Abby Thesis
-
Upload
abby-mellinger-scott -
Category
Documents
-
view
122 -
download
1
Transcript of Abby Thesis
To The University of Wyoming:
The members of the Committee approve the thesis of Abigail J. Mellinger
presented on July 13, 2012.
Benjamin S. Rashford, Chairperson
L. Steven Smutko, Chairperson
Hall Sawyer, External Department Member
Scott N. Lieske
APPROVED:
Dr. Roger Coupal, Head, Department of Agricultural and Applied Economics
Dr. Frank Galey, Dean, College of Agriculture and Natural Resources
1
Mellinger, Abigail, Economic and ecological tradeoffs of targeting conservation easements for habitat protection: A case study of Sublette County, Wyoming, M.S., Department of Agricultural and Applied Economics, August, 2012.
Extensive energy development in Sublette County, Wyoming has prompted land
management agencies to undertake compensatory (off-site) mitigation projects aimed at off-
setting adversely impacted wildlife species, particularly mule deer (Odocoileus hemionus),
pronghorn antelope (Antilocapra americana), and Greater sage grouse (Centrocercus
urophasianus). Agencies have used conservation easements, or purchases of development rights,
as a tool for protecting wildlife habitat on private agricultural lands. To most effectively mitigate
impacts to wildlife from energy development and from expanding rural residential development,
decision-makers must protect lands that offer the most biological value at the least cost. Given
increasing demand for rural, amenity-rich residential properties in Sublette County, I define the
economic value of agricultural lands as the sum of a given parcel’s productive value in
agriculture and its value in residential development.
I use propensity score matching to estimate the unobservable future residential value of
parcels currently in agricultural use and hence, assess each parcel’s economic value. I impute
the median value of residential parcels to their matched agricultural counterparts to calculate an
economic score. Similarly, I calculate a biological score for each parcel based on the parcel’s
acreage of and proximity to critical wildlife habitat. Combined, the economic and biological
scores form a production possibilities frontier that represents economically efficient
arrangements of parcels in either agricultural or residential use across the landscape of Sublette
County. I identify optimal conservation easement purchases according to four different policy
approaches and compare the current Sublette County landscape to my results.
My results indicate that while the economic efficiency of conservation easement
purchases can be improved, opportunities to protect critical biological values are limited by a
lack of key habitat on private agricultural lands. Further, I find that substantial biological values,
including those on already protected lands, are likely to continue in the absence of conservation
easements given my estimate of observing each parcel in a residential rather than agricultural
use. This suggests that resource managers should carefully target conservation easement
purchases based on parcels’ risk of development in addition to increasing efforts to carry out on-
site mitigation on public lands.
ECONOMIC AND ECOLOGICAL TRADEOFFS OF TARGETING CONSERVATION EASEMENTS FOR HABITAT PROTECTION: A CASE STUDY
OF SUBLETTE COUNTY, WYOMING
by Abigail J. Mellinger
A thesis submitted to the University of Wyoming in partial fulfillment of the requirements
for the degree of
MASTER OF SCIENCE in
AGRICULTURAL AND APPLIED ECONOMICS
Laramie, Wyoming August 2012
All rights reserved
INFORMATION TO ALL USERSThe quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a complete manuscriptand there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
Microform Edition © ProQuest LLC.All rights reserved. This work is protected against
unauthorized copying under Title 17, United States Code
ProQuest LLC.789 East Eisenhower Parkway
P.O. Box 1346Ann Arbor, MI 48106 - 1346
UMI 1520119
Published by ProQuest LLC (2012). Copyright in the Dissertation held by the Author.
UMI Number: 1520119
ii
© 2012, Abigail J. Mellinger
iii
TABLE OF CONTENTS
1 Introduction………..…………………………………………….……………………………………. 1
2 Background…….….……………………………………………………..……………………………. 3
2.1 Oil and Gas Development in Sublette County .......................................................................... 3
2.2 Rural Residential Development and Land Use in Sublette County ........................................... 7
2.3 Impacted Species ...................................................................................................................... 8
Sage grouse and energy development ....................................................................................... 9
Mule deer and energy development ......................................................................................... 12
Pronghorn and energy development ....................................................................................... 13
2.4 The Jonah Interagency Office and Pinedale Anticline Project Office ..................................... 15
2.5 Conservation Easements .......................................................................................................... 17
3 Methods………………………………………………………………..……………………………. 19
3.1 Economic Score ....................................................................................................................... 20
Propensity Score Matching ..................................................................................................... 24
Data ......................................................................................................................................... 27
Calculating the Economic Score ............................................................................................. 32
3.2 Biological Score ...................................................................................................................... 34
3.3 Production Possibilities Frontier ............................................................................................. 39
4 Results……………………………...……………………………………………………………....... 40
4.1 Economic Score ....................................................................................................................... 40
Logit Results ............................................................................................................................ 40
Results of sub-class and caliper matching .............................................................................. 47
4.2 Biological Score Results ......................................................................................................... 50
4.3 Estimated Ecological-Economic Tradeoffs ............................................................................. 54
5 Optimal Targeting of Conservation Easements .................................................................................... 59
Targeting Approaches ............................................................................................................. 62
6 Conclusion………………...……………………………………………………..………………….. 69
7 Literature Cited …………………………………………………………………………………….. 74
iv
LIST OF FIGURES
Figure 1. Sublette County, Wyoming. Source: www.wy.blm.gov/jio-papo. .............................................. 4
Figure 2. Sample production possibilities frontier of land use patterns across a landscape. ..................... 19
Figure 3. Land cover in Sublette County, Wyoming ................................................................................. 31
Figure 4. Mule deer habitat and privately-owned lands in Sublette County, Wyoming. ........................... 36
Figure 5. Pronghorn habitat and privately-owned agricultural lands in Sublette County, Wyoming. ....... 37
Figure 6. Sage grouse habitat and privately-owned agricultural lands in Sublette County, Wyoming. .... 38
Figure 7. Frequency of predicted probabilities for agricultural parcels. .................................................... 44
Figure 8. Map of predicted propensity scores on agricultural parcels using AGRES model – Sublette
County, Wyoming. .............................................................................................................................. 45
Figure 9. Map of predicted propensity scores on agricultural parcels using AGRESVAC model – Sublette
County, Wyoming. .............................................................................................................................. 46
Figure 10. Map of predicted propensity scores on agricultural parcels using AGRESRESVAC model –
Sublette County, Wyoming. ................................................................................................................ 47
Figure 11. Estimated biological scores for mule deer on agricultural parcels in Sublette County,
Wyoming............................................................................................................................................. 51
Figure 12. Estimated biological scores for pronghorn on agricultural parcels in Sublette County,
Wyoming............................................................................................................................................. 52
Figure 13. Estimated biological scores for sage grouse on agricultural parcels in Sublette County,
Wyoming............................................................................................................................................. 53
Figure 14. Estimated total biological scores on agricultural parcels in Sublette County, Wyoming. ........ 54
Figure 15. Production possibilities frontier from the alternative propensity score models using median
residential values. ................................................................................................................................ 55
Figure 16. Production possibilities frontiers by species-specific biological scores estimated using
AGRESVAC model. ............................................................................................................................. 57
Figure 17. Production possibilities frontier using the expected biological score and economic score
estimated using the AGRESVAC model. ............................................................................................. 58
Figure 18. Production possibilities frontiers by species using expected biological scores and economic
score estimated using AGRESVAC model........................................................................................... 59
Figure 19. Comparison of existing conservation easements to the efficient production possibilities
frontier - AGRESVAC model. .............................................................................................................. 60
Figure 20. Comparison of existing conservation easements to the efficient production possibilities
frontier disaggregated by species - AGRESVAC model. ..................................................................... 62
v
Figure 21. Total biological scores produced from alternative targeting strategies with a total budget of
$36 million. ......................................................................................................................................... 64
Figure 22. Landscape-level expected biological scores produced from alternative targeting strategies with
a total budget of $36 million. .............................................................................................................. 65
Figure 23. Relationship between parcel-level biological scores and propensity scores. ........................... 66
Figure 24. Relationship between the biological and residential values of parcels. .................................... 67
Figure 25. Ten "best" currently unconserved parcels for conservation according to each targeting
approach. Existing conservation easements designated by cross-hatch pattern. ................................ 69
vi
LIST OF TABLES
Table 1. Expected initial and life-of-project well field components' surface disturbance. .......................... 5
Table 2. Explanatory variables used in binary logit models. ..................................................................... 29
Table 3. Summary statistics of agricultural lands' assessed values in Sublette County. ............................ 33
Table 4. Distribution of habitat across private and public lands in Sublette County. ................................ 38
Table 5. Parameter estimates for AGRES, AGRESVAC, and AGRESRESVAC logit models. ............... 40
Table 6. Average marginal effects. ............................................................................................................ 42
Table 7. Average predicted probabilities by observed land use and model. .............................................. 43
Table 8. Average per acre value of matched residential parcels using each matching approach. ............. 48
Table 9. Predicted residential values calculated for AGRESVAC model using median, minimum, and
average values per acre of matched residential parcels. All values are in dollars per acre. ............... 50
Table 10. Summary statistics of biological scores calculated for agricultural parcels in Sublette County,
Wyoming............................................................................................................................................. 50
Table 11. Difference in biological scores for each species between current easement point and efficient
frontier................................................................................................................................................. 61
1
1 Introduction
Oil and gas development in Sublette County, Wyoming has proceeded at levels relatively
unprecedented since 2002, when developers began exploration and drilling in the Jonah Infill
Drilling Project Area (JIDPA) and the Pinedale Anticline Project Area (PAPA). The
development of the fields has had substantial impacts. Combined, the two fields have
contributed substantially to the Wyoming state and federal governments in severance and other
tax revenues. Despite staggered, geographically strategic development and the implementation
of seasonal drilling restrictions and other best management practices, development has had
environmental consequences. For example, the mule deer population that relies on winter range
in the PAPA declined by nearly 60 percent during the first 10 years of development (Sawyer and
Nielson 2011).
The Bureau of Land Management’s (BLM) Record of Decision (ROD) for the JIDPA
(BLM 2006) and the PAPA (BLM 2000, 2008) mandated the creation of a multi-agency office to
oversee and coordinate mitigation and monitoring efforts for each field, the Jonah Interagency
Office (JIO) and Pinedale Anticline Project Office (PAPO), respectively. Each office is staffed
by representatives of the BLM, U.S. Forest Service (USFS), Wyoming Game and Fish
Department (WGFD), and Wyoming Department of Environmental Quality (DEQ). Each office
uses limited operator-committed funds to carry out its monitoring and mitigation mandates. Both
the JIO and PAPO have undertaken a variety of mitigation and monitoring projects, but based on
spending practices, have favored conservation easements as a tool to protect wildlife habitat, air
quality, and scenic values. Of the $14 million earmarked by the JIO for wildlife mitigation
projects, nearly $8 million has already been committed to mitigation projects that include a
2
conservation easement component. Given limited funding and the substantial need for
mitigation, this thesis research explores the cost minimization problem faced by the JIO and
PAPO: how should agencies best allocate mitigation funds to achieve the most biological benefit
at least cost?
This research examines this cost minimization problem while focusing on three species,
pronghorn, mule deer and sage grouse. Because the limiting habitat (e.g., winter range,
migration routes, lek sites) of these species is directly impacted by oil and gas development,
these species are arguably most in need of off-site (compensatory) mitigation efforts. Given the
habitat requirements of these key species, including functional migration corridors, stopover
sites, and shrub communities in the form of Wyoming big sagebrush, the JIO and PAPO should
seek to place conservation easements on those properties that provide such biological
requirements and mimic habitat lost to development on the JIDPA and PAPA.
The following chapters describe how this research examines conservation easement
purchases using a spatially-explicit propensity score model to estimate optimal placement of
easements, given parcel-specific land characteristics. Economic efficiency is defined as the
optimal placement of conservation easements on those parcels that maximize benefit to mule
deer, pronghorn, and sage grouse for given levels of economic returns to land in non-
conservation uses. This analysis will highlight existing purchases that are economically
efficient, and inform future purchases.
3
2 Background
2.1 Oil and Gas Development in Sublette County
Sublette County, Wyoming, situated in southwest Wyoming’s Upper Green River Basin,
holds vast reserves of natural gas. The geologic composition of the region is unique; below the
stretches of sagebrush rangeland south of Pinedale, Wyoming are trillions of cubic feet of natural
gas trapped within tight-gas fluvial reservoirs (Stilwell and Crockett 2006). Advances in
technology in the mid 1990’s made the extraction of gas held within such formations technically
and economically feasible (Pinedale Anticline Working Group 2005).
The natural gas resources in Sublette County are part of an energy sector growing in
national importance. The dry, tight-sands gas found in Sublette County’s formations are
considered an unconventional resource play; unconventional gas has increasingly contributed to
the United States’ domestic production, offsetting declines in conventional gas production
(Stilwell and Crockett 2006). Unconventional gas increased from 32 percent of total domestic
gas production in 2002 to 40 percent by 2004 (Stilwell and Crockett 2006), and was
approximately 30 percent in 2011 (Secretary of Energy Advisory Board 2011). The Energy
Information Administration anticipates that continued advances in technology and continued
exploration of shale resources will result in increases in natural gas’ share of electricity
generation and increases in other types of natural gas consumption in coming decades (U.S.
Energy Information Administration 2012).
