can groundwater sampling techniques used in monitoring ...espace.inrs.ca/6917/1/p3307.pdf · rivard...

Rivard et al. 2018. Environ. Monit. Assess., doi: 10.1007/s10661-018-6532-7

1

Can groundwater sampling techniques used in monitoring wells 1

influence methane concentrations and isotopes? 2

3

Christine Rivard1*, Geneviève Bordeleau1, Denis Lavoie1, René Lefebvre 2, Xavier 4

Malet1 5

1 Geological Survey of Canada, 490 rue de la Couronne, Québec, Quebec, Canada, G1K 9A9 6

2 Institut national de la recherche scientifique – Centre Eau Terre Environnement, 490 rue de la Couronne, 7

Québec, Quebec, Canada, G1K 9A9 8

*corresponding author :Tel : 418-654-3173, Email : [email protected] 9

10

Abstract 11

12

Methane concentrations and isotopic composition in groundwater are the focus of a 13

growing number of studies. However, concerns are often expressed regarding the integrity 14

of samples, as methane is very volatile and may partially exsolve during sample lifting in 15

the well and transfer to sampling containers. While issues concerning bottle-filling 16

techniques have already been documented, this paper documents a comparison of methane 17

concentration and isotopic composition obtained with three devices commonly used to 18

retrieve water samples from dedicated observation wells. This work lies within the 19

framework of a larger project carried out in the Saint-Édouard area (southern Québec, 20

Canada), whose objective was to assess the risk to shallow groundwater quality related to 21

mailto:[email protected]


2

potential shale gas exploitation. The selected sampling devices, which were tested on 10 22

wells during three sampling campaigns, consist of an impeller pump, a bladder pump, and 23

disposable sampling bags (HydraSleeve). The sampling bags were used both before and 24

after pumping, to verify the appropriateness of a no-purge approach, compared to the low-25

flow approach involving pumping until stabilization of field physicochemical parameters. 26

Results show that methane concentrations obtained with the selected sampling techniques 27

are usually similar and that there are no systematic bias related to a specific technique. 28

Nonetheless, concentrations can sometimes vary quite significantly (up to 3.5 times) for a 29

given well and sampling event. Methane isotopic composition obtained with all sampling 30

techniques are very similar, except in some cases where sampling bags were used before 31

pumping (no-purge approach), in wells where multiple groundwater sources enter the 32

borehole. 33

34

Keywords: groundwater, sampling techniques, dissolved methane, shale gas, monitoring 35

36

Introduction 37

38

Public concerns about shale gas development are largely related to groundwater quality, 39

with fear that hydraulic fracturing fluids, methane, or saline brines from deep hydrocarbon 40

reservoirs could contaminate shallow aquifers (Lefebvre 2017). Over the last decade, the 41

oil and gas industry has started collecting groundwater samples from residential, farm and 42

monitoring wells up to 1 km around unconventional energy wells prior to and following 43


3

drilling. This sampling work is required by many jurisdictions (Jackson and Heagle 2016), 44

but the industry has also carried out such sampling even if the local jurisdiction does not 45

require it, due to public concerns and eventual legal dispute about the impact of their 46

activities on the groundwater quality of surrounding domestic wells. 47

For baseline studies, the type of wells and sampling techniques should be carefully 48

evaluated when trying to obtain a representative picture of methane concentrations in a 49

given area. For instance, Jackson and Heagle (2016) have highly recommended that 50

dedicated observation wells be used for monitoring, due to potential water quality and poor 51

maintenance issues associated with residential wells. The selection of a sampling technique 52

(with respect to both water withdrawal and bottle filling methods) is especially important 53

for water highly charged with dissolved gases (i.e. effervescing samples), as it can impact 54

concentration results (Humez et al. 2016; Molofsky et al. 2016). Indeed, the amount of gas 55

that may dissolve in groundwater at a certain depth depends on the water pressure, which 56

is related to the height of the overlying water column. When downhole gas concentrations 57

are high, as the sample is being lifted to the surface (progressively lowering water 58

pressure), and then poured into containers at atmospheric pressure, some of the dissolved 59

gas will exsolve and thus be lost from the water sample. Factors affecting exsolution 60

include flow type (laminar versus turbulent), pressure changes in the well due to 61

drawdown, the technique used to lift groundwater to the surface, and the bottle filling 62

technique (Gorody et al. 2005 and 2012; Hirshe and Mayer 2009; Coleman and McElreath 63

2012; Molofsky et al. 2016). 64

Issues related to bottle filling procedures for effervescing groundwater have been reported 65

elsewhere (Humez et al. 2015; Smith et al. 2016; and especially in Molofsky et al. 2016). 66


4

A commonly used technique involves filling and capping bottles at the bottom of a larger 67

container filled with purge water, as first published by the USGS 68

(http://water.usgs.gov/lab/chlorofluorocarbons/sampling/bottles/). While being considered 69

a semi-closed system (not in direct contact with atmosphere), this technique still involves 70

potential sample degassing. Such issues are better controlled in closed systems such as the 71

recently developed IsoFlasks® containers (Isotech Laboratories Inc., Champaign, IL), 72

consisting of single-use flexible plastic pouches connecting directly to a sampling tube in 73

the field, and then to a GC-FID at Isotech Laboratories Inc. This technique allows the 74

quantification of both dissolved and free gas phases, which allows computing the original 75

downhole methane concentration. This type of device is currently only available from one 76

supplier, and not many laboratories are set-up to handle such analyses. Consequently, the 77

materials and related analyses are expensive, which is an important limiting factor, 78

especially in studies involving a large number of samples. Noteworthy, in non-effervescing 79

samples, both semi-closed and closed systems have been shown to give comparable results 80

(Molofsky et al. 2016). 81

Issues related to sample lifting have been comparatively less studied, especially for highly 82

volatile compounds such as methane. When sampling dedicated observation wells, several 83

methods can be used for lifting the sample to the surface. These include various types of 84

pumps (e.g., suction, inertial lift, impeller, bladder) or “no-purge” devices (e.g., downhole 85

samplers), which can all target specific intervals within a well. The selection of a purge 86

versus no-purge approach, and even the pump-specific mechanism for lifting the sample, 87

could potentially have an effect on results. 88

http://water.usgs.gov/lab/chlorofluorocarbons/sampling/bottles/


5

The objective of the present study is therefore to compare some commonly used water 89

withdrawal techniques, as well as a purge versus no-purge approach, and to see how these 90

techniques may affect both concentrations and isotopic composition of methane in 91

groundwater. This work was conducted within the framework of a larger environmental 92

project carried out in the St. Lawrence Lowlands (eastern Canada), with the objective to 93

assess the risk to shallow groundwater quality from upward fluid migration related to 94

eventual shale gas development in the Utica Shale (Lavoie et al. 2014; Bordeleau et al. 95

2017; Rivard et al. 2017). Within this larger project, both residential (n=30) and 96

observation (n=14) wells were sampled, often several times, amounting to nearly 250 97

samples. Restrictive conditions in some of these wells (e.g., deep targeted sampling interval 98

or very low yield) precluded the use of a single sampling device for all wells. Additionally, 99

the generally low water yield of open borehole wells in this region led to significant 100

drawdown and long recovery times in some of the observation wells, making a no-purge 101

sampling approach appealing. For these reasons, a specific study to verify whether different 102

sample lifting techniques could be used interchangeably within the project was undertaken. 103

To do so, ten observation wells were selected, and three distinct sampling campaigns were 104

conducted. The selected wells represent a wide range of depths, methane concentrations 105

and isotopic composition. At each of these wells, sampling was done consecutively using 106

three commonly used water withdrawal techniques (impeller pump, bladder pump, and 107

downhole sampling bag). Furthermore, the sampling bag was also used prior to any 108

pumping to study the effect of a purge versus no-purge approach. The first sampling 109

campaign, however, only included the impeller pump and sampling bags after pumping. 110

