Vehicle-based methane surveys for finding natural gas leaks and

Subscriber access provided by Kaohsiung Medical University

Energy and the Environment

Vehicle-based methane surveys for finding natural gas leaks and estimating their size: validation and uncertainty Zachary Weller, Joseph Robert Roscioli, W. Conner Daube, Brian K. Lamb, Thomas Ferrara, Paul E. Brewer, and Joseph C. von Fischer Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.8b03135 • Publication Date (Web): 20 Sep 2018 Downloaded from http://pubs.acs.org on September 21, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 27

Environmental Science & Technology

Vehicle-based methane surveys for finding natural gas leaks and estimating their size: Validation and uncertainty Zachary D. Weller,∗,†,‡ Joseph R. Roscioli,¶ W. Conner Daube,¶ Brian K. Lamb,§ Thomas W. Ferrara,k Paul E. Brewer,⊥ and Joseph C. von Fischer† 1

†Department of Biology, Colorado State University, Fort Collins CO 80523 USA ‡Department of Statistics, Colorado State University, Fort Collins CO 80523 USA ¶Aerodyne Research Incorporated, Billerica, MA 01821 USA §Laboratory for Atmospheric Research, Department of Civil & Environmental Engineering, Washington State University, Pullman, WA 99164 USA kGHD Services Incorporated, Niagara Falls, NY 14304 USA ⊥Smithsonian Environmental Research Center, Edgewater, MD 21037 USA E-mail: [email protected]

2

Abstract

3

Managing leaks in urban natural gas (NG) distribution systems is important for

4

reducing methane emissions and costly waste. Mobile surveying technologies have

5

emerged as a new tool for monitoring system integrity, but this new technology has

6

not yet been widely adopted. Here, we establish the efficacy of mobile methane surveys

7

for managing local NG distribution systems by evaluating their ability to detect and

8

locate NG leaks and quantify their emissions. In two cities, three-quarters of leak indica-

9

tions from mobile surveys corresponded to NG leaks, but local distribution companies’

1

ACS Paragon Plus Environment


10

field crews did not find most of these leaks, indicating that the national CH4 activity

11

factor for leaks in local NG distribution pipelines is underestimated by a factor of 2.4.

12

We found the median distance between mobile-estimated leak locations and actual leak

13

locations was 19 m. A comparison of emission quantification methods (mobile-based,

14

surface enclosure, and tracer-ratio) found that the mobile method over-estimated leak

15

magnitude for the smallest leaks, but accurately estimated size for the largest leaks that

16

are responsible for the majority of total emissions. Across leak sizes, mobile methods

17

adequately rank relative emission rates for repair prioritization, and they are easily

18

deployed and offer efficient spatial coverage.

19

1

Introduction

20

Repairing natural gas (NG) leaks in urban distribution systems has significant environmental,

21

economic, and public safety benefits. Methane (CH4 ) is the primary component of NG

22

and is the second most important anthropogenic greenhouse gas 1 in large part because it

23

has a global warming potential 86 times greater than CO2 over a 20-year period. 2 The

24

economic burdens of lost gas from distribution pipelines are largely carried by rate payers

25

because their rate includes the cost of lost and unaccounted-for gas. Excluding incidents

26

due to excavation or other outside force damage (e.g., motor vehicle damage), there were

27

349 reported significant gas pipeline incidents in the U.S. from January 2010 to June 2018.

28

These incidents led to 44 deaths, 244 injuries, and an estimated $161 million in direct costs; 3

29

on top of this, there were significant litigation expenses and regulatory fines to the utilities.

30

Locating and quantifying NG leaks is the first step in increasing the efficacy of NG distri-

31

bution pipeline repair and replacement efforts. New high-sensitivity, mobile CH4 monitoring

32

platforms offer a novel approach compared to traditional walking surveys for detecting and

33

quantifying NG leaks in low-pressure distribution systems and have been deployed using a

34

diversity of transport vehicles including cars, aircraft, or walking. 4 Data from these mobile

35

sensors can be used to detect and map locations with elevated CH4 concentrations, often 2


Page 2 of 27

Page 3 of 27

36


called leak indications. 5–7

37

Data from mobile platforms have also been used to estimate NG leak rates, 8,9 and their

38

ease of deployment and ability to detect leaks and quickly provide large spatial coverage

39

makes them an attractive approach for prioritizing leak repairs and pipeline replacements to

40

reduce CH4 emissions. Mobile surveys can complement required Department of Transporta-

41

tion (DOT)/Pipeline and Hazardous Materials Safety Administration (PHMSA) safety-based

42

leak surveys, which are still necessary for safety assessments that produce leak grades. Yet

43

this new technology has not yet been widely deployed, in part because there have been ques-

44

tions regarding the correspondence of leak indications with NG leaks as well as the precision

45

and accuracy of estimating leak locations and emissions.

46

We evaluate the efficacy of mobile CH4 surveys for detecting, locating and estimating the

47

size of NG leaks in local distribution systems. We assess the correspondence between leak

48

indications and actual NG leaks using the results from mobile CH4 surveys in parts of five

49

U.S. urban areas 8 and data collected during follow-up field visits to these urban areas. These

50

data also provide insights into activity factors (numbers of leaks) for national CH4 emission

51

inventories. Next, we compare emission rate estimates from mobile surveys to enclosure

52

and tracer quantification methods. Finally, we evaluate the results of our investigation to

53

characterize the value of mobile mapping for informing infrastructure repair and replacement.