Sublette County is largely within the jurisdiction of the Pinedale BLM Field Office. The
U.S. Geological Survey (USGS) has classified 80% of the Field Office’s jurisdictional area as
having “high occurrence potential” – indicating a high likelihood of oil and gas resource
potential (Stilwell and Crockett 2006). The USGS characterizes two of the biggest plays in the
4
region, the Jonah Field (JIDPA) and the Pinedale Anticline PAPA as having high development
potential (Stilwell and Crockett 2006). The Jonah Field and the Pinedale Anticline are among
Wyoming’s largest natural gas fields (Figure 1), contributing to the state’s status as having the
second largest proven dry natural gas reserves in the nation. Wyoming accounts for 12 percent
of the nation’s dry natural gas proven reserves (Stilwell and Crockett 2006).
Figure 1. Sublette County, Wyoming. Source: www.wy.blm.gov/jio-papo.
5
In 2000, the BLM approved exploration and development on the Pinedale Anticline1
(BLM 2000) and in 2008, the BLM completed a supplemental environmental impact statement
for the PAPA (BLM 2008), which detailed a comprehensive plan for the field’s development.
The PAPA ROD approved a maximum of 4,339 wells and 600 well pads, stipulated the
implementation of a liquids gathering system, and adopted a plan for phased development to
minimize impacts to wildlife, water and air quality, and to minimize surface disturbance (BLM
2008). The PAPA ROD details the expected amount of surface disturbance from the
development of the field (Table 1). Life-of-project estimations included in the PAPA ROD
assume a reclamation rate of 60 percent. The field is estimated to contain 25 trillion cubic feet of
recoverable natural gas (Stilwell and Crockett 2006).
Table 1. Expected initial and life-of-project well field components' surface disturbance.
Infrastructure Initial Disturbance
Life-of-Project Disturbance a
Well pads (600) 8,113 acres 3,245 acres
Gas (100 miles) and liquids (471 miles) gathering pipelines
- - 3,157 acres
Local and resource roads (100 miles) 606 acres 484.8 acres
Pipelines, ancillary facilities, compressor sites, stabilizer sites, and other gathering facilities
1008 acres 282 acres
Total well field components 12,885 acres 4,012 acres
a Surface disturbance information shown in Table 1 is adapted from Bureau of Land Management (2008; pp. 36).
In 2002, EnCana Oil & Gas, BP America, and other operators submitted a proposal to the
BLM that would substantially increase drilling in the Jonah Field, located just south of the
1 See the Record of Decision for the Pinedale Anticline Oil and Gas Exploration and Development Project (2000).
6
PAPA. In 2003, the BLM began the public scoping process required under the National
Environmental Policy Act to gather public comment, research, and general involvement. Over
the course of the succeeding three years, the BLM issued draft and final environmental impact
statements, with the BLM’s ROD signed in March of 2006.
The JIDPA covers a surface area of 30,500 acres, of which the ROD stipulates only 46
percent can be disturbed at one time. The 14,030 acres of disturbance allowed under the ROD
can be supplemented with up to 6,304 reclaimed acres, for a cumulative maximum surface
disturbance of 20,334 acres (BLM 2006). The recommended alternative adopted in the ROD
includes the surface disturbance stipulations outlined above, administrative requirements,
operator-committed practices, adaptive management guidelines, and a commitment to limiting
environmental impacts. The Jonah Field is expected to produce more than 8 trillion cubic feet of
natural gas over the next 76 years, enough to heat 4.8 million homes for 20 years (JIO 2009).
Given the significance of the energy resource found within the Jonah Field and the
Pinedale Anticline, the BLM’s decisions are an effort to fulfill its multiple use mandate by
developing energy resources while protecting other valued resources.2 Significant benefits will
accrue to the state of Wyoming from the development of the fields. The Jonah Field alone is
expected to generate up to $6.1 billion in taxes and federal royalties, half of which will go to the
state of Wyoming (JIO 2012). The Pinedale Anticline will also generate significant revenue in
royalties and taxes: it is estimated to generate $232 million in average total federal mineral
royalties alone by 2065 (BLM 2008). Wyoming allocates 25 percent of all severance taxes to the
Permanent Wyoming Mineral Trust Fund (PWMTF), which acts as a savings account for the
state, earning interest and acting as a loan source for other state programs. Natural gas
2 The BLM decision was based on its obligation under the Federal Land Policy and Management Act (FLPMA), the National Energy Policy Act of 2005, the Mineral Leasing Act, and the President Bush National Energy Policy.
7
production also accounts for a larger portion of Wyoming’s gross state product than any other
industry (Bureau of Economic Analysis 2012).
2.2 Rural Residential Development and Land Use in Sublette County
Population growth and land use issues in the Rocky Mountain region have increased in
recent years. Colorado, Utah, Idaho, and Montana have seen increases in population up to three
times the national rate (Taylor and Lieske 2002). Wyoming’s population growth between 2000
and 2010 was 14 percent, 45 percent higher than the national average (Census Bureau 2010), and
some counties within the state have seen even higher increasing rates of growth since the 2000
nationwide census. Sublette County experienced a 22.2 percent increase in population from
1990 to 2000 and another 12.4 percent population growth between 2000 and 2004, much of
which is attributable to the county’s oil and gas development (Pinedale Anticline Working Group
2005). Since 2004, Sublette County’s population continued to boom: the county’s growth rate
between 2000 and 2010 was greater than 73 percent – the highest within the state (Census
Bureau 2010).
While much of the population growth in Sublette County is attributable to an influx of
gas field workers, growth is also due to in-migrants seeking rural and outdoor amenities.
“Second home” growth – seasonal or temporary residences – makes up nearly 20 percent of
Wyoming’s housing units (Census Bureau 2010), increasing from only 5.5 percent in the 2000
census (Taylor and Lieske 2002a). Wyoming falls behind only Montana, Arizona, and Idaho for
second-home growth since 2000 (Census Bureau 2010). Sublette County has the highest
percentage of second homes in the state, with 25 percent of residences considered second homes;
8
in comparison, Teton County, which is widely known for its large number of second homes, only
contained 22 percent second homes (Census Bureau 2010).
Because many second home owners seek outdoor amenities, much of the growth in such
residences occurs outside of municipal boundaries. Sublette County is one of seven counties in
Wyoming where rural growth exceeds urban growth (Census Bureau 2010). Between 2000 and
2010 Sublette County’s rural population nearly doubled (Census Bureau 2010). Such rural
residential development is driven by demand for amenities like recreational access, scenery,
wildlife, and open space, and is the catalyst for conversion of agricultural lands to residential
uses.
2.3 Impacted Species
The oil and gas reservoirs below the JIDPA and PAPA are thin (i.e., may not be in
communication with surrounding reservoirs) and vary in depth, which necessitates dense well
spacing. Drilling densities in the area may be as high as one well per five acres, or as low as one
well per 40 acres (Stilwell and Crockett 2006). The BLM Reservoir Management Group
predicts well spacing in the PAPA of 10 to 20 acres per well and well spacing in the Jonah Field
of five acres per well in productive areas (Stilwell and Crockett 2006). The density of drilling
and associated surface disturbance combined with traffic and human presence have impacted
wildlife in Sublette County (Lyon and Anderson 2003, Ingelfinger and Anderson 2004, Holloran
2005, Holloran et al. 2010, Sawyer et al. 2006, 2009a,b, Gilbert and Chalfoun 2011, Beckman et
al. 2012).
The predominant land cover on the Pinedale Anticline and Jonah Field is Wyoming big
sagebrush grasslands, which have been significantly altered by gas development (Walston et al.
2009). Sagebrush is a key habitat for mule deer (Odocoileus hemionus), pronghorn antelope
9
(Antilocapra americana), and the Greater sage grouse (Centrocercus urophasianus) - hereafter
referred to as “sage grouse”. Among the many sagebrush obligate species affected by oil and gas
development in Sublette County, these three have garnered special attention on both a state and
national scale.
Sage grouse and energy development
Sage grouse populations in Wyoming and in the Rocky Mountain region have declined in
the last century despite management and research efforts that began as early as the 1930’s
(Connelly et al. 2004). Sage grouse range historically covered 296 million acres in the West
(Johnson and Holloran 2010), including Oregon, Wyoming, Montana, Utah, Idaho, Colorado,
Nevada, and Washington (Macsalka 2011). The bird currently occupies only 56 percent of its
historic range (Johnson and Holloran 2010), and its overall decline in population varies from at
least 17 percent to 47 percent throughout its current range (Connelly et al. 2004). Researchers
have identified habitat reduction and fragmentation from rural sprawl, energy development,
agricultural practices, invasive species, and fire as primary causes of the bird’s decline (Johnson
and Holloran 2010).
Researchers believe rural and suburban sprawl is one of the leading causes for the sage
grouse’s regional population decline, as over 60 percent of counties in the Rocky Mountain West
are experiencing some degree of sprawl (Johnson and Holloran 2010). Infrastructure associated
with houses, roads, and utility lines can displace habitat directly or otherwise disturb sage
grouse, effectively limiting or fragmenting habitat (Johnson and Holloran 2010, Walker et al.
2007).
Agriculture has impacted sage grouse habitat and population levels in multiple ways.
Sage grouse have demonstrated a preference for using agricultural lands as habitat for brood-
10
rearing, but have also shown sensitivity to pesticide use on agricultural lands (Connelly et al.
2004, USFWS 2008). Livestock grazing can reduce grass heights and shrub cover, leaving
nesting areas exposed to increased predation (Johnson and Holloran 2010). Livestock grazing
can also reduce forb availability, creating direct competition for food (Johnson and Holloran
2010). Agricultural practices – for crop production or livestock grazing – can directly eliminate
sagebrush habitat through mechanical treatment, herbicide use, or controlled burning, impacting
sage grouse populations (USFWS 2008).
Fires set by humans – to clear sagebrush for agriculture or set by accident – are
detrimental to sagebrush habitat. Fire has reduced some sage grouse populations by more than
80 percent (Connelly et al. 2004). After burning, sagebrush communities can take 100 to 200
years to re-establish (Johnson and Holloran 2010). In the meantime, burned areas are susceptible
to invasion from other species, such as cheatgrass (Rowland 2006). Other surface disturbances
can similarly make sagebrush communities vulnerable to invasive species.
Energy development has the potential to create these or similar impacts to sage grouse
and their habitat. The access roads, utility lines and pipelines, fences, well-pads, holding ponds,
seismic surveys, and noise and human activities associated with energy development have all
been shown to impact sage grouse populations (Johnson and Holloran 2005, Walker et al. 2007).
Roads, fences, and pipeline corridors fragment habitat, directly reducing its availability. Well-
pad construction results in the removal of habitat, and collisions with utility lines and vehicles
result in direct sage grouse mortality. The presence of roads has had a demonstrated impact on
sage grouse populations; one study found that male lek attendance declined within less than 2
miles of an access road when the traffic volume on the road exceeded one vehicle per day
(Johnson and Holloran 2010).
11
Energy development indirectly reduces sage grouse habitat by creating noise, changing
water and habitat quality, and creating opportunities for predation (Walker et al. 2007). Utility
lines and other energy-related infrastructure give raptors a predatory advantage; they are better
able to prey upon sage grouse with the perching opportunities infrastructure provides. Energy
development indirectly affects sage grouse populations by helping the spread of West Nile virus;
mosquitoes that transmit the virus are more abundant near produced-water holding ponds
(Walker et al. 2007).
Holloran (2005) and Naugle et al. (2009) found that impacts to sage grouse from energy
development activities can be detected at distances of 3-5 kilometers and 3-4 miles (5-6
kilometers), respectively. Walker et al. (2007) cite four individual studies that found declines in
sage grouse populations following the introduction of energy development. In the Sublette
County region, specifically, Holloran (2005) found that numbers of displaying males and
recruitment of juvenile males decreased in proximity to gas fields, and that nesting females and
brooding females avoided producing wells. These findings are consistent with Wyoming Game
and Fish Department (WGFD) reports. WGFD monitoring reports indicate decreases in both
active lek counts in Sublette County and decreases in male lek attendance (WGFD 2011).
Given the widespread declines in population, the sage grouse was considered as a
candidate for listing under the Endangered Species Act (ESA) in 2010. The Greater sage grouse
ESA listing decision of 2010 was “warranted but precluded,” meaning that currently the bird is
not receiving protection under the ESA, but could still be listed at a later time. Wyoming state
government and agencies, businesses, and residents have taken steps towards protecting the bird
and its leks to prevent the broad and potentially negative impacts of listing.
12
Mule deer and energy development
Development of the PAPA has negatively impacted Sublette County mule deer
populations, particularly the Mesa herd, through direct and indirect habitat loss. Direct habitat
loss occurs when native habitat is converted to infrastructure (e.g.., well pads, roads, pipelines),
whereas indirect habitat loss occurs when animals avoid infrastructure. Indirect habitat loss has
been shown to be much larger than direct habitat loss, and is concerning because it effectively
reduces the size of the winter range (Sawyer et al. 2006). Although indirect habitat loss can be
reduced by minimizing traffic levels and installing underground liquids gathering systems
(Sawyer et al. 2009b), it cannot be eliminated. Through the first 10 years of development,
monitoring efforts in the PAPA indicate mule deer have declined 56 percent, from 5,228 animals
in 2001 to 2,318 in 2010 (Sawyer and Nielson 2011). This level of decline, when compared with
other herd units near the PAPA, is sufficient to prompt a mitigation response from the BLM,
according to its Wildlife Monitoring and Mitigation Matrix (WMMM): “…changes requiring
mitigation are as follows: 15% population decline in any year, or cumulatively overall years,
compared to the Sublette mule deer herd unit or other mutually agreeable area” (BLM 2012).