Preliminary tests were also conducted with a peristaltic (suction) pump in three wells. The 111


6

characteristics of each sampling technique, along with the results obtained for 112

concentrations and isotopes, are used to recommend the best technique and to discuss the 113

use of other techniques when well conditions are restrictive. 114

It is noteworthy that sampling groundwater containing dissolved gases in various 115

concentrations, and in particular methane, is needed in other situations than those related 116

to hydrocarbon development. For instance, biogenic methane is often produced during 117

anaerobic in situ biodegradation processes that involve the addition of organic substrates 118

to reduce chlorinated volatile organic compounds (CVOCs), nitrate, hexavalent chromium 119

(CrVI) and perchlorate (EPA 2013). Also, landfill gas produced by the decomposition of 120

organic wastes is made up of methane and CO2, so methane can also be found in 121

groundwater adjacent to landfills (Nastev et al. 2001). The recommendations provided in 122

this paper are applicable to any monitoring program involving groundwater with high 123

concentrations of dissolved gases. 124

125

Previous studies on the sampling of groundwater with high dissolved 126

gas content 127

128

Few field studies have assessed the impacts of sampling techniques in groundwater 129

containing high concentrations of dissolved gas, even though it has long been known that 130

sampling gas-charged water from wells is challenging. Several studies starting in the late 131

1980s investigated the effect of sampling devices (both purge and no-purge) on different 132

chemical components, often including volatile organic compounds (VOCs) (e.g., Muska et 133


7

al. 1986; Devlin 1987; Barker and Dickout 1988; Parker and Clark 2004; McHugh et al. 134

2015), but no shorter-chain hydrocarbons such as methane. Early studies were reviewed by 135

Parker (1994) who concluded that significant problems of degassing and loss of VOCs had 136

been encountered with almost all the samplers and that, generally, bladder pumps gave the 137

best overall recovery of sensitive constituents while suction-lift pumps had one of the 138

poorest performances. Suction-lift pumps (such as peristaltic pumps) apply a vacuum to 139

the groundwater sampled that can cause depressurization and degassing of the samples 140

(Parker 1994). It must be emphasized that some of the devices reviewed in Parker (1994) 141

have evolved since then and that low-flow rates, minimizing degassing, is now being 142

routinely used. Of note, McHugh et al. (2015) cited several references where samples 143

collected using no-purge methods showed little or no bias in contaminant concentrations 144

(including VOCs but not methane) compared to samples collected after well purging. 145

The U.S. Interstate Technology Regulatory Council (ITRC 2007) provided guidance for 146

proper deployment and collection of groundwater samples containing a variety of 147

contaminants, including volatile gases, using five no-purge sampling technologies. The 148

ITRC literature review did not include testing. Two grab samplers, namely the 149

HydraSleeve (Las Cruces, NM, USA) bags, and the Snap sampler (ProHydro inc., Fairport, 150

NY), and different types of passive diffusion samplers are discussed in this report. Both 151

the HydraSleeve bags and Snap samplers were recommended for the sampling of dissolved 152

gasses such as methane. A significant disadvantage of the Snap sampler is that it can only 153

collect small samples, the largest being 350 mL for polypropylene bottles (and 40 mL for 154

glass bottles) and only four bottles can be connected in a series. Therefore, this could be a 155

major limitation, as sampling campaigns sometimes require much larger volumes of water 156


8

for a series of analyses, including isotopic ratios. HydraSleeve bags come in different sizes 157

(the maximum being 2.5 L for a 10 cm diameter) and are designed for single use, as they 158

should be perforated using the dedicated straw to transfer the water from the bag into the 159

sampling bottles. However, this procedure can be circumvented and the bag can often be 160

reused a few times in the same well, if larger water volumes are needed (see Section 161

“Groundwater sampling techniques”). Nonetheless, the thin plastic bags are somewhat 162

fragile, and can generally only be reused once or twice before being damaged (pierced) if 163

the borehole walls are not completely smooth. These two sampling devices can either 164

provide an immediate sample when deployed, or be left within a well at the desired 165

sampling depth for a few days. The latter allows sufficient time for the water within the 166

well to re-equilibrate after being “disrupted” by the positioning of the sampler and for 167

concentrations inside the sampler to equilibrate with the in situ chemical constituents. The 168

latter case, also called passive sampling, thus requires two visits, one for the installation 169

and one for the removal of the device, a time- and money-consuming exercise. 170

Furthermore, it is now recognized that passive sampling is not particularly useful for 171

collecting samples for methane concentrations, since the latter were often shown to vary 172

considerably over time, even over short periods (Hirshe and Myer 2009; Gorody 2012; 173

Humez et al. 2015; Smith et al. 2016). 174

175

Description of the study area 176

177


9

The 500 km2 study area is located in the St. Lawrence Lowlands, in southern Quebec, 178

Canada (Figure 1), where shale gas exploration activities targeting the Utica Shale went on 179

from 2006 to 2010, before a de facto fracking moratorium came into force. It is situated 180

about 65 km south-west of Quebec City and extends from the outer edge of the Appalachian 181

piedmont northward to the St. Lawrence River. Most of our study area is located in the St. 182

Lawrence Platform, mainly composed in this region of black shale that contains variable 183

content of organic matter with subordinate siltstone. The shallow bedrock geology is made 184

of three Upper Ordovician clastic units: the Lotbinière, Les Fonds and Nicolet formations. 185

The Lotbinière and Les Fonds formations are time- and facies-correlative with the Utica 186

Shale, which is present at a depth of 1.5 to 2 km in this area. 187

Most residential wells in the Lotbinière area are drilled into bedrock; their depth is 50 m 188

on average. Bedrock is mainly composed of shale and is thus poorly permeable: rock 189

hydraulic conductivities in the region vary between 10-9 and 10-6 m/s (Ladevèze et al. 190

2016). Dissolved hydrocarbons in groundwater originate from the shallow bedrock units 191

(Lavoie et al. 2016). Figure 1 shows the near-surface geology and location of observation 192

wells specifically drilled for this project. The location of observation wells was selected in 193

order to obtain a good spatial distribution over the three geological formations. These 194

observation wells were found to be either under semi-confined or confined conditions 195

(Ladevèze et al., 2016; Ladevèze, 2017). 196

Methane concentrations in groundwater in the Saint-Édouard region, obtained from 14 197

observation wells and 30 residential wells, vary between the detection limit (0.006 mg/L) 198

to more than 80 mg/L (Bordeleau et al. 2017), with a median of about 4 mg/L. Higher 199

methane concentrations are associated with more evolved waters (Na-HCO3, Na-HCO3-Cl 200


10

and Na-Cl water types). Above laboratory-measured concentrations of ~20 to 25 mg/L of 201

methane, the in situ groundwater is usually considered highly charged or even 202

supersaturated (effervescent). This condition leads to obvious issues related to sampling of 203

groundwater downhole (thus under the pressure of a water column) and bringing it to 204

atmospheric pressure to fill sampling bottles. During sampling, gas bubbles were observed 205

in the tubing for many of the wells. Ten observation wells were selected for this study. 206

They are all open to the bedrock and have a sealed metal casing through the overburden. 207

Their total depth varies from 30 to 60 m and their sampling depth ranges from 7.5 to 54 m. 208

The characteristics of observation wells are provided in Table 1 and their locations are 209

shown on Figure 1. 210

211

Field and laboratory methodology 212

213

When sampling the ten selected observation wells, care was taken to minimize drawdown 214

and water disturbance, and samples were always collected at the same targeted depth within 215

a well, where flowing fractures had previously been identified using borehole geophysics. 216