54

Specifically, we assess the ability of mobile CH4 surveys to rank NG pipeline leaks by leak rate

55

as the basis for prioritizing repairs among Grade 3 (non-hazardous) leaks in order to reduce

56

CH4 emissions. The rest of the paper is organized into the following sections: materials

57

and methods; evaluating the efficacy of mobile surveys for detecting and locating leaks using

58

data from on the ground site visits; emission rate rankings and emission factor comparisons

59

among mobile, enclosure and tracer methods; and the value of mobile mapping for informing

60

infrastructure repair and replacement.

3



61

2

Materials and Methods

62

2.1

63

The mobile survey instrumentation and method is described in detail in von Fischer et al. 8

64

Briefly, we installed Picarro high-precision CH4 analyzers in three Google Street View (GSV)

65

cars and a functionally similar Los Gatos high-precision CH4 analyzer in a fourth. All CH4

66

analyzers measured atmospheric CH4 concentrations at approximately 2 Hz with a response

67

time (1/e) of 1 to 2 s. A pump drew air to the instrument via a polyethylene sampling

68

manifold mounted on the front bumper. The GSV vehicles were also outfitted with a global

69

positioning system (GPS) to record vehicle location and speed and an anemometer to record

70

wind speed and direction.

Mobile Survey Instruments and Methodology

71

We used an updated data processing algorithm relative to von Fischer et al. 8 that yields

72

as a key data product a leak indication which consists of an estimated leak location and

73

emission rate. For the mobile method, a leak indication corresponds to a location where

74

elevated CH4 concentrations (exceeding 110% of background levels) are detected during two

75

or more mobile traverses. These two traverses typically occur on different days, as we require

76

a targeted mapping area to be surveyed twice, where a second pass of the area is conducted

77

after the entire area is surveyed once. 8 See the supporting information (SI), Section 6, for

78

further details about our updated algorithm.

79

We deployed GSV cars equipped with CH4 analyzers for sampling in several U.S. cities

80

using the survey protocols described in von Fischer et al. 8 The cities included parts of Los

81

Angeles serviced by the NG distribution company Southern California Gas (SoCal Gas), and

82

four other urban areas anonymized to comply with non-disclosure agreements with local gas

83

distribution companies (LDCs). We refer to these anonymized cities/LDCs as A, B, C, and

84

D.

85

Our mobile surveys resulted in hundreds of leak indications with corresponding location

86

and emissions estimates. In each city we shared leak indication information with the LDCs

4


Page 4 of 27

Page 5 of 27


87

in the form of GPS locations and a categorical estimate of leak size. In the cases reported

88

here, LDCs visited the locations and reported their findings at each location. The leak sizes

89

were categorized as small (0-1.6 g min−1 ), medium (1.6-26 g min−1 ) and large (>26 g min−1 ).

90

These bins were based on the range of controlled CH4 releases used to calibrate an algorithm

91

for estimating leak rates from the mobile survey. For comparison, these bins represent an

92

estimated 94%, 2.4%, and 26 g min−1 ) Total, All Leak Indications

137

LDC A, % no leak found LDC B, % no leak found 88% 85% 77% 50% 0% 50% 86% 80%

indications and NG leaks, as indicated by the results of the survey crews’ investigation:

138

• mobile mapping may have a high incidence of false positives for NG leaks. These could

139

arise from elevated CH4 readings from sources besides NG leaks (e.g., biological source,

140

NG-powered vehicle exhaust),

141

• mobile mapping had instrument error, leading to elevated CH4 levels that are spurious,

142

• the LDC survey crews’ high false negative rate (i.e., failure to find leaks that are

143

present) for finding NG leaks is a result of either equipment that was not sufficiently

144

sensitive or inadequate search effort

145

• leak indications were associated with leaks that were ignored because they were too small to be graded as leaks, and/or

146

147

• the locations of the leak indications provided by the mobile platforms were not close enough to actual leak locations for the survey crews to find them.

148

149

Below we explore each of these possibilities.

150

3.2

151

In cities A and B we visited a random subset of approximately 30 of the mobile-based leak

152

indication locations where LDC survey crews did not discover a NG leak. A representative of

Follow-up Site Visits

7



Page 8 of 27

CH4 (ppm) 5 4 3

Figure 1: A representative map for our ground search to follow up on leak indications from the mobile platform that were not found by a LDC. The above image is from Google, map data: Google Imagery 2018. This map was accessible on a cell phone during our leak indication visits. The colored points represent elevated CH4 concentrations measured during the mobile survey. The white square is the estimated leak location based on the mobile CH4 measurements. The blue triangle is the location where we ultimately found a NG leak.

153

the LDC crew joined us at each of the leak indication locations that we visited, and we spent

154

a maximum of 30 minutes collaboratively surveying each location for a NG leak or other

155

CH4 emitting source. We used data from the mobile survey and a portable, high-sensitivity

156

Los Gatos Research CH4 analyzer to direct our search. Figure 1 provides an example of the

157

maps we used during the location visits. While our CH4 analyzer was sensitive to just a few

158

parts-per-billion (ppb) departures from background, LDCs often use hand-held sensors with

159

parts-per-million (ppm) level sensitivities. Thus, there was more than an order of magnitude

160

difference in the sensitivity of instruments used to measure CH4 levels.