Further development outside of the PAPA has the potential to interfere with seasonal
migration, which could further decrease herd numbers. Between 2,500 and 3,500 mule deer
migrate from winter ranges in the PAPA to summer ranges across western Wyoming (Sawyer et
al. 2005). Despite loss of crucial winter range from development of the PAPA, the migration
corridor has remained functional (Sawyer et al. 2009a). Mule deer and other ungulates survive
by using a migration strategy that allows them to maximize their access to forage of the highest
nutritional value. While capable of completing the spring and autumn migration in as little as
one day, mule deer consistently spend up to three weeks completing their migrations, in order to
13
take advantage of phenological forage gradients along the migration corridor (Sawyer and
Kauffman 2011). Foraging along the migration route is believed to be critical because it allows
animals to recover body condition earlier in the spring and maintain it later into the autumn
(Sawyer et al. 2005).
Migration corridors serve two functions for migrating ungulates: they serve as both a
movement corridor and a series of stopover sites (Sawyer et al. 2009c). Analysis of the
movements of GPS-collared deer show that 95 percent of migrating animals’ time is spent
foraging in stopover sites, and only five percent of migrating animals’ time is spent in movement
corridors (Sawyer and Kauffman 2011). Mule deer show high fidelity to migrations routes,
seasonal ranges, and stopover sites; as young deer learn migration behavior from their mothers
(Sawyer and Kauffman 2011). Strong fidelity to learned migration routes and seasonal ranges is
an important consideration for resource managers because it emphasizes that habitat
improvement or protection efforts must overlap with existing routes and ranges in order to be
effective. Like crucial winter range, direct and indirect impacts from human activities can lessen
the functionality of stopover sites or can make movement corridors impassable to migrating
animals (Sawyer et al. 2009c).
Pronghorn and energy development
Like mule deer, pronghorn antelope complete annual migrations to maximize their access
to seasonal range and nutritional forage (Berger 2004, Sawyer et al. 2005, Berger et al. 2006).
Each fall and spring, between 1,500 and 2,000 pronghorn antelope migrate over 300 miles
roundtrip between summer range in Grand Teton National Park and crucial winter range in the
PAPA and JIDPA – the longest-known terrestrial migration in the 48 contiguous states (Berger
2004, Sawyer et al. 2005). Expanding residential and energy development threaten pronghorn
14
migration corridors in Sublette County. Pronghorn contend with roads, impassable fences, and
human presence and disturbances throughout the migration corridor. Fragmentation of the
migration corridor could have adverse impacts on the area’s pronghorn population (Berger
2004).
Though mule deer and pronghorn share migration behavior, migration corridors, and
winter range, they appear to differ in their response to disturbance association with energy
development (see Beckman et al. 2012). In contrast to mule deer, monitoring efforts in the
PAPA indicate that pronghorn do not avoid well pads or roads (Nielson and Sawyer 2010)
However, roads and highways are particularly troublesome for migrating pronghorn. Roads and
highways pose a risk to migrating pronghorn in two ways: directly, through vehicle-animal
collisions, and indirectly, through the migration corridor fragmentation that right-of-way fences
create (Sawyer and Rudd 2005). Data from GPS-collared pronghorn indicates a bottleneck in the
migration corridor, commonly known as “Trapper’s Point,” which is located just north of
Pinedale, Wyoming (Sawyer et al. 2005, Berger et al. 2006). Trapper’s Point is a natural
bottleneck – a corridor only 1.6 kilometers wide between two plateaus – that is bisected by US
Highway 191. Residential and commercial development near Trapper’s Point has narrowed the
navigable width of the bottleneck from its natural width of 1.6 kilometers to less than 0.8
kilometers (Sawyer et al. 2005). As many as 800 animals have been observed passing through
the area (Riis 2009). Given such high numbers of migrating animals, and increased traffic
related to energy and residential development, Trapper’s Point is one of the leading areas in
Wyoming for frequency of animal-vehicle collisions (WyDOT 2007). In an effort to mitigate
this problem, WYDOT is constructing two over-passes designed specifically for pronghorn.
15
Fences present another impediment to pronghorn migration. Though physically capable
of jumping over fences, pronghorn nearly always pass underneath them, requiring at least 16
inches of space between the bottom wire of the fence and the ground (Sawyer et al. 2005).
Fences lining highway and utility right-of-ways and fences associated with rural residential
development fragment pronghorn migration corridors, and will further fragment migration
corridors as rural development expands.
Monitoring efforts have shown that pronghorn, though facing substantial challenges
along their migration corridor, have continued to use winter range on the PAPA (Nielson and
Sawyer 2010). Nielson and Sawyer (2010) found that during the winter of 2009-2010,
pronghorn were not negatively affected by the energy development in the PAPA (see Beckman
et al. 2012).
2.4 The Jonah Interagency Office and Pinedale Anticline Project Office
Recognizing the potential for negative impacts from natural gas extraction, the JIDPA
ROD established the Jonah Interagency Office (JIO), an amalgamation of agency representatives
from the Wyoming Department of Agriculture, WGFD, Wyoming Department of Environmental
Quality, and the BLM. The JIO’s charge is “overall management of field monitoring and
mitigation activities, both on- and off-site,” (BLM 2006). The office also manages the
approximately $21 million and $3 million that EnCana Oil & Gas and BP America, respectively,
have put towards compensatory (off-site) mitigation of impacts from the Jonah Infill Drilling
Project’s development. Of the total $24.5 million, $16.5 million is specifically earmarked for
compensatory mitigation projects for wildlife.
As of December 2010, $12,888,718 of the $16.5 million earmarked for wildlife
mitigation projects was already committed to conservation easements on private lands in the area
16
and similar measures on BLM grazing allotments, leaving $3,111,282 for other projects. Most of
these conservation easements have been placed on privately-owned agricultural lands that
contain valuable habitat. To accomplish off-site, or compensatory, mitigation, the JIO and
PAPO have purchased the development rights on private agricultural lands that could provide
habitat that would off-set direct and indirect habitat losses from energy development.
Agricultural operations continue under the conservation easements.
Aside from directing funds towards conservation easements, the JIO has approved
projects to minimize human-wildlife conflicts and improve habitat outside of the JIDPA.
Examples of past projects include:
• construction of snow fences to add moisture to lands under reclamation;
• purchase and installation of Dynamic Message Sign boards placed along Sublette County highways to minimize animal/vehicle collisions;
• prescribed burning of upland plant communities to enhance a mosaic vegetation pattern for the benefit of several species, including sage grouse;
• drilling and improvement of water wells, installation of diversions and improvement of existing springs to provide water sources for migrating pronghorn and mule deer;
• construction and placement of nesting platforms for ferruginous hawks, installation of wildlife escape ramps in all BLM range improvement water tanks, and, contribution to the Green River Valley Land Trust’s (GRVLT) Wildlife Friendly Fencing Initiative.
Like the JIDPA ROD, the PAPA ROD established an agency office to oversee and
administer funds for mitigation and monitoring of the effects of developing the Pinedale
Anticline. The PAPO is charged with the “overall management of on-site monitoring and off-
site mitigation activities,” (PAPO 2012). Like the Jonah Interagency Office, the PAPO is staffed
by representatives of several state and federal agencies. Members of the Wyoming Department
of Agriculture, WGFD, Wyoming Department of Environmental Quality, BLM, and Sublette
17
County Commissioners staff the Pinedale Anticline Mitigation Management Board (PAMMB),
an entity that oversees the PAPO’s mitigation and monitoring efforts (BLM 2012).
The PAPO focuses its efforts on the key species discussed previously: mule deer,
pronghorn, and sage grouse. To meet its obligations to monitor impacts from the development
and mitigate those impacts as necessary, the PAPO obtains, collects, stores, and distributes data
intended to inform adaptive management in the PAPA. The PAPO operates on its Monitoring
and Mitigation Fund; the fund is generated through $7,500 contributions from Ultra, Shell, and
Questar for each new well drilled in the PAPA. The fund is currently $16,507,500, and is
expected to reach $36 million during the life of the Pinedale Anticline project (PAPO 2012).
PAPO administrators distribute funds for projects that meet the office’s goals on both federal and
non-federal lands.
Past PAPO-funded projects include annual data collection on mule deer, pronghorn, sage
grouse, pygmy rabbit, raptor, and white-tailed prairie dog populations, and snow and traffic
monitoring. PAPO-funded on-site mitigation projects have included sagebrush fertilization and
mule deer winter habitat improvement on the Anticline. The PAPO has also collaborated with
the JIO to fund conservation easement projects in its off-site mitigation efforts. Therefore, I do
not consider the mitigation funds of the two offices – the JIO and the PAPO – separately in my
analysis; because the JIO and PAPO have similar mitigation goals and similar mitigation needs, I
treat both funds as a single entity in this model.
2.5 Conservation Easements
Conservation easements have become a popular tool for conservation of lands valuable
for wildlife habitat, scenic or recreational amenities, or protection of open space. Conservation
18
easements are a legal construct enabling private landowners to forego their right to develop their
property in exchange for direct payments, tax benefits, and for estate planning purposes. In
essence, a conservation easement can be considered a purchase of development rights (PDR)3.
Conservation easements effectively prevent development of the parcel for residential use
or mineral extraction. Generally, before a conservation easement can be placed on a parcel, a
licensed geologist is required to analyze the potential for oil and gas or other extractive
development on the parcel, should the parcel have a split estate (privately-owned surface and
federally-owned mineral rights) and have the potential for development of federally-owned
minerals in the future. Federally-owned mineral rights in some cases would not be prevented
from development under a conservation easement (Perrigo and Iversen 2002). In the case of split
estate on a property under conservation easement, “the…mineral estate owner generally has the
right to reasonable use of the surface estate to access and extract the minerals” (Benson 2005).
3 The terms ‘conservation easement’ and ‘purchase of development rights’ are often used interchangeably. When land is donated as a conservation easement in exchange for tax benefits and for estate planning purposes, it is considered a ‘conservation easement.’ When land is conserved under an easement in exchange for money or is traded in a market, the term ‘purchase of development rights’ is appropriate (Perrigo and Iversen 2002).
19
3 Methods
To assess the efficiency of conservation easement purchases, I characterize the trade-off
between biological (mule deer, pronghorn, and sage-grouse) and economic value across Sublette
County. I construct a production possibilities frontier (PPF) to represent the trade-off between
economic development and critical wildlife habitat (Polaskey et al. 2008, Polasky et al. 2005,
Lichtenstein and Montgomery 2003). I construct a biological “score” to represent the expected
number of species that can be sustained on the landscape, and an economic “score” which sums
the net present value (NPV) of economic returns for each land parcel under different land uses
(Polasky et al. 2008). To generate an efficient hypothetical pattern of land use, I maximize the
economic score for a given biological score, or vice versa (Figure 2).
Figure 2. Sample production possibilities frontier of land use patterns across a landscape.
Similarly, I generate an economic score for each parcel in Sublette County – its value in
either agricultural or rural residential use – and a biological score using data on mule deer,
pronghorn, and sage-grouse population distributions, and evaluate which lands should be put
under a conservation easement to maximize the biological score at least cost. Given the intent
and contractual stipulations of conservation easements, I consider conservation and agriculture to
Biological Score
Economic Score
PPF: Efficient patterns of land use
Inefficient patterns of land use
20
be the same land use (as opposed to the methods used in Polasky et al. 2008). Because I am
investigating the most efficient placement of conservation easements on private land, where
there is relatively little oil and gas, commercial, or industrial development, I limit parcels to
either agricultural or rural (outside of municipal boundaries) residential use in my model.
3.1 Economic Score
Theoretically, the purchase of a conservation easement, because it prevents development,
compensates the landowner for the profits they could have captured by developing the land – i.e.,
the opportunity cost of future development. Given this concept, I must estimate the value of
future development on parcels currently in agriculture to identify which agricultural parcels have
the least value in future residential development, and therefore the least cost of easement
purchase for a given biological score.
From both a legal and economic perspective, the land value of a parcel can be based on
two components, including: 1) productive use and 2) potential use (Plantinga et al. 2002). I
consider these components for any given agricultural parcel’s value to be the productive
agricultural value and the potential for residential development. Thus, the value of a given
parcel is a function of the net returns to agriculture and the potential returns to future residential
development. To maximize profits, the landowners should convert land from agriculture to
residential development at the optimal time, t*, given market conditions. Assuming conversion
to residential use is irreversible, the current value of a parcel of agricultural land i can be
expressed as:
(1) *
*, ,
0 *
( , ) ( , )t
rt rt rti i ag i i res i
t t
P t z e dt t z e dt Ceπ π∞
− − −
=
= + −∫ ∫ ,
where,
21
πi,ag(t,zi) is the net return to agriculture on parcel i with characteristics z;
πi,res(t,zi) is the net returns to residential use on parcel i with characteristics z; and
C is the one-time cost of converting parcel i from agricultural to residential use.
The first term in (1) is the present value of returns from using the parcel for agricultural
production, from the current time to the optimal conversion time, t*. The second term is the
present value of future returns from residential development less conversion costs, from t* to
infinity.
Equation (1) demonstrates the opportunity cost of future development for which the purchase of
a conservation easement should compensate a private landowner. The value of the conservation
easement, or cost of the purchase of development rights, should be equal to the net present value
(NPV) of foregone residential rents less conversion costs (Plantinga and Miller 2001).
Given this definition of the cost of a conservation easement, I must separate the
residential value from the agricultural value of every agricultural parcel in the county. Hedonic
modeling could provide parameter estimates to show how much of the observed price represents
future residential development (Rosen 1974, Bastian et al. 2002). The standard hedonic
approach, however, requires accurate price observations that capture the true economic value of
each agricultural parcel (i.e., recent sales data) and a set of measureable characteristics for each
parcel to regress over the sales data. Parameter estimates from a hedonic model could then be
t = 0 t*
“NPV of Agricultural Rents”
“NPV of Residential Rents”
22
used to derive the proportion of residential development value in current land price for any
parcel (Plantinga et al. 2001, Bastian et al. 2002).
