Systematically Incorporating Environmental Objectives into Shale Gas

May 3, 2019 - Department of Electrical Engineering, Pontificia Universidad Católica de Chile, Av. Vicuna. Mackenna 4860, Santiago, Chile. d. First St...
0 downloads 0 Views 660KB Size
Subscriber access provided by Bethel University

Energy and the Environment

Systematically incorporating environmental objectives into shale gas pipeline development: A binary integer, multi-objective spatial optimization model Kailin Kroetz, Jhih-Shyang Shih, Juha V. Siikamäki, Vladimir Marianov, Alan J. Krupnick, and Ziyan Chu Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.9b01583 • Publication Date (Web): 03 May 2019 Downloaded from http://pubs.acs.org on May 22, 2019

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 25

Environmental Science & Technology

Systematically incorporating environmental objectives into shale gas pipeline development: A binary integer, multi-objective spatial optimization model

Kailin Kroetza,*, Jhih-Shyang Shiha, Juha V. Siikamäkib, Vladimir Marianovc, Alan Krupnicka, and Ziyan Chud

*Corresponding author. Email: [email protected]. a Resources for the Future, 1616 P St. NW, Washington, DC 20036 b International Union for Conservation of Nature, 1630 Connecticut Ave. NW, Suite 300, Washington, DC 20009 c Department of Electrical Engineering, Pontificia Universidad Católica de Chile, Av. Vicuna Mackenna 4860, Santiago, Chile d First Street Foundation, Brooklyn, NY 11201

Keywords: Binary integer programming, multi-objective, spatial optimization, shale gas, pipeline, systematic conservation planning

1 ACS Paragon Plus Environment

Environmental Science & Technology

Page 2 of 25

1 2

Abstract

3 4 5 6 7 8 9 10 11 12 13 14 15 16

Shale gas pipeline development can have negative environmental impacts, including adverse effects on species and ecosystems through habitat degradation and loss. From a societal perspective, pipeline development planning processes should account for such externalities. We develop a multi-objective binary integer-programming model, called the Multi Objective Pipeline Siting (MOPS) model, to incorporate habitat externalities into pipeline development and to estimate the tradeoffs between pipeline development costs and habitat impacts. We demonstrate the utility of the model using an application from Bradford and Susquehanna counties in northeastern Pennsylvania. We find that significant habitat impacts could be avoided for relatively low cost, but that avoiding additional habitat impacts becomes gradually and increasingly costly. For example, 10 percent of the habitat impacts can be avoided at less than a two percent pipeline cost increase relative to a configuration that ignores habitat impacts. MOPS or a similar model could be integrated into the pipeline siting and permitting process, so oil and gas companies, communities, and states can identify cost-effective options for habitat conservation near shale gas development.

17

2 ACS Paragon Plus Environment

Page 3 of 25

18 19

Environmental Science & Technology

1. Introduction Advances in drilling technologies have led to a boom in oil and gas development in the

20

United States. This growth has provided a number of benefits, including lower energy prices,

21

displaced coal in electricity production (and related health and climate policy benefits1, 2), and

22

increased energy independence. However, shale gas extraction requires intense development (if

23

possibly less intense than conventional well development), including the construction of well

24

pads, gathering pipelines, larger inter- and intra-state pipelines, compressor stations, and roads to

25

transport materials. These infrastructure needs can lead to tradeoffs between shale gas

26

development and environmental impacts.3

27

Environmental impacts of shale gas development depend on the ecosystem in which

28

development is occurring. They also depend on the distribution of pipelines, wellpads, and wells

29

within the ecosystem. Documented impacts include ecosystem disruptions, habitat loss, core

30

forest loss, and forest fragmentation.4-10 These impacts are associated with a decline in forest

31

biodiversity, at both local and regional scales.11 Moreover, forest biodiversity is critical for the

32

stable provision of ecosystem services, including natural resource availability and carbon

33

sequestration.12 Natural gas development can also adversely affects surface water systems by

34

reducing streamflow and introducing chemical and sediment runoff.13 Reduced streamflow may

35

inhibit wetland productivity, and pipelines pose a risk of rupture. Approximately half the global

36

wetland area has already been lost, and thus ecosystem services provided by wetlands (e.g. flood

37

abatement, water quality benefits, and carbon sequestration) are highly valued and threatened by

38

irreversible damage.14

39 40

Recent gas development in Pennsylvania and West Virginia intersects with forested areas, which has spurred a significant body of literature focused on how development has 3 ACS Paragon Plus Environment

Environmental Science & Technology

Page 4 of 25

41

contributed to core forest loss.6-9 Studies within and outside of Pennsylvania have found that

42

shale gas pipeline infrastructure development is a primary contributor to rural land-use change

43

and forest fragmentation. For example, Abrahams, Griffin and Matthews 5 found gathering lines

44

account for 94 percent of incremental fragmentation of core forest in Bradford County,

45

Pennsylvania. Langlois, Drohan and Brittingham 4 determined core forest decreased by 4 percent

46

in Lycoming County, Pennsylvania due to shale development, with pipelines and roads

47

responsible for around 80 percent of that loss. Focusing on seven counties in the Barnett shale in

48

Texas, Jordaan, et al. 15 found 74 percent of total land use change was related to midstream

49

infrastructure (mainly, pipelines).