The goal was to collect groundwater samples from these flowing fractures, which is 217

representative of the surrounding bedrock aquifer. To verify the representativeness of 218

groundwater withdrawn with this technique, physico-chemical profiles were measured at 219

every five meters within four observation wells and they provided very distinct 220

characteristics for pH, electrical conductivity, salinity and dissolved oxygen, indicating 221


11

that low-flow pumping in these wells indeed allowed sampling water from the targeted 222

intervals. 223

224

Groundwater sampling techniques 225

226

The selected devices to withdraw water from wells were the Grundfos (Bjerringbro, 227

Denmark) Redi-Flo2 impeller pump, the Solinst (Georgetown, ON, Canada) model 407 228

bladder pump, and the HydraSleeve single-use downhole sampling bags. The impeller and 229

bladder pump models were also those used by Humez et al. (2015) for their 8-year 230

monitoring of a well located in a region with groundwater containing high dissolved gas 231

concentrations. Some preliminary tests performed with a peristaltic pump on three of our 232

monitoring wells provided concentrations that were significantly lower compared to those 233

obtained with the impeller pump, especially for well F4 that has a water level at ~8 m below 234

the top of the casing and high dissolved methane concentrations. This represents the 235

maximum distance above the water level from which the peristaltic pump can lift water 236

(Parker 1994). The suction effect of the pump, coupled to the low water level, likely caused 237

additional degassing, as was observed from the numerous bubbles visible in the sampling 238

tube. This pump was therefore not tested any further. An inertial-lift (Waterra) pump was 239

not considered, as based on its operating principle, it was expected to lead to excessive 240

degassing, as reported by Barker and Dickout (1988), Devlin (1987) and others cited in 241

Parker (1994). Table 2 presents the wells and techniques used for each sampling event. 242


12

The submersible impeller (Redi-Flo2) pump was selected for its ease of use and sturdiness. 243

It was equipped with a 90-m long, 6.25 mm (¼”) diameter tubing. The pump was slowly 244

lowered into the well to the targeted sampling depth. The flow rate was adjusted so as to 245

cause minimal drawdown in the well, in order to limit degassing. The EPA 246

recommendation for low-flow purge sampling suggests that drawdown should be limited 247

to 10 cm prior to stabilization of field parameters (Puls and Barcelona 1996). However, in 248

most of the wells, especially the least permeable ones, the minimum flow rate that could 249

be achieved with this pump still caused significant drawdown. The average drawdown for 250

all wells from the three sampling campaigns pooled together is 52 ± 41 cm, with a 251

maximum of 183 cm in well F3, where a nearby residential well was in use. Yields were 252

on average 0.32 L/min, with minimum and maximum values of 0.05 and 2.00 L/min. 253

Between 1 and 2 hours of pumping were usually needed for physico-chemical parameters 254

(temperature, pH, conductivity, redox potential, dissolved oxygen) to stabilize. Samples 255

were collected when parameters had been stable for at least 15 minutes. 256

The bladder pump was also equipped with a 90-m long, 6.25 mm (¼”) diameter tubing. 257

This pump allows sampling at a lower flow rate than the impeller pump, which may be 258

desirable in some very low-yielding wells. However, the bladder pump requires some 259

delicate fine-tuning, is less sturdy and requires the use of an air compressor. The pump was 260

carefully lowered in the well, and purging and sampling procedures were identical as those 261

used with the impeller pump. 262

Finally, disposable HydraSleeve bags were used. These bags simply consist of a 263

polyethylene bag that is sealed at the bottom and has a self-sealing check valve at the top. 264

This method was selected because: 1) it appeared simple to use, 2) it allegedly allows the 265


13

collection of representative groundwater samples without the need to purge the wells and 266

3) it was recommended for water containing volatile gasses by ITRC (2007). These 267

sampling bags are especially advantageous for sampling wells that have an extremely low 268

hydraulic conductivity, where purging prior to sampling is not possible without lowering 269

the water level by several meters. In our case, this method also allowed sampling of our 270

deepest well (depth of 147 m), where the targeted sampling interval was too deep for our 271

low-flow pumps. As we wished to test both the bag performance for sampling methane and 272

the purge versus no-purge approach, these bags were used before (no-purge approach) and 273

after the two pumps (purge approach) (Table 2). For sampling, the bags were carefully and 274

slowly lowered in the wells until the targeted sampling depth was reached, and samples 275

were collected without delay, using the recommended standard technique (a description is 276

provided in McHugh et al. (2015) and on the manufacturer website). 277

It was not possible in our study to allow the water to equilibrate for a few days (passive 278

sampling) because: 1) methane concentrations are known to vary significantly over time 279

(Rivard et al. 2017), and were even suspected to vary over very short periods based on the 280

work reported in Hirshe and Mayer (2009), so sampling a few days before or after the other 281

techniques would not have led to comparable values between sampling techniques and 2) 282

the largest HydraSleeve bags available were not large enough to provide the required 283

sample volume for the various analyses. Indeed, the sample volume needed was up to 6 284

times the volume contained in the bag. Furthermore, using 6 disposable bags for each well 285

was not a reasonable option, both for financial and environmental reasons. We therefore 286

developed a way to reuse the bags several times at a single well by carefully inserting a 287

sampling tube through the unsealed end of the bag, instead of piercing the bag with the 288


14

intended sampling straw. The unsealed end of the bag is still air-tight as the water pressure 289

forces the plastic double-wall to remain closed; it is possible, although not easy, to force 290

our way inside the bag with the sampling tube. In the best cases, the same bag could be 291

used 3 or 4 times before being damaged by the borehole walls. The physico-chemical 292

parameters were verified in the water collected with HydraSleeve bags. Although multiple 293

fillings of the bags involved repeated lowering and lifting of the bags through the water 294

column, the physico-chemical parameters were very similar from one bag to the other and 295

to those measured during pumping, which indicates that the water sampled with the bags 296

is representative of the water previously sampled with the pumps. Likewise, major ions 297

and trace metal analyses (not discussed in this paper) provided similar results for samples 298

collected using the HydraSleeve bags after pumping than with the impeller pump. 299

300

Bottle filling, storage and analyses 301

302

The method chosen for bottle filling was the one documented by the USGS 303

(http://water.usgs.gov/lab/chlorofluorocarbons/sampling/bottles/). It was selected as it is 304

widely used, affordable, involves readily available materials and generates minimal waste 305

(bottles and vials are reused, only the septa need to be changed). Vials (for alkane 306

concentrations) or bottles (for methane isotope composition) were held upright at the 307

bottom of a larger container. The sampling tube was inserted at the bottom of the 308

vial/bottle, progressively filling it and then the larger container. Once the container was 309

full, the tube was removed and the vial/bottle was capped underwater to avoid contact with 310

http://water.usgs.gov/lab/chlorofluorocarbons/sampling/bottles/


15

the atmosphere. This technique is similar to the inverted bottle method that has been widely 311

used for sampling groundwater containing gas (Humez et al. 2015; 2016, Siegel et al. 2015; 312

2016; Moritz et al. 2015; Molofsky et al. 2016; Smith et al. 2016). In our case, the upright 313

position was selected because the inverted (upside-down) position seemed more prone to 314

trap gas and create a headspace while being filled with groundwater highly charged with 315

methane. This was recently confirmed by Molofsky et al. (2016) in a comparative study. 316

For each sample, three 40-mL amber glass vials were collected for replicate alkane 317

concentration measurements, along with two 1-L amber glass bottles for single methane 318

isotopic measurements (δ13C and δ2H). The open-top caps were lined with grey butyl septa 319