Because agricultural lands in Sublette County change ownership infrequently, adequate
transactions data to estimate a hedonic model do not exist. In the absence of actual transactions
data, assessed land values are the most readily available proxy for market values. Assessed
values are estimated by city or county assessors to determine property taxes. In ex-urban or rural
areas where conversion to development is largely driven by amenity values, Ma and Swinton
(2012) found that assessed values underestimate or omit amenity values in determining future
residential development values and, thus underestimate current market value. Spahr and
Sunderman (1998) found that assessed values differed from market value as well; they concluded
that agricultural properties with amenity values are undervalued, as assessors underestimate the
economic value of non-agricultural attributes. Moreover, agricultural lands in Wyoming are
assessed strictly on the basis of their productivity for agricultural purposes. Assessors combine
commodity price data and capitalized net income figures to determine agricultural land value
(Wyoming Department of Revenue 2012). This assessment approach explicitly ignores the value
of future development. Thus, hedonic models applied to assessed agricultural land values will be
biased because parameter estimates are based on assessed values that do not account for future
residential development.
Accordingly, I propose an alternative method for deriving the potential future residential
value of agricultural parcels using observed assessed values. Given that assessors only consider
a parcel’s current use, the assessed value (AV) of an agricultural parcel i and a residential parcel j
can be expressed as:
23
(2) , and
(3)
If the characteristics of the two parcels are identical (i.e., zi = zj), then an estimate of the true
economic value of the agricultural parcel i is given by:
(4) .
The estimated value in (4) is a biased approximation of the true value, Pi, because it assumes that
parcel i can receive both agricultural and residential rents in perpetuity – from t to infinity. In
reality, the contribution to rents in perpetuity from agriculture and development depends on the
optimal conversion time (t*), which is unobservable. The bias in is explicitly given by:
(5) .
Since (5) is strictly non-negative, overestimates the true economic value. As t* approaches
infinity (i.e., parcels with little or no foreseeable development pressure), the bias in (5) is given
by:
(6) .
overestimates the true value by giving too much weight to future development rents, which
are highly unlikely. Similarly, as t* approaches zero, the bias in (5) is given by:
(7) .
24
Here overestimates the parcel’s true value by giving too much weight to agricultural rents,
which are soon to disappear.
If parcels are matched well according to the propensity score matching approach I use
(see below), the magnitude of bias in (4) should be relatively small. If a very rural agricultural
parcel with low development pressure is matched with a similarly rural residential parcel, then
the assessed value of the residential parcel should reflect the low market demand for such
residential properties. The assessed residential value will be close to the assessed agricultural
value and the bias will be relatively small (as opposed to matching the rural agricultural parcel
with a parcel under high development pressure). Similarly, for parcels facing high development
pressure, the assessed agricultural value will be small relative to the assessed residential value,
making the bias of including agricultural rents relatively small. I therefore use (4) to define the
economic score of each parcel using assessed values.
Propensity Score Matching
Propensity score matching was developed in the health sciences to reduce selection bias
when comparing non-equivalent groups (Rosenbaum and Rubin 1983). The standard approach
to understanding the effects of a health treatment – a vaccination, for example – is to conduct an
experiment where individuals from a population are chosen at random and assigned to treatment
and control groups. Randomly assigning individuals implies that a comparison of outcomes
between the treatment and control group will provide an unbiased estimate of treatment effects.
With observational data, individuals are not assigned to treatment and control groups
randomly. This means that researchers can observe the outcomes for treated and untreated
individuals, but not the counterfactual – the outcomes of treated individuals had they not
25
received the treatment and vice versa. Differences in outcomes between the treated and
untreated may then be attributable to the treatment, or underlying differences in the individual
characteristics of the experiment subjects, such as gender, income, education, and race. If these
individual characteristics influence who gets treated or how they respond to treatment, then a
comparison of the outcomes of the two groups will produce a biased estimate of the treatment
effect. Propensity score matching can be used to reduce this bias by matching individuals based
on the similarity in their characteristics, and is widely applicable outside of the health science
disciplines. In short, it assumes that if two individuals share every characteristic other than
receiving or not receiving the treatment, then the difference in their outcomes is the treatment
effect.
I employ the propensity score matching approach because I can only observe agricultural
parcels that have not yet been converted to residential use or residential parcels that have already
converted from agricultural use. In each case, I cannot observe a counterfactual. In keeping
with the example of propensity score applications in the health sciences, I consider conversion to
residential use as a “treatment” to understand how this treatment affects the value of untreated
parcels – i.e., the counterfactual. This tells me what the assessed value of an existing agricultural
parcel would be if it were a residential parcel.
However, the nature of assessors’ data for parcel value does not allow for the standard
application of a propensity score matching approach. Assuming that assessors accurately
capitalized future development values into current agricultural land values, the assessed value of
an agricultural parcel i would be given by:
(8) *
*, , ,
0 *
( , ) ( , ) ,t
rt rt rti ag i ag i i res i
t t
AV t z e dt t z e dt Ceπ π∞
− − −
=
= + −∫ ∫
26
and the assessed value of a residential parcel j would be:
(9) *
, ,0
( , ) .rt rtj res j res jAV t z e dt Ceπ
∞− −= −∫
If (8) and (9) were true, I could use propensity scores to match agricultural and residential
parcels, and the residential value of the matched parcel would proxy for the proportion of the
agricultural value that is attributed to future development potential.
Because assessed values do not capitalize future development values into current
agricultural land values, I cannot observe this “treatment.” Therefore, I use a propensity score
matching method to directly estimate (4), by matching agricultural and residential parcels that
share similar physical and geographical characteristics, and use the matched residential value as
a proxy for the future development values of the agricultural parcel to which it is matched.
Specifically, I use the following approach to estimate an economic score using propensity
score matching and assessed values:
i) Estimate propensity scores [p(z)] using a standard binary logit model.
ii) Match agricultural parcels to residential parcels using each parcel’s predicted
propensity score.
iii) Estimate the residential value of agricultural parcels based on the matched
residential parcels:
(10)
where ( ), ( )k res kf AV z is a function (depending on matching method) of the
assessed values of matched residential parcels.
I use a standard binary logit model, which predicts the probability of observing each
parcel in residential use, to estimate propensity scores. Discrete choice regression models are
27
often used to estimate propensity scores (Caliendo and Kopeinig 2008)4. The binary logit model
can be expressed as:
(11) 𝑝(𝑌𝑖 = 1|𝑥) = 𝑒𝑥′𝛽
1+𝑒𝑥′𝛽
where Y is an indicator variable equal to 1 if parcel i is in residential use, x is a vector of
characteristics expected to influence the probability of residential use, and β is a vector of
parameters to be estimated.
Data
I use the Sublette County assessors’ data from the Wyoming Department of Revenue
cadastral dataset to define parcels according to ownership and to characterize their current use.
The data classifies parcels according to an account type describing the parcel’s current use as:
agricultural, commercial, commercial vacant, industrial, industrial vacant, residential, residential
vacant, state assessed, exempt, and other. Because I examine conservation easement purchases
on agricultural lands that will preclude residential development, I narrow the cadastral data to
include only those parcels in “agricultural,” “residential,” or “residential vacant” uses. I consider
three different binary logit models that each use a different combination of the agricultural
parcels and the two residential parcel types: agricultural and residential (AGRES), agricultural
and residential vacant (AGRESVAC), and agricultural, residential and residential vacant
combined (AGRESRESVAC). In each data set, I exclude parcels within municipal boundaries.
Residential and residential vacant parcels within the municipalities of Sublette County face
different market conditions, different types of development, and different motivations for
4 There are several other methods to match observations, including matching based directly on the distribution of characteristics themselves (i.e., covariate matching). The propensity score matching approach avoids the ‘curse of dimensionality’ in cases when there are a large number of characteristics that need to be compared (see Caliendo and Kopeinig 2008).
28
conversion from one land use to another. They may therefore not be explained using the same
variables I include in the binary logit model for predicting ex-urban and rural residential use.
I run the binary logit model (11) on the AGRES, AGRESVAC, and AGRESRESVAC
datasets using the same set of independent variables. These include variables commonly
included in hedonic models, as they are explaining the probability of observing residential use. I
calculate measurements of independent variables for each parcel using spatial data in ArcMap
10® (ESRI 2010) and the CommunityViz® Scenario 360™ (Placeways, LLC 2012) analysis
software package. A review of the hedonic literature informed my choice of independent
variables (see Table 2).
Bergstrom and Ready (2009) and Plantinga et al. (2002) used time-series demographic
and land development trend data to disaggregate the value of future residential development and
explain the contribution of amenity value to overall farmland value. Because demographics in
Sublette County are generally homogenous and the geographic scope of my study is limited to
one county, thereby limiting the availability of demographic data at a fine scale, I do not include
demographic variables in my study. Additionally, I only observe cross-sectional data for one
year, preventing me from including time-series data, such as development rates, population
growth and price trends, as previous studies have (Plantinga et al. 2002; Bergstrom and Ready
2009; and Ma and Swinton 2012).
29
Table 2. Explanatory variables used in binary logit models.
Variable Variable Name Description Distance to town DISTTOWN Distance from parcel boundary to nearest town Distance to road DISTROAD Distance from parcel boundary to nearest road Average standard deviation of slope AVGSTDD_SLOPE Average standard deviation of slope measurements
on parcel to measure roughness of terrain Agricultural land in neighborhood AGNEARVIEW Share of agricultural lands, based on land cover
data, within 1 mile viewshed of parcel Residential land in neighborhood RESNEARVIEW Share of residential lands, based on cadastral data,
within 1 mile viewshed of parcel Commercial land in neighborhood COMMNEARVIEW Share of commercial lands, based on cadastral
data, within 1 mile viewshed of parcel Land cover – wetlands WETLANDCOVER Share of wetlands, based on land cover data, on
parcel Land cover – forest FORESTCOVER Share of forest, based on land cover data, on parcel Land cover – not developable UNDEVCOVER
Share of undevelopable lands – perennial ice/snow, open water, or barren land – based on land cover data, on parcel
Viewshed – mountain peaks MTN_PKS Count of mountain peaks above 13,000 feet
elevation visible from parcel
Distance to town (DISTTOWN) measures the proximity of parcels to services, such as
shopping, schools, and economic activity, and therefore likely influences the value of residential
development (Ma and Swinton 2012; Plantinga et al. 2002; Bastian et al. 2002). Urban spatial
models show that distance from a central business district is negatively related to rents for
developed land (Plantinga et al. 2002). I expect that there will be a negative relationship
between DISTTOWN and the probability of observing residential development in my model, as
the value of future residential development is a decreasing function of a parcel’s total value as its
distance from town increases (Plantinga et al. 2002). To calculate DISTTOWN I use the
“MinDistance” tool in CommunityViz® Scenario 360™ software, which measures the minimum
distance from each parcels’ boundary to the nearest municipality.
Consistent with land value theory, Plantinga et al. (2002) found that a one unit increase in
highway density, as a proxy for increased population density and decreased commuting costs,
30
increased the value of agricultural land substantially. Similar to DISTTOWN I include
DISTROAD as an indicator of the likelihood that a parcel will be converted to residential
development (Ma and Swinton 2012). I expect increased ease of access to a parcel to increase
the probability of observing residential use.
I include two independent variables to represent conversion cost from agricultural land to
residential use: AVGSTDD_SLOPE and UNDEVCOVER. Using the National Elevation Dataset
(USGS 2012), I calculate the standard deviation of the slope on each parcel (AVGSTDD_SLOPE)
to indicate roughness of terrain (Ma and Swinton 2012) and buildability. I also create a shapefile
layer to represent land covers that preclude construction of housing by reassigning the following
land covers from the National Land Cover Database (NLCD): barren, perennial ice/snow, and
open water. Using the Scenario 360™ “GridOverlap” tool, I calculate the combined area of
these land cover categories for each parcel. I include this area measurement as a share of each
parcel’s total area (UNDEVCOVER).
I use the same method to calculate other land cover area measurements by parcel. I
combine NLCD land covers “woody wetlands” and “emergent herbaceous wetlands” to create a
single land cover for wetlands. Finally, I combine NLCD land covers “deciduous,” “evergreen,”
and “mixed forest” to create a single land cover for forest. I calculate these area measurements
as a share of each parcel’s total area to construct the WETLANDCOVER and FORESTCOVER
variables. Both of these variables potentially serve as proxies for on-parcel amenities. Figure 3
shows the distribution of land covers according to agricultural parcels’ status as privately or
publicly owned lands. The distribution of wetlands and agricultural lands follow the pattern of
river beds and bottom lands, indicating the likelihood that soil characteristics and water
31
availability are shared or similar. Bergstrom and Ready (2009), Bastian et al. (2002), Ma and
Swinton (2012), and Spahr and Sunderman (1998) include similar measures of land productivity.
Figure 3. Land cover in Sublette County, Wyoming
View quality rating and scenic beauty modeling is an emerging field that is relevant for
hedonic modeling (Germino et al. 2001). I construct a variable, MTN_PKS, to capture the
viewshed from each parcel in Sublette County. Using a viewshed model constructed in ArcMap
10® to count the number of mountain peaks that exceed 13,000 feet in elevation that are visible
from each parcel. For parcels near the border of the county, I include a count of peaks within a
32
20 mile viewshed of the county line. I expect that the higher the count of mountain peaks visible
from a parcel, the higher the probability of observing residential development.
To capture the nearby scenic values from each parcel, I calculate shares of land use,
according to cadastral data categories (i.e. “commercial,” “industrial,” “residential,” or
“agricultural”) within a one mile viewshed of each parcel. I construct the variable
COMMNEARVIEW by calculating the area of parcels in commercial and industrial in the near
viewshed of each parcel, and combine those areas to calculate the share of commercial and
industrial land within a one mile radius of the center of each parcel. Similarly, I calculate the
share of residential (RESNEARVIEW) and agricultural (AGNEARVIEW) land within a one mile
radius of each parcel. Bergstrom and Ready (2009) found that neighboring land uses influence
amenity values and relative scarcity of agricultural lands influences farmland value.