50

Despite frequent calls for assessment of habitat externalities and for incorporating them

51

into systematic shale gas development planning across larger spatial scales (see e.g. Drohan,

52

Brittingham, Bishop and Yoder 6, Northrup and Wittemyer 16, Mauter, et al. 17, Middleton and

53

Brandt 18), the availability of research on how to integrate these concepts into models that can be

54

used for policy-making is limited. Studies focusing on larger spatial scales, such as the state- or

55

county-level, have mainly focused on retrospective analyses characterizing past land-use change

56

related to shale development or pipelines, or have projected future impacts based on status quo

57

policies and technologies 4, 6-9, 19. Only one study we are aware of, Abrahams, Griffin and

58

Matthews 5, conducts an analysis of a counterfactual shale gas development plan at the county-

59

scale or larger. The Abrahams, Griffin and Matthews 5 model assesses forest impacts under two

60

pre-determined policies: requiring pipelines to follow roads and reducing well pad density.

61

To our knowledge, ours is the first quantitative model to systematically enumerate

62

tradeoffs between habitat impacts and pipeline siting costs on a large spatial scale. Other recent

63

studies have incorporated environmental externalities into modeling shale gas supply chains 20-22, 4 ACS Paragon Plus Environment

Page 5 of 25

Environmental Science & Technology

64

but these studies do not model pipeline siting. Doing so requires a quantitative approach to

65

optimizing pipeline configuration under economic and environmental objectives, including

66

evaluating tradeoffs. Here, we address this gap in the literature by developing a multi-objective

67

binary integer-programming model, the Multi Objective Pipeline Siting (MOPS) model, which

68

incorporates habitat considerations into pipeline development and estimates the tradeoffs

69

between pipeline development costs and habitat impacts. The model takes well locations as

70

fixed, and then solves for the least-cost pipeline configuration, given constraints on habitat

71

impacts and the requirement that each well be connected to a larger intra- or inter-state pipeline

72

via gathering lines. The model formulation is novel in that it accounts for pipeline connectivity,

73

flow direction, and allows merging of upstream pipelines.

74

To demonstrate an application of MOPS to pipeline siting and permitting processes, we

75

apply it using data from Bradford and Susquehanna counties in northeastern Pennsylvania. We

76

find that large amounts of habitat impacts could be avoided for a relatively low cost, but that

77

avoiding additional habitat impacts becomes gradually and increasingly costly. For example, 10

78

percent of habitat impacts can be avoided at less than a two percent cost increase over the

79

privately optimal pipeline configuration.

80

2. Methods

81

Below we discuss the optimization model and our methods for applying it to a case study.

82 83

2.1 Optimization Model While most pipeline design models focus on cost and construction details 23, 24, the MOPS

84

model is designed to calculate the tradeoff between pipeline development costs and

85

environmental outcomes. The optimization problem minimizes a weighted combination of cost 5 ACS Paragon Plus Environment

Environmental Science & Technology

86

and habitat impacts of building a pipeline network. We treat these two objectives as

87

incommensurable, as is often the case in multi-objective optimizations. Each component is

88

assigned a weight such that the weights add up to one and varying the weight on each term

89

allows us to develop a tradeoff curve between cost and habitat impacts. An equivalent

90

representation is a function minimizing habitat impact subject to a cost constraint. Under this

91

formulation a non-binding cost constraint is equivalent to a weight of zero on cost and positive

92

weight on habitat impacts; a cost constraint equal to the cost of a solution associated with the

93

minimum cost pipeline configuration is equivalent to a weight of zero on the habitat term. An

94

alternative approach would be to minimize social costs, which consist of private costs and the

95

external damage to habitat. The challenging nature of valuation of habitat favors starting with a

96

weighting/tradeoff approach.

97

We measure habitat impacts relative to the cost of pipeline construction using MOPS to

98

derive a tradeoff curve for the marginal cost of avoiding habitat impacts, in terms of the cost of

99

pipeline configuration. There are two endpoints, or corner solutions, that arise from the

100

weighting scheme: full weight on pipeline construction costs (minimum cost solution that

101

disregards habitat impacts) and full weight on habitat impacts (minimum impact solution that

102

disregards pipeline construction costs).

103

Page 6 of 25

The model is formulated such that shale gas is collected from all origin wells through

104

gathering pipelines, which then connect to compression stations on the transmission pipeline. A

105

conceptual diagram of the problem and two extreme solutions (minimum cost and minimum

106

impact, respectively) are presented in Figure 1; a full mathematical formulation of the problem is

107

available in the Supporting Information (SI).

6 ACS Paragon Plus Environment

Page 7 of 25

108

Environmental Science & Technology

The basis of the optimization model is the representation of a continuous landscape as a

109

grid, allowing pipeline placement to be modeled as a binary integer program. Integer

110

programming has been used to solve pipeline siting models before (see e.g. Brimberg, et al. 25),

111

and in our case allows for inclusion of both pipeline siting constraints as well as habitat impacts,

112

while still being able to solve the problem. The centroid of each grid cell is a node in the integer

113

programming problem. Within the grid we specifically identify two set of nodes: (1) origins, in

114

this case wells, and (2) destinations, in this case all nodes through which transmission lines pass.

115

Binary variables are used to represent whether a pipeline is present on the path connecting the

116

nodes of two adjacent grid cells (called an arc). Attributes can be associated with an arc, for

117

example, cost, distance, population exposure, habitat area, and types of land use. The objective

118

function of the integer program is the weighted sum of: (1) cost of pipeline construction for each

119

arc where a pipeline is present; and (2) the area of forest and wetland along the arc where

120

pipeline is present (with area along the arc being a proxy for habitat impact).