161

We categorized the findings from our visit to each leak indication location (Table 2).

162

When we failed to find any elevated CH4 levels, we categorized the location as having no

163

source found. If we found elevated CH4 levels during our visit to the location, we collected

164

air samples to be analyzed for CH4 and ethane concentrations. We used the results of this

165

analysis to attribute elevated CH4 levels to biogenic or NG sources 16–18 . See Section S3 of 8


Page 9 of 27


Table 2: Results from the site visit to a subset of leak indications where LDCs had reported no NG leak. The data below are counts and percentages in each category. These results indicate a 72% and 68% false-negative rate, where LDCs failed to find a leak that existed within the leak indication area. Source NG, new actionable NG, recent repair evidence NG, non-actionable Biological CH4 No CH4 Source Found Other source a Totals a b

c

b

18 5 0 5 3 1 32

City A 56% Total NG: 16% 23 (72%) 0% 16% 9% 3% 100%

c

13 2 6 9 1 0 31

City B 42% Total NG: 6% 21 (68%) 19% 29% 3% 0% 100%

One leak indication corresponded to restaurant emissions on the customer side of the meter. Only half were measurable on the Heath RMLD, and two of the actionable leaks were classified as Grade 1. Three of the actionable leaks were classified as Grade 1.

166

the SI for more details on the CH4 /ethane analysis. For each leak indication attributable

167

to NG, we sub-categorized the leak indication as being actionable, non-actionable, being

168

on the customer side of the meter, or having evidence of recent infrastructure repair. We

169

sub-categorized the leak indication as actionable when the LDC field crew determined that

170

a leak indication was the result of an active and gradable (i.e., Grade 1, 2 or 3) NG leak.

171

We sub-categorized leak indications as non-actionable when elevated CH4 was due to a NG

172

leak, but the soil concentrations were low enough (typically below 20% of the lower explosive

173

limit, or 10,000 ppm) for the LDC crew to not grade the leak. Finally, we sub-categorized

174

leak indications that were not attributable to NG sources as being from a biological CH4

175

source or other CH4 source (end use).

176

3.3

177

Using the data from ground visits by both our team (Table 2) and the LDCs (Table 1), we

178

estimated that 76% (95% CI: 60%, 88%) of all mobile leak indications were attributable to

179

NG leaks in city A. That is, the false positive rate for the correspondence of leak indications

Correspondence Between Leak Indications and NG Leaks

9



180

from mobile surveys with actual NG leaks is an estimated 24%. In city B an estimated 74%

181

(95% CI: 59%, 86%) of our reported leak indications are attributable to NG. We calculated

182

these results by considering the percentage of leak indications discovered by the LDC as NG

183

leaks, and the percentage of leak indications that were not discovered by the LDC but are

184

attributable to NG. See SI Section S4 for further details on how we derived these percentages.

185

Weller et al. 9 explore the false negative rate of mobile CH4 surveys to estimate the number

186

of leaks that go undetected.

187

Based on these findings, we conclude that the majority of leak indications from mobile

188

surveys corresponded to NG leaks, and that the mobile survey method had a reasonably

189

low incidence of false positives for detecting NG leaks. These results are similar to those

190

of Phillips et al. 5 and Jackson et al. 6 who both found that the majority of elevated CH4

191

concentrations collected by their mobile monitoring system were attributable to NG sources.

192

We note that at only a very small percentage of leak indication locations did we fail to find

193

any elevated CH4 concentrations with the handheld sensors that we used during our ground

194

visit (3% in city A, 9% in city B), suggesting that spurious false positives from the mobile

195

approach were rare.

196

3.4

197

When we found a leak indication attributable to NG, we considered it a false negative leak

198

report by the LDC survey crew because they had previously reported no NG leak at the

199

location. As we show in Table 2, 70% (44 of 63) of the leak indications that we visited

200

in cities A and B corresponded to false negative leak determinations by the survey crews.

201

Of the NG leaks we found in our ground survey, the overwhelming majority (93% or 41 of

202

44) were attributable to pipeline leaks, with the remaining 7% coming from metering and

203

regulating devices. Thus we estimate that that the false negative rate specific to finding

204

pipeline leaks is 65% (41/63).

205

Leak Activity Factors

Furthermore, 11% (5/44) of these leaks, which were not initially found by the survey 10


Page 10 of 27

Page 11 of 27


206

crews, were determined to be Grade 1 leaks by the LDC following our ground survey. Grade

207

1 leaks are defined as an existing or likely hazard to people or property, 19 thus requiring

208

immediate repair. In city A, there were Grade 1 leaks at two locations and in city B, there

209

were Grade 1 leaks at three locations. It is important to recognize that leak grades are

210

assigned on the basis of proximity to structures and potential hazard as opposed to leak

211

emission rate. Small leaks can be rated as Grade 1, while larger leaks can be rated as Grade

212

2 or 3 depending primarily on location.

213

Ground survey effectiveness is an interaction of both equipment sensitivity and field

214

crew survey protocols. In most instances in both cities, the combined efforts of the LDC

215

representative and our team found leaks within five minutes of arriving at the leak indication

216

location, suggesting that field crews lack sufficiently sensitive equipment, may not have

217

been meticulous in their initial on-the-ground search, or determined that that leak was not

218

gradable. In city A, only half of the 18 actionable NG leaks we visited were measurable

219

on the LDC’s instrument, the Heath Remote Methane Leak Detector (RMLD) which has a

220

detection limit of approximately 3-5 ppm. In city B, the field crews used the Heath DPIR

221

for vehicle-based survey and the Sensit Gold G2 for the walking survey and leak grading.