Calculating the Economic Score
After estimating each parcel’s propensity score (i.e., predicted residential probability)
from the binary logit model, I calculate the economic score for each agricultural parcel by
imputing the value of residential or residential vacant parcel matched to the agricultural parcel. I
consider the assessed value for agricultural land to be its productive value only. The Wyoming
Department of Revenue assesses agricultural land value according to a formula that combines the
Wyoming Agricultural Statistics Service’s commodity price data and capitalized net income
(Wyoming Department of Revenue 2012). Summary statistics (Table 3) of assessed agricultural
values show a wide range of values.
33
Table 3. Summary statistics of agricultural lands' assessed values in Sublette County.
Summary Statistic Value (per acre)
Minimum $9.30
Maximum $21,521.74
Mean $535.31
Median $195.71
I match parcels using two alternative approaches: the sub-class matching approach and
the caliper matching approach; after evaluating each, I use the caliper approach. Sub-class
matching is simply dividing parcels into equal groups according to propensity score and
matching residential parcels with a propensity score that falls into the range of those observed in
each sub-class and imputing the average value of matched residential parcels to agricultural
parcels. I perform the sub-class matching approach in two different ways: first by dividing
agricultural parcels into 10 groups (sub-classes) and then into 20 groups. I arbitrarily chose 10
and 20 groups to gain accuracy in matching; the finer the sub-classes (i.e., more groups), the
fewer the number of parcels in each sub-class. I rank agricultural parcels by their estimated
propensity score from lowest to highest and then divide agricultural parcels (n=1,053) into 10
(n=105) or 20 (n=52) sub-classes with an even number of parcels in each. For example, the 105
agricultural parcels with the lowest propensity scores were assigned to sub-class 1 in the 10-sub-
class matching, and the 52 parcels with the lowest propensity scores were assigned to sub-class 1
in the 20-sub-class matching. Next, I match residential parcels’ propensity scores to agricultural
parcels’ propensity scores and assign residential parcels the same sub-class as the agricultural
parcel to which it is matched. For example, a residential parcel with a propensity score of 0.76
was matched to an agricultural parcel with the same propensity score that was assigned to sub-
34
class 8, so the residential parcel was also given a sub-class number of 8. I then average the
assessed per acre land value5 of all matched residential parcels within each matched sub-class to
impute the potential residential value of agricultural parcels. The sum of the imputed residential
value and the assessed agricultural value constitute the economic score using the sub-class
matching approach (see Equation 10).
The caliper approach matches parcels based on the stratification of their propensity
scores, like the sub-class approach, but defines strata according to a caliper statistic, ε:
(12) 𝜀 ≤ 0.25𝜎𝑝,
where 𝜎𝑝 is the standard deviation of the predicted propensity scores across all parcels.
Agricultural parcels are therefore matched to a residential parcel if the difference between their
propensity scores falls into caliper strata calculated using (12), or one quarter of the standard
deviation of all parcels’ propensity scores (Guo and Fraser 2010). The caliper approach results
in each agricultural parcel being matched to a group of residential parcels that satisfy (12). I
calculate the economic score using this approach by imputing the per acre residential land value
of parcels to the agricultural parcels to which they are matched. I impute the residential values
using the minimum, maximum, average, and median value per acre of matched residential
parcels.
3.2 Biological Score
I construct a simple index to rank habitat quality for the three key species that the JIO and
PAPO have targeted mitigation activities toward: pronghorn, mule deer, and sage grouse.
5 I separate assessed land values from assessed value of improvements and structures within the cadastral dataset before modeling parcel values.
35
I base the biological score for mule deer and pronghorn on three key factors: 1) distance
to crucial winter range and/or area of crucial winter range on parcels, 2) overlap with migratory
stopover sites, and 3) overlap with migratory movement corridors. I weight and sum these
measurements, which I calculated using Scenario 360™ and spatially explicit migration data
(Sawyer and Nielson 2011)6 and winter range data (WGFD 2011):
(13) 𝑀𝐷𝐵𝑖𝑜𝑙𝑆𝑐𝑜𝑟𝑒,𝑖 = ∑ 𝐷𝑖𝑠𝑡𝑘,𝑤,𝑖𝐴𝑐𝑟𝑒𝑠𝑤,𝑖 + 𝐴𝑐𝑟𝑒𝑠𝑠,𝑖𝑘 + (0.5)𝐴𝑐𝑟𝑒𝑠𝑚,𝑖, and
(14) 𝑃𝐻𝐵𝑖𝑜𝑙𝑆𝑐𝑜𝑟𝑒,𝑖 = ∑ 𝐷𝑖𝑠𝑡𝑘,𝑤,𝑖𝐴𝑐𝑟𝑒𝑠𝑤,𝑖 + 𝐴𝑐𝑟𝑒𝑠𝑠,𝑖𝑘 + (0.5)𝐴𝑐𝑟𝑒𝑠𝑚,𝑖,
Where Acresw,i is acres of crucial winter range on parcel i, Acress,i is acres of stopover habitat,
and Acresm,i is acres of movement corridor. Acres of crucial winter range are weighted by the
distance between the parcel and the nearest crucial winter range. Parcels that contain crucial
winter range get a weight of 1 (i.e., Distk,w,i = 1), parcels within 1 km of winter range get a
weight of 0.5 (i.e., Distk,w,i = 0.5), and parcels that are greater than 1 km from crucial winter
range get a weight of zero (i.e., Distk,w,i = 0) (Sawyer, pers. comm. 2012). This metric reflects
the importance of conserving parcels located in habitat already used by mule deer and pronghorn
(Sawyer, pers. comm. 2012). Finally, parcels containing migratory stopover habitat receive a
greater weight than parcels containing movement corridors, due to the ecological importance of
stopover to migrating animals (Sawyer et al. 2011).
To incorporate sage grouse into the biological score, I construct a 5 kilometer buffer
around occupied leks, calculate the area within that buffer, and calculate the number of acres of
resulting sage grouse habitat contained within each parcel:
(15) 𝑆𝐺𝐵𝑖𝑜𝑙𝑆𝑐𝑜𝑟𝑒,𝑖 = 𝐴𝑐𝑟𝑒𝑠𝑆𝐺,𝑖,
6 I only calculate the biological score for parcels north of Hwy 351, because radio-collar data for mule deer and pronghorn migration is limited to that area. The herds impacted by natural gas development in the Jonah Field and Pinedale Anticline are included in the data.
36
Holloran (2005) found that over 64 percent of sage grouse nest within 5 kilometers of the lek
site, so I use this buffer to represent both leking and nesting habitat.
I add the scores for each species without assigning any weighting (i.e., weight of 1 per
acre) to generate a total biological score for each agricultural parcel:
(16) 𝑇𝑜𝑡𝑎𝑙 𝐵𝑖𝑜𝑙𝑜𝑔𝑖𝑐𝑎𝑙 𝑆𝑐𝑜𝑟𝑒𝑖 = 𝑀𝐷𝐵𝑖𝑜𝑙𝑆𝑐𝑜𝑟𝑒,𝑖 + 𝑃𝐻𝐵𝑖𝑜𝑙𝑆𝑐𝑜𝑟𝑒,𝑖 + 𝑆𝐺𝐵𝑖𝑜𝑙𝑆𝑐𝑜𝑟𝑒,𝑖
Figure 4 shows mule deer winter range, migratory stopover sites, and movement corridors
overlaid with agricultural parcels and existing conservation easements in Sublette County.
Figure 4. Mule deer habitat and privately-owned lands in Sublette County, Wyoming.
Similarly, Figure 5 shows pronghorn movement, stopover, and crucial winter range in Sublette
County, and is overlaid with privately-owned agricultural parcels and existing conservation
easements.
37
Figure 5. Pronghorn habitat and privately-owned agricultural lands in Sublette County, Wyoming.
Finally, Figure 6 shows currently occupied sage grouse leks with a five kilometer buffer
surrounding each, overlaid with agricultural parcels and existing conservation easements in
Sublette County.
38
Figure 6. Sage grouse habitat and privately-owned agricultural lands in Sublette County, Wyoming.
Figures 4, 5, and 6 show the distribution of habitat types across existing easements and
agricultural parcels in Sublette County. Importantly, much of the habitat for mule deer,
pronghorn, and sage grouse is located on public lands. Table 4 provides a comparison of the
acreage available to each species on public and private lands, and a maximum biological score
according to my definition.
Table 4. Distribution of habitat across private and public lands in Sublette County.
Acres Sage Grouse
Mule Deer Movement
Mule Deer Stopover
Mule Deer Winter
Pronghorn Movement
Pronghorn Stopover
Pronghorn Winter
Public 926,198 117,244 69,722 395,514 123,682 71,393 167,639 Private 328,127 45,149 17,381 95,868 134,296 97,896 27,959 Percent Private
26.1%
27.8%
19.9%
19.5%
52.0%
57.8%
14.2%
39
Table 4 indicates that while nearly half of pronghorn migration habitat is on private land,
only approximately 20 to 28 percent of sage grouse, mule deer winter and mule deer migration
habitat is located on private lands. Pronghorn winter habitat is limited to only 14 percent on
private lands.
3.3 Production Possibilities Frontier
To understand and visualize the economic-ecological tradeoff, I combine the biological
and economic scores to construct a production possibilities frontier (PPF). I rank parcels
according to total biological score (benefit) and total conservation easement cost, which I define
as the estimated foregone value of future residential development (cost). This arrangement of
parcels according to benefit-cost ranking forms the PPF. Additionally, I rank parcels similarly
according to economic score, biological score, and propensity score to assess alternative
strategies for targeting easements on the landscape. This illustrates the potential cost for
achieving different levels of habitat protection, given different conservation objectives. Chapter
5 provides an in-depth analysis of targeting strategies.
40
4 Results
4.1 Economic Score
Logit Results
As discussed in the sections above, I conducted a review of hedonic modeling literature
to inform my econometric model. The data generally fit each of the binary logit models –
AGRES, AGRESVAC and AGRESRESVAC - well. Independent variables are generally
significant at the 1% level. The likelihood ratio for each model is also significant at the 1%
level. With few exceptions, the logit model gave the expected sign for the estimated coefficients
(Table 5). Additionally, a comparison of parameter estimates between the AGRES, AGRESVAC
and AGRESRESVAC models shows similar signs and magnitude, with a few exceptions.
Table 5. Parameter estimates for AGRES, AGRESVAC, and AGRESRESVAC logit models.
Variable
AGRES a Coefficient
AGRESVAC b
Coefficient AGRESRESVAC c
Coefficient Intercept 11.9612***
(1.9071) 23.5819*** (2.4281)
19.371*** (2.2756)
DistTown -0.00001*** (0.000003256)
-0.00000778** (0.000003483)
-0.00000868*** (0.000002872)
DistRoad -0.00419*** (0.000429)
-0.00130*** (0.000183)
-0.00195*** (0.000184)
AvgStdD_Slope -0.8480*** (0.0567)
-0.6746*** (0.0555)
-0.7729*** (0.0467)
AgNearView -9.9308*** (1.9045)
-22.3094*** (2.4443)
-17.0809*** (2.2812)
ResNearView -3.4804 (3.5155)
-23.5311*** (3.7738)
-13.3267*** (3.8571)
CommNearView -56.8621*** (11.9696)
-59.3578*** (14.0856)
-62.4819*** (11.6260)
WetlandCover -1.3780*** (0.1933)
-2.0502*** (0.222)
-1.5520*** (0.1679)
ForestCover 3.7394*** (0.4143)
3.4315*** (0.3777)
3.6354*** (0.3570)
UndevCover -14.3355*** (5.3951)
1.1253 (3.0179)
-2.9434 (2.804)
Mtn_Pks -0.0088** (0.00396)
0.00123 (0.00388)
-0.00452 (0.00346)
Likelihood Ratio 1248.1685*** (n=3362)
1091.0318*** (n=2939)
1411.4881*** (n=5259)
Note: *,**,*** denote significance at the 10%, 5%, and 1% level, respectively. Standard errors are in parentheses. a AGRES is a binary logit model where is the data includes agricultural and residential parcels. bAGRESVAC is a binary logit model where data includes agricultural and residential vacant parcels. cAGRESRESVAC is a binary logit model where the data includes agricultural, residential, and residential vacant parcels.
41
Coefficients for both AgNearView and ResNearView were much lower in the AGRES
model than the AGRESVAC and AGRESRESVAC models. The estimated sign for WetlandCover
was negative; I hypothesized that it would measure amenity value and/or soil and soil moisture
levels that make it productive land because of its nearness to lands currently used for agricultural
production. If WetlandCover is measuring amenity value, it would have a positive sign –
indicating an increasing probability of observing residential land use. Because WetlandCover is
negative, it is likely a proxy for soil characteristics that make it suitable for agriculture; further, it
could represent high conversion costs of vegetation removal, high water table or flood plain
issues, or other factors that make locations unsuitable for residential development. Therefore, a
negative sign on the parameter estimate for WetlandCover likely is indicating correctly that
parcels containing wetlands are more profitable in agriculture, relatively, than other parcels and
are therefore less likely to be observed in residential use.
Coefficients of UndevCover and Mtn_Pks had different signs in the AGRESVAC model
than the AGRES and AGRESRESVAC models. I expect UndevCover to decrease as the
probability of observing residential use increases. Because Mtn_Pks measures amenity value –
the number of mountain peaks within each parcel’s viewshed – I expect it to be positive.