121

Figure 1: Conceptual Model

7 ACS Paragon Plus Environment

Environmental Science & Technology

Page 8 of 25

122 123

By formulating the gathering pipeline layout design as a multiple origins (wells) and

124

multiple destinations (transmission pipeline connectors) (MOMD) problem, we are able to draw

125

from the solution algorithm for other MOMD network analysis problems, such as logistics

126

planning in optimizing package delivery subject to shipping cost26 to solve the problem. The

127

MOMD problem resembles the well-known minimum Steiner tree problem 27. The Steiner tree

128

problem in graphs can be seen as a generalization of two other famous combinatorial

129

optimization problems: the (non-negative) shortest path problem and the minimum spanning tree

130

problem. If a Steiner tree problem in graphs contains exactly two terminals, it reduces to finding

131

a shortest path. If, on the other hand, all vertices are terminals, the Steiner tree problem in graphs

132

is equivalent to the minimum spanning tree.

133

Conventional network analysis models focused on connecting sources and destinations

134

would require all the sources and destinations in the final layout. However, for the gathering

135

pipeline problem, not all the potential transmission pipeline connectors (destinations) must be in 8 ACS Paragon Plus Environment

Page 9 of 25

Environmental Science & Technology

136

the optimal layout. Therefore, the optimization model must determine how many and which

137

connectors should be in the optimal layout. To allow such flexibility, we augment the network by

138

adding a pseudo (destination) node which connects all the transmission pipeline connectors and

139

treats connectors as intermediate nodes. We also assign zero value to all the attributes for the

140

arcs between transmission line connectors and pseudo destination nodes. This approach provides

141

the flexibility for the optimal pipeline layout to include or exclude potential transmission

142

connectors. We solve the model using a commercial mixed integer programming solver,

143

specifically the CPLEX solver in GAMS (GAMS Development Corporation 2018).

144

To validate the model, we compare the observed configuration of pipelines to the

145

modeled optimal configuration without accounting for habitat impact. To empirically measure fit

146

we calculate, for each grid cell, whether there is a model pipeline, an observed pipeline, or both.

147

Recognize, however, that gathering line siting may, to an unknown extent, already account for

148

the avoidance of habitat impacts in the process of meeting stakeholder concerns that arise during

149

the siting process. Thus, any validation of the model by comparing model results to existing

150

pipeline locations may be of limited utility.

151

2.2 Estimation of Tradeoffs in Bradford and Susquehanna, PA

152

We apply the model to data from Bradford and Susquehanna counties in northeastern

153

Pennsylvania. We choose northeastern Pennsylvania because there has been considerable shale

154

gas pipeline construction over the past decade. Additionally, the counties were selected for study

155

based on data availability, pipeline characteristics, and environmental characteristics. We use

156

two datasets in the study: pipeline location data from Rextag, current as of 2016, and 2001 land

157

cover data from the National Land Cover Database (NLCD). We use the 2001 NLCD land cover

158

data to match the land cover in the region at the beginning of the pipeline construction boom in 9 ACS Paragon Plus Environment

Environmental Science & Technology

Page 10 of 25

159

the region. We searched within Pennsylvania for a study area with a pipeline network relatively

160

dense within the area and relatively disconnected to nearby areas, to reduce the impact of edge-

161

effects on our results (i.e. modeled pipeline design being influenced by a pipeline network

162

outside the study area).

163

The region, as shown in pink in Figure 2, includes 117 wells across an area of about 364

164

square kilometers. The existing gathering (brown) pipeline and transmission (black) pipeline are

165

based on 2016 production well data from the Pennsylvania Department of Environmental

166

Protection and pipeline data from Hart Energy Mapping & Data Services (Figure 2). There are

167

eight connectors for connecting gathering and transmission pipelines. Within the study area,

168

approximately 53 percent of land cover is forest or wetlands, which we identify as previously

169

undeveloped habitat, and is the focus of our modeling effort. Of this previously undeveloped

170

habitat area, 95 percent is forest and 5 percent is wetland.

171

Figure 2.

10 ACS Paragon Plus Environment

Page 11 of 25

Environmental Science & Technology

172 173 174 175 176

(a) MOPS cost minimizing solution. The pipeline configuration represents optimal pipeline placement to minimize pipeline construction costs, with no limit on the quantity of previously undeveloped habitat traversed. Pennsylvania state and county spatial data are from the US Census Bureau. Pipeline and well spatial data are from Rextag. Maps were generated using ArcGIS 10.3 (ESRI, Redlands, CA, USA: http://www.esri.com/software/arcgis).

177 178 179 180 181 182

(b) MOPS habitat impact reduction versus minimum cost solutions. Red translucent coloring is used to represent the pipeline in the cost minimization solution; the grey pipeline configuration represents pipeline placement in the solution with full weight on wetland and forest habitat. Pennsylvania state and county data from the US Census Bureau and land cover is derived from the 2001 NLCD. Maps were generated using ArcGIS 10.3 (ESRI, Redlands, CA, USA: http://www.esri.com/software/arcgis).

183

To apply the model to this region we represent the area as a set of nodes and arcs. There

184

is a tradeoff between computational burden and incorporation of heterogeneity of the land cover.

185

We find that using 1,010 evenly-spaced grid cells across the study area, each about 600 meters

186

by 600 meters, provides adequate treatment of landscape heterogeneity while remaining

187

computationally feasible to solve. We assume that a pipeline, if constructed, will go through and

188

connect the center of a grid cell.

189 190

For each potential connection (arc), we calculate several attributes including distance, forest and wetland area, slope of pipeline connection, and the length of any water crossings. We 11 ACS Paragon Plus Environment

Environmental Science & Technology

Page 12 of 25

191

include the latter two elements because, although data is not easily accessible, literature has

192

claimed that the cost of materials and construction of shale gas pipelines increases if pipelines go

193

through difficult terrain such as sloped land (see e.g. Marcoulaki, et al. 28) or over/under river

194

channels (see e.g. Marcoulaki, Tsoutsias and Papazoglou, 28 Oil and Gas Journal, 29 Racicot, et al.

195

30).