(1-L bottles) or Teflon-coated silicon septa (40-mL vials). Containers were stored on their 320

side (1-L bottles) or upside down (40-mL vials) in a fridge at 4˚C. Water for alkane 321

concentrations and methane C and H isotope ratios was acidified to pH < 2 to avoid 322

proliferation of microorganisms. 323

Concentrations of dissolved C1-C3 alkanes were determined at the Delta-Lab of the 324

Geological Survey of Canada (Quebec City, QC) using a Stratum PTC (Teledyne Tekmar, 325

Mason, OH) purge and trap concentrator system interfaced with an Agilent (Santa Clara, 326

CA) 7890 gas chromatograph equipped with a flame ionisation detector (GC-FID). The 327

method employed was adapted from Pennsylvania Department of Environmental 328

Protection method PA-DEP 3686 (2012) and US Environmental Protection Agency (EPA) 329

method RSK 175 (Kampbell and Vandegrift 1998). Quantification limits on our samples 330

were 0.006, 0.002, and 0.01 mg/L for methane, ethane and propane, respectively. The 331

uncertainty related to sampling, handling and analysis was estimated at 15% of the reported 332

concentration, based on the 90th percentile in replicate samples (Rivard et al. 2017). 333


16

Alkane isotopic composition (δ13C and δ2H) was analyzed at either one of three different 334

laboratories, namely the Delta-Lab of the Geological Survey of Canada (Quebec City), 335

Concordia University (Montréal), or the G.G. Hatch laboratory of the University of Ottawa. 336

The choice of the lab depended on the availability of the analytical instruments, in order to 337

ensure timely analysis. Control samples were sent concurrently to the different labs to 338

ensure that results are comparable. Analysis was performed on a Delta V (Thermo Fisher 339

Scientific, Waltham, MA) isotope ratio mass spectrometer (IRMS) at the Delta-Lab and 340

the G.G. Hatch lab, while a GC Agilent 6890 coupled to an Isoprime 100 (Manchester, 341

UK) was used at Concordia. Results are expressed in the usual per mil notation relative to 342

Vienna Pee Dee Belemnite (V-PDB; δ13C) and Vienna Standard Mean Ocean Water 343

(VSMOW; δ2H). The uncertainty related to sampling, handling and analysis was estimated 344

at 1.7‰ for δ13C and 19‰ for δ2H, based on the 90th percentile in replicate samples (Rivard 345

et al. 2017). 346

347

Results 348

349

Methane concentrations 350

351

Results for methane concentrations obtained through the different sampling techniques 352

during each of the three sampling campaigns are presented in Figure 2 (note that all 353

geochemical results will be available in a public database to be released in 2017). It is 354

noteworthy that a comparison of different techniques could not be done for well F10 in fall 355


17

2014, as the well was artesian shortly after its drilling. Artesian flowing conditions then 356

resorbed, so that the well could be included for the following campaigns. Also, in summer 357

2015, well F7 could not be sampled using the bladder pump due to malfunction of the 358

pump, which caused numerous air bubbles to enter the sampling tube, thus compromising 359

results. 360

Figure 2 reveals that methane concentrations obtained through the different sampling 361

techniques for a given well are, with few exceptions, quite similar with nearly all individual 362

values being close, when considering uncertainty related to sampling, handling and 363

analysis. Furthermore, there are no systematic trends for higher or lower values related to 364

a given method or well. Student and Fisher statistical tests performed on these time series, 365

with a 10% level of significance, confirmed that there is no evidence that their statistical 366

properties (mean and variance) are different, as they indicated that the null hypothesis of 367

population equivalency could not be rejected. 368

However, even though there does not appear to be a systematic bias related to any particular 369

method, there are still important differences in absolute concentrations measured for some 370

samples, including well F4 in May 2015 (concentrations varying between 34.75 and 46.33 371

mg/L), F4 in July 2015 (concentrations between 26.15 and 55.35 mg/L) and F6 in July 372

2015 (concentrations between 6.39 and 20.96). When comparing the largest and lowest 373

values (using the maximum/minimum concentration ratios, hereafter called “max/min 374

ratio”) obtained for a given well on a given sampling campaign, ratios are mostly between 375

1 (i.e. no variation between techniques) and 2 (i.e. a 100 % variation between highest and 376

lowest values), with a few higher ratios exceeding 3 (Figure 3). High ratios (> 2) tend to 377

be either associated with high (e.g. > 20 mg/L) or low (e.g. < 1 mg/L) concentrations; 6 378


18

results (21%) have such high ratios. High concentrations imply important degassing when 379

the sample is lifted to the surface and degassing may occur differently according to the 380

sampling technique (e.g., wells F2 and F4). In contrast, when concentrations are very low, 381

very small absolute variations result in a high max/min ratio (e.g., wells F3 and F8). The 382

two very high max/min ratios for well F6 (containing intermediate methane concentrations) 383

are puzzling; they may be related to mixing of different waters, as this is the case in the 384

nearby well F7 (see Section “Methane isotopic composition”). For the May and July 2015 385

campaigns, max/min ratios always involve HydraSleeve bags, except for two cases (wells 386

F4 and F6 in July 2015), when values obtained with the two pumps appeared to be 387

abnormal compared to the others (see Figure 4). These abnormally low or high 388

concentrations could be due to technical or human error along the process or to an enhanced 389

methane contribution that can sporadically occur under the form of a “slug” (pulses) as 390

described in Dusseault and Jackson (2014). 391

Figure 4 shows that results obtained with the impeller and bladder pumps generally agree 392

very well, except for two points (F4 and F6 in July 2015), where the result from the Redi-393

Flo2 pump was either higher or lower than with the bladder pump. When discarding these 394

two values, the remaining values are almost perfectly aligned and the determination 395

coefficient (R2) becomes 0.98, indicating that methane concentrations are generally similar 396

when using these two types of pump. 397

The comparisons of concentrations obtained from the four sampling methods two by two 398

(matrix plots in Figure 5) also confirms that there is no general bias and that none of the 399

methods systematically underestimates methane concentrations. Values above 20 mg/L 400

often show more disparity in absolute values of concentrations. At such high 401


19

concentrations, it is likely that the sum of dissolved gases exerts a pressure above 402

atmospheric pressure, causing effervescence; such conditions were also reported by 403

Molofsky et al. (2016) to cause disparity in measured methane concentrations, depending 404

on the bottle-filling method. However, when methane concentrations are compared using 405

concentration ratios or relative “errors” ([C1-C2]/C1 × 100%, C1 being the concentrations 406

obtained with a given method taken as a reference and C2 the concentrations corresponding 407

to one of the other three methods), a bias for higher concentrations is not obvious. Figure 408

6 provides an example of one of these graphs using the concentrations obtained with the 409

impeller pump as a reference, which is representative of the other similar graphs. The only 410

abnormally high values (e.g. > 70% and < -70%) correspond either to a case for which 411

concentrations were very low (below 1 mg/L, for instance in wells F3 and F8) or to a case 412

for which anomalous values were obtained with one of the sampling techniques in July 413

2015 (wells F4 and F6). 414

Very little ethane and even less propane were found in these 10 wells. C2+ hydrocarbons 415

are rarely present in significant quantity in groundwater, as was noted in many other studies 416

(e.g. Baldassare et al. 2014; Molosfky et al. 2016; Humez et al. 2015; 2016; Currell et al. 417