Though the signs of the parameter estimates from the logit model are informative, given
that the binary logit model is a non-linear function, I cannot directly interpret their magnitude. I
therefore calculate marginal effects of each parameter:
(17) 𝜕𝑃𝑖𝜕𝑋𝑘
= 𝜕𝑒𝑥𝛽
1+𝑒𝑥𝛽
𝜕𝑥𝑘,
where 𝑃𝑖 is the probability given by the logit model for observing residential use on parcel i and
𝑋𝑘 is the value observed for a characteristic k. Equation (17) can be re-written as:
42
(18) 𝜕𝑃𝑖𝜕𝑋𝑘
= [𝛽𝑘𝑃�𝑖(1 − 𝑃�𝑖)].
I calculate the marginal effect of each variable on each parcel’s propensity score and then
average them to get an estimated marginal effect (Table 6).
Table 6. Average marginal effects.
Variable AGRES AGRESVAC AGRESRESVAC DistTown -0.000001 -0.000001 -0.000001 DistRoad -0.000594 -0.000203 -0.000225 AvgStdD_Slope -0.120296 -0.105303 -0.088984 AgNearView -1.408769 -3.482419 -1.966519 ResNearView -0.493724 -3.673122 -1.534299 CommNearView -8.066374 -9.265545 -7.193523 WetlandCover -0.195481 -0.320029 -0.178681 ForestCover 0.530466 0.535645 0.418543 UndevCover -2.033613 0.175655 -0.338873 Mtn_Pks -0.001248 0.000192 -0.000520
The marginal effects for the independent variables in Table 6 are interpreted as the
change in the probability of observing residential use on a given parcel for a one unit change in
the explanatory variable. Therefore, a one unit increase (1 foot) in the distance between a given
parcel and the nearest municipality decreases the probability of observing that parcel in
residential use by 0.0001%; DistRoad can be interpreted similarly. The near viewshed variables,
AgNearView, ResNearView, and CommNearView can be interpreted in the following way: an
increase of 1% in the share of the near view in the land use of interest (agricultural, residential,
or commercial, respectively) decreases the probability of observing a parcel in residential use by
1.4%, 0.49%, or 8.06%, respectively, in the AGRES model. Marginal effects for the AGRESVAC
and AGRESRESVAC models can be interpreted in the same way. Marginal effects for the other
variables that measure shares – WetlandCover, ForestCover, and UndevCover – of land cover on
a given parcel can also be interpreted in this way: an increase of 1% in the share of forest land
cover on a given parcel increases the probability of observing a parcel in residential use by
43
0.53% in the AGRES model, for instance. Finally, my calculation of marginal effects indicates
that an increase of one mountain peak within a given parcel’s viewshed increases the probability
of observing that parcel in residential use by approximately 0.02% in the AGRESVAC model and
decreases that probability by 0.12% and 0.05% in the AGRES and AGRESRESVAC models,
respectively.
The AGRES model predicts the broadest range of propensity scores; the AGRESRESVAC
model predicts the narrowest range of propensity scores (Table 7). The frequency of predicted
probabilities (propensity scores) shows that predicted probabilities of observing residential use
on agricultural parcels is skewed to the right in both the AGRES and AGRESRESVAC models,
and is more evenly distributed in the AGRESVAC model (Figure 7). Because residential parcels,
as opposed to only residential vacant parcels are included in the former datasets, the AGRES and
AGRESRESVAC predict more high residential probabilities. Figures 8, 9, and 10 display
predicted propensity scores from each model on agricultural parcels in Sublette County.
Table 7. Average predicted probabilities by observed land use and model.
Observed Use AGRES AGRESVAC AGRESRESVAC
Agriculture 0.4472 0.4233 0.5642 Residential/Residential Vacant 0.7961 0.7640 0.8587
44
Figure 7. Frequency of predicted probabilities for agricultural parcels.
0
20
40
60
80
100
120
140
160
180
0.05 0.
10.
15 0.2
0.25 0.
30.
35 0.4
0.45 0.
50.
55 0.6
0.65 0.
70.
75 0.8
0.85 0.
90.
95 1
Freq
uenc
y
Predicted Probability
AGRES AGRESVAC AGRESRESVAC
45
Figure 8. Map of predicted propensity scores on agricultural parcels using AGRES model – Sublette County, Wyoming.
46
Figure 9. Map of predicted propensity scores on agricultural parcels using AGRESVAC model – Sublette County, Wyoming.
47
Figure 10. Map of predicted propensity scores on agricultural parcels using AGRESRESVAC model – Sublette County, Wyoming.
Results of sub-class and caliper matching
I evaluate the sub-class and caliper matching procedures by how closely each matches
agricultural parcels to residential parcels according to their estimated propensity scores. The
caliper matching approach generally provides closer matches than the sub-class approach, though
each approach gives similar average predicted values (measured in dollars per acre) between
models (Table 8).
48
Table 8. Average per acre value of matched residential parcels using each matching approach.
Model
Sub-class
10 Groups
Sub-class
20 Groups
Caliper
(n=1053)
AGRES $17,984.14 $12,731.15 $15,062.88 AGRESVAC $18,131.59 $12,179.17 $14,945.91 AGRESRESVAC $17,466.79 $12,575.00 $14,969.90
While average predicted values between matching approaches and models are similar, predicted
values for each parcel differ substantially between the matching approaches. The accuracy of
the sub-class matching approach is less consistent than the caliper approach. The range of
matched propensity scores is 0.26 and 0.20 for sub-class 10 and sub-class 20, respectively. Such
a large range implies that an agricultural parcel with a propensity score of 0.99 and a parcel with
a propensity score of 0.73 are matched to the same group of residential parcels. Alternatively,
this can be interpreted as assigning the same future residential value to a parcel with a 99 percent
probability of being observing in residential use and a parcel with only a 73 percent probability
of observing in residential use.
49
Conversely, within some of the sub-classes, the match between agricultural and
residential parcels’ propensity scores is closer than the match using the caliper approach. The
caliper approach matches residential and agricultural parcels based on a consistent caliper of
0.068, 0.068, and 0.053 for the AGRES, AGRESVAC, and AGRESRESVAC models, respectively.
Where the range of propensity scores is less than 0.068, in the AGRESVAC model for example,
the sub-class approach results in a closer match. Given these tradeoffs between approaches, I
use the results of the caliper matching approach in my analysis because it more consistently
matches agricultural parcels with residential parcels with a similar propensity score and
generates more heterogeneity in residential parcels matched to agricultural parcels.
Based on the results of the logit model and caliper matching, I continue my analysis using
only the AGRESVAC model. Propensity scores and estimated future residential development
value are most accurately captured using the AGRESVAC model, as including parcels currently
in residential use could over-estimate future residential value for two reasons: parcels already
developed might add additional bias to assessed values, and there may be variables omitted from
the logit model associated with residential parcels that are not associated with residential vacant
parcels. Further, I use the median value per acre of caliper-matched residential parcels to impute
future residential development value for agricultural parcels when calculating the economic
score. The median value is the best representation of the range of values of residential parcels
matched to a given agricultural parcel because it is less affected by outlier residential values of
matched parcels (high or low). Table 9 shows summary statistics of the economic score
calculated using the median residential value of residential parcels matched to agricultural
parcels using the caliper matching approach.
50
Table 9. Predicted residential values calculated for AGRESVAC model using median, minimum, and average values per acre of matched residential parcels. All values are in dollars per acre.
Imputed Values
Economic Score Average Minimum Maximum
Median $10,434.77 $7,010.00 $31,972.00
Minimum $1,723.09 $153.47 $22,285.75
Average $17,997.45 $8,035.27 $51,734.28
4.2 Biological Score Results
Conservation easements are limited in their effectiveness as mitigation mechanisms by
the relatively small amount of critical habitat located on private lands in Sublette County (see
Table 4). Table 10 shows summary statistics of the biological score calculations for privately-
owned7 agricultural parcels.
Table 10. Summary statistics of biological scores calculated for agricultural parcels in Sublette County, Wyoming.
Total Biological Score
Mule Deer Biological Score
Pronghorn Biological Score
Sage Grouse Biological Score
Average 549.93 132.46 110.81 306.66
Minimum 0 0 0 0
Maximum 9,492.36 3,975.52 9,132.64 5,925.25
7 Some conservation easement projects funded by the JIO or PAPO include components wherein conservation practices occur on the landowners’ grazing allotments on federally-owned lands. Biological scores are not calculated for these lands.
51
These statistics illustrate the heterogeneity among agricultural parcels in their estimated total
biological score, and in the estimated biological scores of individual species. Figures 11 - 13
show the estimated biological scores in total and for each species on private agricultural parcels.
Figure 11. Estimated biological scores for mule deer on agricultural parcels in Sublette County, Wyoming.
52
Figure 12. Estimated biological scores for pronghorn on agricultural parcels in Sublette County, Wyoming.
53
Figure 13. Estimated biological scores for sage grouse on agricultural parcels in Sublette County, Wyoming.
54
Figure 14. Estimated total biological scores on agricultural parcels in Sublette County, Wyoming.
4.3 Estimated Ecological-Economic Tradeoffs
The production possibilities frontiers (PPFs) that I construct using predicted economic
and biological scores depict the economic and ecological tradeoffs associated with protecting
impacted species’ habitat using conservation easements to mitigate for habitat loss on the Jonah
Field and Pinedale Anticline. Regardless of the propensity score model used (AGRES,
AGRESVAC, or AGRESRESVAC), the predicted PPFs show an increasing rate of product
transformations between the economic value of land in Sublette County (i.e., the economic
score) and the biological value of the land for mule deer, pronghorn, and sage grouse (Figure
15). As expected, the highest landscape (total of all parcels) economic score occurs when no
55
parcels are placed under a conservation easement and thus no habitat is guaranteed to be
conserved. Parcels efficiently conserved are those with the highest “bang per buck,” or highest
biological score per dollar of foregone economic value (i.e., predicted future residential
development value). By selecting the most efficient parcels first, the biological score increases
rapidly for relatively low cost. As more parcels are conserved, each additional parcel is
successively less efficient (i.e., protects less biological score per dollar), giving the PPF the
traditional concave form.
Figure 15. Production possibilities frontier from the alternative propensity score models using median residential values.
Figure 15 also demonstrates that the alternative propensity score models generate similar
results. In each case, the biological scores of conserved parcels are identical because they are not
calculated used estimated propensity scores. The propensity scores determine which residential
0
1,000
2,000
3,000
4,000
5,000
6,000
0 100 200 300 400 500 600 700
Eco
nom
ic S
core
(mill
ions
)
Total Biological Score (1000s)
AGRESVAC AGRES AGRESRESVAC
56
parcels are matched with each agricultural parcel, which generates differences in the assigned
residential values (and thus economic scores) across models. As noted in Section 4.1, the models
that include parcels currently in residential use (AGRES and AGRESRESVAC) systematically
predict higher residential value and hence higher economic scores than the AGRESVAC model,
which only includes residential vacant parcels.
I also construct production possibilities frontiers that separate components of the total
biological score to demonstrate species-specific tradeoffs (Figure 16). While Figure 16 only
shows PPFs generated from using the AGRESVAC model, the other propensity score models
generate the same pattern. Figure 16 shows that the tradeoff between the economic and
biological scores is much steeper for mule deer and pronghorn. This is because most mule deer
and pronghorn habitat does not overlap with private agricultural lands. Accordingly, the potential
to mitigate impacts to mule deer and pronghorn through conservation easements is limited.
Additionally, optimal conservation targeting depends on which species are targeted or weighted
most heavily.
57
Figure 16. Production possibilities frontiers by species-specific biological scores estimated using AGRESVAC model.
The PPFs shown in Figure 16 assume that habitat is only protected on parcels with
conservation easements (i.e., the minimum landscape biological score is equal to zero). This
assumption underestimates the biological score of the total landscape since many agricultural
parcels have a low propensity of being observed in residential use and may not convert to
residential use, thereby maintaining their function as habitat. Because even in the absence of a
conservation easement many agricultural parcels will remain in agriculture and continue to
provide mule deer, pronghorn, and sage grouse habitat, I also generate PPFs using an expected
biological score. To calculate the expected biological score, I assume that agricultural parcels
without easements will generate biological scores equal to their biological score multiplied by
their probability of being in agriculture (i.e., 1- propensity score). The total expected biological
score can be interpreted as what the landscape is likely to produce in the long-run as agricultural
parcels convert to residential use according to my predicted propensity scores.
0
500
1,000
1,500
2,000
2,500
3,000
3,500
4,000
4,500
0 100 200 300 400 500 600 700
Eco
nom
ic S
core
(mill
ions
)
Landscape Biological Score (1000s)
Total Score MD Score PH Score SG Score
58
PPFs using the expected biological scores no longer have a minimum value of zero, since
even if no parcels are placed under a conservation easement some parcels will continue to
provide habitat (see Figure 17). The tradeoffs in Figure 17, the total biological score, are similar
to those shown in Figure 15, but the PPF simply begins with a positive biological score,
demonstrating that some biological score is achieved in the absence of conservation easements.
Figure 17. Production possibilities frontier using the expected biological score and economic score estimated using the AGRESVAC model.
Figure 18 shows PPFs by species using the expected biological score, and compared to
Figure 16, suggests some important differences. Mule deer and pronghorn biological scores are
substantially lower in expectation than the biological score for sage grouse since sage grouse
habitat is widely dispersed and thus more sage grouse habitat is present even in the absence of
conservation easements. In contrast, mule deer and pronghorn scores show a very steep tradeoff.
The steep tradeoff and relatively low minimum expected biological scores again indicate that
although parcels with mule deer and pronghorn habitat have relatively low propensity to be
0
500
1,000
1,500
2,000
2,500
3,000
3,500
4,000
4,500
0 100 200 300 400 500 600 700
Eco
nom
ic S
core
(mill
ions
)
Landscape Biological Score (1000s)
Total Score
59
residential, there are so few of them that even without conservation easements much of the mule
deer and pronghorn biological scores are not lost in expectation.