196

the grid cells the arc connects. If the slope is greater than a pre-specified number, we assume

197

construction along this arc would be cost-prohibitive and constrain the arc to be an infeasible

198

connection. Given that we only measure the difference in slope between nodes and not the

199

maximum slope at any point along the connection route, the measured slope of the arc may

200

underestimate the maximum slope.

201

We define the slope of an arc to be the difference in elevation between the two centroids of

We also account for water crossings in our model. Through correspondence with multiple

202

shale gas industry contacts we established “ballpark” estimates of the increased cost of routing

203

shale gas gathering pipelines across water crosings as a function of three main categories of

204

crossing width: small (less than 10 feet), medium (10-100 feet), and large (greater than 100 feet).

205

For small crossings we assume a cost increase of 10 percent for the arc (which is around 600

206

meters or .4 miles), for medium crossings we assume an increase of 20 percent, and for large an

207

increase of 150 percent.

208

We follow previous models and use distance as a proxy for private pipeline infrastructure

209

cost to the developer (see e.g. Abrahams, Griffin and Matthews 5). Following Lade and Rudik 31

210

we assume that gathering lines are 4 inches in diameter. Then, using an estimate of the cost per

211

inch-mile from the Interstate National Gas Association of America 32, and the PPI to convert the

212

estimate to $2016, we estimate the cost per mile of gathering line as $105,936 ($2016).

12 ACS Paragon Plus Environment

Page 13 of 25

Environmental Science & Technology

213

Forest and wetland area affected is used to approximate the habitat impact as an

214

environmental externality. Coverage is calculated as the average percentage of the area of the

215

two connected grid cells covered by forest or wetland. There are no protected areas in our

216

selected study area, so, as an important simplification, we treat all forest and wetland areas with

217

equal weight.

218

3. Results

219

3.1 Model Validation

220

We first estimate and validate the model by running it with the objective of minimizing

221

the total pipeline construction cost and compare the optimal pipeline network from the model

222

results with the existing pipeline network. Figure 2a shows the cost minimizing pipeline network

223

in red and existing gathering pipeline network in brown for our preferred model specification. In

224

evaluating model performance, we focus on the extent to which the placement of pipelines by the

225

model overlaps with the observed pipeline configuration. To quantify the overlap, we calculate

226

the total number of grid cells where the model places a pipeline and then calculate the number of

227

these grid cells that also contain an existing pipeline within a one-cell radius of the model

228

pipeline. We view this as a rough estimate of model fit, as it abstracts from important pipeline

229

properties such as direction (we do not have data on pipeline direction) and influence of the

230

pipeline network outside of the study area on study-area pipeline siting. Note that we do not

231

measure whether the existing pipeline matches the model pipeline. As evident in Figure 2, some

232

existing pipelines do not connect to a well. We use a snapshot of pipeline placement, and

233

therefore do not have information on wells that are expected to be drilled in the future.

234

Furthermore, it is possible that some pipeline placement is to facilitate network robustness. We

235

leave the dynamics of pipeline and well siting for future work. 13 ACS Paragon Plus Environment

Environmental Science & Technology

236

Page 14 of 25

Our preferred model avoids steeply sloped land, specifically any connections with an

237

estimated slope greater than 20 percent, assigns a higher cost to cells that require water

238

crossings, and places a pipeline in the same grid cell as existing pipeline for 94 percent of model-

239

pipeline grid cells. By and large, the model replicates actual pipeline development. However, in

240

some regions, especially in the northwest region and particularly around the edge of the modeled

241

area, the actual and modeled pipeline network differ from each other. This indicates that factors

242

in addition to the shortest path (cost), slope, and water crossings determine the pipeline layout.

243

For example, topography, soils, connections needed to pipelines outside the study area, pipeline

244

and well ownerships, and even habitat concerns may factor into the observed layout. The

245

reasonable possibility that habitat concerns factored into the pipeline layout, albeit not with the

246

use of a model such as ours, reduces the accuracy of our validation approach. But with the high

247

fit statistic for the layout minimizing private costs, and with a variety of factors omitted from our

248

model that affect such costs, it is likely habitat concerns had little influence. See the SI for

249

detailed statistics on slope characteristics of grid cells with pipelines.

250

3.2 Model Results

251

We bound the model results by two cases: the previously described cost minimization

252

solution and a second case where all wells must be connected, but the area of previously

253

undeveloped habitat traversed is minimized. The second bounding case is solved by putting a

254

weight of zero on the pipeline cost term and a weight of one on habitat value in the objective

255

function.

256

In Figures 2a and b we contrast results of the scenario that places a full weight on habitat

257

impact with the cost minimization solution previously described. In Figure 2b, the green color

258

represents forest and wetland land cover. The red line represents the least cost pipeline layout, 14 ACS Paragon Plus Environment

Page 15 of 25

Environmental Science & Technology

259

while the grey lines represents the optimal pipeline layout with a full weight on habitat impact.

260

As is evident from the figure, the grey optimal pipeline layout of this scenario avoids many areas

261

covered by forest and wetland.

262

In addition to exploring the two bounding cases, we develop a trade-off curve depicting

263

the relationship between pipeline cost and habitat impacts as we vary the weights in the model’s

264

objective function. The weights are associated with the decision makers’ preference over the two

265

objectives. To develop the curve, we run our multi-objective optimization model using different

266

sets of weights on pipeline cost and area of forest and wetland lost and report the model output as

267

points in Figure 3. Specifically, we report (x,y) pairs where the x-coordinate is the avoided

268

habitat impact relative to the habitat impact in the cost minimizing solution and the y-coordinate

269

is the cost per acre of habitat impact avoided, calculated as the change in pipeline construction

270

costs (proxied by pipeline distance) relative to the cost-minimization case divided by the acres of

271

habitat impacted.