2017). Ethane and propane concentrations above 10 μg/L occurred in only three wells and 418

one well, respectively, and not in all sampling campaigns. Therefore, they could not be 419

used as a basis for comparison of the sampling methods in this paper. 420

421

Methane isotopic composition 422

423


20

Methane stable carbon (δ13C-CH4) and hydrogen (δ2H-CH4) isotope ratios were also 424

analysed to investigate whether they could be impacted by the sampling technique. The 425

insight of the effect of sampling technique on methane C and H isotope ratios is essential, 426

as these isotopic ratios can be used as a potential indicator for methane migration (either 427

natural or anthropogenic) in groundwater. Figures 7 and 8 present δ13C-CH4 and δ2H-CH4 428

for the ten observation wells from the three sampling campaigns. This section only 429

discusses similarities or dissimilarities obtained with the different sampling techniques; a 430

discussion on the isotopic results in relation to methane source will be presented in an 431

upcoming paper. 432

Figure 7 confirms that δ13C-CH4 values are very similar for all sampling techniques in most 433

wells, with significant overlaps of individual values, considering the ±2‰ uncertainty 434

associated with sampling, handling and analysis. However, in well F7 (May and July 2015), 435

there is a marked difference of approximately 10‰ between the values obtained with 436

HydraSleeve bags before pumping, and the other techniques. This well has an upward flow 437

bringing in some very old, saline groundwater from the bottom of the well, resulting in 438

uncommonly high salinity in the water column when the well is resting (18-25 PSU or 439

practical salinity unit). The well is very sensitive to pumping, with salinity quickly 440

decreasing as freshwater from the shallow aquifer invades the well under pumping. It is 441

therefore not surprising to obtain markedly different isotopic values in a sample that was 442

collected prior to pumping, as the methane present in the very old groundwater is of a 443

different origin than that in the shallow aquifer. Interestingly, concentrations of methane 444

in both sources of groundwater seem to be similar, as this purge-related effect went 445

unnoticed with methane concentrations (Figure 2). 446


21

In a few other wells (F3 in May and July 2015, F4 in July 2015, F8 in July 2015, and F10 447

in July 2015), δ13C-CH4 values also exhibit a spread exceeding the uncertainty (maximum 448

spread of 7‰), but in these cases the variability is not clearly related to a specific sampling 449

technique. This higher variability may be due to well-specific factors, such as small 450

variations of different sources of methane in given well when methane concentrations are 451

very low (wells F3 and F8), or an upward flow in the well (well F10). In well F4, the 452

isotopic spread is relatively small (max. spread 4.9‰). The difficulty in stabilizing the 453

physico-chemical parameters in this very low-yield well and the fact that very different 454

concentrations were found may indicate contributions of water from different depths or of 455

gas slugs. 456

Compared to methane carbon isotope ratios, hydrogen stable isotope ratios are known to 457

be subject to a greater uncertainty, which was estimated at ±19‰ in this project. They also 458

span a much larger isotopic range than do carbon stable isotope ratios and are less 459

diagnostic with regards to methane origin (Whiticar 1999). Nonetheless, for most of our 460

wells, the δ2H-CH4 values from the different techniques are very similar, and overlap when 461

considering the uncertainty (Figure 8). The only exceptions are wells F3 (all three sampling 462

campaigns; maximum spread of 72‰) and F10 (July 2015, spread of 54‰), which can be 463

explained by the same mechanisms discussed above for carbon isotopes. More specifically, 464

the very low methane concentration in well F3 is significantly affected by methane 465

oxidation, which causes important variations in concentrations over time (Rivard et al. 466

2017). Oxidation causes stronger isotopic fractionation on the hydrogen than on the carbon 467

atoms (Alperin et al. 1988; Kinnaman et al. 2007), which likely explains the observed wider 468

spread in δ2H-CH4 values for well F3. The water in this well may also be affected by the 469


22

sporadic pumping of a nearby well, which is deeper (76 m) and which was in use at the 470

time of sampling in July 2015. Contrary to oxidation, late-stage methanogenesis causes 471

isotopic fractionation on the carbon isotopes, but not on the hydrogen isotopes. This occurs 472

because the carbon used by methanogens comes from a limited carbon pool, which may 473

become exhausted over time, when not replenished due to isolated groundwater conditions. 474

In contrast, the hydrogen comes from the ambient water, which is a comparatively very 475

large pool where the supply of “light” hydrogen, preferred by the microbes, is unlimited 476

(Martini et al. 1998). This could thus explain why the purge-related isotopic effect in wells 477

F4 and F7 are visible on the carbon isotope ratios but not on the hydrogen isotope ratios. 478

Figure 9 presents box plots for both δ13C-CH4 and δ2H-CH4 according to each sampling 479

technique when results from the three field campaigns are pooled together. These graphs 480

confirm that median isotopic values are very close for all sampling techniques, being within 481

the uncertainty range of one another. This is particularly surprising for δ2H-CH4 values, 482

which are naturally more variable and have a higher uncertainty related to sampling, 483

handling and analysis; in spite of this, the median δ2H-CH4 values for each sampling 484

technique only vary between -249.0 and -250.8‰. It is worth mentioning that δ13C-CH4 485

and δ2H-CH4 values from the three groundwater samples collected with the peristaltic 486

pump also provided similar results to those of the other two types of pumps. 487

Although the δ13C-CH4 and δ2H-CH4 values do not seem sensitive to the sampling method, 488

results suggest that no-purge methods could sometimes provide different values, for 489

instance if an upward flow is present in the well. Therefore, if the objective of the sampling 490

campaign is to identify the gas origin, one does not have to worry much about the sampling 491

technique as long as the well is pumped long enough to have representative water from the 492


23

surrounding aquifer. Nonetheless, other sampling devices than those selected for this study 493

should be tested to make sure that they do not result in isotopic fractionation, such as 494

inertial-lift pumps for instance because its mechanism can entrain water turbulence. 495

496

Discussion and recommendations 497

498

The comparison of methane concentrations and stable isotope ratios (δ13C-CH4, δ2H-CH4) 499

made in the present study did not show any systematic bias related to one of the selected 500

sampling techniques (impeller pump, bladder pump, sampling bags). However, important 501

differences (in absolute values) between concentrations obtained via different techniques 502

could sometimes be observed for a given well and sampling date, especially (but not 503

systematically) when concentrations were high, (i.e. when degassing was significant). 504

Nonetheless, unlike McHugh et al. (2015) who had tested different sampling techniques to 505

compare VOC concentrations and had found that HydraSleeve bags provided lower and 506

more variable VOCs concentrations, our results did not show that dissolved methane 507

concentrations were systematically lower nor were especially more variable with 508

HydraSleeve bags than with the other selected techniques. 509

Furthermore, results obtained before and after pumping were usually similar. This suggests 510

that in many cases, using a no-purge, fast method could be appropriate. However, in 511

particular cases, such as when there is an upward flow bringing more evolved water into a 512

well, a no-purge method (such as HydraSleeve bags used before pumping) could provide 513

different results compared to the other techniques, simply because the water being sampled 514


24

is not the same with and without pumping. In such cases, the choice of no-purge or purge 515

method will depend on the water source that needs to be sampled, and it is crucial to follow 516

the same approach every time to obtain comparable values over time. Generally, unless 517

one is specifically trying to sample a water source that recedes upon pumping, we 518

recommend using a method that involves low-flow pumping until stabilization of the 519

physicochemical parameters, at the depth where flowing fractures have been previously 520

identified through borehole logging. In cases where there is more than one source of 521

groundwater in a well, isotopic results may also be affected by pumping. Unless very 522

detailed geochemical and hydrogeological characterization of each well in a study area 523

have been conducted, it is likely that such mixing of different water sources would go 524

unnoticed. Therefore, low-flow pumping seems the most prudent choice in most cases. 525

Borehole logging can provide important clues regarding the presence of different types of 526

groundwater from fractured intervals of a well open to a rock aquifer. 527

While none of the tested sampling techniques caused significantly more degassing of 528

samples compared to the others, it is known that the selected “water bucket” bottle-filling 529

technique (corresponding to a “semi-closed” system) will cause some degassing in 530

samples with high gas concentrations (Molofsky et al. 2016). “Closed systems”, such as 531

those using Isoflasks®, are promising in that they allow the collection and analysis of both 532

free and dissolved gas phases. This technique is relatively new and is still very costly, as 533

not many laboratories (outside of the corporation’s internal lab) are set-up to conduct theses 534

analyses. Also, the use of disposable pouches (in some cases, several pouches at a well if 535

large sample volumes are needed), is an environmental downside. If this technique keeps 536