222

The Sensit device was able to detect elevated CH4 levels at all the locations where we found

223

NG leaks; our work with the DPIR was not sufficiently systematic to allow comment on its

224

effectiveness.

225

Our findings suggest that the number of pipeline leaks in local distribution systems could

226

be significantly greater than indicated by current NG LDC inventories that are derived using

227

traditional survey methods alone. Our data indicates that LDC crews successfully locate 35%

228

of the pipeline leaks (65% false negative rate) that are found by the mobile survey method.

229

Interestingly, two previous studies by the NG LDC Pacific Gas & Electric (PG&E) found that

230

mobile surveys and walking surveys each find a different subset of leaks, 20,21 with relatively

231

little overlap between the two approaches (5% and 12% overlap in the two studies). Like

232

our study, these PG&E studies found an average 65% false negative rate (i.e., 35% find rate)

11



Page 12 of 27

233

for traditional walking surveys as compared to an average 32% false negative rate (i.e., 68%

234

find rate) when using the Picarro Surveyor as the mobile platform.

235

These false negative rates are approximately 4x higher than current EPA estimates, which

236

assume that LDC surveys find 85% of distribution pipeline leaks. 10,22 Our results and the

237

PG&E study indicate that the find rate for traditional LDC leak surveys is approximately

238

35%. If this detection rate is applied at the national scale instead of the currently assumed

239

85% find rate, then the national inventory for the number of pipeline leaks in NG distribution

240

systems would increase by a factor of 2.4 (see the SI, Section S5 for details on this calculation).

241

These results mirror other recent studies that estimate greater NG leakage than reported by

242

the EPA 23–25 across the NG supply chain and demonstrate how mobile surveys can inform

243

national emission inventories by improving estimates of activity factors.

244

3.5

Location Errors of Mobile-Based Leak Indications (ii): Utility D

(iii): Utilities C & D Combined

20 15 10

Frequency

4

0

50

100

Location Error (m)

150

0

0

0

1

5

2

3

Frequency

10 5

Frequency

5

6

15

7

(i): Utility C

0

50

100

150

Location Error (m)

0

50

100

150

Location Error (m)

Figure 2: The distribution of location errors for NG leaks surveyed in areas serviced by LDCs C and D. Location error is the distance between the center of the leak indication and the location of the leak as observed during the ground visit. For 75% of the leak indications shown, the mobile location error was less than 42 m.

245

We used the leak expression GPS coordinates recorded in Cities C and D to evaluate

246

the precision of mobile-based location estimates. The average distance (error) between the

247

mobile-based leak indication locations and the actual leak expression location was 31 m. The 12


Page 13 of 27


248

median error was 19 m with a standard deviation of 30 m (Figure 2). This error in location

249

estimates may encompass multiple NG service lines but offers a reasonably constrained search

250

area for discovering NG leaks. The mobile-generated leak indication location is derived from

251

a weighted average of the multiple observations of elevated CH4 concentrations (see the SI

252

Section S6.1 for more details).

253

Because many NG leaks occur off the roadway where our mobile CH4 mapping efforts

254

are constrained, we expect there to be at least some location error for leaks emanating from

255

pipes off the roadway. For example, the estimated leak indication location would be placed

256

on the roadway for a leaking service line in the adjacent front yard of a home. Despite this

257

limitation, results in Figure 2 indicate that most leak indication locations are within 42 m of

258

a NG leak, a distance that corresponds to one or two lot widths in typical urban residential

259

areas.

260

4

261

In the following subsections, we compare estimated NG leak emission rates from mobile,

262

tracer-ratio, and enclosure-based methods. Our comparisons include analyses of the differ-

263

ences as well as the ratios of the estimates from the various methods. We use these two

264

comparisons to explore both the additive (differences) and multiplicative (ratios) dissimilar-

265

ities between the different methods.

266

4.1

267

Estimated emission rates from the mobile platform were significantly positively correlated

268

with rates estimated by tracer methods used by Aerodyne Research Inc. (ARI) and GHD

269

(Figure 3). The correlation with mobile-based estimates was stronger for the GHD emission

270

rates (r = 0.92, 95% CI: 0.71, 0.98, n = 10) than for ARI (r = 0.47, 95% CI: 0.22, 0.66,

271

n = 49). The combined correlation was also significantly positive (r = 0.48, 95% CI: 0.26,

Quantifying Natural Gas Emissions

Leak Flux and Ranking Estimates: Tracer Method

13



0.1

0.3

1

3.2

10

Tracer Estimate, g min−1

31.6

31.6 ● ●

● ●

●

0.1

0.3

1

3.2

10


31.6

10

● ●● ● ● ● ● ● ● ● ●● ●● ●● ● ● ●● ● ●●

3.2

●

● ●

●

1

10

●

● ●

● ●●

●

●

● ●● ●●●● ●● ● ● ● ● ● ● ● ● ● ●

● ● ● ●

0.3

●

● ●

●

●

0.1

●

0.3

1

● ●

● ●

Mobile Estimate, g min−1

● ● ●●●● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ●● ●

3.2

3.2

● ●● ● ● ● ●

1

●

●

0.3

●

●

0.1

10

●


●

●

(iii): Utilities C & D Combined

31.6

(ii): Utility D

●

0.1


31.6

(i): Utility C

Page 14 of 27

●

0.1

0.3

1

3.2

Utility C Utility D

10


31.6

Figure 3: Comparison of emission rate estimates from the mobile platform and the tracer method; lines are 1:1 lines. Figure (i) compares the flux estimates from the mobile platform to the ARI tracer method at 49 NG leaks. Figure (ii) compares the flux estimates from the mobile platform to the GHD tracer method at 10 NG leaks.