Figure 18. Production possibilities frontiers by species using expected biological scores and economic score estimated using AGRESVAC model.
5 Optimal Targeting of Conservation Easements
The PPFs derived in the previous section identify the theoretically efficient allocation of
conservation easements for any given biological or economic score. In other words, for every
possible biological score, the associated efficient point on the PPF identifies the highest
economic score (i.e., lowest reduction in potential residential values) that can be achieved.
These efficient allocations may not be achievable in reality for many practical reasons.
Conservation easements are voluntary and thus efficient parcels may not be available. Economic
theory suggests that every landowner should be willing to accept a conservation easement if
offered the full opportunity cost of foregone development; however, landowner preferences and
0
500
1,000
1,500
2,000
2,500
3,000
3,500
4,000
4,500
0 50 100 150 200 250 300 350
Eco
nom
ic S
core
(mill
ions
)
Landscape Biological Score (1000s)
MD Score SG Score PH Score
60
uncertainty about the future development rights makes it difficult to determine individual
landowners’ maximum willingness to accept conservation easements and hence difficult to target
precisely on the efficient frontier. Nonetheless, from a social welfare perspective, achieving a
landscape that is represented by a point on the efficient frontier should be the goal. I use GIS
data on currently existing conservation easements to calculate the current landscape’s economic
and biological score and compare it to the PPF (Figure 19).
Figure 19. Comparison of existing conservation easements to the efficient production possibilities frontier - AGRESVAC model.
The current landscape arrangement and conservation easements are less efficient than
what is theoretically possible. The existing easements have an associated economic score of
approximately 3.59 million and a total biological score of 51,361. This biological score is
approximately 125,000 points lower than the efficient point with the same economic score (i.e.,
moving horizontally from the current easement point to the efficient frontier). The same
economic score and a much higher biological score could be achieved by re-allocating the
0
500
1,000
1,500
2,000
2,500
3,000
3,500
4,000
4,500
0 100 200 300 400 500 600 700
Eco
nom
ic S
core
(mill
ions
)
Landscape Biological Score (1000s)
Total Score All Existing Projects
61
easements to parcels that generate a higher total biological score per dollar than those included in
current easements. Alternatively, the same approximate biological score could be achieved with
a higher economic score (3.95 million) through more efficient targeting (i.e., moving vertically
from the current easement point to the efficient frontier).
I also calculate this relative efficiency by species (Table 11). Current conservation
easements are closest to being efficient for sage grouse and furthest from being efficient for
pronghorn. This is likely because of the wider distribution of sage grouse habitat (see Figure 6).
Figure 20 separates the total biological score shown in Figure 19 by individual species.
Table 11. Difference in biological scores for each species between current easement point and efficient frontier.
Mule Deer Pronghorn Sage Grouse Difference in relative efficiency (biological scores) -27,268 -87,287 -10,787
62
Figure 20. Comparison of existing conservation easements to the efficient production possibilities frontier disaggregated by species - AGRESVAC model.
Targeting Approaches
The preceding PPFs identify efficient allocations of conservation easements for every
possible economic score by iteratively selecting parcels with the highest benefit-cost ratio, or
highest biological score per dollar of foregone residential value. This approach is only one way
to target conservation easements on the landscape. Newburn et al. (2005) discuss four common
alternative methods for targeting conservation, each of which have different policy implications:
1) Benefit targeting: select parcels to maximize ecological benefits for a given conservation
budget;
0
500
1,000
1,500
2,000
2,500
3,000
3,500
4,000
4,500
0 50 100 150 200 250 300 350
Eco
nom
ic S
core
(mill
ions
)
Landscape Biological Score (1000s)
MD Score
PH Score
SG Score
Existing Projects -MD ScoreExisting Projects -PH ScoreExisting Projects -SG Score
63
2) Benefit-cost targeting: select parcels with the highest benefit-cost ratio until a given
conservation budget is exhausted;
3) Benefit-loss targeting: select parcels to minimize the expected loss of biological benefits
for a given budget;
4) Benefit-loss-cost targeting: select parcels to minimize the expected loss of biological
benefits per dollar for a given budget.
Applying different targeting strategies (for a fixed budget of $36 million) significantly affects
which parcels are optimally conserved and the resulting biological scores. I identify parcels that
are the “best” candidates for conservation under each targeting strategy by ranking parcels
according to each strategy and then iteratively selecting parcels until the budget of $36 million is
exhausted. For example, to select parcels that should be targeted first according to the benefit
targeting strategy, I ranked parcels according to their biological score and selected them
iteratively, from highest biological score down the ranking, until the budget was exhausted.
Considering only the biological scores of conserved parcels (i.e., assuming un-conserved
parcels produce no biological values), benefit-cost targeting produces the highest biological
score (Figure 21). This result is consistent with theory since benefit-cost targeting maximizes
the “bang per buck” of selected parcels. Benefit targeting performs relatively poorly because it
targets parcels with high biological scores regardless of cost. Considering cost – defined as the
foregone future residential development value of each parcel – in the benefit-cost targeting
strategy gives more biological benefit (captures a greater total biological score) than targeting
only according to biological benefit. This is because given a budget of $36 million, resource
managers could hypothetically purchase more parcels with slightly lower biological value than
64
they could high-biological benefit parcels that are much more expensive, resulting in a greater
total of biological benefit captured.
Figure 21. Total biological scores produced from alternative targeting strategies with a total budget of $36 million.
Benefit and benefit-cost targeting do not consider the relative risk of losing parcels,
however, so the expected landscape biological score is lower (i.e., they ignore the expected
biological score of un-conserved parcels). When risk of losing parcels to development is
considered across the entire landscape, benefit-loss-cost targeting produces the best conservation
outcome (Figure 22). This approach performs best because it targets parcels with the highest
expected loss of biological value while also accounting for cost of conservation. It includes
parcels that have relatively low expected biological losses if they are sufficiently inexpensive,
and only includes relatively expensive parcels if they have sufficiently high expected losses. In
contrast, benefit-loss targeting performs poorly because it includes high risk (and often
expensive) parcels even if their biological score is too low to be cost-effective.
0
5,000
10,000
15,000
20,000
25,000
30,000
35,000
40,000
Benefit Benefit-Cost
Tota
l Bio
logi
cal S
core
65
Figure 22. Landscape-level expected biological scores produced from alternative targeting strategies with a total budget of $36 million.
Figure 22 highlights the importance of considering risk of development when targeting
easements. Parcels with low propensity scores (i.e., low probability of being residential) are
likely to continue producing biological value regardless of whether they are placed in a
conservation easement. Benefit-loss-cost targeting performs better than other approaches so long
as there are low propensity (i.e., low expected biological loss) parcels with high biological value.
Accounting for expected loss, these parcels should not be conserved since they are unlikely to be
lost – rather, limited budgets should be targeted to higher risk-higher biological value parcels.
The relationship between propensity scores and biological scores for parcels in Sublette
County shows that there are many relatively low risk parcels with relatively high biological
scores (Figure 23). Figure 23 illustrates that this trend exists for each individual species as well:
for mule deer, pronghorn, and sage grouse, there are parcels that provide high biological value
and have a low probability of being observed in residential use. There is a general positive
relationship between propensity scores and residential value (i.e., opportunity cost of the
easement) and thus a positive relationship between biological scores and residential value
280,000
282,000
284,000
286,000
288,000
290,000
292,000
294,000
Benefit Benefit-Cost Benefit-Loss Benefit-Loss-Cost
E[T
otal
Bio
logi
cal S
core
]
66
(Figure 24). As a result, accounting for risk of development, cost of easement purchases, and
level of biological value generates the best conservation outcome given a limited budget.
Figure 23. Relationship between parcel-level biological scores and propensity scores.
67
Figure 24. Relationship between the biological and residential values of parcels.
The relationship between risk and biological value has important implications for the
effectiveness of conservation easements as a conservation tool. As shown in Figure 23, there are
many low risk, high biological score parcels, and the highest risk parcels tend to have relatively
low biological value. This suggests that there is little competition between species’ critical
habitat and parcels that are likely candidates for residential development. This also suggests that
the landscape is likely to produce substantial biological value even in the absence of
conservation easements, and suggests that conservation or mitigation efforts may be better
directed at minimizing impacts to species on public lands, where much of species’ critical habitat
is located. Because so many agricultural parcels are likely to remain in agriculture, by placing
them under a conservation easement, resource managers would be paying for a service (provision
of critical habitat) that would otherwise still be provided at no expense.
The expected biological score across all parcels is 377,570, while the total biological
score if all parcels were placed under conservation easement is 587,618. Thus, approximately 64
percent of the biological value on the landscape is likely to remain even if there were no
01,0002,0003,0004,0005,0006,0007,0008,0009,000
10,000
0 20,000,000 40,000,000 60,000,000 80,000,000
Tota
l Bio
logi
cal S
core
Residential Value
68
conservation easements. Current conservation easement projects are relatively inefficient
because they do not appear to account for risk8. Parcels within current easements have a total
biological score of 51,361, with an expected biological score of 34,953. Thus, nearly 70 percent
of the biological score on currently conserved parcels was likely to remain in the absence of the
easements. In contrast, a similar calculation and comparison of efficient conservation plans
generated with benefit-loss-cost targeting have ratios less than 50 percent – meaning that less
than 50 percent of the conserved biological score is likely to remain without easements. This
illustrates the potential gains in biological value that exist given careful targeting. Species
specific targeting, as discussed above, is another policy consideration given the level of observed
impacts to species like mule deer.
Finally, I identify, according to each of the targeting strategies discussed above, the 10
next best parcels to put under a conservation easement as a hypothetical policy recommendation
(see Figure 25). I identified these parcels by ranking all parcels according to each targeting
strategy and selecting the top ten in each ranking (regardless of the budget constraint). Parcels
that I target as next-best purchases are relatively small and widely distributed. Though smaller
than already-purchased easements, they provide high biological value.
As discussed previously, there are many reasons for theoretically optimal purchases to
differ from purchases that are practically feasible. Landowners and resource managers have
differing information regarding the other’s willingness to pay and willingness to accept for a
purchase of development rights. Landowners face uncertainty surrounding future residential
markets, costs associated with converting agricultural land to residential properties, and when the
optimal conversion time (t*) might be. These dynamics are among many reasons for what I
8 Easement buyers may have additional, parcel-specific information on relative risk, such as landowner plans to retire, that are not captured in my model.
69
identify as optimal and what is practically feasible to differ; however, identifying the
theoretically next best candidate parcels for conservation provides a starting point for resource
managers. Importantly, identifying next best candidate parcels, as shown in Figure 25, illustrates
the limitations to mitigating on private lands the loss of habitat on public lands.
Figure 25. Ten "best" currently unconserved parcels for conservation according to each targeting approach. Existing conservation easements designated by cross-hatch pattern.
6 Conclusion
This study examines the tradeoff between biological value and economic value on
agricultural lands in Sublette County, Wyoming. Extensive natural gas production on mule deer
and pronghorn crucial winter range, increasing fragmentation of mule deer and pronghorn
70
migration corridors, and sage grouse habitat degradation attributable to many changes on the
landscape have prompted compensatory mitigation efforts to benefit these impacted species. I
explore ways in which agencies charged with overseeing these mitigation efforts can best
allocate their limited funds to achieve the most biological benefit at least cost, which I define as
the opportunity cost of future residential development on privately-owned agricultural lands. My
key findings are: opportunities for conservation are limited by a lack of critical habitat on private
agricultural lands, and the risk of development for a given parcel must be a consideration when
targeting conservation easements.
To explore which parcels provide the most biological value at least cost, I use a method
for estimating the opportunity cost of future residential development that enables me to correct
for limitations of assessors’ data. Specifically, I estimate a binary logit model using parcel-level
physical and geographical characteristics commonly included in hedonic models to generate a
propensity score – the probability of observing a given parcel in residential use – for each parcel
in the county. Using a matching procedure, I then impute the potential residential value of each
agricultural parcel. Hence, I am able to consider this disaggregated future use value as the
amount of compensation needed to purchase a conservation easement on a given agricultural
parcel. The sum of the assessed agricultural value and the median residential value resulting
from matching is the economic “score” (i.e., total value) of each parcel.
I combine economic scores with a simple biological score that I estimate by weighting
parcels’ distance to crucial mule deer and pronghorn winter range and their acreage of winter
range, stopover range, and movement corridors. I combine these individual species scores with
the acreage of sage grouse habitat (a five kilometer buffer around currently occupied leks) to
form a total biological score for each parcel. Constructing production possibilities frontiers from
71
these total and disaggregated biological scores and economic scores shows that the economic-
biological tradeoff is particularly steep for mule deer and pronghorn. This is likely because
much of these species’ critical habitat is on public land rather than privately-owned agricultural
lands, which limits the amount of mitigation that can be achieved using private land conservation
easements.
To avoid underestimating the total biological score available across the entire landscape,
I calculate an expected biological score for each parcel that accounts for both the parcel’s
biological value and the likelihood of observing that parcel in a residential use. This identifies
parcels that currently provide habitat and that are unlikely to become residential in the future,
and thus will continue to provide biological value regardless of whether they are placed in a
conservation easement. Even incorporating expected biological values, the tradeoff between
economic and biological value remains steep for mule deer and pronghorn.
I identify multiple approaches for targeting conservation easement purchases: benefit,
benefit-cost, benefit-loss, and benefit-loss-cost targeting. I find that benefit-loss-cost targeting
performs the best of each strategy for conserving the most biological value at the least cost.
