affected by upward-migrating Marcellus-derived brines subsequent to hydraulic fracturing over the ∼15 month duration of the 9870
dx.doi.org/10.1021/es501099k | Environ. Sci. Technol. 2014, 48, 9867−9873
Environmental Science & Technology
Article
from Upper Devonian conventional wells39,42 and Middle Devonian Marcellus produced water4−6,8,22,35,39 and compared it in a series of mixing models to GSC groundwater. Elemental data used in the mixing models are summarized in Table S5. The Sr isotope endmember for the modeled Upper Devonian brine (+157) is the average of Upper Devonian brines from the GSC, and the Marcellus produced water endmember (+35) is in the average of measured values from well MW-1. The groundwater εSW Sr value reflects that of the spring at the GCS (+59); this could vary considerably from site to site, which emphasizes the importance of obtaining baseline data prior to possible environmental impacts of gas exploration. The change in elemental concentrations (relative to the groundwater baseline) and Sr isotope composition as a function of produced water added are shown in Figure 7. Of the element
Figure 6. Variation in spring water Sr isotope ratio over the period of the study. The shaded vertical bar represents the period of hydraulic fracturing of horizontal Marcellus wells MH-D, E, and F. The dashed line represents the mean isotopic composition of the spring samples over the course of this study.
ideal case, shallow groundwater would be sampled above or near the hydraulically fractured zone with 2−3 seasons of prefrac baseline data to determine background variations that could be caused by changes in recharge or by surface sources of Sr such as agricultural amendments.40,41 The one well that showed a potentially significant change in εSW Sr subsequent to hydraulic fracturing is the vertical Marcellus well MW-1 that taps into the hydraulically fractured unit. Two prefrac measurements yielded values of +33.5 and +34.0, while the postfrac values jumped to an average of +35.9 (ranging from +35.6 to +36.1; Figure 3). The relatively low Mann− Whitney p-value (0.028; Table S4) also suggests a significant shift, again recognizing the limitations of the small number of prefrac values. One possibility is that UD/LM waters (high εSW Sr ) penetrated downward into fractures in the underlying Marcellus Formation. However, this is not consistent with mixing calculations; introduction of UD/LM water sufficient to affect this change (∼10%) would decrease the Sr concentration of the Marcellus produced waters because UD/LM water has almost an order of magnitude lower Sr content. In fact, the average postfrac MW-1 Sr concentrations are ∼200 mg/L higher than the prefrac values. We suggest that hydraulic fracturing within the Marcellus opened up new pathways for formation water within or below the shale to enter the vertical well, which would be consistent with the observed net increase in Sr concentration following hydraulic fracturing. Sensitivity of Sr Isotopes to Subsurface Brine Migration. Monitoring of ground and surface water for contamination from drilling and hydraulic fracturing requires sensitive tracers that can provide an early warning of unexpected fluid migration. Given the multiple potential sources of high TDS brine, knowledge of the source of exogenous fluid is also critical. The differences in chemistry between produced waters and fresh waters allow for potential early detection of produced water incursions, where elements with the highest produced water−fresh water concentration ratio are more likely to be detected first. Elements that fit this description include Ba, Sr, Br, Cl, Na, Ca, and Ra (activity). The high concentrations of Sr in produced waters, when combined with a difference in isotope ratio from the fresh water, make Sr isotopes a very sensitive natural tracer that can also identify provenance.5,34 To evaluate the sensitivity of elemental concentrations, ratios, and isotopic tracers for monitoring produced water migration, we compiled elemental data for produced water
Figure 7. Modeled changes in selected elemental concentrations, element ratios, and Sr isotope composition (εSr) of fresh groundwater relative to baseline values at the GSC as a function of the percent infiltration of Upper Devonian (top) or Marcellus (bottom) produced water. The vertical dashed line shows a hypothetical mixture of 0.001% produced water and 99.999% groundwater; in this case, Sr isotope ratios are shifted by up to 4 ε units, while element concentrations change by