272

0.66).

273

Mobile-based estimates were generally greater than tracer-based estimates, but the mag-

274

nitude of difference depended on leak size. Emissions estimates from the mobile platform

275

averaged 1.5 g min−1 larger than the tracer method relative to a mean leak rate of 2.1 g

276

min−1 . This mean difference in emissions estimates was significantly different from 0 (95%

277

CI 0.44, 2.48, n = 59), indicating a positive bias in the mobile based estimates. We found

278

that the over-estimation of small leaks led to a large and statistically significant average

279

multiplicative bias where leak emissions estimates were 4.4 times greater than tracer esti-

280

mates (p1.3 g min−1 .

311

We found a moderate, positive correlation (r = 0.43, 95% CI 0.10, 0.68, n = 33) between

312

mobile and surface enclosure estimates (Figure 5ii). The mobile rates were, on average, 1.7 g

313

min−1 greater than the enclosure rates (95% CI 1.2 to 2.2). When evaluated multiplicatively,

314

enclosure rates were, on average, only 15.4% the size of mobile based rates (Figure 5ii,

315

p < 0.001). This general pattern is retained for larger leaks where the mobile estimate is

316

>1.3 g min−1 .

317

There are several possible explanations for why the tracer and mobile methods are larger

318

than the surface enclosure estimates. If the seal on the enclosure is not sufficient, there

319

may be additional dilution which is not measured with the sampling system which would

320

lead to an underestimation of the leak rate. Similarly, if the enclosures failed to cover the

321

whole surface expression, it would lead to an underestimate of the total emission. It is also

322

possible that the ambient measurement methods are affected by nearby leaks not captured

323

with the enclosure. In cases where the surface expression of the leak is large (crosswind width

324

of 5 to 10 m), it is plausible that use of a point source for the tracer would not simulate

325

the distributed CH4 source, but previous tracer release studies using a second tracer as an

326

internal standard found no significant bias, even when the tracers are spatially separated. 26

327

In contrast to our findings, a previous surface enclosure study of controlled CH4 releases

328

agreed to within 12% of the metered release rates (see the SI) over the range 1.6 to 4.8 g

329

min−1 . Repeated enclosure measurements of the same set of actual leaks over a 3-week period

330

showed variability up to 60% from week to week. 10 In our study, most mobile measurements

331

were not conducted at the same time as the tracer ratio and enclosure measurements.

332

SoCal Gas found similar magnitudes of disagreement between enclosure estimates and

17



333

mobile estimates of leak rates. 27 Following our mobile-based estimation of leak rates in their

334

Los Angeles service area, SoCal Gas hired GHD to conduct enclosure-based estimates of

335

leak rates. SoCal Gas shared these results with us and later made them public in a legal

336

proceeding. 28 SoCal reported that the mobile-based emissions estimates were 13 times greater

337

than the enclosure estimates, 27 which is equivalent to enclosure results averaging 7.7% the

338

size of the mobile-based emission estimates. Our results suggest that this difference emerged

339

from an overestimate of small leaks by the mobile method, as well as the underestimate of

340

rates by the enclosure method.

341

5

342

5.1

343

We have demonstrated that a large majority of leak indications from mobile CH4 monitoring

344

platforms correspond to NG leaks, that leak indication locations are sufficient for finding

345

NG leaks, and that leak flux estimates are effective for ranking leak sizes. Because NG LDC

346

field crews were unable to locate these NG leaks, our findings suggest that NG distribution

347

systems have many more leaks than LDCs are able to find using their existing leak survey

348

equipment and methods. This may partly reflect differences in how a leak is defined by

349

LDCs, the sensitivity of the instrumentation used for surveys, and/or the breadth of area

350

associated with mobile leak indications vs single address leak designations. From a public

351

safety perspective, this is concerning because some of the leaks we discovered during our

352

ground visit were categorized as Grade 1.

Value of Mobile Mapping for Repair Prioritization Finding Natural Gas Leaks

353

These results raise questions about why field crews were unable to find leaks indicated

354

by the mobile platform. We suspect differences in instruments are partly responsible. The

355

Picarro and Los Gatos instruments on the mobile platform are able to measure CH4 con-

356

centrations at a very high frequency (2 times per second) and are highly sensitive (detecting

357

changes down to 0.001 part-per-million), while the instruments used by LDCs are generally 18


Page 18 of 27

Page 19 of 27


358

less sensitive. It is also possible that crews using hand-held sensors are only able to detect

359

leaks at ground level and may not detect small atmospheric plumes from a nearby source. We

360

hypothesize that using more sensitive instruments coupled with additional training would

361

reduce the false negative rate and improve leak detection.