Two key observations can be made from this analysis: cost of conservation easements and risk of
future development should inform conservation easement purchases. In most cases, targeting
conservation easements towards large, expensive parcels does not generate as much biological
value as targeting smaller, less expensive parcels with high biological value (see Figure 25).
However, a limitation of my study is that I do not consider the possibility of dividing parcels for
either conservation use (placing a conservation easement on only part of the agricultural parcel)
or for residential use (subdividing only part of the agricultural parcel). It may be possible for
72
resource managers to conserve only key portions of larger parcels, thereby capturing more
biological value at a lower cost when compared to the cost of purchasing the entire parcel.
It is critical that policy makers consider the likelihood of losing agricultural lands to
development when making conservation easement purchases; I found that many agricultural
parcels have both a low probability of converting to residential use and a high biological value.
Conservation easements would be better placed on parcels that offer a high biological value and
are at higher risk for residential development (accounting for conservation costs).
Finally, I compare what is optimal according to my analysis with conservation easement
purchases that have already been made. While these purchases are not efficient according to my
model, I cannot account for many factors influencing purchase decision. I am unable, for
example, to observe landowners’ individual preferences and therefore their minimum willingness
to accept. There is likely a self-selection bias in the application process to obtain conservation
easements through the available funding mechanisms. Landowners with strong preferences for
preserving agricultural production or the environment are the most likely to seek out easement
agreements and may be willing to accept less than the opportunity cost. Because conservation
easements are strictly voluntary, I cannot draw any certain conclusions regarding the actual cost
of the easement (I can only estimate what the cost should be, according to land value theory) and
I cannot predict which lands are actually likely to be placed under an easement. Each of these
may substantially affect the efficient frontier and the optimal targeting strategy. Though less
efficient than what is theoretically possible, considering these challenges of targeting
conservation easement purchases, existing easements appear to be relatively good purchases.
Future research could resolve several such limitations of my study, such as correcting for
spatial autocorrelation in the econometric model, and calculating a more sophisticated biological
73
score. A more sophisticated biological score should reflect contiguity between parcels’ locations
and reward habitat agglomeration (Parkhurst et al. 2002). Additional econometric analysis of
highly-valued parcels that appear to be statistical outliers should also be considered. Some
parcels that had extremely high assessed values could be indicating a structural change in the
market for land in Sublette County or could indicate an omitted variable problem.
My research suggests that while some compensatory mitigation is achievable, it is limited
by the lack of critical habitat on private agricultural lands – only approximately 20 – 30 percent
of sage grouse and mule deer habitat exists on private lands, and while nearly half of pronghorn
migration habitat exists on private lands, all but 14 percent of pronghorn winter range exists on
public lands. Thus, limiting the impacts to species will require some level of habitat protection
on public lands. In conclusion, my results suggest it may be more effective for land managers to
invest less in conservation easement purchases – except where carefully targeted to consider
parcels’ risk of development – and instead focus efforts and funding on habitat reclamation and
minimizing on-site impacts.
74
7 Literature Cited
Bastian, C.T., D.M. McLeod, M.J. Germino, W.A. Reiners, and B.J. Blasko. 2002. “Environmental amenities and agricultural land values: a hedonic model using geographic information systems data.” Ecological Economics 40(3): 337-349.
Beckman, J. P., K. Murray, R. G. Seidler, and J. Berger. 2012. “Human-mediated shifts in animal habitat use: sequential changes in pronghorn use of natural gas field in Greater Yellowstone.” Biological Conservation doi:10.1016/j.biocon.2012.01.003.
Benson, M. 2005. Supplement: Recent Questions on Conservation Easements. William D. Ruckelshaus Institute of Environment and Natural Resources, University of Wyoming.
Berger, J. 2004. “The last mile: how to sustain long-distance migration in mammals.” Conservation Biology 18: 320–331.
Berger, J., S.L. Cain, and K.M. Berger. 2006. “Connecting the dots: an invariant migration corridor links the Holocene to the present.” Biology Letters 22: 528–531.
Bergstrom, J. and R. Ready. 2009. “What Have We Learned from Over 20 Years of Farmland
Amenity Valuation Research in North America?” Review of Agricultural Economics 31(1): 21-49.
Bureau of Land Management and State of Wyoming. 2012. Memorandum of Agreement. Accessed September 2011.
Bureau of Land Management. 2012. Pinedale Anticline Wildlife Monitoring and Mitigation Plan 2012 Update. Pinedale, Wyoming, February.
Bureau of Land Management [BLM]. 2000. Record of Decision: Environmental Impact Statement for the Pinedale Anticline Natural Gas Field Exploration and Development Project. Pinedale Field Office, Wyoming.
Bureau of Land Management [BLM]. 2006. Record of Decision: Environmental Impact
Statement for the Jonah Infill Drilling Project. Pinedale Field Office, Wyoming. Bureau of Land Management [BLM]. 2008. Record of Decision: Final Supplemental
Environmental Impact Statement for the Pinedale Anticline Oil and Gas Exploration and Development Project. Pinedale Field Office, Wyoming.
Caliendo, M. and S. Kopeinig. 2008. “Some practical guidance for the implementation of
propensity score matching.” Journal of Economic Surveys 22(1): 31-72.
Connelly, J., M. Schroeder, A. Sands, and C. Braun. 2004. “Guidelines to manage sage grouse populations and their habitats.” Wildlife Society Bulletin 28(4): 967-985.
ESRI (Environmental Systems Resource Institute). 2010. ArcMap 10. ESRI, Redlands, California.
75
Germino, M., W. Reiners, B. Blasko, D. McLeod, and C. Bastian. 2001. “Estimating visual properties of Rocky Mountain landscapes using GIS.” Landscape and Urban Planning 53(1-4): 71-82.
Gilbert, M. M., and A. D. Chalfoun. 2011. “Energy development affects populations of sagebrush songbirds in Wyoming.” Journal of Wildlife Management 75:816-824.
Guo, Shenyang, and M. Fraser. 2010. Propensity Score Analysis: Statistical Methods and Applications. Thousand Oaks, California: SAGE Publications, Inc.
Hartman, T. 2011. Sage-Grouse: 2011 Monitoring Update. Wyoming Game and Fish Department: Pinedale Anticline Project Office.
Holloran, M. 2005. “Greater Sage-Grouse (Centrocercus urophasianus) population response to natural gas field development in western Wyoming.” PhD dissertation, University of Wyoming.
Holloran, M. J., R. C. Kaiser, and W. A. Hubert. 2010. “Yearling greater sage-grouse response to energy development in Wyoming.” Journal of Wildlife Management 74:65–72.
Ingelfinger, F., and S. Anderson. 2004. “Passerine response to roads associated with natural gas extraction in a sagebrush steppe habitat.” Western North American Naturalist 64:385–395.
Johnson, G. and M. Holloran. 2010. Greater sage-grouse and wind energy development a review
of the issues. Western EcoSystems Technolgy Inc. Cheyenne, WY.
Jonah Interagency Office. 2012. About the Jonah Infill Drilling Project. Accessed September 2011.
Jonah Interagency Office. 2009. Jonah Interagency Mitigation and Reclamation Office Strategic Plan. Accessed September 2011.
Kail, C. et al. 2005. SocioEconomic Task Group Report & Monitoring Plan. Pinedale, WY: Pinedale Anticline Working Group.
Lichtenstein, M. and C. Montgomery. 2003. “Biodiversity and Timber in the Coast Range of Oregon: Inside the Production Possibility Frontier.” Land Economics 79(1): 56-73.
Lyon, A. G., and S. H. Anderson. 2003. “Potential gas development impacts on sage-grouse nest initiation and movement.” Wildlife Society Bulletin 31:486-491. Ma, S. and S. Swinton. 2012. “Hedonic valuation of farmland using sale prices versus appraised
values.” Land Economics 88(1): 1-15. Macsalka, N. 2011. “Assessing the conflict between wind energy development and sage-grouse
conservation in Wyoming: An application using a spatial explicit wind development model.” MS thesis, University of Wyoming.
76
Nielson and Sawyer 2010. Pronghorn monitoring in the Pinedale Anticline Project Area: 2010 Annual Report. Western Ecosystems Technology, Inc. Laramie, Wyoming.
Parkhurst, G., J. Shogren, C. Bastian, P. Kivi, J. Donner, and R. Smith. 2002. “Agglomeration bonus: an incentive mechanism to reunite fragmented habitat for biodiversity conservation.” Ecological Economics 41: 305-328.
Perrigo, A., and J. Iversen. 2002. Conservation Easements: An Introductory Review for Wyoming. William D. Ruckelshaus Institute of Environment and Natural Resources, University of Wyoming.
Pinedale Anticline Project Office. 2012. What’s Happening at JIO-PAPO? Accessed September 2011.
Placeways, LLC. 2012. CommunityViz® Scenario 360
Plantinga, A.J., R. Alig, and H. Cheng. 2001. “The supply of land for conservation uses: evidence from the conservation reserve program.” Resources, Conservation, and Recycling 31: 199-215.
Plantinga, A., R. Lubowski, and R. Stavins. 2002. “The effects of potential land development on agricultural land prices.” Journal of Urban Economics 52: 561-581.
Polasky, S., et al. 2008. “Where to put things? Spatial land management to sustain biodiversity and economic returns.” Biological Conservation 141: 1505-1524.
Polasky, S., E. Nelson, E. Lonsdorf, P. Fackler, and A. Starfield. 2005. “Conserving species in a working landscape: land use with biological and economic objectives.” Ecological Applications 15(4): 1387-1401.
Riis, J. 2009. Pronghorn Passage. National Geographic Wild Chronicles. Accessed March 2010.
Rosen, S. 1974. “Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition. Journal of Political Economy 82(1): 34-55.
Rosenbaum, P. and D. Rubin. 1983. “The central role of the propensity score in observational studies for causal effects.” Biometrika 70(1): 41-55.
Rowland, M., M. Wisdom, L. Suring, and C. Meinke. 2006. “Greater sage-grouse as an umbrella species for sagebrush-associated vertebrates.” Biological Conservation 129: 323-335.
Sawyer, H. and B. Rudd. 2005. “Pronghorn Roadway Crossings: A Review of Available Information and Potential Options.” Western Ecosystems Technology, Inc. Cheyenne, Wyoming.
Sawyer, H., F. Lindzey, and D. McWhirter. 2005. “Mule deer and pronghorn migration in Western Wyoming.” Wildlife Society Bulletin 33(4): 1266-1273.
77
Sawyer, H., R. M. Nielson, F. G. Lindzey, and L. L. McDonald. 2006. “Winter habitat selection of mule deer before and during development of a natural gas field.” Journal of Wildlife Management 70:396–403.
Sawyer, H., R. Nielson, and D. Strickland. 2009a. Sublette Mule Deer Study: Phase II Final Report. Western EcoSystems Technology, Inc. Cheyenne, Wyoming.
Sawyer, H., M. J. Kauffman, and R. M. Nielson. 2009b. “Influence of well pad activity on the winter habitat selection patterns of mule deer.” Journal of Wildlife Management 73: 1052–1061.
Sawyer, H., M. Kauffman, R. Nielson, and J. Horne. 2009c. “Identifying and prioritizing ungulate migration routes for landscape-level conservation.” Ecological Applications 19(8): 2016-2025.
Sawyer, H. and M. Kauffman. 2011. “Stopover ecology of a migratory ungulate.” Journal of Animal Ecology 80: 1078-1087.
Sawyer, H. and R. Nielson. 2011. Mule Deer Monitoring in the Pinedale Anticline Project Area: 2011 Annual Report. Western EcoSystems Technology, Inc. Cheyenne, Wyoming.
Spahr, R. and M. Sunderman. 1998. “Property Tax Inequities on Ranch and Farm Properties.” Land Economics 74(3): 372-389.
Stilwell, D. and F. Crockett. 2006. Reasonable Foreseeable Development Scenarios for Oil and Gas Activities on Federal Land in the Pinedale Field Office, Wyoming. Cheyenne, WY: U.S. Department of Interior, Bureau of Land Management.
Taylor, D.T., and S. Lieske. 2002. Population change in Wyoming, 1990-2000. William D. Ruckelshaus Institute of Environment and Natural Resources, University of Wyoming.
Taylor, D.T., and S. Lieske. 2002. Second Home Growth in Wyoming, 1990-2000. William D. Ruckelshaus Institute of Environment and Natural Resources, University of Wyoming.
U.S. Bureau of Economic Analysis. 2012. Regional Data: Gross Domestic Product by State. Accessed June 2012.
U.S. Census Bureau. 2010. 2010 Census Data. Accessed July 2012.
U.S. Department of Energy, Energy Information Administration. 2012. AEO2012 Early Release Overview. Accessed May 2012.
U.S. Department of Energy, Secretary of Energy Advisory Board. 2011. The SEAB Shale Gas Production Subcommittee Ninety-Day Report – August 11, 2011. Accessed June 2012.
U.S. Department of Interior, Fish and Wildlife Service. 2008. Greater Sage-Grouse Interim Status Update. Accessed April 2012.
U.S. Geological Survey. 2012. National Elevation Dataset. Accessed March 2012.
U.S. Geological Survey. 2012. National Land Cover Dataset. Accessed March 2012.
78
Walker, B., D. Naugle, and K. Doherty. 2007. “Greater sage-grouse population response to energy development and habitat loss.” Journal of Wildlife Management 71(8): 2644-2654.
Walston, L. J., B. L. Cantwell, and J. R. Krummel. 2009. “Quantifying spatiotemporal changes in a sagebrush ecosystem in relation to energy development.” Ecography 32:943-952.
Wyoming Department of Revenue. 2012. Property Tax. Accessed May 2012.
Wyoming Department of Transportation. 2007. Wyoming’s Comprehensive Report on Traffic Crashes 2007. Accessed May 2012.