272

15 ACS Paragon Plus Environment

Environmental Science & Technology

273 274

Page 16 of 25

Figure 3. Per-acre cost of avoided forest and wetland impact

275 276

We interpolate between points to develop a curve representing the marginal pipeline cost

277

of avoiding an additional acre of habitat impact using Matlab’s Piecewise Cubic Hermite

278

Interpolating Polynomial (PCHIP). The slope of the curve is relatively flat for low quantities of

279

avoided habitat impact, but increases sharply with more avoided habitat. This suggests that

280

relatively large amounts of habitat impact can be avoided for relatively low cost, but that

281

avoiding additional units of habitat impact become costlier the more habitat is avoided. For

282

example, 10 percent of the habitat impact can be avoided at less than 2 percent increase in costs

283

over the privately optimal pipeline configuration. Exploring the relationship between costs and

284

habitat impact for additional habitat avoidance – further to the right on the curve – we find that

285

approximately 20 percent of habitat impact can be avoided for about a 11 percent increase in

286

cost. Avoiding additional habitat impacts becomes gradually and increasingly costly.

16 ACS Paragon Plus Environment

Page 17 of 25

287 288

Environmental Science & Technology

4. Discussion Since the beginning of the boom in the mid-2000s, shale gas production has quickly

289

become the main source of natural gas production in the US. Shale accounted for 60 percent of

290

total US gas production in 2017 and resulted in the US becoming a net exporter of natural gas in

291

2017 for the first time in at least sixty years. Furthermore, US shale gas production is expected to

292

nearly double by 2040 33, 34. This development has been accompanied by a large increase in

293

pipeline construction, with over 0.4 million miles of gathering, transmission, and distribution

294

natural gas pipelines added to the US network, totaling 2.2 million miles in 2015 35. Of those

295

added pipelines and associated natural gas infrastructure, gathering pipelines, which bring

296

produced natural gas from the wellhead to midstream infrastructure, are responsible for nearly 80

297

percent of total land use over the entire life cycle of natural gas-fired electricity 15.

298

Our results contribute to growing research showing that state regulations and policies

299

have not led to efficient land use and that energy development could occur with fewer

300

environmental impacts at relatively low cost 36-38. Klaiber, Gopalakrishnan and Hasan 36 used

301

satellite land cover data to find that policies encouraging consolidation on well pads in

302

Pennsylvania would have conserved almost 113,000 acres over a nine-year period. Abrahams,

303

Griffin and Matthews 5 find that requiring pipelines to follow existing roads in forested areas

304

would prevent additional fragmentation at a cost of $0.005 to $0.02 per Mcf of natural gas. Milt,

305

Gagnolet and Armsworth 37, using 20 case study sites, found that conservation-oriented policies

306

could reduce surface impacts from gathering pipelines and access roads without excessively

307

altering current industry practices.

308 309

Our work augments studies assessing alternative policies and/or more comprehensive or systematic planning that have thus far generally focused on small spatial scales. For example, 17 ACS Paragon Plus Environment

Environmental Science & Technology

Page 18 of 25

310

Milt, Gagnolet and Armsworth 38 and Milt and Armsworth 39 assess the costs and benefits of

311

planning to reduce habitat impacts of pipelines but only within leaseholds. Our model allows for

312

quantification of tradeoffs between shale gas pipeline development costs and habitat, a particular

313

contribution to policymaking aimed at coordinating shale gas development across spatial scales.

314

Although a number of studies call for such tradeoffs to be estimated, this paper is the first to do

315

so, particularly on a large a spatial scale 6, 16, 17.

316

The tradeoff curve could be used in conjunction with ecosystem service values to guide

317

policy-making, aiding in the selection of optimal habitat impact avoidance. Specifically, given a

318

value of mitigating forest and wetland impact, the value could be compared with the cost (the y-

319

axis in Figure 3) to determine the optimal and additional expenditure on pipeline construction to

320

avoid disturbing habitat. The cost of avoiding habitat impact ranges between roughly $2,000 and

321

$14,000 ha-1 (Figure 3), which corresponds to $100-$700 ha-1 year-1 when annualized (5 percent

322

discount rate, perpetuity). As long as the economic benefits of avoiding habitat impact exceed its

323

costs, from a societal perspective, investing in habitat protection is economically desirable.

324

Depending on the source, the literature includes estimates of ecosystem service values for forests

325

and wetlands that fall within or go beyond or below the above cost range (see e.g. Petrolia, et al.

326

40,

327

habitat area varies depending on the location and type of habitat (see e.g. de Groot, et al. 42, who

328

screened 300 case studies on the value of ecosystem services, many of them in tropical and

329

temperate forests). Accordingly, further work is needed to locally determine the magnitude of

330

benefits from avoided habitat impacts on average and to account for the heterogeneity of habitat

331

vulnerability and protection value across space.

Jenkins, et al. 41). This is intuitive, as the value of ecosystem services supported by specific

18 ACS Paragon Plus Environment

Page 19 of 25

Environmental Science & Technology

332

Our results also expand the literature related to mitigation of shale gas impacts more

333

broadly. For example, a growing body of work examines the potential gains from mitigating

334

water-related shale development impacts by using more systematic, forward-looking planning

335

and by planning across space. There have been both qualitative and quantitative studies

336

examining environmental and health benefits, as well as private costs from more systematic

337

planning of shale development to minimize water impacts. Rahm and Riha 43 assess several

338

water management policy options and argue in favor of regional impact analysis to ensure the

339

most efficient use of water resources. Bartholomew and Mauter 44 study several well pads to

340

assess the tradeoffs between water management strategy costs (e.g. acquisition, transportation,

341

storage, and treatment) and the benefits of minimizing impacts on human and environmental

342

health. Gao and You 45 developed a model that covers the life cycle of electricity generated from

343

shale gas, consisting of a number stages including freshwater acquisition, shale well drilling,

344

hydraulic fracturing and completion, shale gas production, wastewater management, shale gas

345

processing, electricity generation, transportation, and storage. Shih, et al. 46 developed a multi-

346

objective programming model for shale gas water and wastewater management that incorporates

347

the objectives of four types of decision makers: oil and gas well developers and operators,

348

centralized wastewater treatment facility planners and operators, environmental regulators, and

349

social planners.