25

developing and becomes more affordable, degassing issues during sample lifting will 537

become less of a concern. 538

Until then, researchers must still acknowledge that their reported concentrations in 539

effervescing samples likely underestimate true methane concentrations in the aquifer when 540

an open or semi-closed sampling system is used; underestimations are expected to increase 541

along with concentrations and sampling depth. However, methane concentrations are 542

known to vary naturally over time in many regions. For instance, in our study area, 543

concentrations in a given well could reach up to six times the smallest recorded value 544

(Rivard et al. 2017), which exceeds by far any dissimilarities observed among results 545

obtained with the different sampling techniques (in the present study and, for example, in 546

Molofsky et al. 2016). Due to such natural variations, methane concentrations are generally 547

not a very robust diagnostic tool of methane provenance compared to isotopic composition. 548

Although all of the sampling devices tested in this study provided similar results, practical 549

considerations must be taken into account when choosing a technique. We strongly 550

recommend the use of an impeller pump, which is easy to use and very robust, and unlike 551

the bladder pump, cannot pump water when damaged; a similar recommendation was also 552

made by Muska et al. (1986). The bladder pump can, indeed, allow entrance of air in the 553

tubing when the bladder is defective, thereby resulting in much further degassing, which 554

compromises the sample. It is also more fragile and requires more fine-tuning than the 555

impeller pump. However, an advantage of the bladder pump is that it can usually achieve 556

very low pumping rates (lower than the impeller pump), which may be critical in wells 557

with a very low yield. Despite their high initial purchase cost, the advantage of using pumps 558

over HydraSleeve bags is that they can be used for a large number of wells and sampling 559


26

events over several years. HydraSleeve bags may be an interesting option when sampling 560

only a few wells, or when sampling deep wells or wells with extremely low yield. However, 561

the financial (and environmental) costs of these disposable bags rise quickly, especially if 562

large sample volumes (and thus several bags) are required for a suite of analyses. 563

564

Conclusions 565

566

Three groundwater sampling techniques were compared to evaluate their suitability and 567

interchangeability for collection of samples in open bedrock wells to analyze 568

concentrations and stable carbon and hydrogen isotope ratios of methane, which is the most 569

volatile and abundant hydrocarbon in groundwater. The selected techniques were an 570

impeller (Redi-Flo2) submersible pump, a bladder submersible pump, and disposable 571

sampling HydraSleeve bags, which were used both before and after pumping. The latter 572

procedure was performed to examine the effect of purging the wells on methane 573

concentration and isotopic composition. These sampling techniques were tested over three 574

sampling campaigns in 10 observation wells in the Saint-Édouard area, located ~65 km 575

south-west from Quebec City (eastern Canada). In this region, dissolved methane is 576

naturally present in groundwater and concentrations are usually highly variable spatially 577

and temporally. 578

Results showed that methane carbon and hydrogen stable isotope ratios were not sensitive 579

to the selected sampling techniques, with all four techniques usually providing similar 580

results. Methane concentrations were comparatively more sensitive and significant 581


27

differences were observed in a few wells. However, no systematic technique-related bias 582

was observed. As for the no-purge approach, it was appropriate in some wells but not in 583

others, depending on the hydrogeological conditions, in particular in the presence of 584

vertical hydraulic or salinity gradients within the well. 585

Based on this work, we therefore recommend the following approach for every 586

groundwater sampling program aiming to characterize methane concentrations and stable 587

isotope ratios: 1) carry out a purging period until stabilization of groundwater 588

physicochemical parameters at the depth where flowing fractures are documented; 2) pump 589

the well at a low flow that will keep drawdown to a minimum, to avoid groundwater 590

pressure changes that result in degassing; 3) remain consistent in sampling depth and bottle 591

filling procedure, as well as for the sampling device; and 4) preferably use a low-flow 592

impeller submersible pump, such as the Redi-Flo2 pump, as this kind of device is simple 593

to use and very reliable, and does not involve the use of disposable materials. 594

595

Acknowledgments 596

597

The authors would like to thank Dr. Mathieu Duchesne of the GSC and Pr. Erwan Gloaguen 598

of INRS for their advices and contribution related to the representation of data with Matlab. 599

Authors would like to acknowledge funding support from the Energy Sector (Eco-EII and 600

PERD programs) and the Earth Science Sector (Environmental Geoscience Program) of 601

Natural Resources Canada. Our gratitude goes out to Mrs Marianne Molgat, formely of 602

Talisman Energy, without whom this project would likely not have taken place. We would 603


28

also like to deeply thank the Ministère du Développement durable, de l’Environnement et 604

de la Lutte contre les Changements climatiques (MDDELCC), land and well owners that 605

allowed work to be performed on their property, the Municipality of Saint-Édouard, the 606

MRC de Lotbinière and the Ministère des Forêts, de la Faune et des Parcs du Québec. The 607

authors also want to sincerely thank Nicolas Benoit of the GSC for his internal review and 608

two anonymous reviewers for their careful review (to be completed). This paper is GSC 609

contribution # 31812. 610

611

References 612

613

Baldassare, F.J., McCaffrey, M.A., Harper, J.A. (2014) A geochemical context for stray 614

gas investigations in the northern Appalachian Basin: Implications of analyses of natural 615

gases from Neogene-through Devonian-age strata. AAPG Bulletin, 98(2): 341-372. 616

Barker, J.F., Dickhout, R. (1988) An evaluation of some systems for sampling gas-charged 617

groundwater for volatile analysis. Ground Water Monitoring Review, 8, 112-120. 618

Bordeleau, G., Rivard, C., Lavoie, D., Lefebvre, R., Malet, X., Ladevèze, P. (2017) 619

Geochemistry of groundwater in the Saint-Édouard area, Quebec, Canada, and its influence 620

on the distribution of methane in shallow aquifers, Applied Geochemistry, in press. 621

Coleman, N. P., McElreath, D. (2012) Short-term Intra-Well Variability in Methane 622

Concentrations from Domestic Well Waters in Northeastern Pennsylvania, AAPG Search 623

and Discovery Article #90154©2012, AAPG Eastern Section Meeting: Stray Gas 624


29

Incidence & Response Forum, Cleveland, Ohio, 24-26 July 2012, 625

http://www.gwpc.org/sites/default/files/event-sessions/Coleman_Nancy.pdf. 626

Currell, M., Banfield, D., Cartwright, I., Cendón, D.I. (2017) Geochemical indicators of 627

the origins and evolution of methane in groundwater: Gippsland Basin, Australia. 628

Environmental Science and Pollution Research, 1-16. doi:10.1007/s11356-016-7290-0. 629

Devlin, J.F. (1987) Recommendations concerning materials and pumping systems used in 630

the sampling of groundwater contaminated with volatile organics. Water Pollution 631

Research Journal of Canada, 22(1): 65-72. 632

Dusseault, M., Jackson R. (2014) Seepage pathway assessment for natural gas to shallow 633

groundwater during well stimulation, in production, and after abandonment. 634

Environmental Geosciences, 21(3):107-126. 635

EPA. (2013) Introduction to in situ bioremediation of groundwater, Office of solid wastes 636

and emergency response, EPA 542-R-13-018, 86 pages. https://nepis.epa.gov/Exe. 637

Gorody, A. W., Baldwin, D., Scott, C. (2005) Dissolved methane in groundwater, San Juan 638

Basin, La Plata County Colorado: Analysis of data submitted in response to COGCC orders 639