362

5.2

Prioritization Method Tracer Estimate Mobile Estimate Random Repair

0.6 0.4 0.2

Cumulative Emissions

0.8

1.0

Using Leak Size Information

Emissions Reduction Comparisons

0.0

Mobile vs. Random Mobile vs. Tracer

0

5

10

15

20

25

30

35

40

45

Leak Ranking (largest to smallest)

Figure 6: Cumulative emissions curves (CECs) from the ARI tracer and mobile quantification methods. The solid red and black lines show the estimated cumulative emissions curves and the shading shows 95% confidence intervals. Despite imprecision in the mobile estimates, the effectiveness of prioritizing leak repair for approximately the top 15% of leaks is equivalent for the mobile and tracer methods.

363

The mobile-based leak rate estimates are useful for prioritizing leak repair and pipeline

364

replacement, despite uncertainty in estimates of absolute leak emission rates. To illustrate

365

the relative benefit of using mobile quantification as compared with tracer methods and a

366

random repair approach, we constructed cumulative emission curves for each approach. We

367

construct the cumulative emission curves by first ranking the leaks from largest to smallest 19



368

based on estimated emission rate, then computing the proportion of the estimated total

369

emissions attributable to the largest x leaks (Figure 6 for city C). We computed the uncer-

370

tainty in the cumulative emissions curves using a bootstrap approach. See SI Section S7 for

371

more details on the bootstrap analysis.

372

The tracer-based cumulative emission curve for city C indicates that the top 9 leaks

373

(18% of leaks) account for an estimated 67% (64%-71%) of the total emissions (Figure 6).

374

Other NG leak studies have similarly found that a small proportion of leaks account for the

375

majority of emissions. 9,29,30 When compared to the tracer-based cumulative emission curve,

376

it is clear that the mobile approach does an effective job of characterizing the emissions,

377

while also being much quicker and much lower cost than the tracer method. Despite the

378

uncertainty in both the tracer ratio and mobile flux estimates, the mobile-based cumulative

379

emission curve is not significantly different from the tracer curve for the largest leaks (i.e.,

380

top 9). However, a gap emerges between the tracer and mobile curves due to the aforemen-

381

tioned over-estimate of small leaks by the mobile approach. Both tracer and mobile-based

382

curves lie well above the random repair line, indicating that the mobile-based rankings offers

383

an effective approach, in terms of spatial coverage and acceptable quantification accuracy,

384

for reducing NG system leakage. This efficacy results from the mobile-based rankings suc-

385

cessfully identifying the largest leaks, which contribute disproportionately to total emissions.

386

Further improvements in mobile leak rate estimation methodologies will increase precision

387

and accuracy and therefore the efficacy of ranking leaks for repairs, but it is clear that cur-

388

rent precision is more than sufficient for effectively characterizing the population of leaks

389

and cost-effective prioritization within leak repair and pipeline replacement programs such

390

that large reductions in methane emissions can be achieved quickly and easily.

391

In light of differences among quantification methods, there remain open questions about

392

the accurate quantification of NG leak emission rates. For example, evidence here suggests

393

that the surface enclosure method tends to under-estimate leak rates compared to estimates

394

obtained from downwind ambient measurements and this may be more problematic at higher

20


Page 20 of 27

Page 21 of 27


395

leak rates. Further work is required to test both approaches in complex urban environments

396

and for the type of dispersed surface expressions associated with below ground pipeline leaks.

397

Likewise, it appears that mobile methods over-estimate the size of small leaks. Understanding

398

these discrepancies and their associated uncertainties is important for an accurate accounting

399

of emissions from and management of NG distribution systems, yet they do not undercut the

400

value of mobile methods for detecting and ranking leaks for effective system management.

401

Finally, mobile surveys are not a substitute for the regular walking surveys conduct

402

by LDCs. Previous work has revealed that some NG leaks are detected only by mobile

403

surveying but not by walking surveys, and vice-versa. 20,21 Of course, the process of grading

404

leaks for safety requires boots on the ground, but the co-deployment of vehicle-based surveys

405

can help find more leaks and also provides space- and time-effective monitoring. Currently,

406

mobile surveys cannot substitute for walking leak surveys, because mobile surveys can fail to

407

detect NG leaks under some environmental conditions and at some locations. 9 Despite these

408

limitations, mobile surveys are a powerful new tool available for effectively and responsibly

409

managing NG distribution systems.

410

Acknowledgement

411

This work was funded by grants from the Alfred P Sloan Foundation and Robertson Foun-

412

dations to the Environmental Defense Fund. The authors thank Steven P. Hamburg, Ramon

413

Alvarez, and Jennifer Hoeting for helpful feedback that improved the manuscript. The au-

414

thors also thank Duck Keun Yang and Sam Chamberlain for helping with data collection

415

and management.

416

Supporting Information Available

417

The following file is available online as supporting information:

21



418

• Supporting Information for Vehicle-based methane surveys for finding nat-

419

ural gas leaks and estimating their size: validation and uncertainty. The SI

420

contains further information on data collection, processing, and analysis.

421

References

422

(1) Boucher, O.; Friedlingstein, P.; Collins, B.; Shine, K. P. The indirect global warming

423

potential and global temperature change potential due to methane oxidation. Environ.

424

Research Letters 2009, 4, 044007.