350

Our main finding -- that there are gains from coordination of land use changes across

351

larger spatial scales -- arises in many contexts in the systematic conservation literature focused

352

on incorporating biodiversity and cost attributes into the planning process 47-49, and further work

353

could be done to better understand conservation outcomes and integrate with the systematic

354

conservation planning literature. Potential extensions to representation of habitat include

19 ACS Paragon Plus Environment

Environmental Science & Technology

Page 20 of 25

355

consideration of habitat fragmentation (e.g. Saunders, et al. 50), size of reserves relative to

356

species habitat needs (e.g. Marianov, et al. 51), and complementarity of the sites left undisturbed

357

(see e.g. Margules and Pressey 52). In the extreme, connectivity in the form of habitat corridors

358

could be considered using model elements similar to those used in the model now to ensure

359

pipeline connectivity (see e.g. Conrad, et al. 53).

360

Although the model developed in this paper captures essential elements necessary to

361

identify the tradeoff curve, additional work could be done to develop a more realistic

362

representation of both pipeline cost and placement and habitat impacts. Potential modifications

363

to the pipeline model include allowing for diagonal connections across grid squares, inclusion of

364

flow constraints, and extension of the geographic reach. Other extensions include considering

365

ownership of wells and pipelines, allowing for estimation of gains for coordination between

366

firms, and building in constraints and/or cost modifications to allow for pipelines following an

367

existing transportation network. Habitat values may also be heterogeneous across space and

368

therefore the ecosystem service valuation methods previously discussed could be applied to

369

develop arc-specific values. Additionally, there may be uncertainty over one or more inputs

370

and/or the future. Although our model is deterministic, conducting multiple runs varying

371

uncertain inputs could be used to gain insight into sensitivity of the model results to these inputs.

372

Well siting is now a given in the model, and although it would add significant

373

complexity, well siting could be made endogenous to the model. Including well siting could be

374

more realistic in some settings as shale gas drilling methods do have the potential to mitigate

375

surface environmental impacts. New drilling technologies allow more production from a smaller

376

area of land than would conventional development, as well as flexibility in well and pipeline

377

siting decisions. Companies generally drill four to six wells (potentially up to several dozen) and 20 ACS Paragon Plus Environment

Page 21 of 25

Environmental Science & Technology

378

have also increased the lateral length of these wells by thousands of feet, with this efficiency

379

having increased over time throughout the shale development boom 54.

380

In conclusion, incorporation of habitat impacts associated with both well and pipeline

381

siting encourages more efficient land use in shale gas development. MOPS or a similar model

382

could be integrated into the pipeline siting and permitting process, so oil and gas companies,

383

communities, and states can identify cost-effective options for habitat conservation near shale

384

gas development.

385 386 387 388

5. Supporting Information

389 390 391 392 393 394 395 396 397

6. Acknowledgements

Key attributes of the MOPS model, model definition and problem statement, formulation of a solution to the multi-objective constrained optimization, information on study site slopes and water crossings

The authors gratefully acknowledge the support of the Alfred P. Sloan Foundation and the Stephen D. Bechtel Foundation, Jr. Foundation for RFF’s work on a range of issues related to shale gas development, including this research. Marianov acknowledges support from the Complex Engineering Systems Institute through grant CONICYT PIA FB0816 and grant FONDECYT 1160025. The authors thank Paul Armsworth, Michael Griffin, and Jan Mares for helpful insight and feedback and Matthew Ashenfarb, Jessica Blakely, Isabel Echarte, and Justine Huetteman for research assistance. The authors also thank anonymous shale gas industry stakeholders for sharing their insights into pipeline construction cost determinants.