112-156 & 112 -157, November 2005 IPEC Conference, Houston, TX, 14 pages. 640

http://ipec.utulsa.edu/Conf2005/Papers/Gorody_DISSOLVED_METHANE_IN_GROUNDWATER.pdf 641

Gorody, A.W. (2012) Factors affecting the variability of stray gas concentration and 642

composition in groundwater. Environmental Geosciences, 19(1): 17–31. 643

Hirsche, T., Mayer, B. (2009) A comprehensive literature review on the applicability of 644

free and dissolved gas sampling for baseline water well testing. Report prepared for 645

http://www.gwpc.org/sites/default/files/event-sessions/Coleman_Nancy.pdf

https://nepis.epa.gov/Exe

http://ipec.utulsa.edu/Conf2005/Papers/Gorody_DISSOLVED_METHANE_IN_GROUNDWATER.pdf


30

Alberta Environment, 47 pages, 646

http://www.waterforlife.alberta.ca/documents/ApplicabilityFreeDissolvedGas-Mar2009.pdf 647

Humez, P., Mayer B., Inga, J., Nightingale M., Becker V., Kingston A., Akbilgic O., 648

Taylor, S. (2016) Occurrence and origin of methane in groundwater in Alberta (Canada): 649

Gas geochemical and isotopic approaches. Science of the Total Environment 541: 1253-650

1268. 651

Humez, P., Mayer, B., Nightingale, M., Ing, J., Becker, V., Jones, D. (2015) An 8-year 652

record of gas geochemistry and isotopic composition of methane during baseline sampling 653

at a groundwater observation well in Alberta (Canada). Hydrogeology Journal, DOI: 654

10.1007/s10040-015-1319-1. 655

ITRC (Interstate Technology & Regulatory Council). (2007) Protocol for Use of Five 656

Passive Samplers to Sample for a Variety of Contaminants in Groundwater, Technical and 657

Regulatory Guidance, 121 pages. http://www.itrcweb.org/GuidanceDocuments/DSP-5.pdf 658

Jackson, R.E. and Heagle, D.J. (2016) Sampling domestic/farm wells for baseline 659

groundwater quality and fugitive gas. Hydrogeology Journal, doi 10.1007/s10040-016-660

1369-z. 661

Kampbell, D.H., Vandegrift, S.A. (1998) Analysis of Dissolved Methane, Ethane, and 662

Ethylene in Ground Water by a Standard Gas Chromatographic Technique. Journal of 663

Chromatographic Science, 36: 253-256. 664

Ladevèze, P. (2017) Aquifères superficiels et ressources profondes : le rôle des failles et 665

des réseaux de fractures, Ph.D. thesis, INRS-ETE, 220 pages. 666

http://www.waterforlife.alberta.ca/documents/ApplicabilityFreeDissolvedGas-Mar2009.pdf

http://www.itrcweb.org/GuidanceDocuments/DSP-5.pdf


31

Ladevèze, P., Rivard, C., Lefebvre, R., Lavoie, D., Parent, M., Malet, X., Bordeleau, G., 667

Gosselin, J.S. (2016) Travaux de caractérisation hydrogéologique dans la plateforme 668

sédimentaire du Saint-Laurent, région de Saint-Édouard-de-Lotbinière, Québec, Dossier 669

public 8036, 2016; 112 pages, doi:10.4095/297891. 670

Lavoie, D., Rivard, C., Lefebvre, R., Sejourne, S., Theriault, R., Duchesne, M. J., Ahad, 671

J. M. E., Wang, B., Benoit, N., Lamontagne, C. (2014) The Utica Shale and gas play in 672

southern Quebec: Geological and hydrogeological syntheses and methodological 673

approaches to groundwater risk evaluation. International Journal of Coal Geology, 126: 674

77–91. 675

Lavoie, D., Pinet, N., Bordeleau, G., Ardakani, O.H., Ladevèze, P., Duchesne, M.J., 676

Rivard, C., Mort, A., Brake, V. Sanei, H., Malet, X. (2016) The Upper Ordovician black 677

shales of southern Quebec (Canada) and their significance for naturally occurring 678

hydrocarbons in shallow groundwater. International Journal of Coal Geology, 158, 44-64. 679

Lefebvre, R. 2017. Mechanisms leading to potential impacts of shale gas development on 680

groundwater quality. WIREs Water, 4(1), 15 pages. doi: 10.1002/wat2.1188. 681

Martini, A.M., Walter, L.M., Budai, J. M., Ku, T.C.W., Kaiser, C.J., Schoell, M. (1998) 682

Genetic and temporal relations between formation waters and biogenic methane: Upper 683

Devonian Antrim Shale, Michigan Basin, USA. Geochimica et Cosmochimica Acta, 62 684

(10): 1699–1720. 685

McHugh, T.E., Kulkarni, P.R., Beckley, L.M., Newell, C.J., Zumbro, M. (2015) Negative 686

bias and increased variability in VOC concentrations using the HydraSleeve in monitoring 687

wells. Ground Water Monitoring & Remediation, doi:10.1111/gwmr.12141. 688


32

Molofsky, L.J., Richardson, S.D., Gorody, A.W., Baldassare, F.,Black, J.A., McHugh, 689

T.E., Connor, J.A. (2016) Effect of Different Sampling Methodologies on Measured 690

Methane Concentrations in Groundwater Samples. Ground Water, 24: 1-12, doi: 691

10.1111/gwat.12415. 692

Moritz, A., Helie, J.F., Pinti, D.L., Larocque, M., Barnetche, D., Retailleau, S., Lefebvre, 693

R., Gelinas, Y. (2015) Methane Baseline Concentrations and Sources in Shallow Aquifers 694

from the Shale Gas-Prone Region of the St. Lawrence Lowlands (Quebec, Canada). 695

Environmental Science & Technology, 49, 4765−4771. 696

Muska, C.F., Colven, W.P., Jones, V.D., Scogin, J.T., Looney, B.B., Price, V.Jr. (1986) 697

Field evaluation of ground water sampling devices for volatile organic compounds, In: 698

Proceedings of the sixth national symposium and exposition on aquifer restoration and 699

ground water monitoring, Columbus, Ohio, May 19-22. 700

Nastev, M., Therrien, R., Lefebvre, R., Gélinas, P.J. (2001) Gas production and migration 701

in landfills and geological materials. Journal of Contaminant Hydrology, 52 (1-4): 187-702

211. 703

PA-DEP. (2012) Light Hydrocarbons in Aqueous Samples via Headspace and Gas 704

Chromatography with Flame Ionization Detection (GC/FID). PA Dept. of Environmental 705

Protection, PA-DEP 3686, Rev. 1, 13 pages. 706

Parker, L. V. (1994) The effects of groundwater sampling devices on water quality: A 707

literature review. Ground Water Monitoring and Remediation, 14(2): 130–141, 708

doi:10.1111/j.1745 -6592.1994.tb00108.x. 709


33

Parker L.V., Clark, C.H. (2004) Study of five discrete-interval-type ground water sampling 710

devices. Ground Water Monitoring and Remediation, 24(3): 111-123. 711

Puls, R.W., Barcelona, M.J. (1996) Low-flow (minimal drawdown) groundwater sampling 712

procedures. EPA Ground Water Issue, EPA/540/S-95/50, pp 12. 713

Rivard, C., Bordeleau, G., Lavoie, D., Lefebvre, R., Malet, X. (2017) Methane variations 714

in groundwater over time. Hydrogeology Journal, on line: http://rdcu.be/yjMl. 715

716 Siegel, D.I., Azzolina, N.A., Smith, B.J., Perry, AE, Bothun, R.L. (2015) Methane 717

concentrations in water wells unrelated to proximity to existing oil and gas wells in 718

Northeastern Pennsylvania. Environmental Science & Technology, 49(7): 4106–4112. 719