425

(2) Myhre, G.; Shindell, D.; BrÂŐon, F.-M.; Collins, W.; Fuglestvedt, J.; Huang, J.;

426

Koch, D.; Lamarque, J.-F.; Lee, D.; Mendoza, B.; Nakajima, T.; Robock, A.;

427

G. Stephens, T. T.; Zhang, H. In Climate change 2013: The Physical Science Basis.

428

Contribution of working group I to the fifth assessment report of the intergovernmen-

429

tal panel on climate change; Stocker, T., Qin, D., Plattner, G., Tignor, M., Allen, S.,

430

Boschung, J., Nauels, A., Xia, Y., Bex, B., Midgley, B., Eds.; Cambridge University

431

Press, 2013.

432

(3) PHMSA, Distribution, Transmission & Gathering, LNG, and Liquid Accident and

433

Incident Data. 2018; https://www.phmsa.dot.gov/pipeline/library/data-stats/

434

pipelineincidenttrends.

435

(4) Eapi, G. R.; Sabnis, M. S.; Sattler, M. L. Mobile measurement of methane and hydrogen

436

sulfide at natural gas production site fence lines in the Texas Barnett Shale. J. Air &

437

Waste Manag. Assoc. 2014, 64, 927–944.

438

(5) Phillips, N. G.; Ackley, R.; Crosson, E. R.; Down, A.; Hutyra, L. R.; Brondfield, M.;

439

Karr, J. D.; Zhao, K.; Jackson, R. B. Mapping urban pipeline leaks: Methane leaks

440

across Boston. Environ. Pollut. 2013, 173, 1–4.

22


Page 22 of 27

Page 23 of 27


441

(6) Jackson, R. B.; Down, A.; Phillips, N. G.; Ackley, R. C.; Cook, C. W.; Plata, D. L.;

442

Zhao, K. Natural gas pipeline leaks across Washington, DC. Environ. Sci. & Technol.

443

2014, 48, 2051–2058.

444

(7) Gallagher, M. E.; Down, A.; Ackley, R. C.; Zhao, K.; Phillips, N.; Jackson, R. B. Natural

445

gas pipeline replacement programs reduce methane leaks and improve consumer safety.

446

Environ. Sci. & Technol. Letters 2015, 2, 286–291.

447

(8) von Fischer, J. C.; Cooley, D.; Chamberlain, S.; Gaylord, A.; Griebenow, C. J.; Ham-

448

burg, S. P.; Salo, J.; Schumacher, R.; Theobald, D.; Ham, J. Rapid, Vehicle-Based

449

Identification of Location and Magnitude of Urban Natural Gas Pipeline Leaks. Envi-

450

ron. Sci. & Technol. 2017, 51, 4091–4099.

451

(9) Weller, Z. D.; Hoeting, J. A.; von Fischer, J. C. A Calibration-Capture-Recapture Model

452

for Inferring Natural Gas Leak Population Characteristics Using Data From Google

453

Street View Cars. to appear, Environmetrics 2018, https://doi.org/10.1002/env.2519 .

454

(10) Lamb, B. K.; Edburg, S. L.; Ferrara, T. W.; Howard, T.; Harrison, M. R.; Kolb, C. E.;

455

Townsend-Small, A.; Dyck, W.; Possolo, A.; Whetstone, J. R. Direct measurements

456

show decreasing methane emissions from natural gas local distribution systems in the

457

United States. Environ. Sci. & Technol. 2015, 49, 5161–5169.

458

(11) Allen, D. T.; Torres, V. M.; Thomas, J.; Sullivan, D. W.; Harrison, M.; Hendler, A.;

459

Herndon, S. C.; Kolb, C. E.; Fraser, M. P.; Hill, A. D. Measurements of methane

460

emissions at natural gas production sites in the United States. Proc. Natl. Acad. Sci.

461

U.S.A. 2013, 110, 17768–17773.

462

(12) Mitchell, A. L.; Tkacik, D. S.; Roscioli, J. R.; Herndon, S. C.; Yacovitch, T. I.; Mar-

463

tinez, D. M.; Vaughn, T. L.; Williams, L. L.; Sullivan, M. R.; Floerchinger, C. Measure-

464

ments of methane emissions from natural gas gathering facilities and processing plants:

465

Measurement results. Environ. Sci. & Technol. 2015, 49, 3219–3227. 23



466

(13) Lamb, B. K.; McManus, J. B.; Shorter, J. H.; Kolb, C. E.; Mosher, B.; Harriss, R. C.;

467

Allwine, E.; Blaha, D.; Howard, T.; Guenther, A. Development of atmospheric tracer

468

methods to measure methane emissions from natural gas facilities and urban areas.

469

Environ. Sci. & Technol. 1995, 29, 1468–1479.

470

(14) Shorter, J. H.; McManus, J. B.; Kolb, C. E.; Allwine, E. J.; Siverson, R.; Lamb, B. K.;

471

Mosher, B. W.; Harriss, R. C.; Howard, T.; Lott, R. A. Collection of leakage statistics

472

in the natural gas system by tracer methods. Environ. Sci. & Technol. 1997, 31, 2012–

473

2019.

474

(15) Lamb, B. K.; Cambaliza, M. O.; Davis, K. J.; Edburg, S. L.; Ferrara, T. W.; Flo-

475

erchinger, C.; Heimburger, A. M.; Herndon, S.; Lauvaux, T.; Lavoie, T. Direct and

476

indirect measurements and modeling of methane emissions in Indianapolis, Indiana.

477

Environ. Sci. & Technol. 2016, 50, 8910–8917.