21 ACS Paragon Plus Environment

Environmental Science & Technology

Page 22 of 25

7. References 1. Chen, L.; Miller, S. A.; Ellis, B. R., Comparative Human Toxicity Impact of Electricity Produced from Shale Gas and Coal. Environmental Science & Technology 2017, 51, (21), 13018-13027. 2. Newell, R. G.; Raimi, D., Implications of Shale Gas Development for Climate Change. Environmental Science & Technology 2014, 48, (15), 8360-8368. 3. Small, M. J.; Stern, P. C.; Bomberg, E.; Christopherson, S. M.; Goldstein, B. D.; Israel, A. L.; Jackson, R. B.; Krupnick, A.; Mauter, M. S.; Nash, J.; North, D. W.; Olmstead, S. M.; Prakash, A.; Rabe, B.; Richardson, N.; Tierney, S.; Webler, T.; Wong-Parodi, G.; Zielinska, B., Risks and Risk Governance in Unconventional Shale Gas Development. Environmental Science & Technology 2014, 48, (15), 8289-8297. 4. Langlois, L. A.; Drohan, P. J.; Brittingham, M. C., Linear infrastructure drives habitat conversion and forest fragmentation associated with Marcellus shale gas development in a forested landscape. Journal of Environmental Management 2017, 197, 167-176. 5. Abrahams, L. S.; Griffin, W. M.; Matthews, H. S., Assessment of policies to reduce core forest fragmentation from Marcellus shale development in Pennsylvania. Ecological Indicators 2015, 52, 153-160. 6. Drohan, P. J.; Brittingham, M.; Bishop, J.; Yoder, K., Early trends in landcover change and forest fragmentation due to shale-gas development in Pennsylvania: a potential outcome for the Northcentral Appalachians. Environmental Management 2012, 49, (5), 1061-1075. 7. Slonecker, E.; Milheim, L.; Roig-Silva, C.; Malizia, A.; Marr, D.; Fisher, G., Landscape Consequences of Natural Gas Extraction in Bradford and Washington Counties, Pennsylvania. US Geological Survey Open-File Report 2012, 1154. 8. Slonecker, E. T.; Milheim, L.; Roig-Silva, C.; Malizia, A.; Gillenwater, B. Landscape consequences of natural gas extraction in Fayette and Lycoming Counties, Pennsylvania, 2004– 2010; 2331-1258; US Geological Survey: 2013. 9. McGunegle, M. L., The effects of oil and gas development on forest fragmentation and breeding bird populations in the Allegheny National Forest. Master's Thesis, Pennsylvania State University 2009. 10. Donnelly, S.; Cobbinah Wilson, I.; Oduro Appiah, J., Comparing land change from shale gas infrastructure development in neighboring Utica and Marcellus regions, 2006–2015. Journal of Land Use Science 2017, 12, (5), 338-350. 11. Fahrig, L., Effects of Habitat Fragmentation on Biodiversity. Annual Review of Ecology, Evolution, and Systematics 2003, 34, 487-515. 12. Balvanera, P.; Pfisterer, A. B.; Buchmann, N.; He, J.-S.; Nakashizuka, T.; Raffaelli, D.; Schmid, B., Quantifying the Evidence for Biodiversity Effects on Ecosystem Function and Services. Ecology Letters 2006, 9, 1146-1156. 13. Entrekin, S.; Evans-White, M.; Johnson, B.; Hagenbuch, E., Rapid Expansion of Natural Gas Development Poses a Threat to Surface Waters. Frontiers in Ecology and the Environment 2011, 9, (9), 503-511. 14. Zedler, J. B.; Kercher, S., Wetland Resources: Status, Trends, Ecosystem Services, and Restorability. Annual Review of Environment & Resources 2005, 30, 39-74.

22 ACS Paragon Plus Environment

Page 23 of 25

Environmental Science & Technology

15. Jordaan, S. M.; Heath, G. A.; Macknick, J.; Bush, B. W.; Mohammadi, E.; Ben-Horin, D.; Urrea, V.; Marceau, D., Understanding the life cycle surface land requirements of natural gas-fired electricity. Nature Energy 2017, 2, (10), 804. 16. Northrup, J. M.; Wittemyer, G., Characterising the impacts of emerging energy development on wildlife, with an eye towards mitigation. Ecology Letters 2013, 16, (1), 112-125. 17. Mauter, M. S.; Palmer, V. R.; Tang, Y.; Behrer, A. P., The next frontier in United States shale gas and tight oil extraction: Strategic reduction of environmental impacts. Energy Technology Innovation Policy (ETIP) Research Group: Belfer Center for Science and International Affairs 2013, 4. 18. Middleton, R. S.; Brandt, A. R., Using Infrastructure Optimization to Reduce Greenhouse Gas Emissions from Oil Sands Extraction and Processing. Environmental Science & Technology 2013, 47, (3), 1735-1744. 19. Johnson, N.; Gagnolet, T.; Ralls, R.; Zimmerman, E.; Eichelberger, B.; Tracey, C.; Kreitler, G.; Orndorff, S.; Tomlinson, J.; Bearer, S., Pennsylvania energy impacts assessment report 1: Marcellus Shale natural gas and wind. Harrisburg, PA, US: The Nature ConservancyPennsylvania Chapter 2010. 20. He, L.; Chen, Y.; Li, J., A three-level framework for balancing the tradeoffs among the energy, water, and air-emission implications within the life-cycle shale gas supply chains. Resources, Conservation and Recycling 2018, 133, 206-228. 21. He, L.; Chen, Y.; Zhao, H.; Tian, P.; Xue, Y.; Chen, L., Game-based analysis of energywater nexus for identifying environmental impacts during Shale gas operations under stochastic input. Science of The Total Environment 2018, 627, 1585-1601. 22. Chen, Y.; He, L.; Guan, Y.; Lu, H.; Li, J., Life cycle assessment of greenhouse gas emissions and water-energy optimization for shale gas supply chain planning based on multilevel approach: Case study in Barnett, Marcellus, Fayetteville, and Haynesville shales. Energy Conversion and Management 2017, 134, 382-398. 23. Borraz-Sánchez, C., Optimization methods for pipeline transportation of natural gas. Dep. of Informatics, Univ. of Bergen, Bergen, Norway 2010. 24. Dilaveroglu, S. Optimization for Design and Operation of Natural Gas Transmission Networks. 2012. 25. Brimberg, J.; Hansen, P.; Lin, K.-W.; Mladenović, N.; Breton, M., An oil pipeline design problem. Operations Research 2003, 51, (2), 228-239. 26. Barták, R.; Dovier, A.; Zhou, N.-F. In Multiple-origin-multiple-destination path finding with minimal arc usage: complexity and models, Tools with Artificial Intelligence (ICTAI), 2016 IEEE 28th International Conference on, 2016; IEEE: 2016; pp 91-97. 27. Garey, M. R.; Johnson, D. S., Computers and intractability: a guide to NP-completeness. In WH Freeman and Company, San Francisco: 1979. 28. Marcoulaki, E. C.; Tsoutsias, A. V.; Papazoglou, I. A., Design of optimal pipeline systems using internal corrosion models and GIS tools. Industrial & Engineering Chemistry Research 2014, 53, (29), 11755-11765. 29. More construction, higher costs in store for US pipelines. Oil & Gas Journal 9/4/2000, 2000, pp 1-23. 30. Racicot, A.; Babin-Roussel, V.; Dauphinais, J.-F.; Joly, J.-S.; Noel, P.; Lavoie, C., A Framework to Predict Impacts of Shale Gas Infrastructures on the Forest Fragmentation of an Agroforest Region. Environmental Management 2014, 53, 1023-1033.