Smith, B., Becker, M., Siegel, D. (2016) Temporal variability of methane in domestic 720

groundwater wells, northeastern Pennsylvania. Environmental Geosciences, 23(1): 49–80. 721

Whiticar, M.J. (1999) Carbon and hydrogen isotope systematics of bacterial formation and 722

oxidation of methane. Chemical Geology, 161: 291–314. 723

724

http://rdcu.be/yjMl

Table 1 Characteristics of the observation wells used in this study

Site Drilling

type

Drilling

year

Total

drilled

depth (m)

Static water

level (m

below TOC)

Overburden

thickness

(m)

Sampling

depth (m

below TOC)

Conditions

F-1 Diamond 2013 50 0.63 2.44 7.5 Confined




F-5 Hammer 2014 50 2.12 9.75 14.4 Confined


F-7 Diamond 2014 50 4.485 11.43 17.7 Semi-

confined



F-11 Hammer 2014 50 1.97 6.4 10.3 Semi-

confined

Notes: Diamond: Diamond-drilled well with a 100 mm (4 in.) diameter; Hammer: Hammer-drilled well with a 152

mm (6 in.) diameter. TOC: top of casing.

Table 2 Characteristics of the three sampling campaigns

Sampling event Sampling technique used Sampled observation wells

November 2014 1) Impeller (Redi-Flo2) pump

2) HydraSleeeve bags (after pumping)

F1, F2, F3, F4, F5, F6, F7, F8

and F11

May 2015

1) HydraSleeve bags (before pumping)

2) Impeller (Redi-Flo2) pump

3) Bladder pump

4) HydraSleeve bags (after pumping)

F1, F2, F3, F4, F5, F6, F7, F8,

F10 and F11

July 2015

1) HydraSleeve bags (before pumping)

2) Impeller (Redi-Flo2) pump

3) Bladder pump

4) HydraSleeve bags (after pumping)

F1, F2, F3, F4, F5, F6, F7, F8,

F10 and F11

1

Fig. 1 Location of the study area and the observation wells

2

Fig 2 Comparison of methane concentrations obtained using different sampling

techniques for nine wells in November 2014 (top) and ten wells in May 2015 (middle)

and July 2015 (bottom). Uncertainty of ± 15% is shown with error bars

0

10

20

30

40

50

60

F1 F2 F3 F4 F5 F6 F7 F8 F10 F11

Met

han

e (m

g/L

)

November 2014

0

10

20

30

40

50

60

F1 F2 F3 F4 F5 F6 F7 F8 F10 F11

Met

han

e (m

g/L

)

May 2015

0

10

20

30

40

50

60

F1 F2 F3 F4 F5 F6 F7 F8 F10 F11

Met

han

e (m

g/L

)

July 2015

Redi-Flo 2

Bladder pump

Hydra-Sleeve before pumping

Hydra-Sleeve after pumping

3

Fig. 3 Maximum differences corresponding to ratios of maximum over minimum

methane concentrations obtained at ten sites using either two methods (fall 2014) or four

methods (spring and summer 2015) for groundwater sampling. Note: Concentrations in

well F8 were very low in May 2015 and one value (the one from the HydraSleeve bag

before pumping) fell below the detection limit and was thus attributed half the detection

limit (i.e., 0.003 mg/L). This resulted in a very high max/min ratio of 11.7

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

F1 F2 F3 F4 F5 F6 F7 F8 F10 F11

Max

imu

m d

iffe

ren

ce b

etw

ee

n m

eth

od

s (m

ax/m

in r

atio

)

Nov./Dec. 2014 (2 methods) May 2015 (4 methods) July 2015 (4 methods)

well F8 in May 2015 max/min = 11.7

4

Fig. 4 Comparison between methane concentrations obtained with the impeller (Redi-

Flo2) and the bladder pumps. A few tests were also done outside the three field

campaigns for a total of 24 data pairs (blue dots). The solid line integrates all samples

(R2 = 0.813), while the dotted line excludes two abnormal values (R2 = 0.979)

R² = 0.8133

R² = 0.979

0

10

20

30

40

50

60

0 10 20 30 40 50

[CH

4] w

ith

th

e b

lad

de

r p

um

p (

mg/

L]

[CH4] with the RediFlo2 pump (mg/L)

F4, July 2015

F6, July 2015

When rejecting the

two abnormal values

(i.e. F4 and F6 in July)

5

Fig. 5 Plots of methane concentrations obtained with the four sampling techniques,

presented in pairs, along with each method’s statistical distribution. The 45° line represents

the perfect match. SLV: HydraSleeve bags; Redi-Flo2: impeller pump; Bladder: bladder

pump.

6

Fig. 6 Comparisons of dissolved methane concentrations obtained through HydraSleeve

bags and a bladder pump with those obtained with the impeller (Redi-Flo2) pump

(considered here as a reference)

-250%

-200%

-150%

-100%

-50%

0%

50%

100%

150%

0 5 10 15 20 25 30 35 40 45R

ela

tive

dif

fere

nce

(%

)

Methane concentration (impeller pump)

SLV before

Bladder

SLV after

7

Fig. 7 Comparison of δ13C -CH4 values obtained using different sampling techniques for

nine wells in November 2014 (top) and May 2015 (middle) and ten wells in July 2015

(bottom). Uncertainty of 1.7‰ is shown with error bars. Notes: For November 2014, well

F1 does not have a value for the HydraSleeve technique as the bottle broke. In May 2015,

well F8 did not have enough methane to run isotopic analyses. For July 2015, well F7

does not have a value for the bladder pump as it did not function well

-110

-90

-70

-50

F1 F2 F3 F4 F5 F6 F7 F8 F10 F11

Met

han

e δ

13C(‰

)

November 2014

-110

-90

-70

-50

F1 F2 F3 F4 F5 F6 F7 F8 F10 F11

Met

han

e δ

13C(‰

)

May 2015

-110

-90

-70

-50

F1 F2 F3 F4 F5 F6 F7 F8 F10 F11

Met

han

e δ

13C(‰

)

July 2015

Redi-Flo 2Bladder pumpHydra-Sleeve before pumpingHydra-Sleeve after pumping

8

Fig 8 Comparison of δ2H -CH4 values obtained using different sampling techniques for

ten wells in November 2014 (top), May 2015 (middle), and July 2015 (bottom).

Uncertainty of ± 19‰ is shown with error bars. Notes: For November 2014, well F4 does

not have results for the Redi-Flo2 pump due to broken bottles. For May 2015, wells F2

and F6 do not have results for HydraSleeve bags before pumping again due to broken

bottles. Well F8 did not have enough methane to run isotopic analyses. For July 2015,

well F7 does not have a value for the bladder pump as it did not function well

-400

-300

-200

-100

F1 F2 F3 F4 F5 F6 F7 F8 F10 F11

Met

han

e δ

2H(‰

)

November 2014

-400

-300

-200

-100

F1 F2 F3 F4 F5 F6 F7 F8 F10 F11

Met

han

e δ

2H(‰

)

May 2015

-400

-300

-200

-100

F1 F2 F3 F4 F5 F6 F7 F8 F10 F11

Met

han

e δ

2H(‰

)

July 2015 Redi-Flo 2Bladder pumpHydra-Sleeve before pumpingHydra-Sleeve after pumping

9

Fig. 9 Box plots of the values obtained for δ13C-CH4 (left) and δ2H-CH4 (right) with the

four sampling techniques over the three field campaigns. The band inside the box

corresponds to the 50th percentile (median), the bottom and top of the box correspond to

the 25th percentile (1st quartile, Q1) and 75th percentile (3rd quartile, Q3), while the

whiskers provide the minimum and maximum value. “SLV” stands for HydraSleeve

bags.