478

(16) Yacovitch, T. I.; Herndon, S. C.; Roscioli, J. R.; Floerchinger, C.; McGovern, R. M.;

479

Agnese, M.; PeÌĄtron, G.; Kofler, J.; Sweeney, C.; Karion, A. Demonstration of an

480

ethane spectrometer for methane source identification. Environ. Sci. & Technol. 2014,

481

48, 8028–8034.

482

(17) McKain, K.; Down, A.; Raciti, S. M.; Budney, J.; Hutyra, L. R.; Floerchinger, C.;

483

Herndon, S. C.; Nehrkorn, T.; Zahniser, M. S.; Jackson, R. B. Methane emissions from

484

natural gas infrastructure and use in the urban region of Boston, Massachusetts. Proc.

485

Natl. Acad. Sci. U.S.A. 2015, 112, 1941–1946.

486

(18) Karion, A.; Sweeney, C.; Kort, E. A.; Shepson, P. B.; Brewer, A.; Cambaliza, M.; Con-

487

ley, S. A.; Davis, K.; Deng, A.; Hardesty, M. Aircraft-based estimate of total methane

488

emissions from the Barnett Shale region. Environ. Sci. & Technol. 2015, 49, 8124–8131.

489

(19) California Public Utilities Commission, Survey of Natural Gas Leakage Abate-

490

ment Best Practices. 2015; https://www.socalgas.com/regulatory/documents/ 24


Page 24 of 27

Page 25 of 27


491

r-15-01-008/R.15-01-008%20SED%20Report%20Survey%20of%20Natural%20Gas%

492

20Leakage%20Abatement%20Practices%20031815%20Rpt%20Only.pdf.

493

494

(20) Clark, T.; Conley, E.; Kerans, M.; Piazza, M. Picarro Surveyor Leak Detection Study: Diablo Side-by-Side Study. Pacific Gas and Electric 2012, Internal Study Report.

495

(21) Kerans, M.; Clark, T.; Conley, E.; Crosson, E.; Piazza, M. Picarro Surveyor Leak De-

496

tection Study: Sacramento Side-by-Side Study. Pacific Gas and Electric 2012, Internal

497

Study Report.

498

(22) Campbell, L. M.; Campbell, M. V.; Epperson, D. L. Methane Emissions from the

499

Natural Gas Industry, Volume 9: Underground Pipeline. 1996; https://www.epa.gov/

500

sites/production/files/2016-08/documents/9_underground.pdf.

501

(23) Miller, S. M.; Wofsy, S. C.; Michalak, A. M.; Kort, E. A.; Andrews, A. E.; Biraud, S. C.;

502

Dlugokencky, E. J.; Eluszkiewicz, J.; Fischer, M. L.; Janssens-Maenhout, G. Anthro-

503

pogenic emissions of methane in the United States. Proc. Natl. Acad. Sci U.S.A. 2013,

504

110, 20018–20022.

505

(24) Wunch, D.; Toon, G. C.; Hedelius, J. K.; Vizenor, N.; Roehl, C. M.; Saad, K. M.;

506

Blavier, J.-F. L.; Blake, D. R.; Wennberg, P. O. Quantifying the loss of processed

507

natural gas within California’s South Coast Air Basin using long-term measurements

508

of ethane and methane. Atmosph. Chem. and Physics 2016, 16, 14091.

509

(25) Alvarez, R. A.; Zavala-Araiza, D.; Lyon, D. R.; Allen, D. T.; Barkley, Z. R.;

510

Brandt, A. R.; Davis, K. J.; Herndon, S. C.; Jacob, D. J.; Karion, A. Assessment

511

of methane emissions from the US oil and gas supply chain. Science 2018, eaar7204.

512

(26) Roscioli, J.; Yacovitch, T.; Floerchinger, C.; Mitchell, A.; Tkacik, D.; Subramanian, R.;

513

Martinez, D.; Vaughn, T.; Williams, L.; Zimmerle, D. Measurements of methane emis-

514

sions from natural gas gathering facilities and processing plants: measurement methods.

515

Atmos. Meas. Techniques 2015, 8, 2017. 25



516

(27) Newton, E. Southern California Gas Company’s Verification Study of the Methane

517

Mapping of Four California Cities by the Environmental Defense Fund and Colorado

518

State University. Online, 2016; https://socalgas.com/regulatory/documents/

519

r-15-01-008/EDF_4-Cities_Methane_Mapping_Report_Final_081916.pdf.

520

(28) Hovsepian, M. Supplemental Replay Comments of Southern California Gas Company

521

(U 904 G) and San Diego Gas and Electric Company (U 902 G) On Best Practices

522

Recommendations. 2015; http://docs.cpuc.ca.gov/PublishedDocs/Efile/G000/

523

M166/K474/166474903.PDF.

524

(29) Brandt, A. R.; Heath, G.; Kort, E.; O’sullivan, F.; Pétron, G.; Jordaan, S.; Tans, P.;

525

Wilcox, J.; Gopstein, A.; Arent, D. Methane leaks from North American natural gas

526

systems. Science 2014, 343, 733–735.

527

(30) Balcombe, P.; Anderson, K.; Speirs, J.; Brandon, N.; Hawkes, A. The natural gas sup-

528

ply chain: the importance of methane and carbon dioxide emissions. ACS Sustainable

529

Chem. & Eng. 2016, 5, 3–20.

26


Page 26 of 27

Page 27 of 27

530


Graphical TOC Entry

531

27


Vehicle-based methane surveys for finding natural gas leaks and

Recommend Documents