23 ACS Paragon Plus Environment

Environmental Science & Technology

Page 24 of 25

31. Lade, G. E.; Rudik, I. Costs of Inefficient Regulation: Evidence from the Bakken; National Bureau of Economic Research: 2017. 32. INGAA Foundation Inc. North America Midstream Infrastructure through 2035: Capitalizing Our Energy Abundance; INGAA Foundation Final Report No. 2014.01; 26 March, 2014. 33. U.S. EIA, Today In Energy: Shale gas production drives world natural gas production growth. U.S. Energy Information Administration 2016. 34. Malik, N. S., U.S. Becomes a Net Gas Exporter for the First Time in 60 Years. Bloomberg 10 January, 2018. 35. Department of Transportation Pipeline Mileage and Facilities. https://www.phmsa.dot.gov/data-and-statistics/pipeline/pipeline-mileage-and-facilities 36. Klaiber, A. H.; Gopalakrishnan, S.; Hasan, S., Missing the forest for the trees: balancing shale exploration and conservation goals through policy. Conservation Letters 2017, 10, (1), 153-159. 37. Milt, A. W.; Gagnolet, T.; Armsworth, P. R., Synergies and tradeoffs among environmental impacts under conservation planning of shale gas surface infrastructure. Environmental Management 2016, 57, (1), 21-30. 38. Milt, A. W.; Gagnolet, T. D.; Armsworth, P. R., The costs of avoiding environmental impacts from shale‐gas surface infrastructure. Conservation Biology 2016, 30, (6), 1151-1158. 39. Milt, A. W.; Armsworth, P. R., Performance of a cap and trade system for managing environmental impacts of shale gas surface infrastructure. Ecological Economics 2017, 131, 399406. 40. Petrolia, D. R.; Interis, M. G.; Hwang, J., America’s Wetland? A National Survey of Willingness to Pay for Restoration of Louisiana’s Coastal Wetlands. Marine Resource Economics 2014, 29, (1), 17-37. 41. Jenkins, W. A.; Murray, B. C.; Kramer, R. A.; Faulkner, S. P., Valuing ecosystem services from wetlands restoration in the Mississippi Alluvial Valley. Ecological Economics 2010, 69, (5), 1051-1061. 42. de Groot, R.; Brander, L.; der Pleog, S.; Costanza, R.; Bernard, F.; Braat, L.; Christie, M.; Crossman, N.; Ghermandi, A.; Hein, L.; Hussain, S.; Kumar, P.; McVittie, A.; Portela, R.; Rodriguez, L. C.; ten Brink, P.; van Beukering, P., Global Estimates of the Value of Ecosystems and their Services in Monetary Units. Ecosystem Services 2012, 1, (1), 50-61. 43. Rahm, B. G.; Riha, S. J., Toward strategic management of shale gas development: Regional, collective impacts on water resources. Environmental Science & Policy 2012, 17, 1223. 44. Bartholomew, T. V.; Mauter, M. S., Multiobjective Optimization Model for Minimizing Cost and Environmental Impact in Shale Gas Water and Wastewater Management. ACS Sustainable Chemistry & Engineering 2016, 4, (7), 3728-3735. 45. Gao, J.; You, F., Shale Gas Supply Chain Design and Operations toward Better Economic and Life Cycle Environmental Performance: MINLP Model and Global Optimization Algorithm. ACS Sustainable Chemistry & Engineering 2015, 3, (7), 1282-1291. 46. Shih, J.-S.; Swiedler, E.; Krupnick, A. J. A Model for Shale Gas Wastewater Management; Resources for the Future: 2016/10/17/, 2016. 47. Naidoo, R.; Balmford, A.; Ferraro, P. J.; Polasky, S.; Ricketts, T. H.; Rouget, M., Integrating economic costs into conservation planning. Trends in Ecology & Evolution 2006, 21, (12), 681-687. 24 ACS Paragon Plus Environment

Page 25 of 25

Environmental Science & Technology

48. Joseph, L. N.; Maloney, R. F.; Possingham, H. P., Optimal allocation of resources among threatened species: a project prioritization protocol. Conservation Biology 2009, 23, (2), 328338. 49. Newbold, S. C.; Siikamäki, J., Conservation prioritization using reserve site selection methods1. Handbook on the economics of natural resources 2015, 358. 50. Saunders, D. A.; Hobbs, R. J.; Margules, C. R., Biological consequences of ecosystem fragmentation: a review. Conservation biology 1991, 5, (1), 18-32. 51. Marianov, V.; ReVelle, C. S.; Snyder, S. A., Selecting Compact Habitat Reserves for Species with Differential Habitat Size Needs. Computers & Operations Research 2008, 35, 475487. 52. Margules, C. R.; Pressey, R. L., Systematic conservation planning. Nature 2000, 405, (6783), 243. 53. Conrad, J. M.; Gomes, C. P.; van Hoeve, W.-J.; Sabharwal, A.; Suter, J. F., Wildlife corridors as a connected subgraph problem. Journal of Environmental Economics and Management 2012, 63, (1), 1-18. 54. U.S. EIA, Trends in US Oil and Natural Gas Upstream Costs. U.S. Energy Information Administration 2016.

25 ACS Paragon Plus Environment